I
DISSOLVED GAS ANALYSIS FOR FAULT
DIAGNOSIS OF POWER TRANSFORMERS
A thesis
Submitted
For the award of the Degree of
DOCTOR OF PHILOSOPHY
By
SUKHBIR SINGH
Electrical Engineering Department
National Institute of Technology Kurukshetra
Kurukshetra, India
Registration No.: Ph. D. -1127E/2006
2012
II
DISSOLVED GAS ANALYSIS FOR FAULT
DIAGNOSIS OF POWER TRANSFORMERS
A thesis
Submitted
For the award of the Degree of
DOCTOR OF PHILOSOPHY
By
Registration No.: Ph. D. -1167E/2006
SUKHBIR SINGH
Under the supervision of
Dr. DHEERAJ JOSHI
Department of Electrical Engineering
National Institute of Technology Kurukshetra
Kurukshetra, India
2012
I
DECLARATION
I hereby declare that;
i) The present work in thesis is my own and it has been carried out under
the able guidance of my supervisor.
ii) The work has not been submitted to any other institute/university for
the award of degree or diploma.
iii) I have adhered to the institute guidelines/regulations while preparing
the contents of the thesis.
iv) Wherever I have used material (reports, data, book, journal, text, and
figure) from other source, I have given due credits by citing in the text
of the thesis with the details in the references.
Date: .2012 (SUKHBIR SINGH)
Registration No.: Ph. D.-1167E/2006
II
CERTIFICATE
Certified that the thesis entitled, “DISSOLVED GAS ANALYSIS FOR
FAULT DIAGNOSIS OF POWER TRANSFORMERS”, submitted by
Mr. SUKHBIR SINGH is in fulfillment of the requirements for the award
of the degree of DOCTOR OF PHILOSOPHY in ELECTRICAL
ENGINEERING from NATIONAL INSTITUTE OF TECHNOLOGY,
KURUKSHETRA. The candidate has worked under my supervision. The
work presented in this thesis has not been submitted for the award of any
other degree/diploma.
(Dr. DHEERAJ JOSHI)
Assistant Professor
Electrical Engineering Department
National Institute of Technology
Kurukshetra-136119, India
III
ACKNOWLDGEMENTS
Like all faithful researchers, it is my prime duty to acknowledge all the books, research
journal, periodicals and other sources of knowledge from whom, I have sought
knowledge and inspiration for the completion of this present stupendous research work.
I am beholden beyond words in gratitude to Dr. Dheeraj Joshi, Assistant Professor, EED,
NIT, Kurekshetra for his valuable guidance and cooperation to complete this thesis work.
I am also grateful to Dr. M. N. Bandyopadhyay, Director, NIT, Calicut, erstwhile
Director, NIT, Kurukshetra for initiating me to this area of research.
For extending critical counsels and periodic encouragement, my heartfelt thanks are due
to Prof. Akhilesh Swarup and Prof. K. S. Sandhu, EED, NIT, Kurukshetra. I am highly
thankful to Prof. (Mrs.) L. Deewan, HOD, EED, NIT, Kurukshetra for facilitating the
process of my research work at various stages.
I am sincerely grateful to my children Er. Monika Bainsala, Er. Sukhanshu Singh and
Master Kaushal Singh, who always have been constant inspiration to continue my
research work even during the movements of my drudgery, tiresomeness and nervous
break-down. For my dear wife, I extend my sincere gratitude for her untiring patience and
unlimited fore bearance without any complaint.
Special thanks to M. Tech. students of NIT Kurukshetra, namely Er. Mithlesh Verma, Er.
Kavita Jaiswal, Er. Reshmi PR. for rendering a number of jobs during the course of
pursuit of this research work for the last one year.
For providing precious learning, testing, materials and reports, etc for this thesis work, I
am thankful to Messers R & D NTPC HQ. NOIDA, CPRI Testing Laboratory NOIDA
and TIFAC CORE, NIT Hamirpur.
I also acknowledge help of those, whom I could not inadvertently mention here, but who
contributed to the completion of this work directly or indirectly.
To all my critics and debunkers, I also thank for giving courage and strength to tolerate
their bitter criticism.
Last but not the least, I acknowledge the help and encouragement given to me by Prof. S.
D. Sharma, former Professor & Dean Dept. of English, G B Pant Nagar University, Pant
Nagar for bringing about remarkable language competence.
In the end I bow my head in reverence to ALMIGHTY GOD, who has given me such a
fine life to live by.
(SUKHBIR SINGH)
IV
ABSTRACT
Power transformers are designed to distribute and transmit electrical power. In case of
failure, replacement cost of major power transformer becomes very high in the power
industry. Catastrophic failure not only results in unavailability of equipment and revenue
loss, but also the damage of peripheral equipment, environmental damage, and alternative
power resources. There could be two causes, include external factors such as short circuit,
overloading of system lightning strikes, system switching operations and internal factors
such as overheating, insulation deterioration, loss of winding clamping, presence of solid
contaminations, moisture, oxygen, and winding resonance.
Various off-line and on-line oil tests for fault diagnoses of power transformers are
conducted periodically as per expert recommendation. A number of standards have
evolved over the time on transformer loading and power transformer fault diagnosis to
minimize unplanned outages. Among all the existing diagnostic techniques, dissolved gas
analysis (DGA) is an important technique to assess incipient fault in a power transformer
by analyzing ratios or/and contents of specific dissolved gas concentrations arising from
the deterioration of transformer liquid/solid insulations.
DGA interpretations use methods such as key gases, Roger‘s ratio, Dornenburg‘s ratio,
CIGRE‘s standards, IEC/IEEE standards, Duval triangle, etc. Still, there is no universally
accepted interpretation technique for DGA. Further its acceptance becomes need based
depending upon economy, environment or geographical conditions of a continent,
country, authority, organization, utility, etc. An expert opinion is always sought whenever
any damaging trend is occurring as a result of DGA fault diagnostic techniques.
Dissolved gas analysis for fault diagnosis of power transformers is done periodically
including other tests as per the expert‘s recommendation. DGA analysis shows the
distribution of key gases dissolved in oil. Generation of these key gases are due to thermal
and electrical faults, and provide an indication of type of faults, presence and severity of
the faults in the power transformer, However, sometimes other offline or invasive tests
such as sweep frequency response analysis, thermography, UV-tests, furan analysis,
degree of polymerization are required to confirm the exact fault. A huge amount of cost is
involved in fault diagnostic. Therefore, it becomes equally important to use better
diagnostic method or a combination of methods to detect probable failure and its root
V
causes. So fault diagnostic capability also includes a list of faults. Preparation for
necessary corrective action can be planned in advance and executed quickly.
Several methods for DGA interpretations in power transformers in service are available in
IEC Standard 60599, the IEEE Guide C-57-104 as well as in published reviews on the
topic. The Duval triangle method is also available in IEEE standards. However, usesr are
not familiar or not quite at ease with the use of triangular graphical analysis. An effort has
been made in this thesis work to simulate Duval triangle method 1 to Duval triangle
method 5 for dissolved gas analysis for fault diagnosis of power transformers. Duval
triangle1 method uses three hydrocarbon gases only (methane, ethylene and acetylene).
These gases are decomposition of transformer oil corresponding to the increasing levels
of energy required to generate gases under thermal and electrical stresses. In addition to
six mentioned faults (PD, D1, D2, T1, T2 and T3) an intermediate zone DT is available in
Duval triangle method 1 as compared to the other methods. More than 60 fault reports
collected from various authorities and DGA test laboratories, which are analyzed for fault
interpretations by graphical triangular coordinates, calculated manually and the same are
cross verified with existing faults.
All the faults reports were simulated in Duval triangle method 1. These are again cross
verified with existing faults and manual faults analysis with simulated faults analysis. It
revealed that DGA fault analysis in power transformer with Duval triangle methods
provide better overall diagnosis results at ease with improved efficiency and accuracy.
Artificial intelligence methods can be used for enhancing resolution of faults
identification by fusion on information from multiple tests and available transformer
information. Depending on the type of diagnostic technique and available knowledge, the
development of shallow models of artificial intelligence technique required are as
follows:
A. Diagnostic rules (used standard)
B. Data driven model (FUZZY,ANN)
C. Case history data (for expert opinion)
To increase the efficiency and accuracy in DGA analysis with Duval triangle method for
fault diagnosis of power transformer is achieved in this thesis work with application of
artificial intelligence techniques (Fuzzy logic and artificial neural network).
Fuzzy logic (FL) is applied for all the Duval triangle methods. In this process, all the zone
boundaries of the faults in each Duval triangle are arranged and faults are assigned at
VI
each vertex. Ranges for three input percentages and outputs are assigned to create
membership function. Fuzzy rules using fuzzy membership function are decided with
range of input percentages and output as fault using crisp boundaries. Fuzzy logic
approaches of Duval triangle1 faults diagnosis on reported faults showed better crisp
faults and reduced the uncertainties between the faults, and the increasing accuracy.
Artificial neural network (ANN) is also applied for all Duval triangles. Polygon
coordinates of all Duval triangles are developed to large string of input percentages for
triangular coordinates. Some test inputs are also assigned. A feed-forward, back
propagation (BP) ANN for this thesis work has been assigned with input and output
patterns. For ANN configuration and training, the Duval triangles are applied with tan-
sigmoid (TANSIG) transfer function with Levenberg- Marquardt (TRAINLM) training.
The result of DGA fault diagnosis of power transformer shows that superior learning
capability of ANN gives quicker analysis as compared to other faults diagnosis method.
VII
CONTENTS
Page No.
DECLARATION (I)
CERTIFICATE (II)
ACKNOWLDGEMENTS (III)
ABSTRACT (V)
CONTENTS (XI)
LIST OF FIGURES (XV)
LIST OF TABLES (XX)
CHAPTER 1
INTRODUCTION 1-23
1.1 INTRODUCTION 1
1.2 TRANSFORMER 1
1.3 TYPES OF TRANSFORMERS 2
1.4 CHOICE OF VOLTAGES 5
1.5 SPECIFICATIONS WITH DESIGN PRACTICES OF POWER TRANSFORMER 6
1.5.1 Basic Objectives of Specifications 8
1.5.1.1 Specification for power transformers 9
1.5.2 Design Practices of Power Transformers 9
1.5.2.1 Basic physical technologies 11
1.5.2.2 Requirements 11
1.6 POWER LOSSES 12
1.6.1 Technical Loss 12
1.6.2 Commercial Loss 12
1.7 POWER SYSTEM RELATED LOSSES IN POWER TRANSFORMERS 13
1.7.1 Distribution Transformers (DTs) not Located at Load Center on the
Secondary Distribution System 13
1.7.2 Over-Rated Distribution Transformers and hence their Under-Utilization 14
1.8 SELECTION OF POWER TRANSFORMERS 14
1.9 POWER TRANSFORMER FAILURES AND PROBLEMS 15
1. 10 NECESSITY OF FAULT DIAGNOSIS 17
1.11 BENEFITS OF FAULT DIAGNOSIS 19
1.12 ORGANISATION OF THESIS 20
VIII
1.13AUTHOUR‘S CONTRIBUTION 22
CHAPTER 2
LITERATURE SURVEY 24-37
2.1 INTRODUCTION 24
2.2 FAILURE SURVEY ON TRANSFORMERS 24
2.3 FAULT DIAGNOSIS OF POWER TRANSFORMERS 25
2.4 TESTING OF SOLID INSULATIONS 26
2.4.1 Degree of Polymerization 27
2.4.2 Furan Contents 27
2.5 TESTING OF TRANSFORMER INSULATION OIL FOR FAULT DIAGNOSIS IN
POWER TRANSFORMERS 28
2.5.1 Break Down Voltage (BDV) 28
2.5.2 Acidity 28
2.5.3 Flash Point 29
2.5.4 Loss Factor 29
2.5.5 Viscosity 29
2.5.6 Tan-delta, Resistivity and Dielectric Constant 29
2.6 DGA FAULT DIAGNOSIS OF POWER TRANSFORMERS 30
2.6.1 On-line/Off-site DGA 30
2.6.2 DGA Duval Triangle Methods 30
2.6.3 Standards and Guides for Fault Interpretations 31
2.7 ARTIFICIAL INTELLIGENCE (AI) APPLICATION
IN DGA OF POWER TRANSFORMERS 34
2.8 ADVANCEMENTS IN FAULT DIAGNOSIS METHODS OF POWER
TRANSFORMERS 34
2.8.1 Thermography 34
2.8.2 Sweep Frequency Response Analysis (SFRA/FRA) 35
2.8.3 Thermal UV- Response Test on Transformer Oil 35
2.8.4 Fault Detection (winding deformation) by Transformer Tank Vibration 35
2.8.5 Alternative Insulating Fluids and their DGA 36
IX
CHAPTER 3
TRANSFORMER INSULATING MATERIALS AND OIL TESTING 38-57
3.1 INTRODUCTION 38
3.2 INSULATING MATERIALS 38
3.3 PROPERTIES OF INSULATING MATERIAL 38
3.4 TRANSFORMER PAPER INSULATION 39
3.5 TRANSFORMER INSULATING OIL 40
3.6 INSULATING OIL QUALITY 41
3.7 FACTORS AFFECTING INSULATING OIL QUALITY 42
3.7.1 Effect of Oxygen on Oil 42
3.7.2 Moisture in Oil 42
3.7.3 Absorption of Moisture by Insulating Materials 44
3.7.4 Absorption of Nitrogen by Oil 44
3.8 IMPORTANT OIL SPECIFICATIONS 44
3.9 TESTING OF TRANSFORMER OIL 46
3.9.1 Non Traditional Tests on Transformer Oil 48
3.9.2 Traditional Tests for Fault Diagnosis 50
3.10 IFT-NN RELATIONSHIP 51
3.11 THERMAL DEGRADATION TEST ON PURE TRANSFORMER OIL 52
3.12 THERMAL DEGRADATION TEST ON OIL WITH PRESS-BOARD AND
KRAFT PAPER 53
3.13 DEGRADATION TESTS DUE TO ACCUMULATION
OF WATER CONTENT 55
3.14 SUMMARY 57
CHAPTER 4
DISILVED GAS ANALYSIS OF POWER TRANSFORMERS 58-92
4.1 INTRODUCTION 58
4.2 DECOMPOSITION OF OIL 58
4.3 FORMATION OF GASES IN TRANSFORMER OIL 59
4.4 SOLUBILITY OF GASES IN TRANSFORMER OIL 60
4.5 DECOMPOSITION OF CELLULOSE INSULATION 61
X
4.6 OTHER SOURCES OF GASES IN TRANSFORMERS 61
4.7 DISSOLVED GAS ANALYSIS 63
4.8 FAULTS IN PAPER AND FAULTS IN TRANSFORMER OIL 64
4.9 SAMPLING OF LIQUID INSULATION AND GAS EXTRACTION FROM OIL 64
4.9.1 Sampling Procedures 64
4.9.2 Taking Gas Samples for Gas Chromatograph Analysis 65
4.10 INFLUENCE OF STORAGE OF OIL SAMPLES ON DGA 65
4.10.1 Effect of Air Bubbles 66
4.10.2 Effect of Temperature 66
4.10.3 Effect of Light 67
4.11 TAKING OIL SAMPLES FOR DGA 68
4.12 TECHNIQUES FOR SAMPLING INSULATING LIQUIDS 68
4.13 GAS EXTRACTION FROM SAMPLE 70
4.14 NEW DGA SYSTEM 71
4.15 VACUUM DEGASSING 71
4.16 MOBILE GAS CHROMATOGRAPHY 72
4.17 ONLINE DGA SYSTEM 73
4.18 GASES ANALYZED BY DGA 75
4.19 FAULTS DETECTION BY DGA 76
4.19.1 Partial discharges (PD) of the corona-type (PD) 76
4.19.2 Discharges of low energy (D1) 76
4.19.3 Discharges of high energy (D2) 76
4.19.4 Thermal faults of temperatures < 300°C (T1) 76
4.19.5 Thermal faults of temperatures between 300 °C and 700°C (T2) 77
4.19.6 Thermal faults of temperatures > 700 °C (T3) 77
4.20 STANDARDS AND CODES FOR DGA INTERPRETATIONS ADOPTED IN
INDIA 77
4.21 OTHER TRANSFORMER CHEMISTRY SERVICES
DURING INTERPRETATION 78
4.22 DIFFERENT DGA METHODS 81
4.23 KEY GASES METHOD 81
4.23.1 Evaluation of Possible Fault Type by the Key Gas Method 81
4.24 EVALUATION OF TYPES OF POSSIBLE FAULT BY ANALYSIS
OF THE SEPARATE COMBUSTIBLE GASES GENERATED 83
XI
4.25 THE DORNENBURG‘S RATIO METHOD 83
4.26 THE MODIFIED ROGER‘S RATIO METHOD 86
4.27 DUVAL TRIANGLE METHOD 87
4.28 IEC 60599 88
4.28.1 EC 60599 Ratio-Fault Diagnosis 89
4.29.2 Rates of Gas Increase 89
4.29.3 IEC 60599 Typical Gas Levels 89
4.30 NOMOGRAPH METHOD 90
4.31 DENKYOKEN METHOD 90
4.32 CIGRE‘S METHOD 91
4.33 NBR7274 METHOD 91
4.34 IS 10593:2006 METHOD 91
4.35 COMPARATIVE ANALYSIS OF DIFFERENT DGA METHODS (KEY GAS,
DORNENBURG‘S, ROGER‘S RATIO & DUVAL TRIANGLE METHOD) 92
4.36 SUMMARY 92
CHAPTER 5
FAULT DIAGNOSIS WITH DUVAL TRIANGLE 93-134
5.1 INTRODUCTION 94
5.2 FAULT DIAGNOSIS BY DUVAL TRIANGLE METHOD 94
5.3 IMPROVED VERSION OF DUVAL TRIANGLE 1 METHOD 95
5.4 THE INFLUENCE OF LABORATORY ACCURACY ON FAULT DIAGNOSIS 96
5.4.1 Influence of Factors Affecting Transformer Reliability 98
5.4.2 Calculation of Own Typical Values 98
5.4.3 Adjustment of Pre-failure and Alarm Values 99
5.5 APPLICATION OF DUVAL TRIANGLE METHOD 1 (DTM 1) 100
5.5.1 Graphical Fault Interpretation 100
5.5.2 Simulated Fault Interpretation with DTM1 101
5.6 DUVAL TRIANGLE 2 FOR LOAD TAP CHANGERS OF THE OIL TYPE 104
5.6.1 Implementation and Results from Duval Triangle 2 (DTM 2) for OLTCs 106
5.7 THE DUVAL TRIANGLES 3 FOR NON-MINERAL OILS 108
5.7.1 Implementation and results from Duval triangle methods 3 (DTMs 3) for Non-
mineral oils 110
5.7.1.1 Silicone oil 110
XII
5.7.1.2 Midel oil 112
5.7.1.3 FR 3 oil 114
5.7.1.4. Bio-Temp oil 115
5.8 DUVAL TRIANGLES 4 AND 5 FOR LOW-TEMPERATURE FAULTS IN
TRANSFORMERS 117
5.8.1 Duval Triangle 4 for Low-Energy Gases 117
5.8.2 Implementation and Results for Duval Triangle Method 4 118
5.8.3 Duval Triangle 5 (DTM5) for Low-Temperature Faults 120
5.8.4 Implementation and Results for Duval Triangles Method 5 122
5.9 SUMMARY 134
CHAPTER 6
APPLICATION OF ARTIFICIAL INTELLIGENCE IN DUVAL TRIANGLE
METHODS 135-202
6.1 INTRODUCTION 135
6.2 ARTIFICIAL INTELLIGENCE TECHNIQUES 135
6.3 FUZZY LOGIC TECHNIQUE (FLT) 135
6.4 FUZZY LOGIC APPROACH IN DTM1 138
6.5 FUZZY LOGIC APPROACH FOR DTM2 144
6.6 FUZZY LOGIC APPROACH FOR DTM 3 147
6.6.1 Fuzzy Logic Approach for DTM3, Silicone oil 147
6.6.2 Fuzzy Logic Approach for DTM3, Midel Oil 153
6.6.3 Fuzzy Logic Approach for DTM3, FR3 Oil 157
6.6.4 Fuzzy Logic Approach for DTM3, Bio-Temp Oil 162
6.7 FUZZY LIGIC APPROACH FOR DTM4, LOW ENERGY FAULTS 167
6.8 FUZZY LIGIC APPROACH FOR DTM5, LOW TEMPERATURE FAULTS 170
6.9 ARTIFICIAL NEURAL NETWORK TECHNIQUES (ANNT) 175
6.10 ANN APPROACH FOR DTM1 178
6.11 ANN APPROACH FOR DTM2 181
6.12 ANN APPROACH FOR DTM3 184
6.12.1 ANN Approach for DTM3, Silicone Oil 184
6.12.2 ANN Approach for DTM3, Midel Oil 187
6.12.3 ANN Approach for DTM3, FR3 Oil 190
6.12.4 ANN Approach for DTM3, BioTemp Oil 195
XIII
6.13 ANN APPROACH FOR DTM4 196
6.14 ANN APPROACH FOR DTM5 198
6.15 SUMMARY 202
CHAPTER 7
CONCLUSIONS AND FUTURE SCOPE OF WORK 203
7.1 CONCLSIONS 203
7.2 FUTURE SCOPE OF WORK 203
REFERENCES 204-213
APPENDIX--A 214-219
APPENDIX-B 220-243
PUBLICATIONS FROM THESIS WORK 244-245
XIV
LIST OF FIGURES
Figure 2.1: Causes of transformer failure 24
Figure 2.2: Percentage of defective components of transformers 25
Figure 2.3: Percentage of failure detection methods 25
Figure 3.1: Dielectric Strength Vs Water in Oil 43
Figure 3.2: Water Solubility Vs Temperature 43
Figure 4.1: Gas Generation Chart 60
Figure 4.2: Relative Solubility (Y-Axis) as Function of Temperature OC (X-Axis) 61
Figure 4.3: Influence of air bubbles of different volume on dga results 66
Figure 4.4: Stray gassing effect at 90 0C 67
Figure 4.5: Influence of sunlight and darkness on concentration of
H2 and CO2 on new oil 67
Figure 4.6: Influence of sunlight and darkness on concentration of
H2 and CO2 on aged oil. 68
Figure 4.7: Toepler pump for laboratory use 70
Figure 4.8: Schematic view of the gas extraction nit 71
Figure 4.9: Photo of the gas extraction unit 71
Figure 4.10: Mobile GC 73
Figure 4.11: Online DGA system 74
Figure 4.12 Key gases (Over Heating, Oil) 82
Figure 4.13: Key gases (Over Heating, Paper) 82
Figure 4.14: Key gases (Partial Discharge) 82
Figure 4.15: Key gases (Arcing) 82
Figure 4.16 : Dornenburg‘s ratio method flow chart 86
Figure 4.17: Modified Roger‘s ratio method Flow-chart 87
Figure 4.18: Duval Triangle 1 87
Figure 4.19: Duval Triangle indicating regions of different faults 88
Figure 4.20: Logarithmic nomograph 90
Figure 5.1: Duval triangle 93
Figure 5.2: Duval triangle 1 graphical plot 95
Figure 5.3: Actual cases of faults visually inspected in different transformers 96
XV
Figure 5.4: Uncertainty on diagnoses for many cases 97
Figure 5.5: Diagnosis uncertainties to laboratory analytical accuracy
of +- 15, 30, 50 and 75% PFS, in % 97
Figure 5.6: Probability of Having a Failure-Related Event(PFS, in %) as a
Function of C2H2 Concentration in Service of in ppm and Norm in % 99
Figure 5.7: Graphical analyses on DTM 1 101
Figure 5.8: Flow-chart of DTM 1 103
Figure 5.9: Duval Triangle 1 simulation analysis 104
Figure 5.10: The Duval Triangle 2 for LTCs of Oil Type Transformers 105
Figure 5.11: Inspected Cases of Faults in the Duval Triangle 2 for OLTCs of oil-type
Transformers 105
Figure 5.12: Fault analysis by DTM 2 for OLTC 108
Figure 5.13: Duval Triangle 3 for Non-mineral oils with their zone
boundaries in colors 109
Figure 5.14: Fault analysis results of DTM 3 for non-mineral oil (Silicon oil) 112
Figure 5.15: Fault analysis results by DTM3 (Midel oil) 113
Figure 5.16: Fault analysis results by DTM3 (FR3 oil) 115
Figure 5.17: Fault analysis result by DTM 3 (BioTemp oil) 117
Figure 5.18: Duval triangle 4 transformers filled with mineral oil 118
Figure 5.19: Fault analysis results by DTM 4 for low-intensity faults 120
Figure 5.20: Duval triangle 5 for transformers filled with mineral oil
(Low-Temperature Gases) 120
Figure 5.21: Fault analysis results by DTM 5 for low temperature fault 123
Figure 5.22: Generalized flow-chart for complete solution by Duval Triangles 124
Figure 6.1: Basic fuzzy logic control 136
Figure 6.2: Fuzzy logic approach in DTMs 138
Figure 6.3(a): %CH4 membership-function 139
Figure 6.3(b): %C2H4 member functions 140
Figure 6.3(c): % C2H2 member functions 140
Figure 6.4: membership function of DTM1 faults 142
Figure 6.5: FIS analysis of fuzzy rules for DTM 1 144
Figure 6.6 (a): DTM 2 Membership functions of CH4 145
Figure 6.6 (b): DTM 2 Membership functions of C2H4 145
Figure 6.6 (c): DTM 2 Membership functions of C2H2 145
XVI
Figure 6.7: OLTC‘s faults output ranges 146
Figure 6.8: FIS analysis of fuzzy rules for DTM2 147
Figure 6.9(a): Membership functions for CH4 DTM3, Silicone oil 148
Figure 6.9(b): Membership functions for C2H4 DTM3, Silicone oil 148
Figure 6.9(c): Membership function for C2H2 DTM3, Silicone oil 148
Figure 6.10: Output membership functions of silicone oil 150
Figure 6.11: FIS analysis of fuzzy rules for DTM3, Silicone oil 152
Figure 6.12 (a): Fuzzy membership function of CH4 for DTM3, Midel oil 153
Figure 6.12 (b): Fuzzy membership function of C2H4 for DTM3, Midel oil 153
Figure 6.12 (c): Fuzzy membership function of C2H2 for DTM3 Midel oil 153
Figure 6.13: Output membership functions for DTM3, Midel oil 155
Figure 6.14: FIS analysis of fuzzy rules for DTM3, Midel oil 157
Figure 6.15 (a): Fuzzy membership functions for DTM3, FR3 oil 158
Figure 6.15 (b): Fuzzy membership functions for DTM3, FR3 oil 158
Figure 6.15 (c): Fuzzy membership functions for DTM3, FR3 oil 158
Figure 6.16: Output membership functions for DTM3 FR3 oil 160
Figure 6.17: FIS analysis of fuzzy rules for DTM3, FR3 oil 162
Figure 6.18 (a): Input membership function CH4 in DTM3, BioTemp oil 163
Figure 6.18 (b): Input membership function C2H4 in DTM3, Biotemp oil 163
Figure 6.18 (c): Input membership function of C2H2 in DTM3, BioTemp oil 163
Figure 6.19: Fuzzy outputs of DTM3, BioTemp oil 165
Figure 6.20: FIS analysis of fuzzy rules for DTM3, BioTemp oil 166
Figure 6.21 (a): Fuzzy membership functions of H2 for DTM4 167
Figure 6.21 (b): Fuzzy membership functions of CH4for DTM4 168
Figure 6.21 (c): Fuzzy membership functions of C2H6 for DTM4 168
Figure 6.22: FIS analysis of fuzzy rules for DTM4 170
Figure 6.23 (a): Fuzzy membership functions of CH4 in DTM5 171
Figure 6.23 (b): Fuzzy membership functions of C2H4 in DTM5 171
Figure 6.23 (c): Fuzzy membership functions of C2H6 in DTM5 172
Figure 6.24: Fuzzy output membership function for DTM5 173
Figure 6.25: FIS analysis of fuzzy rules for DTM5 174
Figure 6.26: Two-layered feedforward neural Network 175
Figure 6.27: Two-layer BP-ANN for DTM1 180
Figure 6.28: Validation performance chart 181
XVII
Figure 6.29 (a): Mu and Gradient plot 181
Figure 6.29 (b): Target plots 181
Figure 6.30: Two-layer BP-ANN for DTM2 183
Figure 6.31: Validation performance chart 183
Figure 6.32 (a): Mu and Gradient plot 184
Figure 6.32 (b): Target plots 184
Figure 6.33: Two-layer BP-ANN for DTM3, silicone oil 186
Figure 6.34: Validation performance chart Figure 186
Figure 6.35 (a): Mu and Gradient plot 187
Figure 6.35 (b): Target plots 187
Figure 6.36: Two-layer BP-ANN for DTM3, silicone oil 188
Figure 6.37: Validation performance chart 189
Figure 6.38 (a): Mu and Gradient plot 189
Figure 6.38 (b): Target plots 189
Figure 6.39: Two-layer BP-ANN for DTM3, FR3 oil 191
Figure 6.40: Validation performance chart 192
Figure 6.41 (a): Mu and Gradient plot 192
Figure 6.41 (b): Target plots 192
Figure 6.42: Two-layer BP-ANN for DTM3, BioTemp oil 194
Figure 6.43: Validation performance chart 195
Figure 6.44 (a): Mu and Gradient plot 195
Figure 6.44 (b): Target plots 195
Figure 6.45: Two-layer BP-ANN for DTM4 197
Figure 6.46: Validation performance chart 197
Figure 6.47 (a): Mu and Gradient plot 198
Figure 6.47 (b): Target plots 198
Figure 6.48: Two-layer BP-ANN for DTM5 199
Figure 6.49: Validation performance chart 200
Figure 6.50 (a): Mu and Gradient plot 200
Figure 6.50 (b): Target plots 200
XVIII
LIST OF TABLES
Table 1.1: Material used for Transformer Manufacturing 7
Table 1.2: General Information & Technical Specification 10
Table 1.3: Calculation of AT&C Losses 13
Table 1.4: Main causes of failure of transformer in service 15
Table 1.5: Components causing failure in service 16
Table 1.6: Typical causes of transformer failures 16
Table 3.1: Electric Grade Papers Specification as per IEC 554-3-5 1984 39
Table 3.2: Characteristics Requirement of IS, IEC, and BS Specification for uninhibited
transformer oil 44
Table 3.3: Standards and Testing 47
Table 3.4: Relationship between IFT vs NN 51
Table 3.5: Permissible limits as per IS 1866-2000 52
Table 3.6: Initial samples parameters 52
Table 3.7: Thermal degradation test on pure transformer oil 53
Table 3.8: Thermal degradation test on oil along with Press board and Kraft paper 54
Table 3.9: Influence of water content 55
Table 3.10: Significance of test 56
Table 4.1: Recommended safe value of various gases in [µl/L(ppm)] 62
Table 4.2: Permissible concentration of dissolved gases on ageing 62
Table 4.3: Key Gases generated by particular fault 63
Table 4.4: Dissolved gases in transformer oil 75
Table 4.5: California State University Sacremento Guidelines for Combustible Gas 78
Table 4.6: California State University Sacremento Guidelines for Combustible Gas 79
Table 4.7: CEGB/ANSI/IEEE Guide for gas concentration limits in ppm V/V 79
Table 4.8: Other International Gas Concentration Limits in ppm V/V 80
Table 4.9: SERC – Japan limiting values in ppm V/V 80
Table 4.10: EDF – France Generator/Transmission transformers without OLTCs 81
Table 4.11: Concentration limits of dissolved gas 84
Table 4.12: Dorenburg‘s ratios for key gases 84
Table 4.13: Roger‘s ratios for key gases 86
XIX
Table 4.14: IEC 60599 Ratio-Fault diagnosis 89
Table 4.15: IEC 60599 Typical gas levels 89
Table 4.16: Relative Performance of Different Methods 92
Table 5.1: Uncertainty on laboratory results 97
Table 5.2: Ranges of 90% typical values for power transformer in ppm 98
Table 5.3: Ranges of 90% typical rate of gas increase for power transformer
in ppm/year 99
Table 5.4: Pre-failure gas concentration values at CIGRE for core- type transformers 100
Table 5.5; Numerical zone boundaries 100
Table 5.6: Triangular coordinates for DTM 1 zones 101
Table 5.7: Identification of Fault Zones in above OLTC figure 105
Table 5.8: Numerical zone boundaries of Duval triangle 2 106
Table 5.9: Numerical zones of DTM 2 106
Table 5.10: Triangular coordinates for Duval triangle 2 zones 107
Table 5.11: Zone Boundaries (% C2H4) for Duval Triangle 3 (non-mineral oils) 110
Table 5.12: Numerical zone boundaries for DTM 3 (Silicone oil) 110
Table 5.13: Triangular coordinates for DTM 3 zones (Silicone Oil) 110
Table 5.14: Numerical zone boundaries for DTM 3 (Midel oil) 112
Table 5.15: Triangular coordinates for DTM 3 zones (Midel Oil) 112
Table 5.16: Numerical zone boundaries for DTM 3 (FR3 oil) 114
Table 5.17: Triangular coordinates for DTM3 zones (FR3 Oil) 114
Table 5.18: Numerical zone boundaries for DTM 3 (BioTemp oil) 115
Table 5.19: Triangular coordinates for DTM3 zones (BioTemp oil) 116
Table 5.20: Duval triangle method 4 fault zones identifications 118
Table 5.21: Duval triangle method 4 numerical zone boundary 118
Table 5.22: Numerical zone boundaries for DTM 4 (low intensity faults) 119
Table 5.23: Triangular coordinates for DTM 4 zones 119
Table 5.24: Duval Triangle method 5 fault zones identifications 121
Table 5.25: Duval Triangle method 5 numerical zone boundaries 121
Table 5.26: Numerical zone boundaries for DTM 5 122
Table 5.27: Triangular coordinates for DTM 5 Zones 122
Table 5.28: Comparative analysis between simulated and graphical fault interpretations in
Duval triangle 1 126
XX
Table 5.29: DTM1 simulated fault diagnosis compared with Roger‘s and Dornenburg‘s
ratio fault interpretations 130
Table 6.1: Membership function ranges for three gas inputs of DTM1 139
Table 6.2: Transformer fault outputs DTM 1 140
Table 6.3: Fuzzy Rules for DTM 1 142
Table 6.4: DTM2 ranges of input gases 144
Table 6.5: Types of faults in OLTCs 145
Table 6.6: Fuzzy rules for OLTC faults 146
Table 6.7: Input Ranges for DTM3, Silicone oil 147
Table 6.8: Fault output membership functions of DTM3, Silicone oil 149
Table 6.9: Fuzzy Rules for DTM3, Silicone oil 150
Table 6.10: Ranges of input for DTM3, Midel oil 152
Table 6.11: Output ranges for DTM3, Midel oil 154
Table 6.12: Fuzzy rules for DTM3, Midel oil 155
Table 6.13: Input ranges for DTM3, FR3 oil 157
Table 6.14: Fault output for DTM3, FR3 oil 158
Table 6.15: Fuzzy rules for DTM3, FR3 oil 160
Table 6.16: Input ranges for DTM3, BioTemp oil 162
Table 6.17: Fault output for DTM3 BioTemp 163
Table 6.18: Fuzzy rules for DTM3, BioTemp oil 165
Table 6.19: Input ranges for DTM4 167
Table 6.20: DTM4 output ranges 168
Table 6.21: Fuzzy rules for DTM4 169
Table 6.22: Input ranges of DTM5 170
Table 6.23: Fault outputs in DTM5 172
Table 6.24: Fuzzy rules for DTM5 173
Table 6.25: Parameters for ANN Duval triangle method 1 179
Table 6.26: Parameters for ANN Duval triangle method 2 182
Table 6.27: Parameters for ANN Duval triangle method 3, silicone oil 185
Table 6.28: Parameters for ANN DTM 3, Midel oil 187
Table 6.29: Parameters for ANN Duval triangle method 3, FR3 oil 191
Table 6.30: Parameters for ANN Duval triangle method 3, BioTemp oil 194
Table 6.31: Parameters for ANN Duval triangle method 4 196
Table 6.32: Parameters for ANN Duval triangle method 5 199
XXI
Table 6.33: Comparative results of DTM1 simulation and Fuzzy logic DTM1 201
Table 6.34: Comparison of results simulation and ANN 202
Table 6.35: Comparison of FL and ANN results 203
Table A-1: Gas concentrations 214
Table A-2: Gas generating rate 214
Table A-3: Gas concentrations 215
Table A-4: H2 generating rate 215
Table A-5: Roger‘s Ratio fault diagnosis 215
Table A-6: Dornenburg‘s ratio fault diagnosis 216
Table A-7: Gas concentration 217
Table A-8: Gas concentrations 219
XXII
ABRIBEATIONS AND NOTATIONS
AEG- Allgemeine Elektrizitäts-Gesellschaft (AEG) (literally General Electricity
Company)
AT&C- Aggregate Technical & Commercial
ASTM- American Society for Testing and Materials
ANSI- American National Standards Institute
AI- Artificial Intelligence
ANN- Artificial Neural Network
AWNN- Fuzzy Wavelet Neural Network
BBC- Brown, Boveri & Cie (BBC), a Swiss group of electrical engineering
BP- Back Propagation
BP-ALM- BP-Adaptive Learning and Momentum
BS- Bureau of Standards
CT- Current Transformer
CEA- Central Electricity Authority of India
CO- Carbon Monoxide
CO2- Carbon Dioxide
CH4- Methane
C2H4- Ethylene
C2H6- Ethane
C2H2- Acetylene
CEBG- Central Electricity Generating Board in UK
CMAC- Cerebellar Model Articulation Controller
CIGRE- International Council on Large Electric Systems in France
CRGO- Cold Rolled Grain Oriented
CCR- Continuous Casting Rolling
DT- Distribution Transformer
DT- Thermal Electrical Fault
DTM- Duval Triangle Method
DGA- Dissolved Gas Analysis
DP- Degree of Polymerization
DF- Dissipation Factor
D1- Discharge of Low Energy
D2- Discharge of High Energy
ES- Expert System
EDF- Energias de Portugal, S.A
ECRA- Electrical Cooperative Research Association
EHV- Extra High Voltage
ET- Evidence Theory
eV- Electron Volt
FCM- Fuzzy C Mean
FIT- Flame Ionization Detector
GCA- Grey Clustering Analysis
GRNN- Generalized Regression Neural Network
GC- Gas Chromatography
XXIII
GM- Grey Model
GA- Genetic Algorithm
GDR- Generalized Delta Rule
HV- High Voltage
H2O- Hydrogen Oxide
H2- Hydrogen
H- Hydrogen Molecule
HED- High Energy Discharge
IEC- International Electrotechniical Commission
IEEE- Institute of Electrical & Electronics Engineering
IS- Indian Standard
ISO- Organization of International Standard
IFT- Interfacial Tension
KV- Kilo Volt KOH- Potassium Hydroxide
LV- Low Voltage LT- Low Tension
LTC - Load Tap Changers
LS- Least Square
LED- Low Energy Discharge
MVA- Mega Volt Ampere
MU- Mega Unit
µL/L- Micro Litre / Litre
M*- Molecule
NN- Acid Number
NBR- Brazil Standards: Associação Brasileira de Normas Técnicas
(ABNT/NBR)
NEMA- National Electrical Manufacturers Association, U.S.
OH- Hydroxyl
O2- Oxygen
OLTC- On-line Tap Changer
PT- Potential Transformer
PF- Power Factor
PD- Partial Discharge
ppm- Parts Per Million
PSO- Particle Swarm Optimization
PFGC- Pre- Failure Gases Theory
PFS- Pre- Failure values
RF- Radio Frequency
FIS- Fuzzy Inference Systems
RS- Rough Set
RBFNN- Radial Basis Fuzzy Neural Network
RTR- Refined Three Ratio
SECR- Japan
SSC- Sum Square Error
SVM- Support Vector Machine
SVMG- Support Vector Machine with Genetic Algorithm
T&C- Technical & Commercial
UHV- Ultra High Voltage
TPMO-
XXIV
T1- Thermal Fault <300o
T2- Thermal Fault 300-7000c
T3- Thermal Fault >7000c
TCG- Total Combustible Gases
TDCG- Total Dissolved Combustible Gases
UAT- Unit Auxiliary Transformers TCD- Thermal Conductivity Detector
WNN- Wavelet Neural Network
WNNGA- WNN Genetic Algorithm
Special Abbreviations Used for fault comparative
tables
Cir. C Circulating Current
NF No Fault
Ind. Individual
PD Partial Discharge
N Normal
SG Gases
IG Individual Gases
S OK Sample OK
NA No Abnormality
DZD Denominator Zero Diagnosis
DZD B DZD Bypassed
MF Mixed Fault
DUVAL Duval Triangle
SRQ Sample req. quarterly
T Testing Required
Reco. Recommendation
AD Arc Discharge
Comments Comments
NI Not Identified
NG Not Given
TF Thermal Fault
DA Discharge Arcing
NF No Faults
1
CHAPTER-1
INTRODUCTION
1.1 INTRODUCTION
Electrical energy is the backbone of the economy of a country. The trend towards a
deregulated global electrical energy market has put the electric utilities under severe
stress to reduce operating cost, increase in the availability of the generation, transmission
and distribution sub-assemblies and to improve the supply from power system. Hence the
function of power system is to supply electric energy to its user with an acceptable degree
of reliability, quality and economy. Power transformer plays a critical role in efficient and
effective transmission, distribution and utilization of electrical power. It makes possible
electric generation at the most economical generator voltage, power transfer at the
economical transmission voltage and power utilization at the most suitable voltage for the
particular utilization device. Power transformer is the heart of the power system network.
Power system reliability can improve when transformer incipient faults are detected and
eliminated before it leads to transformer failures.
1.4 TRANSFORMER
Transformer is an electrical device used to transfer an alternating current or voltage from
one electric circuit to another by means of electromagnetic induction. The simplest type
of transformer consists of two coils of wire, electrically insulated from each other and
arranged in much a way that a change in the current in one coil (the primary) will produce
a change in voltage in the other (the secondary). In many transformers, the coils are
wound on a core made of a material with high magnetic permeability. This intensifies the
magnetic field induced by the current in the primary, increasing the transformer's
efficiency. Neglecting power losses, which are made small by careful design, the ratio of
primary voltage to secondary voltage is the same as the ratio of the number of turns in the
primary coil to the number of turns in the secondary coil. The primary and secondary
currents are in inverse proportion to the number of turns in the coils. The primary and
secondary impedances are in the same ratio as the squares of the numbers of turns in the
primary and secondary coils. Transformers are frequently classified according to their
uses; the details of construction depend on the intended application. Power transformers
are generally used to transmit power at a constant frequency.
2
1.5 TYPES OF TRANSFORMERS
Power Transformers are the main types of transformers, and among these the most
representative group is the double-wound power transformers, which are used on power
lines. Power transformers raise the voltage of the current generated at electric power
plants from 10–15 kilovolts (kV) up to 220–750 kV, thus making possible the
transmission of electric power over aerial transmission lines for several thousand
kilometers. At locations where electric power is consumed, the high voltage is stepped
down to low voltages (220 V, 380 V, and so on) by other power transformers. The
repeated transformation of the electric power requires a large number of power
transformers, and therefore the total power of the transformers in the system is several
times that of the sources and power users. High-capacity power transformers have an
efficiency of 98–99 percent. Their windings are generally made of copper, and their
magnetic circuits of cold-rolled sheets of electrical steel 0.5–0.35 mm thick that have high
magnetic permeability and low hysteresis and eddy-current losses. The magnetic circuit
and windings of power transformers are usually mounted in a tank filled with mineral oil,
which is used to insulate and cool the windings. Such transformers—called oil-filled
transformers—are generally located in the open, which require improved insulation of the
leads and hermetic sealing of the tank. Transformers without oil cooling are called dry
transformers. For better heat dissipation, power transformers are equipped with a tubular
radiator around which the air or, in some cases, water circulates. In lightning-proof
transformers the windings are designed so as to avoid the appearance of dangerous
voltages across the insulation. Sometimes two or three transformers are connected in
series. In many cases transformers with load tap changing are used. Among the dry
transformers, low-power transformers with a large number of secondary windings (multi-
winding transformers) make up a broad class.
In addition to power transformers, there are many types designed to measure high
voltages and currents (for example, instrument transformers, voltage transformers, and
current transformers), to reduce the level of interference on wire communication lines
(negative booster transformers), to convert a sine voltage to a pulse voltage (peaking
transformers), to convert current and voltage pulses (peak transformers), to isolate an AC
current component, and to decouple and match parts of an electric circuit.
Power transformers are used on primary transmission lines for transmission and
distribution of relatively large amounts of power. The power transformer should
withstand stresses occurring during system disturbances and should be designed to satisfy
3
the requirements of the power system. They should show good correlation between actual
and optimum values of cost ratio. Power transformers are designed for maximum
efficiency at near or full load. Thus power transformers are simplest equipments in power
systems, which required great reliability for its life tenure of 30-40 years. Power
transformers are designed to be operated in very convenient manner and easy to maintain
or repair without removing from the service. Therefore, it is required to design
transformers with rare troubles.
Various types of power transformers are as follows:
i. Generator transformer
Transformer connected to generators usually supply power to a transmission line which
runs from generating plant to a bulk power load centre located at considerable distance
away, some of the general requirements of a transformers are as follows:
a) No voltage- regulating winding, because the voltage is regulated by the field of the
generator. Generally, it is provided with off-circuit tap changer with a small variation in
voltages. However in certain cases OLTCs are required for reactive power control of the
system.
b) Fairly, uniform load, the units of high efficiency in particular are kept loaded to the
maximum capacity.
c) Generally provided with unit cooler/ water heat exchanger cooling arrangement.
d) It must take over excitation during operation.
e) Special care to avoid any hot spot in the conducting metallic parts in its vicinity; the
epoxy bonded completely transposed conductor (CTC) is preferred for LV winding to
minimize eddy current losses and provide greater short-circuit strength.
ii. Unit auxiliary transformers (UAT)
These are step-down transformer with primary is connected directly to generator output.
The secondary voltage is of the order of 6.6 KV for supplying to various auxiliary
equipments in the generating station. The voltage rating should be specified in such a way
that the secondary voltages are about 5% more than the nominal voltages to compensate
the transformer regulation. The capacity of these transformers is based on the auxiliary
load of the particular unit with the 10% margin.
iii. Station transformers
These transformers are required to supply auxiliary equipments during setting-up a
generating station and subsequently during starting-up operation. The rating of these
transformers is small, and their winding is connected to a high voltage transmission line.
4
The split secondary winding is often employed to have economical circuit breaker ratings.
These transformers are extensively used in thermal power plants to feed power to boiler
feeding pump motors, induced drought fans etc. other characteristics are similar to
network transformers. The capacity of station transformers should meet the following
requirements:
a) All running auxiliaries of station services at full station load,
b) The running auxiliaries of one of the unit bus at full load in the event of outages of
one UAT at time in the power station and
c) The auxiliaries starting of other unit which may be taken as 40% of the normal
auxiliary load.
iv. Inter-connected transformers
Normally, autotransformers are used to interconnect two-grids/systems operating at two
different voltages. They are located in the transmission system between generating
transformers and receiving transformers and they reduce transmission voltage to sub-
transmission voltage level. They are characterized by wide tapping range and additional
tertiary windings, which may be loaded or unloaded. Synchronous condenser or the shunt
reactors are connected to tertiary windings for reactive power compensation. In case of
loaded secondary conductor area and proper supporting arrangement are required for
withstanding S.C. forces under asymmetrical fault conditions
v. Receiving station transformer
These transformers are normally step-down transformer reducing transmission / sub-
transmission voltage to primary feeder level (33kV). Some of these may be directly
supplying power to industries. Loads on these transformers vary under wider limits and
the losses are expensive. If these transformer, when they close to residential area, it is
desirable to have lower noise level.
vi. Distribution transformer
Distribution transformers are used for distribution of power transmission lines and
network for local consumption. These are used for domestic / industrial uses. A lower
value of no-load loss is desirable to improve all-day efficiency. They are designed with
smaller values of iron losses relative to the full load losses. They show that the actual cost
ratio are high and mass ratio are lower than optimum. A great variety of transformers fall
into this category due to many arrangements and connections. Load on these transformers
varies widely and frequently overloaded.
5
vii. Earthing or grounding transformers
These are used to get a neutral point that facilitates the grounding and detection of earth
faults in an undergrounded part of a network (e.g. delta connected system). The sole duty
of a grounding transformer is to pass ground current during earth fault. The windings are
usually connected in inter-connected star (zigzag). These transformers (zigzag) have the
advantage that that are not affected by D.C magnetisation. The addition functions are:
a) Holding the neutral shift within limits,
b) Permitting the circulation of unbalanced load in the neutral,
c) To limit the current during line to earth fault,
d) To earth the system.
viii. Phase shifting transformers
These are used to control the power flow over the transmission line by varying the phase
angle between the input and output voltage transformer. Through a proper tap change, the
output voltage can be made to either lead or lag the input voltage. The amount of phase
shift required directly affects the rating and size of the transformer.
1.3 CHOICE OF VOLTAGES
Electric supply on its way from generating station to consumer end transformed several
(5-8) times. This energy demands not only needs best design and manufacturing
processes of the transformers to reduce the losses but also needs high voltage handling
capacities. Transformers having nearly constant ratio at all loads are generally used for
transmission and distribution system. These are said power transformers. The loads and
various components of total power are operated on their suitable voltage levels.
Therefore, high voltage transformer voltage choices for techno- economic considerations
are as follow:
i. Generation transformers
Generation transformer are connected to the generators to supply to transmission lines,
which is running between generating plant to the power load centre situated at a place
away. The generation voltages are (2.2, 3.3, 6.6, 11, 22, 33KV).
ii. Transmission transformers
Transmission needs step-up transformers with primary connected to directly to generator
output and secondary connected to transmission lines. These step-up transmission
voltages are (66, 132, 220, 440, 765KV).
6
iii. Distribution transformers
Distribution systems need step-down transformers to reduce transmission/sub-
transmission voltages to primary feeder levels. Further distribution transformers are used
for distribution of power from transmission lines network for local consumption. These
voltages are in the ranges of (3.3, 6.6, 11, 22, 33KV).
iv. Utilization transformers
Utilization generators are generally less than 1000 volts, 433 volts and 250 volts for
power and lighting. These voltages are used for domestic and industrial applications.
1.5 SPECIFICATIONS WITH DESIGN PRACTICES OF POWER
TRANSFORMER
Reliability of the power transformer is linked with failure rate and mode of failure.
Various parts of the transformers viz magnetic circuit, insulation system, termination,
cooling equipments etc, are designed as per specification details. Thus the magnetic
circuit is designed based on continuous and short time power frequency over voltages and
its suitability depends on the design, material selection, the quality of manufacture,
careful handling etc. For the average transformer used in a power station, the conductor
used for the windings consists of paper insulated copper bar or wire. In assembling the
transformer, great care is taken to ensure windings are well insulated both from the iron
core and from each other. The assembled transformer has its winding and iron core
assembly usually contained in a tank and immersed in transformer oil. The oil is used for
further insulating purposes plus the removal of heat from the windings. The assembly of
the windings on the core allows gaps to enhance the oil circulation around the windings.
The tank is constructed with fins or tubes to allow better circulation of the oil and to
provide a greater surface area for contact with the cooling air.
Active materials of the power transformers are conductors and magnetic materials serving
to create the most favorable condition for the electromagnetic process-taking place in the
equipment. In power transformers, transformer action takes place through core and coils
(active materials). Transformer components are to be compared to human body like core-
bone, winding-heart, insulation-flesh, oil-blood, and tank-skin. All the parts of the
transformer must function properly to its best performance. Thus major materials to
fabricate the parts of the transformer must be given proper attention and quality checks.
Some of the major materials in Table 1.1 are as follows;
7
Table 1.1: Material used for transformer manufacturing
Sl. No. Item Materials
1 Magnetic
core
CRGO of various types
Amorphous iron(Advantageous for distribution
transformers)
2 Conductor Copper and aluminum in the form of wire, strip or sheet
Better quality of conducting materials are produced by
continuous casting rolling (CCR process), which can
produce strip/sheet or wire in longer length, which avoids
welded or brazed joints. Completely transposed coils are
essential in certain cases
Replacing strip of LV winding
3 Coolant Transformer oil
4 Insulating
materials
Cellulose pres-board, raft paper, synthetic, self-bonding
paper
Classification of insulating materials: The recommended
thermal classes are 90, 105, 120, 130, 155, 180 and above
(in oC
) [Y, A, E, B, F, H & C]
The different properties of materials are divided into three groups:
i. Technology properties
Power transformer should possess the reasonably good mechanical properties which are
high castability, forgeability, weldability and machineability. All the materials used for
making the parts of the power transformers are required to have high wear resistance and
cold workability. Since transformer temperature always goes above the couples of
hundreds of degrees for the solid and liquid materials used in it. Thus, the material used
also posses the high degree of hot strength and hot rigidity.
ii. Physical properties
Physical properties of transformer includes mechanical properties, thermal properties,
electrical properties, magnetic properties and optical properties. The thermal properties a
power transformer should have high melting and boiling points, low thermal expansions,
high thermal capacity and good thermal conductivity. As a part of mechanical properties
the materials used will have good elasticity and plasticity, high tensile and compressive
8
strength, and very high hardness, low brittleness to withstand any kind of external
damages. Among the electrical properties conductors used must have high conductivity.
All the insulating materials used in transformer manufacturing are required to possess
high dielectric strength and good trekking resistance. Core in the transformers occupies
the maximum area and weight, thus core material would have high permeability, coercive
strength but it should have low resonance power. In optical properties materials should
hold the good colour, gloss and transparency wherever such materials are required and
applicable.
iii. Chemical properties
Since the temperature of the transformers rapidly changes and also exposed to
environmental changes thus it becomes desirable to have high degree of chemical
properties those are high corrosive resistance, high acid resistance, alkali resistance and
scaling resistance for the entire working life of the transformers.
Since core in the transformer is the biggest portion, therefore basic requirements of the
material used in transformer core which are to be taken care off includes:
i. Low hysteresis loss
ii. Low eddy-current loss
iii. High saturation flux density
iv. Long life
v. Easy to manufacture
vi. Low specific apparent –power input for no-load current
vii. Low magnetostriction for low noise level
viii. High grade insulation
ix. Good mechanical processing properties
x. Suitability for high flux densities
1.5.1 Basic Objectives of Specifications
Technical specification is a document which lays down characteristic of a power
transformer or a service as a level of quality, performance, safety or dimensions. It may
include or deal exclusively with the terminology, symbols, testing or test methods,
packaging, marking or labeling requirements. An important benefit of specification is
improvement and suitability of equipment and services for their intended purposes.
Transformers specifications shall necessary include design review / procedure and
process for manufacture and testing requirements. National and international standards
are performance required and prescribe the testing needed to confirm that standards. In
9
power transformers mainly two specifications ie standard specifications and consumer‘s
specifications decide the specifications of power transformer as follows:
i. To achieve optimum overall economy and efficiency.
ii. To establish proper co-ordination with other related equipments.
iii. To reduce cost, human efforts and conserve essential materials.
iv. To increase productivity, eliminate waste and reduce inventories.
v. To achieve countywide interchangeability of parts.
vi. To define quality and to recommend the mode of its operation.
Thus adhering to specifications at all levels becomes important. Some of the importance
considerations are as follows:
a) Increase efficiency, productivity and quality approved by National and International
standards.
b) Promote better applications and understandings of principles and techniques.
1.5.1.1 Specification for power transformers
The specification forms the basis on which the manufacture designs, manufactures and
test a power transformer. With the growth in size of power system and complexities of
modern system, the adequacy of transformer specifications plays an important role on its
reliability. A good specification must properly consider the effect of application, system
operating condition, environmental conditions site details, service conditions, termination
details etc so that the power transformer is designed for the conditions under which it has
to operate. Thus the adequacy of specification of power transformers particularly that of
high voltage and larger rating assumes tremendous importance and an inadequate
specification can lead to a deficient design and poor reliability. Mostly the following
standards are accepted in India:
i. IEC: 60404 – 8 – 7:1998
ii. IEC: 60076 part 1-5
iii. Indian standard IS 3024:2002 and
iv. IS: 2026 part 1-5
v. IEEE C57-12.00
1.5.2 Design Practices of Power Transformers
The basic materials of construction of oil immersed transformers have not changed
dramatically in the last century, but there are large improvements in quality and their
performance qualities. Designers are taking full advantage of these developments in
power transformers with the state of the art of design methodologies. However, there
10
shall not be any relaxation in quality margins. On the operational side also, rapidly
development power system have identified more areas for concerns such as:
i. Very fast fronted transients specially for GIS application
ii. Part winding resonance
iii. Static electrification
iv. Geomagnetic disturbances
v. Quality and purity of supply
Despite the presence of international and national standards defining the performance and
the testing requirement, the actual service requirement are not reflected in these standards.
A power transformer designed to meet the international and national specifications may
not be completely appropriate in the actual operating conditions. The power transformers
specifications should include
i. Environmental condition
ii. Anticipated system condition, voltage and frequency profile
iii. Loadability limits under varying frequency trend
iv. External short-circuit scenario and
v. Asthetics.
Power transformer design involves many interdependent parameters. It becomes very
difficult to optimize a power transformer design. Most power transformer designers use
an electrical model that allows them to approximate a transformer design. The
preliminary approximate design will be evaluated, and then adjusted as needed to
achieve desired objectives. Table 1.2 shows the technical specification of a power
transformer. Here core size is optimized for minimum material price as prime objective in
design.
Table 1.2: General Information & Technical Specification
Input voltage 3 x 35000/20230V, star sine wave
Transformer output voltage 3 x 690/400V, star
Line output current 3 x 1340A, continuous operating mode
Frequency 50Hz
Average oil temperature 55°C
Max. temperature rise and/or
max. Cu-winding losses at
75°C
25°K
18000W => 1.125%
11
Short-circuit voltage 6.5%
°Short-circuit voltage 6.5%
Max. core losses 3200W => 0.2%
Max. no-load current 1.3%
Test Voltage at 50Hz, 1
minute
Primary 85kV, outside Secondary 4kV,
inside Steel & Core Assembly M5, annealed, strips for alternated stacking
(4x45°+3x90° per shape), "round" cross
section with 8 steps
Core Size Optimized for minimal material price for:
Cu_Price/Fe_Price = 2 with Cu-winding
1.5.2.1 Basic physical technologies
To make reliable designs for large power and high voltage/extra high voltage, there must
be good theoretical understanding of physics. The basic physical areas like to:
i. Dielectrics control the complete insulation structures (for EHV / UHV );
ii. Acoustics achieve lower sound level;
iii. Load losses, thermal performances control of losses at load and the temperatures in
the transformer created by those losses;
iv. Short circuit strength make mechanical stable design due to all kinds of external
short circuits; and
v. Transients and fields ( using sophisticated 2D and 3D tools) to simulate lighting
impulse phenomena and electromagnetic fields from high currents.
1.5.2.2 Requirements
The main design requirements for a good transformer are:
i. Primary inductance should be as high as possible to reduce no-load current to a small
value;
ii. A high coefficient of coupling between primary and secondary winding to ensure
small voltage regulation under load;
iii. Core losses as low as possible by;
a) Choice of suitable core materials.
b) Lamination of core.
iv. Winding losses to be as low as possible by;
a) Use of high conductivity winding copper wire of adequate cross sectional area.
b) Keeping the number of turns in the primary winding as small.
12
v. Size of transformer to be as small as possible by;
a) Running it at as high a temperature as allowable.
b) Use of conductor and core materials of good space factor.
vi. Efficiency to be high by keeping losses low (No.3 and 4);
vii. Weight to be low;
viii. Reliability to be high by;
a) Choice of materials and types of construction.
b) Use of adequate space factors.
c) Adequate sealing against moisture.
d) Robust mechanical construction.
1.6 POWER LOSSES
Whenever the voltage applied to induction motor varied from rated voltage, its
performance is affected. Within permissible voltage of +/- 6% of the affect practice, the
supply voltage varies by more than 10% in many distribution systems. A reduced voltage
in case of induction motor results in higher currents drawn for the same output. For a
voltage drop of 10%, the full load current drawn by the induction motors increase by
about 10% to 15%. The starting torque decreases by nearly 19% and the line losses in the
distributor increases by about 20%. As the bulk load of rural areas and small scale
industrial areas consist of induction motors, the line losses in the concerned distribution
systems may even touch 20%. Power losses in transmission and distribution systems are
as follows:
1.6.1 Technical Loss
Every element in a power System (a line or a Power transformer etc) offers resistance to
power flow and thus consumes some energy while performing the duty expected of it.
The cumulative energy consumed by all these elements is classified as Technical Loss.
1.6.2 Commercial Loss
Losses occur on account of non-performing and underperforming meters, wrong
applications of multiplying factors, defects in CT & PT circuitry, meters not read,
pilferage by manipulating or by passing of meters, theft by direct tapping etc. These are
all due to non-metering of actual consumption and are called commercial losses. The
total of Technical and Commercial losses are termed as T&D loss. It is unfortunate that
in addition to the above, there is also a loss in revenue due to non-realisation of billed
demand. This is in addition to commercial losses and the aggregate of T&D loss and
revenue loss due to non-realisation is termed as AT&C loss (Aggregate Technical and
13
Commercial loss). Therefore, AT&C loss to the utility is the sum total of technical loss,
commercial losses and shortage due to non-realisation of total billed demand. Table 1.3
shows an example of AT&C losses.
Table 1.3: Calculation of AT&C Losses
1. Units Input: 100MU: Revenue collection with reference to billed demand 90%
2. Unit Billed: 70MU: This means out of 70 MU sales realisation of 90% 63%
3. T & D Losses: 39MU: ATC losses 37%
Power distribution systems in developing countries had to face phenomenal and rapid
growth of load in the last two decades. Distribution systems were expanded on adhoc and
haphazard basis keeping minimum investments in view.
Power systems are highly cost intensive and the investments needed to reduce technical
losses by every 1% are too high. Computer aided load flow studies are to be made to
arrive at peak power loss and also to arrive at technical losses based on load-duration
curves. The Transmission system planning is done on a separate footing that is keeping
in view system adequacy, system security, reliability etc for n-1 and n-2. Conditions as
per guidelines are given by CEA. Losses in transmission systems are too small and loss
reduction is not an issue of focus in Transmission system planning.
1.7 POWER SYSTEM RELATED LOSSES IN POWER TRANSFORMERS
Power transformer losses are reduced to minimum by observing proper design practices,
manufacturing processes, observing proper standards at all the stages. Still there are many
losses associated with power transformers during distribution process. Some of the losses
are as follows:
1.7.1 Distribution Transformers (DTs) not Located at Load Center on the Secondary
Distribution System
Often DTs are not located centrally with respect to consumers. Consequently, the farthest
consumers obtain an extremity low voltage even though a reasonably good voltage levels
maintained at the transformers secondaries. This again leads to higher line losses.
Therefore in order to reduce the voltage drop in the line to the farthest consumers, the
distribution transformer should be located at the load center to keep voltage drop within
permissible limits.
14
1.7.2 Over-Rated Distribution Transformers and hence their Under-Utilization
Studies on 11 KV feeder has revealed that often the rating of distribution transformers is
much higher than the maximum KVA demand on the feeder. Over rated transformers
draw an unnecessary high iron loss. In addition to these iron losses in over rated
transformers the capital costs locked up is high.
From the above it is clear that the rating of distribution transformers should be judiciously
selected to keep the losses with the permissible limits.
For an existing distribution system the appropriate capacity of distribution transformer
may be taken as very nearly equal to the maximum KVA demand at good PF (say 0.85)
Such an exercise has been carried out for a number of distribution systems and
transformers with capacity of 25, 63, 100, 160, 315 KVA and standardized for different
with power factors and diversity factors.
1.8 SELECTION OF POWER TRANSFORMERS
For selection of power transformers, Indian standards code of practice for selection,
installation and maintenance of transformers IS:10028 (part 1) selection should be
adhered to. There are several basic considerations which must be included by the system
engineer to select and design the best voltage which should supply power to both present
and future loads economically. Some of these are:
i. Safety It is desirable to achieve high degree of safety of equipment and the personnel
working on power transformers and the adjoining area.
ii. Reliability Power transformers are required to give high degree of reliability even in
stipulated operating conditions during its stipulated service span of at least 40 years
within working specifications. Power transformers are designed and manufactured to
meet the reliability and offer trouble free service.
iii. Maintenance Power transformers regular maintenances are required to ensure
reliable service throughout its life, put back a transformer which has broken-down into
satisfactory working condition in a minimum possible life and also to keep the
maintenance cost minimum. Thus transformer needs routine, corrective, preventive,
proactive and emergency maintenances during its service life.
iv. Voltage regulation There are possibilities of recurring overloads and exceptional
overloads in the working of power transformers. Thus under any overload conditions, it is
desirable from the power transformers to maintain high voltage regulation to keep the
output voltage constant.
15
v. Initial investment Every utility choose to keep its economy at the priority. Thus it
becomes necessary for every manufacturer to keep the cost of a power transformer in
competitive limits and to keep other required specifications within the ranges.
vi. Simplicity in operation Each user develops his own operation needs appropriate to
his requirements. Utilities also would want to operate power transformers with ease for
the entire life of it. Thus simplicity in operation of power transformer becomes prime
important.
vii. The overall system economics Starting from design, manufacturing, installation,
maintenance, repair and overhauls (if needed) involve economy. Though continuity of
power supply is needed from any transformer. The total cost of a transformer can be
considered as the assured annuity for the estimated life of the transformer obtained from a
capital sum invested. Therefore every efforts are made to compromise on the issues and
overall cost of the system tried to keep low.
1.9 POWER TRANSFORMER FAILURES AND PROBLEMS
Transformer failure can occur as a result of different causes and conditions. Generally,
transformer failures can be defined as follows:
i. Forced outage due to damage in service transformer (e.g. winding damage, tap
changer failure).
ii. Trouble that requires removal of the transformer for return to a repair facility, or
which requires extensive field repair (e.g., excessive gas production, high moisture
levels).
Transformer failures can broadly be categorized as electrical, mechanical, or thermal. The
cause of a failure can be internal or external. Table 1.4 lists typical causes of failures.
Table 1.4: Main causes of failure of transformer in service
No. Part Failures
1 36% Design defects
2 28% Manufacturing problems
3 13% Material defects
4 5% Poor Maintenance
5 4% Lightening surges
6 2% Short circuit
In addition to failures in the main tank, failures can also occur in the bushings, in the tap
changers, or in the transformer accessories. Published data based on failure statics of
16
transformers on Indian network is not readily available. A wide range survey by working
group of CIGRE study committee 12, led by Bossi, published [1, 2] the results of survey.
Failure surveys have revealed that failure in the transformers are because their major
component shown in Table 1.5.
Table 1.5: Components causing failure in service
No. Part Failure
1 29% Winding
2 29% Terminals
3 13% Tank & Dielectric fluid
4 13% OLTCs
5 11% Magnetic circuit
6 5% Other accessories
Causes of transformer failures further categories between two major sections as internal
causes and external causes. These are listed in Table 1.6 below.
Table 1.6: Typical causes of transformer failures
Internal Causes External Causes
Loss of Winding Clamping System Switching
Operations/Thermal
Insulation Deterioration Lighting Strikes/Temperature
Oxygen System faults (short circuit)
Overheating System Overload
Moisture
Solid contamination in the Insulating Oil
Partial Discharge
Design & Manufacturing Defects
Winding Resonance
17
The failure pattern of transformers follows a ―bath-tub‖ curve. The first part of the curve
is failure due to infant mortality; the second part of the curve is the constant failure rate;
and the last part of the curve is failure due to old age. In addition to normal ageing, a
transformer may develop a fault that results in faster-than-normal ageing, resulting in a
higher probability of failure.
Power transformers have proven to be reliable in normal operation.. The end of life of a
transformer is typically defined as the loss of mechanical strength of the solid insulation
in the windings. These power transformers are at the last stage of the ―bath-tub‖ curve.
They are expected to have an increasing failure rate in the next few years.
A survey indicates that the main causes (51% of transformer failures in a five year period)
are due to the following problems:
i. Moisture, contamination and ageing which causes the transformers internal dielectric
strength to decrease,
ii. Damage to the winding or decompression of the winding under short Circuit forces, or
iii. Damage to the transformer bushings caused by loss of dielectric strength of the
internal insulation.
An American utility reported four single-phase EHV autotransformer failures due to
transformer winding resonance. All of the failures involved the breakdowns of the no load
tap changers immediately after the transmission system were energized.
1. 10 NECESSITY OF FAULT DIAGNOSIS
Generally speaking, the term ―monitoring‖ describes a basic parameter measurement with
threshold alarms. The term ―diagnostics‖ indicates the addition of sophisticated analysis,
such as an expert system capable of providing an assessment of equipment condition and
suggested actions.
There are a variety of tools available to evaluate the condition of transformers. They can
be separated into traditional diagnostic methods that have seen widespread use for many
years and nontraditional methods that range from methods that are starting to be used to
methods that are still in the research stage. Fault diagnosis of power transformers are
required for the reasons:
i. Early detection of incipient faults:
a) Avoid catastrophic outage
b) Provide basis for economic repair decision
ii. Efficiency and Management:
a) Maintenance management based on measurement and trend analysis
18
b) Ageing process and residual life under control
The ultimate goal of on-time monitoring and diagnostic is continually to and accurately
diagnose the current condition and capability of the equipment, provide an alert ―just in
time‖ the maintenance action as and when required and remaining useful life of the
equipment.
Conditioning monitoring of power transformers by the combination of ―on and off‖ line
methods strongly reduces the risk of severe failures thereby provides a reliable electric
power supply. The most recent development of sensor technology such as electronic
Buchholz relay, OLTC monitoring, based on power consumption of the motor drive, on-
line PD measurement and anhydron with dissolved gas analysis are contributing very
much and gaining field application for early warning.
A comprehensive diagnostic however needs that in failure, results from on-line and off-
line measurement have to be put together for better proven evaluation methods /
procedures which can be achieved by proper coordination between utility and
manufactures.
sufficient money has been invested in the installed power equipment in the electrical
energy systems. Power transformers are an important part of it. They represent a valuable
asset for the utilities, and are expected to be functional over tens of years. In earlier days
it could happen that power transformers with long operational years and considerable
‗weak‘ or a potential source of trouble were ‗simply‘ replaced. Money was ‗no problem‘
and the final energy customer had to pay the bill in the end. Today the paradigm has
changed and considerable efforts are directed into finding new approaches using
transformer monitoring, diagnosis and life assessment.
The stronger constraints that have faced transformer users stem from two major factors:
The deregulation of the energy market and advancement in science and technology.
The deregulation pressures are reflected in efforts for:
i. Costs reduction, which is affecting maintenance, replacement, repair and upgrade
philosophies
ii. Increase power transformer availability, in order to assure power quality and customer
satisfaction
iii. Handling unprecedented power flow patterns in the system
Consequences of the changing situation on the market have been a trend to allow higher
transformer overloads, to work power transformers longer and to take all necessary
additional measures to achieve that. These two facts imply that the transformer user
19
accepts a higher risk in operation but at the same time a high availability is required.
However, the operation of power transformer units is desired to occur with as low as
possible costs and by many utilities there is a continuous reduction in staff and repair
facilities. These aspects bear an intrinsic contradiction.
This cannot be solved without improved means of transformer life assessment and
condition evaluation, monitoring and diagnosis. Another important aspect influencing the
need for condition evaluation is that there is an ageing population of transformers
installed at the industries and in the transmission and distribution networks. Several of
these were installed during a fairly narrow time span, say 25-30 years ago, when large
investments were made in the electric infrastructure. These transformers have been
exposed to various accumulative deterioration stresses and will eventually attain a higher
risk of failure and will approach their expected end of life. This fact may cause the owner
of the population to see an increased number of transformers being in the need of
refurbishment, repair or replacement if not jeopardizing the reliability or availability of
power. The life assessment and condition evaluation techniques thus take on the task of
assessment of the transformer on a broader basis in a holistic view.
1.11 BENEFITS OF FAULT DIAGNOSIS
Some of the benefits of the fault diagnosis of any equipment in general are as follows:
i. Timely filled measurements
ii. Confirmation, on the spot, of the presence of the faults gases in the event of alarms
iii. On time proactive decision making
iv. Reducing unplanned outages
v. More predictable and reliable maintenance schedules
vi. Prevention of catastrophic failure and destruction of peripheral equipment
vii. Reduced maintenance costs
viii. The results provide a quality control feature, limiting the probability of destructive
failures
This leads to improvements in operator safety and quality of supply, limiting the severity
of any damage incurred elimination of consequential repair activities and identifying the
root causes of failures. Information is provided on the plant operating life enabling
business decisions to be made either on plant refurbishment or replacement, safer work
environment. As transformers age, their internal condition degrades, which increases the
risk of failure.
20
Failures are usually triggered by severe conditions, such as lightning strikes, switching
transients, short-circuits, or other incidents. When the transformer is new, it has sufficient
electrical and mechanical strength to withstand unusual system conditions. As
transformers age, their insulation strength can degrade to the point that they cannot
withstand system events such as short-circuit faults or transient over voltages.
To prevent these failures and to maintain transformers in good operating condition is a
very important issue for utilities. Traditionally, routine preventative maintenance
programs combined with regular testing were used. With deregulation, it has become
increasingly necessary to reduce maintenance costs and equipment inventories. This has
led to reductions in routine maintenance. The need to reduce costs has also resulted in
reductions in spare transformer capacity and increases in average loading. There is also
trend in the industry to move from traditional time based maintenance programs to
condition-based maintenance. These changes occur at a time when the average age of the
transformers in service is increasing and approaching the end of nominal design life.
The change to condition-based maintenance has resulted in the reduction, or even
elimination, of routine time-based maintenance. Instead of doing maintenance at a regular
interval, maintenance is only carried out if the condition of the equipment requires it [1].
Hence, there is an increasing need for better no intrusive diagnostic and monitoring tools
to assess the internal condition of the transformers. If there is a problem, the transformer
can then be repaired or replaced before it fails.
Many testing and monitoring techniques have been used by utilities.
1.12 ORGANISATION OF THESIS
This thesis work has been organized as follows:
Chapter 1 briefly describes different transformers with their types, uses, choice of
voltages, general specifications and design practices in manufacturing, power losses in
distribution system and transformer itself for proper material, design and manufacturing
standards. Choice of transformer selection is also specified. Necessity of fault diagnosis
of power transformers is discussed.
Chapter 2 provides the detailed information regarding research and developments
relevant to research area of Ph. D. work. It involves the progress made for conventional
testing of transformer (liquid and solid) insulations to detect faults at early stage or pre
warnings. It also shows the advancements in fault diagnostics methods. Further it presents
the application of artificial intelligence (AI) in dissolved gas analysis for fault diagnosis
of power transformers.
21
Chapter 3 emphasizes the importance, quality, specifications and periodic monitoring of
transformer insulating oil. The following tests have been performed for thermal
degradation of oil in power transformer:
i. Acidity
ii. Water contents
iii. Breakdown Voltage (BDV)
iv. Volume resistivity
v. Tan-Delta
vi. Dielectric constant and
vii. Flash point
After analysis, following conclusions are drawn;
i. All tests can be performed for fault diagnosis (periodic test) as pre-warnings of the
faults.
ii. Type of faults and reasons for the faults are correlated to confirm the faults.
iii. Most of the solutions are regeneration of oil such as filtering, removal of water
contents and sludge removal.
Chapter 4 classifies various faults that can be detected with the help of DGA. Guidelines
for interpretation of faults adopted in various countries have been presented. The different
diagnosis standards and methods for interpretation of dissolved gases have been
explained. Some methods used by countries other than INDIA for fault diagnosis are also
listed with brief descriptions. Comparison of different DGA methods for fault diagnosis
of power transformers (key gas ratios, Doernenburg‘ ratio, Roger‘s ratios and Duval
triangle) have been discussed. Case studies have been carried out to show the advantages
of Duval triangle method (DTM) over other methods.
Chapter 5 describes the interpretation of DGA in power transformers using the DTMs. It
is also shown that the accuracy of DGA laboratory results can affect the reliability of
DGA diagnosis. In addition to the six zones of individual faults an intermediate zone of
fault DTM has been developed for varieties of electrical and thermal faults in the power
transformer. The following conclusions are drawn from the DTMs:
i. Duval triangle fault interpretation methods can be used graphically in a very easy
way and the evaluation of faults is also easy.
ii. Available eight Duval triangles are capable of fault diagnosis separately; such as
main transformer, online load tap-changers (OLTCs), faults with low energy, faults with
low temperatures and synthetic oils.
22
Chapter 6 deals with the application of artificial intelligence (AI) in fault diagnosis of
power transformers. Fuzzy logic (FL) and artificial neural network (ANN) have been
used in DTM for better interpretation of results. Comparison of simulated results using
FL, ANN and Fault diagnosis reports available from various organizations gives the
following conclusions:
i. Percentage of gases based rules in FL gives flexibility in Duval triangle methods to get
accurate analysis.
ii. Superior learning capabilities of ANN gives better and quicker analysis as compared to
other fault diagnosis methods.
Chapter 7 includes the conclusions and the suggestions for future work. Major
conclusions are as follows;
i. Critical specifications of transformer oil can be performed for fault
diagnosis/condition monitoring (periodic test) as pre-warnings of the faults (if any) after
thermal degradation of transformer oil test analysis.
ii.Type of fault and reasons for the faults are correlated to confirm the faults from the test
reports of critical specifications of oil.
iii. Most of the solutions are regeneration of oil such as filtering, removal of water
contents and sludge removal.
iv. After comparison of different fault diagnosis methods of power transformers, it is
found that Duval Triangle Methods (DTMs) give better results.
v. DTMs have satisfied the fault diagnosis for more than 95% accuracy compared to the
other fault diagnosis methods.
vi. Traces of one of the three gases can be helpful for early fault diagnosis in the power
transformers.
vii. Application of FL and ANN in DTMs can provide higher accuracy, low uncertainty
and quicker analysis in fault diagnosis of power transformers.
viii. It has been proved that ANN based DTMs can provide full solutions in fault diagnosis
of power transformers.
Further chapter gives following suggestions for future work.
i. Optimization techniques may be attempted for improvement of accuracy in fault
diagnosis of power transformers.
ii. Intelligent techniques may be applied to reduce the uncertainties among faults.
1.13 AUTHOUR’S CONTRIBUTION
For the completion of this Ph. D. thesis, the following sequence of work was carried out:
23
i. Since transformer oil is easily available to ascertain faults without removing the
transformer from service, testing of oil parameters has been carried out to establish
correlation among them.
ii. Collection and analysis of dissolved gas analysis (DGA) reports have been done from
different utilities and authorities.
iii. Comparisons of fault diagnosis methods have been done.
iv. Investigations of Duval triangle methods have been explored for fault diagnosis of
power transformers.
v. Applications of the fuzzy logic and artificial neural networks in Dual triangle methods
have been done and their comparison with conventional methods to diagnose the
faults.
24
CHAPTER 2
LITERATURE REVIEW
2.1 INTRODUCTION
Literature survey is the sine aqu non of the research. Literature survey provides idea of
the recent developments in the area of research and also the future scope of the work on
the subject. This chapter includes the complete study of relevant literature pertaining to
the Ph. D. area of work. The systemic analysis of research papers results into specific
observations for the purpose of defining the objectives for the Ph. D.
2.2 FAILURE SURVEY ON TRANSFORMERS
There are a number of failure mechanisms which affect the life expectancy of
transformers, and transformer failure can occur as a result of different causes and
conditions. Zhang et al [1] developed a methodology to use data acquisition derived from
condition monitoring and standards diagnosis for rehabilitation purposes of transformers.
There are a number of failure mechanisms which affect the life expectancy of
transformers, and transformer failure can happen as a result of different causes and
conditions shown in Fig. 2.1. Failure factors in transformers include electrical breakdown,
lightning, dielectric fault, loose connection, incorrect maintenance, moisture, excessive
overloading, and other causes [1]. Contamination, thermal aging, repetitive excessive
voltage stress, and mechanical deformation lead to electrical breakdown. Dielectric
failure is a common failure occurrence and can have a profound effect on useful life.
Contamination and thermal aging can be monitored through testing. Voltage stress can be
controlled by design of the transformer protection and operating procedures.
Figure 2.1: Causes of Transformer failure [1]
Failure statistics of the defective components shown in Fig. 2.2 can identify the areas
where failure reducing efforts will produce the better outcome. When analyzing the
failure causes, information on faults is given in load tap changer (LTC), bushing,
winding, tank, core, and relay [1-2].
25
Figure 2.2: Percentage of defective components of transformers [1, 2]
Fig. 2.3 shows the most common detection methods of transformer failures and the
percentage of all the detection methods representing studies conducted for many years:
relay test, inspection, turns ratio test (TTR), dissolved gas analysis (DGA), resistance test,
total combustible gas test (TCG), current test, power factor test, temperature test,
capacitance test [1].
Figure 2.3: Percentage of failure detection methods [2]
2.3 FAULT DIAGNOSIS OF POWER TRANSFORMERS
Gibeault et al [3] suggested that fault diagnosis of transformers creates the opportunity to
statistically plan and schedule outages and to manage equipment utilization and
availability. The transformer can be kept in service, sometimes under load condition,
when the fault is evolving at a moderate and predictable rate. This condition will prevent
loss of revenue and provide time to plan sequential repair or replace the transformer.
Thus on site repairs can improve the availability. Monitoring a fast developing fault and
evaluating its progress provides the necessary information to marshal all the essential
resources to react on time and reduce the overall diagnostic. Managing and extending the
life of power transformer requires continuous and reliable monitoring as the validity of
any action taken to affect the life of transformer, needs to be tested periodically.
Successful life management and extension of power transformers produce the financial
benefits to the organization.
26
Detection and monitoring of incipient faults work with all maintenance concepts. There
are Reliability Centered Maintenance (RCM) and Condition Monitoring Maintenance.
These are two emerging maintenance concepts that require detection of incipient faults
and monitoring of a development of failure condition.
Thermal and electrical faults may arise following short circuits, local overheating at hot
spots or leakage of flux and eddy current in the core, and PD or arcing at areas of high
stress. Decomposition products from breakdown of the oil, kraft paper and press board,
glues, etc are transported through the transformer by the coolant/insulation oil [4]. The
end of life of a transformer is typically defined as the lost of mechanical strength of the
solid insulation in the windings and decaying of insulating oil.
J. Lapworth [5] suggested new method of interpretation DGA scoring algorithm, which
has proved to be robust and effective in screening out normal DGA results, and shows
promise in improving the detection of developing problems. Sparling et al [6-7] have
shown, the need of transformer monitoring and diagnostic online [6], multi-gas online
monitoring has been incorporated in transformer nursing unit (TNU) [7], which is a fast
fourier infrared spectrometer analyses gases extracted by permeable membrane. John
Sabau et al [8] has shown the progress in DGA testing and provided evidence that in
certain cases the reclamation of oil aged reduces the level of gases generated. Thus by
selectively removing decay products from transformer oil, the DGA interpretation can be
significantly improved and the cost effectiveness of reclamation technology. Shahsiah et
al [9] proposed a new diffusion dynamic model for propagation of characteristic gases in
transformer oil-cellulose structure due to temperature variation. Sabau et al [10] again
suggested the new laboratory technique developed for hidden side effort. It describes the
relationship between the parity of oil and its stability under electrical stresses, and the
necessity for maintaining the liquid insulation in pristine condition through the entire life
time of a power transformer.
2.4 TESTING OF SOLID INSULATIONS
Cellulose is a natural polymer of glucose and it degrades slowly as the polymer chains
breakdown during service, releasing degradation products into the oil. The paper
eventually degrades to such an extent that it loses all its mechanical strength and becomes
susceptible to mechanical damage, which puts the electrical integrity of the equipment at
risk. The transformer is subjected to mechanical forces due to transportation,
electromagnetic forces caused by system short circuits, and inrush current. Vibration and
27
thermal forces generated by different thermal expansion rates in different materials, cause
long-term degradation of the paper.
Faults in paper are generally considered as more serious than faults in oil, because paper
is often located in areas of high electric field (in the windings, or as voltage barriers), and
the destruction of paper insulation may lead to short circuits or severe arcing. The faults
in paper insulation are rare. Their presence can be seen by any DGA techniques for
dissolved gases in transformer oil. Therefore, for fault diagnosis during service or after
failures and/or life estimations of working transformers purpose the degree of
polymerization (DP) of solid paper insulation sample from the transformer and the furanic
analysis of the oil are performed. The solid insulation testings are as follows:
2.4.1 Degree of Polymerization (DP)
Sans et al [11] made a large scale survey of furanic compounds from the transformers
service and tried to correlate the remaining life of the transformer from content of furfural
hydrates dissolved in the oil. DP offline test is useful to confirm the deterioration of paper
insulation to determine the remaining life of power transformers. The DP tests the
cellulose molecules, which are made up of long chain of glucose rings and responsible
for mechanical strength of the paper insulation. To perform D P test remove a sample of
the paper insulation about one centimeter square. Carry out the DP test in laboratory and
analyses the results for remaining life with the standard table used by the utility.
2.4.2 Furan Test
Cellulose insulation decomposes due to overheating and chemical reactions get dissolved
in the oil. To estimate paper deterioration online and fault analysis of power transformers,
the furan analyses are conducted along with DGA test on oil insulation sample in the
laboratories. R. Blue et al [12] constructed a novel optoelectronic sensor for the
determination of furfuraldehyde (FFA) in the transformer oil. This sensor is compact and
portable that can be used by a non specialist operator. Again, R. Blue et al [13] reported
first time the used the of fluorescence based measurement of concentration in transformer
oil. In healthy transformer, the furanic compounds are less than 100 ppb and for
significant damage to paper insulation from heat up to 70,000 ppb show the end of life of
paper insulation or of the transformer. Testing is done for five different furans which are
caused by different problems.
28
2.5 TESTING OF TRANSFORMER INSULATION OIL FOR FAULT
DIAGNOSIS IN POWER TRANSFORMERS
The development of several new laboratory testing procedure for mineral insulating oils
over past years has been the rewarding result of a cooperative research project design to
extend the life expectancy of aging power transformers by eliminating the causes of
premature oil deterioration. Han et al [14], in this literature survey emphasized on the
importance of condition monitoring (CM) to reduce operating cost, enhancing the
reliability of operation and improve power supply. This paper also pointed out the
potential benefits through the utilization of advanced signal processing and artificial
intelligence technique to develop C.M. schemes for transformers through gas in oil
analysis. Kohtoh et al [15, 16] shown in their study that electrical, physical and chemical
characteristics of transformers oils are deteriorating by increasing age an usage. The
electrical characteristics such as volume resistivity, loss tan. Breakdown voltage
interfacial tension, and values gets more affected. Abu-Siada et al [17] shown that
ultraviolet to visible spectral response based equipments are cheap, and correlation
between oil spectral and its furan contents. Authors introduced fuzzy logic approach the
establish transformer ageing oil UV- Vis spectral response.
Some of the oil characteristics are tested as per the standards shown below [18-22]:
2.5.1 Break-Down Voltage (BDV)
Break down voltage is the ability of oil to withstand stresses without failure. Breakdown
strength analysis of transformer oil gives effective results through which suitable
dielectric material for the related high voltage applications can be explored. The
Breakdown Voltage (BDV) test kit consists of two electrodes mounted on horizontal axis
with 2.5mm gap and enclosed in a glass chamber. Electrodes used in test apparatus are of
sphere-sphere electrode configuration. The kit is provided with standard oil test set up is
powered through a step up transformer capable of offering up to 60 kV. The tests are
carried as per Indian standard IS-6792 [18] specifications.
2.5.2 Flash Point
Flash point of a volatile liquid is the lowest temperature at which it can vaporize to form
an ignitable mixture in air. The Fire point is defined as the temperature at which the
vapour continues to burn after being ignited. It is the lowest temperature at which, on
further heating beyond the flash point, the sample support and combustion for five
seconds. Flash and fire point test is carried out in Pensky-Martens open cup apparatus as
per ASTM D-3828 [19].
29
2.5.3 Acidity
Acidity is a measure of free organic and inorganic acid present in the oil and expressed in
terms of milligrams of Potassium Hydroxide (KOH) required to neutralize the total free
acids in one gram of oil. Presence of these acid contents in oil causes detrimental to
insulation system and leads to corrosion of tank. Acids usually originate from oil
decomposition, oxidation products and external atmospheric sources. Acidity level of
transformer oil is measured as per British Standard BS2000 Part1 [20].
2.5.4 Viscosity
Viscosity of oil is the measure of the transformer oil resistance to shear. Viscosity is more
commonly known as resistance to flow for fluids. If lubricating oil is considered as a
series of fluid layers superimposed on each other, the viscosity of the oil is a measure of
the resistance to flow between the individual layers. A high viscosity implies high
resistances to flow while a low viscosity indicates a low resistance to flow. Viscosity
varies inversely with temperature. Viscosity is measured using Red wood viscometer at
the temperature of 300 C. The test is carried out as per ASTM D 2162-06, IP 70 [21].
2.5.5 Tan-delta, Resistivity and Dielectric Constant
For measuring these characteristics, measurement test kit consists of three terminal oil
cell with heating chamber for oil cell. The tests are performed as per IEC 60247 Standard
[22].
2.5.6 Loss Factor
Loss factor is a measure of imperfection of dielectric nature of oil. For a perfect dielectric
when applied with a sinusoidal ac voltage, the current flowing through it should lead the
voltage by 900. But it is not in the case of reality, the angle by which it is short of 90
0 is
called the loss angle. The Loss tangent is also known as dielectric dissipation factor,
which measures the leakage current through oil. It is the measure of contamination
products like moisture and oxidation products. Dielectric dissipation value should be low
for good transformer Oil. Resistivity is the most sensitive property of oil requiring highest
care for its proper determination. Resistivity in ohm cm is numerically equivalent to the
resistance between opposite faces of a centimeter cube of the liquid. Insulation resistance
of windings of a transformer is also dependent upon the resistivity of oil. A low value
indicates the presence of moisture and conductive contaminants. Tested as per standards
[22].
30
2.6 DGA FAULT DIAGNOSIS OF POWER TRANSFORMERS
Based on the relative` proportions of the gases formed, several DGA interpretation
methods are available for identification of faults (low and high energy electrical discharge
faults, PD, arcing, corona or hot spots, etc). acceptable and dangerous gas levels in
transformers thus appears to be significantly influenced by the types of faults involved as
well as by the type and the age of the oil-filled transformers.
2.6.1 On-line/Off-site DGA
Continuous on-line monitoring of the acknowledged key gas indicator or parameter of the
operation of the transformers, the dissolved gases involved are generally CO, CO2,H2,
CH4, C2H6, C2H4, C2H2. The solubility of these gases is dependent on the gas, the gassing
tendency, temperatures. For any given sample the absolute and relative concentration of
fault gas can be indicate the type, intensity and location of the fault. The laboratory
DGA interpretations may rely upon defined critical levels of gases or/and rates increase in
gas level. The various conventional fault interpretation methodologies, i.e. key gas ratios,
Dornenburg‘s ratios, Roger‘s ratios, and some other diagnostic procedures are adopted
around the world. Ruijin et al [23] have shown in their research, the application of back
propagation ANN learning on the ratios of six combustible gases from DGA can diagnose
the transformer faults. McGrail et al [24] suggested various gases analyzing sensors, both
infrared and oxide utilized to develop portable system for continuous monitoring. Samat
et al [25] used gas chromatography, allow the determination of the principle gases, then
DGA interpretation with on revise publication 599 for nature of faults with different
methods (LCIE-LABORELEC – ROGER – DEVIS - DORNENBURG) for monitoring
oil filled equipment. SJ Farrito indicated that DGA can now be accomplished without
high vacuum extraction of the gases from the oil, the alternate method suggest that the
direct injection available alternate technique that offers many advantages and a valuable
tool of evaluating the condition of power transformer [26]. Inoue et al [27] noted the
hydrogen gas in oil is always one of the decomposed gases, a portable oil dissolved
hydrogen detector was developed for quick field screening for oil filled equipment for
detailed inspection.
2.6.2 DGA Duval Triangle Methods
The main gases formed as a result of electrical and thermal faults in transformers, and are
evaluated by DGA, whose concentrations depends upon the type of fault. Duval triangle
is triangular representations of three gases CH4, C2H4 and C2H2 only, which represents 7
faults zones within an equilateral triangle (PD, D1, D2, DT, T1, T2 and T3). M. Duval et
31
al [28, 29] narrated DGA analysis with IEC/IEEE codes for interpretation, sampling
procedure, extraction of gases and fault evaluation based on relative concentrations. Their
other publication [30] suggested different acceptable gas-in- oil levels in generation and
distribution transformers to the 90% & 95% respectively of the values given in IEEE
guide (1978). The advantages of metal deactivators [31] can effectively increase the
thermal resistance of mineral transformer oils but keep the electrical properties
unchanged for longer period and also reduces its charging tendency. Duval et al [32]
introduced the influence of paper insulation in acceptable gas-in-oil levels in DGA
analysis of transformers which involves the contents of carbon oxide and carbon
monoxide in oil. Duval et al [33] emphasis the popularity of interpretations of transformer
faults using new IEC publication 605999, 1999 contains the description of five main
faults and revision of IEC TC 10 databases for changed typical concentration values,
alarm values and rate of gas increase with previous IEC publication 599 issues.. Duval
[34] came out with a complete review of faults detectable by through DGA in
transformers on to improved Duval triangle 1 method for simulated samples for thermal
faults (T1, T2, T3), low-energy discharges (D1), PDs of the corona type, high-energy
discharges (D2) and separate DGA corresponds to a discharges of low- energy, which
will alter their normal gas composition specially the relative percentage of C2H4. M.
Duval again shown some new techniques for dissolved gas analysis in transformers [35].
Duval et al [36] suggested improving the reliability of transformer DGA analysis by
adding proper allowances to the laboratory inaccuracies. This helps to improve
transformer condition assessment and diagnosis by DGA of power transformers. M.
Duval presented a review [37] of previous researches in a conference on technical paper
in a CIGRE‘s report in 2006. For calculation of DGA limit values and sampling intervals
[38], has suggested the authorities and utilities to allow calculation of their own specific
DGA gas limits and required sampling intervals as a function of gas concentration and
also the rates of gas increasing during service. M. Duval [39] proposed new versions of
DTM 1 to DTM5 for all types of oil filled transformers may be filled with mineral/non-
mineral oils, OLTC oil analysis, low-temperature faults and low-energy faults.
2.6.3 Standards and Guides for Fault Interpretations
Various standards and guides are available for interpretation of faults of power
transformers through DGA analysis. These are adopted by the organizations/countries
according to their geographical, environmental conditions, need based, etc. CIGRE Group
2000 issued its guidelines for life management of power transformers [40]. IEEE
32
Standards [40, 41] are used as guide for interpretation of faults in power transformers
through DGA analysis. IEC publications [42-45] are said to be service guide to the
interpretation of dissolved and free gases analysis, which kept on amending the guides
time to time and need basis. IS- 10593 is also available for fault analysis for gas-in-oil
filled electrical equipments in service [46]. Other standards and guides [47-49] are used
as American standards, instructions, facilities, and techniques to help in fault diagnosis of
power transformers. R. R. Roger [50] used gas ratios to establish faults from extracted
gases and only by the use of IEEE and IEC codes for incipient fault interpretation.
2.7 ARTIFICIAL INTELLIGENCE (AI) APPLICATION IN DGA OF POWER
TRANSFORMERS
Computers applications have found wide spread applications ‗human-like abilities‘,
capabilities to make judgments, guesses, change of opinions in fault diagnosis of power
transformers in last two decades. Computer has reduced vagueness, uncertainties, analysis
times, but increased accuracy quicker remedial actions during off-line and on-line fault
diagnostic on power transformers. Data of the dissolved gases in oil insulation of a power
transformer can be incorporated into expert systems to facilitate decision making for fault
diagnosis. Due to the diverse gas content of transformer oil, computer based Artificial
Intelligence (AI) techniques and expert systems have been applied by the various
researchers, scientists, different organizations and utilities. Various AI techniques applied
by researchers for DGA in power transformers are as follows:
The fuzzy logic [51-65] analysis consists of three parts: fuzzification, fuzzy inference and
defuzzification. Fuzzification is the process of transforming crisp input values into grades
of membership for linguistic terms of fuzzy sets. The membership function is used to
associate a grade to each linguistic term. A chosen fuzzy inference system (FIS) is
responsible for drawing conclusions from the knowledge-based fuzzy rule set of if-then
linguistic statements. Fault types are first listed and then form the fuzzy rule set for the
diagnosis system. Defuzzification then converts the fuzzy output values back into crisp
output actions.
Artificial Neural Networks [66-76] is a massively parallel distributed processor, having a
natural tendency to acquire sufficient experimental knowledge and making it available for
use. Among these AI methods, the ANN is widely designed to diagnose transformer
faults. An important advantage of fault diagnosis based on the ANN is that it can learn
directly from the training samples and update its knowledge when necessary. The high
nonlinear mapping capability of neurons always provides a comparable and superior
33
performance over a fuzzy system solution. However, although the computational
complexity of the ANN is not too high, especially in the fault diagnostic process, it
involves certain problems including slow convergence, oscillation and so on, all of which
must be resolved before it is practically applied.
Albus proposed a neural model called CMAC (Cerebellar Model Articulation controller)
[77], which like the models of human memory, perform a reflexive processing. The
CMAC, in a table look-up fashion, produced a vector output in response to a state vector
input. Figure 3 shows a basic configuration of CMAC network for Key gas ratio. Refer to
the IEC std. 60599 the key gas ratio of C2H2/C2H4, CH4/H2, and C2H4/C2H6 are used as
the input states. The diagnosis system contains 9 parallel memory layers and every
memory layer has one output node. Every memory layer remembers one fault type
feature.
Bin et al [78] develop an assistance tool, the diagnostic algorithm must be easy to
implement in the portable device and hardware device with a compact configuration. The
grey theory provides the applications of clustering analysis is useful tool in fault
diagnosis, relational analysis, predication, and decision for the grey system. The so-called
‗‗grey‖ means that system information is incomplete, unclear, and uncertain. It is a useful
method to deal with the problems of limited, deficient, and or no rules available for data
processing. Its analysis makes use of minor data and does not demand strict statistical
procedures and inference rules. Lin et al [79] utilized GCA for fault analysis as a
function of mathematical operation for processing numerical data without adjusting any
parameter.
Due to incompleteness and complexity of condition estimate for power transformer, a
specific model based on rough set (ST) theory is presented. After the statistic analysis on
the collected fault examples of oil-immersed power transformer, Zhang et al [80] using
rough set theory to reduce result, estimate rules are acquired and they could be used to
improve the condition assessment of power transformer. The condition estimate inference
model was built based on the advantage of effectively simple decision rules and easy
reality of rough sets. Huang et al [81] shown significant advantage of the new method is
that it can discriminate the indispensable alarm signals from dispensable ones that would
not affect the correctness of the estimate results even if they are missing or erroneous.
In last two decades, some more AI techniques have been proposed by researchers in
experimental stages. Tang et al [82] presented Parzen–window (PW) based classifier for
fault diagnosis of transformers, which is able to interpret transformer DGA analysis with
34
a probabilistic scheme on the evidence of various gas ratios. [83, 84] have used wavelet
networks with NN and FL to compare and analyzing the network training process on
showing the accuracy and efficiency in fault diagnostics. Yang et al [85] a new self-
organizing polynomial networks (SOPNS) modeling technique is proposed and
implemented to serve as an intelligent decisions support for the transformer fault
diagnosis in comparison to conventional DGA and ANNs. Thang et al [86] illustrated a
novel approach for analysis and interpretation of DGA data which is more reliable
condition monitoring of power transformer based on self –organizing map (SOM)
algorithm. Liu et al introduced support vector machine (SVM), machine learning
algorithm based on statistical learning theory for fault diagnosis in power transformer
along with fault signature [87]. Jeong et al proposed a fault diagnosis on distribution
transformer using graphical information of frequency responses [88] for a group of
transformers for similar rating and capacities. [89-92] are some practiced artificial
intelligence method for DGA of power transformer into experimental stages to introduce
as fault diagnosis tool. Review of intelligent diagnostic method [93-97] have been
presented on conventional DGA method such as IEEE/IEC standards/codes, Roger‘s or
Dornenburg‘s ratios used for fault diagnosis of power transformers and others oil filled
electrical equipments.
2.8 ADVANCEMENTS IN FAULT DIAGNOSIS METHODS OF POWER
TRANSFORMERS
There are many advanced fault diagnosis methods are coming in the market and few of
them are still in the testing stages. Some of them are as follows:
2.8.1 Thermography
Infrared emission testing is used to check the external surface temperature of the
transformer on-line. It is useful for detecting thermal problems in a transformer, such as
cooling system blockages, locating electrical connection problems, and for locating hot
spots by Duke [98]. Infrared imagers ―see‖ the surface heat radiation from objects. It
cannot look ―inside‖ the transformer tank. Black and white/colourful thermograms (heat
pictures) show hot areas in white and cold areas in black. For colour thermograms, white
and red areas are usually hotter, while black and blue areas are colder. Infrared
thermography provides the heating patterns for the load that was on the equipment at the
time that the scan was performed. Any abnormal conditions can be located from the scan.
35
2.8.2 Sweep Frequency Response Analysis (SFRA/FRA)
In the sweep frequency response analysis test, the transformer is isolated from the system
and the impedance or admittance of the transformer is measured as a function of
frequency (typically to at least 2 MHz). This gives a ―fingerprint‖ of the transformer. The
test is repeated over time and the ―fingerprints‖ from two or more tests are compared.
There are two different test methods commonly used to carry out the FRA test; the swept
frequency test and the pulse test. The swept frequency method applies a variable
frequency voltage or a white noise voltage to the high-voltage winding and records the
response in another winding or terminal. This technique is more widely used in Europe
than in North America. This technique is picking up pace in India slowly. With this
technique a pulse signal is applied to the high-voltage winding, and the response is
recorded in another winding or terminal. In general, the greater the difference between the
two ―signatures‖, the greater movement in the transformer winding. The test requires
experienced personnel to compare the two signatures and evaluate the severity of the
movement by Singh J. at al [99-100].
2.8.3 Thermal UV- Response Test on Transformer Oil
U.V. visible spectrophotometer is used to analyse the transmission and absorption
capability of oil. As per IS, IEC, ASTM standards refractive property analysis is not a
mandatory test for liquid dielectrics, as initiative work absorption capabilities of
transformer oil is analysed. Experiment is carried out by UV spectrometer [17] having
single beam facility. The test is conducted for various hours of heating of pure
transformer oil. Based on the test results like wavelength and Peak absorption it is
possible to determine the level of thermal degradation of oil as an indirect methodology.
The wave length of pure transformer oil lies in the UV region (200 nm-300 nm).
2.8.4 Fault Detection (winding deformation) by Transformer Tank Vibration
Garcia et al [101-102], a model is developed for a transformer monitoring system to
estimate transformer tank vibration to detect faults. The model calculated vibration on
transformer tank started from some input variables that can be easily measured on the
transformer. In case of winding deformation winding vibration and, consequently, that of
the tank, changes and a big difference between the estimated and measured vibration
appear. The models were tested experimentally on a test transformer fitted with internal
and external accelerometers. The deformations were provoked in transformer windings to
test the model ability to detect. This detection method was experimental verification of
the proposed model that calculated vibration on the transformer tank.
36
2.8.5 Alternative Insulating Fluids and their DGA
In the present scenario, there has been an increase in the use of environmental-friendly
synthetic fluids (non-mineral oils) such as Silicone, Midel, FR3, Biotemp, `synthetic
esters, and natural esters. Although these non-mineral oils are environmental friendly that
results from their use with no compromise on safety and reliability. These synthetic fluids
are biodegradable unlike mineral oil. In order to their environmental performance, esters
higher flash point and fire point. The esters have a lower relative humidity than mineral
oil, this means that moisture has less of an impact on the dielectric strength of the esters
then mineral oils. When esters are used in conjunction with kraft papers and press boards,
cellulose is kept in a drier condition and the rate of cellulose degrades slower then
mineral oil [103-104]. Most of the conventional diagnostic methods (key gas ratios,
Roger, ratios, Doernenburg‘s ratios) are suitable for mineral oils filled in transformer
tanks but not suitable for the DGA diagnostics to equipment filled with non-mineral oils
(natural and synthetic esters, silicones). M. Duval conducted DGA tests separately on
these fluids and gave different Duval triangles 3 for each type of synthetic and natural
fluids (silicones, midel, FR3 and biotemp) [39].
Different proposal for improvement in transformer diagnosis using DGA might be fault in
order new transformers as laboratory test for routine condition monitoring are proposed
[105-108] literature. Petterson et al [109] tried to estimate the remaining life of power
transformers with insulations degradation during the service. Radigk et al [110]
contributed towards explaining problem of DGA and its evaluation. This also has
recommended how to analysis an oil sample and how to evaluate the results with the
focus on different equipments. [111] I A R Gray, Transformer chemistry service issued a
guide for transformer fault analysis through insulation degradations. Digiorgio et al [112]
issued through NTT (Northern Technology & Testing) Technical Bulletin for dissolved
gas standards/procedures of DGA used in different countries in the world. CIGRE
surveys [113-114] conducted in 2008 and 2007 for showing the influences of various
factors during the sampling process and its preservation before dissolved gas analysis are
carried out in laboratory. If these factors are not taken care-off, there likely to be errors in
interpretations for during fault analysis in transformers.
Literature survey on the proposed topic yields the important outcomes as follows;
i. There is need for an appropriate, efficient and cost effective non-invasive condition
monitoring and fault diagnosis method to test and detect faults at an early stage.
37
ii. Most of the faults and failures in transformers occur because of weak liquid or/and
solid insulation. Thus, it is desired to test transformer and transformer oil regularly to
detect fault at an early stage.
iii. Fault diagnosis methods of power transformers need to be compared for exact
identification of faults.
iv. It is observed that little work has been reported on the applications of Duval triangle
methods (DTMs) for fault diagnosis of power transformers.
v. It is also noticed that applications of fuzzy logic and artificial neural network have
not been reported so far on the applications of DTMs for fault diagnosis of power
transformers.
38
CHAPTER 3
TRANSFORMER INSULATING MATERIALS AND OIL TESTING
3.1 INTRODUCTION
It has been found that most of transformers fail during the service because of the
insulation failures chiefly because either liquid or solid insulating materials. Moreover, it
has further analyzed that many transformers get damaged due to transformer oil failure
than solid insulation failures. In this chapter main emphasis is given for quality and the
periodic monitoring of transformer oil. Typical transformer oil (mineral oil) specifications
are given in 3.1. All possible causes for weakening the insulation properties along with its
periodic testing/monitoring during service have been discussed. Traditional and non-
traditional testing of oil in the laboratories are performed to detect incoming fault
warnings at early stage. Thermal degradation of transformer oil at 1000C for 5, 7, 10, 15,
20, 24, 30 hrs are conducted and variations on some of the critical electrical
characteristics: i.e. BDV, IFT, water content, flash point, resistivity, tan-delta, loss factor,
dissolved gas analysis (DGA), degree of polymerization (DP), etc are observed at Central
power Research Institute (CPRI), NOIDA transformer oil testing laboratory. These tests
are performed on pure oil, mixed with kraft paper and press board and mixed with water.
The variations in critical electrical characteristics for these three cases have also shown.
3.2 INSULATING MATERIALS
These are the materials, which permit only a negligible current (order of pA) to flow in
phase with the applied voltage. Insulating materials or insulants are extremely diverse in
origin and properties. They are essentially non-metallic, organic or inorganic; uniform or
heterogeneous in composition; natural or synthetic. Many of them are of natural origin as,
for example, paper, cloth, paraffin wax and natural resins. Wide use is made of many
inorganic insulating materials such as glass, ceramics and mica. Many of the insulating
materials are man-made products manufactured in the form of resins, glass, ceramics, etc.
In the recent years wide use is made of new materials whose composition and properties
place them in an intermediate position between inorganic and organic substances. These
are the synthetic organo-silicon compounds, generally termed as silicones.
3.3 PROPERTIES OF INSULATING MATERIAL
There are many properties, which characterize the insulating materials, e.g. resistivity,
breakdown voltage, permittivity and dielectric loss, etc. An ideal insulating material
should have:
39
i. High dielectric strength sustained at elevated temperature
ii. High resistivity
iii. Good thermal conductivity
iv. High tensile and shear strength of solid insulation
v. High degree of thermal stability
In addition to the above properties, the material should have other mechanical properties
such as ability to withstand moisture (it should be non-hygroscopic), vibration, abrasion
and bending. Also, it should be able to withstand chemical attack, heat and other adverse
conditions of service.
3.4 TRANSFORMER PAPER INSULATION
Insulation grade paper is made by the delignification of wood pulp by the Kraft process. It
contains about 90% cellulose, 6-7% lignin and, the balance is hemi cellulose. The natural
humidity of paper is 4-5% by weight and the insulation is dried after winding to less than
0.5%. The dried paper is impregnated with insulating oil, which increases its dielectric
strength and also serves to cool the windings. Table 3.1 represents the specification of
electrical grade paper. The major constituent of paper is cellulose, which is a natural
polymer of glucose. Power transformer conductor windings are insulated by paper
impregnated with insulating oil, which is expected to last the life of the transformer (25
years minimum at an operating temperature of 65-95°C). A typical 600 MVA transformer
contains 12 t of paper, 30-120 μm thick (density 0.7-0.8 g/cm3), and 45 t (40,000 lts) of
oil. Heat, water and oxygen degrade (depolymerize) the cellulose, reducing the polymer
molecular chain length and with it the mechanical strength of the material.
Table 3.1: Electric Grade Papers Specification as per IEC 554-3-5 1984
S.No. Properties Specifications
1. Substances (g/m2) 30.4-40.0 ± 10%
2. Apparent Density (g/cc) 0.80 ± 0.05
3. Thickness (Micron) 38-50 ± 10%
4.
Tensile Index (Nm/g)
Min MD
CD
93
34
5. Water Absorption (mm)
MD
CD
8
8
40
6. Tear Index (MNm2/g)
Min MD
CD
5
6
7. Electric Strength (In Air at 90 oC) Min.(kV/mm) 7.0
8. Ash Contents Max. ( %) 1.0
9. Air Permeability (µm/Pa.Sec.) 0.5-1.0
10. Conductivity of Aqueous Extract Max. (mS/m) 10.0
11. Moisture Contents Max. (%) 8.0
12. pH of Aqueous Extract 6-8
13. Heat Stability Max. % (mS/m) at Max.120oC, 168 Hrs.
a. Decrease in Burst Strength
b. Increase in Conductivity
20
23
Local mechanical failure could lead to short the turns or paper fragments or fibers in the
oil ducts and hence to dielectric or thermal failure of the insulation system. Changes in
the compliance and tension of the windings with time, may result in distortion and an
increased susceptibility to short-circuit force failure of the aged insulation or even of the
winding itself. Water is a product of ageing. Its presence in the insulation increases
conductivity and the likelihood of gas bubble formation, reducing the thermal stability of
the insulation system during overload conditions.
3.5 TRANSFORMER INSULATING OIL
Oil used for insulation in transformers is mineral oil and it is obtained by refining crude
petroleum. Animal oils and vegetable oils are not used for this purpose as these form fatty
acids on heating which are corrosive for the cellulose paper used in insulation. Mineral
oils were in use as liquid dielectrics in electrical equipment for over hundred years now.
Despite the availability of a variety of synthetic oils, with far more superior properties,
mineral oils held its way, due to their abundant availability and economy. Three
properties, which are fundamental to use of mineral oil as dielectric are:
i. High insulating property,
ii. Good oxidative and ageing stability,
iii. Good heat transfer ability.
41
The reliable performance of mineral insulating oil in an insulation system depends upon
certain basic oil characteristics, which can affect the overall performance of the electrical
equipment [6]:
i. High electric strength to withstand the stresses in service
ii. Sufficiently low viscosity so that its ability to circulate and transfer heat is not
impaired
iii. Adequate low temperature properties down to the extreme of climatic conditions
expected at the installation site
iv. Proper oxidation resistance to ensure long life in service
v. Good resistance to emulsion to prevent holding up ofwater in suspension in it
vi.Free from inorganic acid, alkali and corrosive sulphur which causes corrosion of metal
parts and insulation and accelerate the production of sludge
vii. Free from sludging under normal separating conditions
viii. High flash/fire points
ix. Low pour point
x. Low water content
3.6 INSULATING OIL QUALITY
The condition of the oil greatly affects the performance and the service life of
transformers. A combination of electrical, physical and chemical tests is performed to
measure the change in the electrical properties, extent of contamination, and the degree of
deterioration in the insulating oil. The results are used to establish preventive
maintenance procedures, to avoid costly shutdowns and premature equipment failure, and
extend the service life of the equipment.
There is a multitude of tests available for insulating oil. Threshold levels for these tests
are specified in ASTM D3487 for new oils and IEEE Guide 637-1985 for service oils. As
paper degrades, a number of specific furanic compounds are produced and dissolved in
the oil. The presence of these compounds is related to the strength of the paper as
measured by its degree of polymerization (DP). Furan and phenol measurement in oil is a
convenient, noninvasive method to assess the condition of the paper insulation.
Transformer oil samples should be analyzed for furans and phenols when one or more of
the following conditions exist:
i. Overheating or overloading of the transformer,
ii. High levels of carbon monoxide or carbon dioxide,
42
iii. Rapid decrease of interfacial tension without a corresponding increase in acid
number,
iv. Sudden darkening of the oil
v. Sudden increase of the moisture content of the oil, transformers over 25 years old.
3.7 FACTORS AFFECTING INSULATING OIL QUALITY
There are various factors which are responsible to affect the quality of the insulating oil
filled in power transformers [112, 115]. These are as follows:
3.7.1 Effect of Oxygen on Oil
Moisture contamination is one of the most common causes of deterioration in the
insulating quality of oil. This contamination can be eliminated by purification. Less rapid
but more serious deterioration, the formation of acids and sludge is caused by oxidation.
Thus, the exclusion of oxygen is of prime importance. In open-breather transformers, the
oxygen supply is virtually unlimited and oxidative deterioration is consequently faster
than in sealed transformers. Atmospheric oxygen is not the only source of oxygen
available for the oxidation of insulating oils; water also serves as a source of oxygen and;
therefore, leaky gaskets increase the rate of oxidation and moisture contamination. The
rate of oxidation also depends on the temperature of the oil; the higher the temperature,
the faster the oxidative breakdown. This fact points to the importance of avoiding
overloading of transformers, especially in the summer time.
3.7.2 Moisture in Oil
Water can be present in oil (a) in a dissolved form, (b) as tiny droplets mixed with the oil
(emulsion), or (c) in a free state at the bottom of the container holding the oil.
Dissolved water is not removed by centrifugal treatment; the filtration process can
partially remove dissolved water if the filter papers are thoroughly dried before filtration
and are replaced frequently. The dissolved water is best removed by complete treatment
of the oil (oil reclamation process). Emulsification occurs when the tiny droplets unite to
form larger drops, which sink to the bottom and form a pool of free water. Emulsified
water or water in the Free State may be readily removed by filtering or centrifugal
treatment. The effect of moisture on the insulating properties of oil depends upon the
form in which the moisture exists. A very small amount of emulsified water has a marked
influence in reducing dielectric strength of oil shown in Fig. 3.1 whereas dissolved water
has little or no effect on the dielectric strength.
43
Water-ppm by volume
Figure 3.1: Dielectric strength Vs Water in oil
The amount of moisture, which can be dissolved in oil, increases rapidly as the oil
temperature increases Fig. 3.2. Therefore, insulating oil purified at too high a temperature
may lose a large percentage of its dielectric strength on cooling because the dissolved
moisture is then changed to an emulsion.
Figure 3.2: Water solubility Vs Temperature
In transformers, sludge sticks to the surfaces through which heat is dissipated; the sludge
forms a barrier to the flow of heat from the oil to the coolant and from the core and coils
to the oil. If allowed to continue, the sludge may block off the flow of oil through the
cooling ducts. As a result, the transformer insulating oil becomes hotter and the
transformer could be damaged, particularly between the coil turns of the windings.
Deterioration of the coil-turn insulation may eventually lead to short circuits between coil
turns and the breakdown of the transformer. When oxidation progresses to the point
where sludge is being precipitated, the first step should be to remove the sludge from the
transformer by a high-pressure stream of oil and to either replace the sledged oil or treat it
44
with activated clay to remove the acid and sludge precursors. Complete treatment of the
oil is normally less costly than replacing it with new oil.
3.7.3 Absorption of Moisture by Insulating Materials
Solid insulation in transformers is very porous and absorbs much water. Some of the
water that is dissolved in the oil is absorbed from the oil by the insulation. As more water
is dissolved in the oil, more water is absorbed by the insulation. Once water is absorbed
by insulation, it is difficult to remove. The most effective method for drying out the
insulation in transformers is with heat and vacuum. Sometimes a vacuum cannot be
applied in the field. In this case, the transformer insulation must be dried by circulation of
hot, dry oil. This oil should then be cooled and dried. Since the dielectric strength of
insulation is reduced by moisture, it is important that steps are taken to prevent the
introduction of moisture.
3.7.4 Absorption of Nitrogen by Oil
Special precaution should be taken when operating nitrogen-blanketed transformers to
avoid bubbling of the oil, due to the release of dissolved nitrogen when the pressure
drops. Experience has shown that the automatic gas-pressure regulating system should be
adjusted to limit the nitrogen pressure range from +3.4 to +21-kPa (+ 1/2- to + 3-lb/in2)
gauge to avoid formation of these bubbles and subsequent troubles due to corona
deterioration.
3.8 IMPORTANT OIL SPECIFICATIONS
The main requirements of transformer oil as listed in various national and international
standards only state the minimum requirements for transformer oil as shown in Table 3.2
whereas many transformer producers and electricity companies and boards have their
own stringent specifications based on the required conditions.
With advanced refining techniques, it is possible to produce transformer oil tailor-made to
suit the exact requirements for such specification.
Table 3.2: Characteristics requirement of IS, IEC, and BS specification for uninhibited
Transformer Oil
S.No. Characteristics IS-335
(1993)
IEC-296
(Class I)
IEC-II
(Class)
BS-148
(Class I)
1. Appearance The oil shall be clear, Transparent & free from suspended
matter or sediments.
2. Density (g/cm2) 0.89 at 0.895 at 0.89 5at 0.895 at
45
29.5oC(max) 20
oC(max) 20
oC(max) 20
oC(max)
3. Kinematic
Viscosity (cSt)
27 max at 270C 16.5 max at
400C 800
max at-150C
11.0 max at
400C
16.5 max at
400C 800
max at 150C
4. Flash Point oC
(min)
140 140 130 140
5. Pour Point, 0C
(max)
-6 -30 -45 -30
6. Interfacial
Tension N/m
(min)
0.04 - - -
7. Electric Strength
(BDV), kV
i.New Unfiltered
oil
ii.After filtration
30(min)
60 (min)
30 (min)
50 (min)
30 (min)
50 (min)
30 (min)(As
delivered)
8. Tan Delta at 900C 0.002 (max) 0.005 (max) 0.005 (max) 0.005 (max)
9. Resistivity (Ω m)
at
900C
270C
35X1012
(min)
1500X1012
(min)
- - -
10. Oxidation
Stability 164 hrs.
1 Neutralization
value mgKOH/g
2 Sludge content
%by wt
0.4 (max)
0.1 (max)
0.4 (max)
0.1 (max)
0.4 (max)
0.1 (max)
1.5 (max)
1.0 (max)
11. Neutralization
Value
1 Total acidity
mgKOH/g
2 Sludge content
% by wt.
0.03
Nil
0.03
-
0.03
-
0.03
-
46
12. Corrosive
Sulphur
Noncorrosive Noncorrosive Noncorrosive Noncorrosi
ve
13. Water content
(ppm)
50 (max) 40 (max) 40 (max) -
14. Oxidation
Inhibitor
0.05%max. is
treated as
absence of
inhibitor
Non
detectable
Non
detectable
Non
detectable
15. S.K. Value Under
considerations
- - -
16. Accelerated
Ageing Test
(open beaker
method with
copper catalyst)
i. Resistivity at
270C
ii.Resistivity at
900C
iii.Tan Delta at
900C
iv.Total acidity
mgKOH/g
v.Total sludge
25x1012
Ωcm
0.2x1012
Ωcm
0.2 (max.)
0.05 (max.)
0.05% by
wt(max.)
-
-
-
17. Gassing
Tendancy
- - - +5 (max.)
3.9 TESTING OF TRANSFORMER OIL
In order to maintain stability in the electrical system, condition monitoring of the
sensitivity equipments in power system is very essential. Among those equipments in
which higher level of concentration required is power transformer. Power transformers
are the most critical components of power system, in which mineral oil plays an
47
important role of insulation and moreover of its thermal properties it acts as a cooling
medium. Transformer life time gets reduced because of deterioration of insulation, which
may be solid or liquid insulation. To determine the performance of transformer oil it is
necessary to have investigations on ageing due to thermal and water content.
Experiments are conducted as per American Society of Testing and Materials (ASTM),
Indian standards (IS) and International Electrotechnical Committee (IEC) standards.
Various standards adopted for testing of each parameter is given in the Table 3.3.
Table 3.3: Standards and testing
S. No. Parameters Testing standards adopted
1 Acidity IS 6792
2 Water Content IEC 814
3 BDV IS6 792
4 Volume Resistivity IEC 60247
5 Tan-delta/Loss Factor IEC 60247
6 Dielectric constant IEC 60247
7 Flash Point ASTM D 3828
Basically to determine the thermal degradation, different tests can be conducted as per the
transformer construction and types of liquid and solid insulations are used. Those are
listed as follows:
i. Thermal Degradation test on transformer oil.
ii. Thermal Degradation test on transformer oil along with Press board and paper.
iii. Degradation test due to accumulation of water content on transformer oil.
Five basic tests on insulating oil, when considered collectively, give a reasonably accurate
diagnosis with respect to the serviceability of insulating oil. The tests are:
i. Acidity
ii. Water content
iii. Break down voltage (BDV)
iv. Volume resistivity
v. Tan-delta
vi. Dielectric constant
vii. Flash point
48
Although these oil tests evaluate the insulating characteristics of the oil, the information
gained from the analysis of these tests, coupled with DGA and DP provide excellent
diagnostic tools for monitoring and evaluating the operating conditions of the oil-filled
equipment.
Samples of insulating oil are used for different thermal ageing tests, with presence of
press boards, Kraft papers and water contents. Oil characteristics are analysed to find the
degradation and to correlate with types of faults. Various trends changes in the critical
characteristics of oil, interfacial tension (IFT) and acid number (NN) of oil were also
examined. Transformer oil testing for fault diagnosis is divided in two categories:
3.9.1 Non Traditional Tests on Transformer Oil
Some of them are as follows:
i. Acidity (Acid number—NN)
Acidity is a measure of free organic and inorganic acid present in the oil and expressed in
terms of milligrams of Potassium Hydroxide (KOH) required to neutralize the total free
acids [18] in one gram of oil. Presence of these acid contents in oil causes detrimental to
insulation system and leads to corrosion of tank. Acids usually originate from oil
decomposition, oxidation products and external atmospheric sources. Acidity level of
transformer oil is measured as per British Standard BS-2000 Part1 [18]. The acid attracts
metal inside the tank and form soaps (sludge). Sludging has been found to begin when
NN reaches 0.40; it is obvious that he oil should be reclaimed long before it reaches
0.40.it is recommended that oil be reclaimed when NN reaches 0.20 mg KOH/gm
ii. Water Content (Test Method IEC 814)
Water in minute quantity is harmful in power equipments because it is attracted to the
places of greatest electrical stress and this is where it is most dangerous. Water accelerate
the deterioration of both the insulating oil and paper insulations, librating more water in
the process (heat catalyze).
This is never ending circle and once a paper insulation has been degraded (loss of
mechanical strength) it can never (unlike the oil) be returned to its original condition.
Internal sources:- paper degradation produces water, oil degradation produces water, wet
insulation contaminates the oil (temperature dependent).
iii. Break down voltage (BDV)
Break down voltage is the ability of oil to withstand stresses without failure. Breakdown
strength analysis of transformer oil gives effective results through which suitable
dielectric material for the related high voltage applications can be explored. The
49
Breakdown Voltage (BDV) test kit consists of two electrodes mounted on horizontal axis
with 2.5 mm gap and enclosed in a glass chamber. Electrodes used in test apparatus are of
sphere-sphere electrode configuration. The kit is provided with standard oil test set up is
powered through a step up transformer capable of offering up to 60 kV. The tests are
carried as per Indian standard IS 6792 [16] specifications. Test result of BDV is taken an
average of at least of 5 readings.
iv. Tan-delta, Resistivity and Dielectric Constant
These tests can be performed on the measurement test kit consists of three terminal oil
cell with heating chamber for oil cell. The tests are carried out as per IEC 60247 Standard
[20].
v. Flash Point
Flash point of a volatile liquid is the lowest temperature at which it can vaporize to form
an ignitable mixture in air. The Fire point is defined as the temperature at which the
vapour continues to burn after being ignited. It is the lowest temperature at which, on
further heating beyond the flash point, the sample support and combustion for five
seconds. Flash and fire point test is carried out in Pensky-Martens open cup apparatus as
per ASTM D 3828 [17].
vi. Loss Factor
Loss factor is a measure of imperfection of dielectric nature of oil. For a perfect dielectric
when applied with a sinusoidal ac voltage, the current flowing through it should lead the
voltage by 900. But it is not in the case of reality, the angle by which it is short of 90
0 is
called the loss angle. The Loss tangent is also known as dielectric dissipation factor,
which measures the leakage current through oil. It is the measure of contamination
products like moisture and oxidation products. Dielectric dissipation value should be low
for good transformer Oil. Resistivity is the most sensitive property of oil requiring highest
care for its proper determination. Resistivity in ohm cm is numerically equivalent to the
resistance between opposite faces of a centimeter cube of the liquid. Insulation resistance
of windings of a transformer is also dependent upon the resistivity of oil. A low value
indicates the presence of moisture and conductive contaminants.
vii. Viscosity
Viscosity of oil is the measure of the transformer oil resistance to shear. Viscosity is more
commonly known as resistance to flow for fluids. If lubricating oil is considered as a
series of fluid layers superimposed on each other, the viscosity of the oil is a
measurement of the resistance to flow between the individual layers. A high viscosity
50
implies high resistances to flow while a low viscosity indicates a low resistance to flow.
Viscosity varies inversely with temperature. Viscosity is measured using Red wood
viscometer at the temperature of 300 0C. The test is carried out as per ASTM D 2162-06,
IP 70 [19].
viii. Interferential Tension (IFT)
The IFT measures the tension at the interface between two liquids (oil and water). Which
do not mix and is expressed i.e. dyne/cm the test is sensitive to the presence of oil decay
product & soluble polar contaminants from solid insulating materials. Good oil will have
an IFT of between 40 -50 dynes /cm. Oil oxidation products lower the IFT and have an
affinity for both water (hydrophilic) and oil. This affinity for both substances lowers the
IFT. The greater the concentration of contaminants the lower the IFT, with a badly
deteriorated oil having an oft of 18 dynes/cm or less. IFT is carried out as per standards
ASTM – D 971.
3.9.2 Traditional Tests for Fault Diagnosis
These tests are used for the fault diagnosis of transformers during service as well to find
the reasons of failures, like a postmortem carried out on human body. These are:
i. Dissolved gas analysis (DGA)
Dissolved Gas Analysis is the most important tool in determining the condition of a
transformer. It is the first indicator of a problem and can identify deteriorating insulation
and oil, overheating hot spot, parallel discharge, and arcing. The health of the oil is
reflective of the health of the transformer itself. DGA consist of sending transformer oil
samples to a commercial laboratory for testing. The most important indicators are the
individual and total dissolved combustible gases (TDCG) and total combustible gases
(TCG) generation rates based on IEC 60599-1999 and IEEE C- 57-104 TM
2008.
ii. Degree of polymerization (DP)-( Standards ASTM D 5837-1996)
One of the most dependable means of determining paper deterioration and remaining life
is the degree of polymerization (DP) test of the cellulose. The cellulose molecule is made
up of long chain of glucose ring which form the mechanical strength of the molecule of
the paper. DP is the average number of these rings in the molecule. As paper ages or
deteriorates from heat, acids, oxygen on water, the number of these rings decrease .when
the insulation is new, the DP is typically between 1,000-1,400, reaches around 200, the
insulation has reached the end of the life. All mechanical strength of the insulation has
been lost; the transformer must be replaced. When cellulose insulation decomposes due to
overheating chemicals in addition to CO2 and CO are released and dissolved in the oil,
51
these chemical compound are known as furanic compound or furans. The most important
one, for our purposes is 2-frfuraldehyde. On healthy transformer there are no detectable
furans in the oil, or they are less than 100 ppb. In case where significant damage to paper
insulation from heat has occurred, furans level has been found to be at least 100 ppb and
up to 70,000 ppb.
3.10 IFT-NN RELATIONSHIP
Studies have shown in Table 3.4 that a definite relationship exists between acid number
(NN) and Interfacial Tension (IFT) [111]. An increase in NN should normally be
followed by a drop in IFT. The IFT test is a powerful tool for determining how insulating
oil has performed and how much life is left in the oil before maintenance is required to
prevent sludge. The IFT provided an excellent back up test for the NN. IFT not
accompanied by a corresponding increase in NN indicates polar contamination which has
not come from normal oxidation.
Although a low IFT with a low NN is an unusual situation, it does occur because of
contamination such as solid insulation materials, compounds from leaky pot heads or
bushings, or from a source outside the transformer.
Table 3.4: Relationship between IFT Vs NN
CORRELATION BETWEEN IFT Vs NN IN OIL FILLED TRANSFORMERS
Neutralization Number Vs sludge
NN(mg/KOH/g) %(out of 500 transformers) Units Sludged
0.00-0.10 0 0
0.11-0.20 38 190
0.21-0.60 72 360
0.61 and above 100 500
Interfacial tension Vs Sludge
IFT (Dynes/cm) %(out of 500 transformers) Units Sludged
Below 14 100 500
14-16 85 425
16-18 69 345
18-20 35 175
20-22 33 165
22-24 30 150
Above 24 0 0
52
3.10 THERMAL DEGRADATION TEST ON PURE TRANSFORMER OIL
Permissible critical parameter specifications as per IS 1866-200 are shown in Table 3.5.
Table 3.5: Permissible limits as per IS 1866-2000
S. No Parameters Highest equipment voltage, KV
<72.5 72.5 to 170 >170
1 Electric Strength KV/mm, Min >30 >40 >50
2 Dielectric dissipation factor @ 90
oC, Max
1.0 1.0 0.2
3 Specific resistance, 90 oC 10
12
ohm.cm, Min
0.1 0.1 0.1
4 Water content, ppm, Max ≤50 ≤40 ≤20
5 Neutralization value, mgKOH/g,
Max
0.3 0.3 0.3
6 Interfacial tension mN/m, Min 15 15 15
7 Flash point, oC, Min Min 125
oC or Max decrease by 15
oC
8 Sediment & Sludge Nil Nil Nil
The Initial sample parameters for pure transformer oil are given in the Table 3.6.
Table 3.6: Initial samples parameters
No. Parameters Values
1 Acidity (mgKOH/g) 0.0269
2 Viscosity (centistoke) 20.06
3 Breakdown
Voltage (KV/2.5mm)
15
4 Flash point (0C) 160
5 Volume
Resistivity(ohm-cm)
339X109
6 Dielectric constant 1.08
7 Tan-delta(900C) 0.1444
Various changes in the critical characteristics like acidity BDV, flash point, loss factor,
volume resistivity, dielectric constant, and viscosity may be examined. Test results for
thermal degradation of oil are given in Table 3.7. Various parameters changes are noticed.
Acidity value slowly increase at the initial stage of heating up to 10 hours of heating, then
53
after 20 hours of heating it remains unchanged. It shows that acidity value after 30 hours
of heating was within the accepted level.
BDV gets increases at the initial stage; the reason is the moisture content in oil gets
evaporated at starting stage. On increasing the hours of heating it trends to decrease
rapidly. Flash point variation shows a steady level up to 18 hours of heating, then its
value gets sharp increase in 24 hours of heating, but finally the characteristics shows a
steady margin value of 160 0C to 161
0C. Fire point value characteristics shows a rapid
increase for 10 hours of heating and then it tends to lower, finally after 18 hours of
heating the value gets saturated in the range of 230 0C.
Table 3.7: Thermal degradation test on pure transformer oil
Sample
No of
Hrs of
heating
Total Acid
value
(mgKOH/g)
BDV
kV/mm
Flash
Point
(oC)
Volume
resistivity
X109
Loss
tangent
Dielectric
constant
1 5 0.02940 25 160 119 0.385 1.56
2 7 0.03205 33 160 452.36 0.117 0.56
3 10 0.03457 23 160 145.76 0.35 1.54
4 12 0.03478 25 160 123.77 0.36 1.49
5 14 0.03479 20 160 219.89 0.087 0,38
6 16 0.03478 19 160 201 0.265 0.76
7 18 0.0336 20 159 198 0.301 0.80
8 20 0.03501 18 160 195 0.372 0.81
9 24 0,03490 17 161 188 0.388 1.28
10 30 0.03491 16 160 180 0.396 1.51
Volume resistivity at initial stage shows a peak increase, but when the oil is heated the
value of volume resistivity trends to decrease. It is inferred that resistivity value of oil
reduces on days of heating. Dissipation factor value at initial stage shows a higher value
but on further heating the dissipation factor value gets decreased. Dielectric constant at
initial stage gets lowered and then it shows a steady increase.
3.12 THERMAL DEGRADATION TEST ON OIL WITH PRESS-BOARD AND
KRAFT PAPER
Another set of tests with press board and craft paper, all the test samples were heated up
to 5, 7, 9, 12, 15, 18, 20 Hours. Test is conducted to simulate the real time thermal
54
degradation of oil along with deterioration of solid insulation like press board and craft
paper. The total weight of press board and kraft paper mixed along with oil is of 300gm.
Maximum temperature maintained inside the apparatus is of 100 0C. Various changes in
the critical characteristics like BDV, Flash point, Fire point, Loss factor, Resistivity,
Dielectric constant, and Acidity were examined. Test results for Thermal Degradation test
on oil along with Press board and paper is given in Table 3.8.
Table 3.8: Thermal Degradation test on oil along with Press board and Kraft paper
Sample
No of
Hr of
heating
Total
Acid
value
BDV Flash
Point
Volume
resistivity
X109
Loss
tangent
Dielectric
constant
1 5 0.5142 17 160 506.39 0.078 0.33
2 7 0.8530 10 170 2242.23 0.008 0.02
3 9 0.7462 16 165 4002.14 0.006 0.01
4 12 0.7831 15 154 4000.58 0.005 0.01
5 15 0.8310 13 163 3780.00 0.008 0.02
6 18 0.8309 13 164 3500.27 0.008 0.02
7 20 0.8308 11 162 2289.91 0.009 0.01
Various observations are made from Table 3.8. Acidity value slowly increase at the initial
stage of heating up to 10 hours of heating, then after 20 hours of heating it remains
unchanged. It shows that acidity value after 30 hours of heating was within the accepted
level. Viscosity value shows a sharp increase at 7 hours of heating and then trends to
decrease down rapidly therefore flow rate of oil gets increased. BDV gets increases at the
initial stage the reason is the moisture content in oil gets evaporated at starting stage. On
increasing the hours of heating BDV trends to decrease rapidly. Flash point to hours of
heating. Flash point values at the initial stage gets sharp increase and then further heating
it will gets reduced. Fire point value shows increasing trend characteristics and more or
less it attains a saturated value after 15 hours. At the initial stage the peak absorption
value sharply increases up to 9 hours of heating and then it shows a gentle increase of
value. Transmission property of oil shows a variation of up and down values, at one stage
(18 hours) the value starts to decrease. Volume resistivity value obtained in the sample of
oil along with press board and craft paper is very high compared to characteristics value
of pure oil. Volume resistivity at initial stage shows a peak increase, but when the oil is
55
heated the value of volume resistivity trends to decrease. It is conferred that resistivity
value of oil reduces on days of heating. Loss factor value obtained in the sample of oil
along with press board and Kraft paper is very low compared to characteristics value of
pure oil. Dissipation factor value at initial stage shows a higher value but on further
heating the dissipation factor value gets decreased. Shows the variation of Dielectric
constant to hours of heating, the value slowly degrades on increase of hours of heating.
3.12 DEGRADATION TESTS DUE TO ACCUMULATION OF WATER
CONTENT
In this test all the test samples were mixed in the proportionate of 10%, 15%, 20%, and
25% of 700 ml of water. Trends of changes in the characteristics were analysed. Test is
conducted to simulate the characteristics changes due to availability of higher
concentration of water content (ml) in oil. Various changes in the critical characteristics
like BDV, Flash point, Fire point, UV response, Loss factor, Resistivity, Dielectric
constant, Acidity and Viscosity were examined. Test results for Degradation due to
accumulation of water content are given in Table 3.9.
Table 3.9: Influence of water content
Sample
Water
content
(ml)
Total
Acid
value
BDV Flash
Point
Fire
Point
Volume
resistivity
X109
Loss
tangent
Dielectric
constant
1 70 0.02944 9 180 200 Not obtain3ed because of water
contents 2 105 0.112705 16 Not obtained
because of
water
contents
3 140 0.16229 10
4 175 0.034735 4
Some of the observations are made from the test results given in Table 3.9. Acidity value
slowly increases on increase of water content, but whenever the level of water content
increases beyond certain percentage of limit the acidity value trends to decrease.
Viscosity value shows a sharp increase at 20 % of water content and then trends to
decrease down rapidly, therefore flow rate of oil gets increased. On increasing the water
content values the BDV value reduces also rapidly.
The following significant results are concluded from the oil parameter testing and
correlation between parameters test values as in Table 3.10.
56
Table 3.10: Significance of test
Sl. No. Test Method Significance
1. Interfacial
tension
Measures the
attraction force
between unlike
molecules.
Indicates polar contaminants
including deterioration products.
Index of deterioration/
contamination.
2. Flash point Measures the
minimum temperature
at which momentary
flash occurs
To find flammable mixture in
transformer oil.
3. Neutralisation
value
Measures Acidic
constituents in oil.
Acids are highly corrosive,
polymerize to form insoluble
sludge. Acids are undesirable.
4. Electric
strength
Measure of minimum
voltage at which oil
breaks.
Indicates free water, dirt, fiber or
conducting particles.
5. Dielectric
dissipation
factor or Tan-
delta
Measure of dielectric
loss in oil- energy
dissipated as heat.
Very sensitive to moisture,
conducting impurities
contaminates & products of
oxidation like acid.
6. Specific
resistance
Measure of resistance
per unit volume.
Sensitive to moisture,
contaminates & products of
oxidation like acid.
Gauge of Purity of oil.
7. Water content Measure the total
water present on oil.
Moisture cause metal corrosion
reduces electric strength &
increase the rate of deterioration
of oil / paper / press board.
8. Sludge
content
Measure the sludge
formed by
combination of
degradation products
of oxidation.
Sludge deposits on core, coil &
cooling ducts lower heat
dissipation & increases aging.
57
3.13 SUMMARY
The cost to repair and replace a power transformer is very high. The time taken to repair
or to replace transformer results in loss and inconvenience to the utilities. Since
transformer oil occupies maximum portion and performs various functions in the
transformers, its quality get deteriorated quickly by hydrolytic, oxidation and thermal
effects, which further degrades or weakens the dielectric property of the oil and stars
degrading the properties of sold insulation of the transformers.
Critical parameters of oil has been analyzed for various levels of thermal degradation
with combinations of insulation materials. Ageing characteristics analyses of transformer
oil using refractive property give some idea towards level of degradation. Correlations
among the electrical characteristics were examined and the results were inferred.
Effective condition monitoring of transformer oil surely paves the way for increased life
time of insulation, as a result catastrophic failure and forced outages shall be minimized.
58
CHAPTER 4
DISSOLVED GAS ANALYSIS FOR FAULT DIAGNOSIS
4.1 INTRODUCTION
It is has been established that transformer oil degrades by three factors; hydrolytic,
oxidative and thermal. Transformer oil is a long chain of hydro-carbon molecules. The
molecular chain breaks into smaller size particles under these three degrading factors.
Thus several gases are generated in transformer oil and keep dissolving in the oil. Access
gases get collected in Buckotz‘s relay of the transformers. Therefore, it is necessary to
monitor and diagnose the health condition of the transformer through dissolved gas
analysis periodically. This chapter describes the mechanism of gas formation due to
decomposition of oil, decomposition of cellulose insulation and various other sources of
gases in the transformer. It also involves the limits of the gases contents and associated
faults. Further sampling procedures along with effects of various factors affecting the
dissolved gas analysis (DGA) interpretations wrongly has been described. Some of the
recent conventionally used DGA analyzers are described briefly. Gas concentration levels
and rates of gas increase in oil (typical, alarm and pre-failure values) are indicated as a
function of time. The chapter also gives recommended safe values of various gases.
The present chapter classifies various faults which can be detected with the help of DGA.
Guidelines for interpretation of faults adopted in various countries and standards are
given in tables of this chapter. Different diagnostic standards/methods for interpretation
of dissolved gases such as IEEE standards ( Dornenburg‘s, Roger‘s and Key gas methods
), IEC ratio codes, Duval Triangle method are described. Most of the interpretation
techniques given in IEEE standards C-57.104TM
1997 & 2008, which are also basis for
the Bureau of Indian Standards (BIS) are illustrated to the large extant. This IEEE
standard is also more or less drawn from IEC 60599 codes. Methods used by other
countries for fault diagnosis are also listed with brief descriptions. Some case studies on
various transformers in different organizations show comparative analysis by Rogers ratio
and Duval triangle. Sampling intervals and action taken based on total combustible gases
(TCG) and total dissolved combustible gases TDCG are also given in the tables.
4.2 DECOMPOSITION OF OIL
Mineral Insulating oils are made of a blend of different hydrocarbon molecules containing
CH3, CH2 and CH chemical groups linked together by carbon-carbon molecular bonds.
Scission of some of the C-H and C-C bonds may occur as a result of electrical and
59
thermal faults, with the formation of small unstable fragments, in radical or ionic form,
such as H, CH3, CH2, CH OR C (among many other more complex forms), which
recombine rapidly, through complex reactions, into gas molecules such as hydrogen (H-
H), methane (CH4), ethane (C2H6), ethylene (C2H4) or acetylene (C2H2). C3 and C4
hydrocarbon gases, as well as solid particles of carbon and hydrocarbon polymers, are
other possible recombination products. The gases formed dissolve in oil, or accumulate as
free gases if produced rapidly in large quantities, and may be analyzed by DGA according
to IEC 60567.
Low energy faults such as partial discharges of the cold plasma type (corona discharges),
favour the scission of the weakest C-H bonds (338 kJ/mole) through ionization reactions
and the accumulation of hydrogen as the main recombination gas. More and more energy
and/or higher temperatures are needed for the scission of the C-C bonds and their
recombination into gases with a C-C single bond (607 kJ /mole), C=C double bond (720
kJ/mole) or CºC triple bond (960 kJ/mole), following processes bearing some similarities
with those observed in the petroleum oil-cracking industry.
4.3 FORMATION OF GASES IN TRANSFORMER OIL
Mineral oils (transformer oil) are composed of saturated hydrocarbons called paraffins,
whose general molecular formula is CnH2n+2 with n in the range of 20-40. The cellulosic
insulation material is a polymeric substance whose general molecular formula is
[C12H14(OH)6]n with n in the range of 300-750. Gases are formed inside an oil-filled
power transformer, in that various gases begin forming at specific temperatures [5] shown
in Fig. 4.1. Hydrogen and methane begin to form in small amounts around 150 °C. Notice
from the chart that beyond maximum points, methane (CH4), ethane and ethylene
production goes down as temperature increases. At about 250°C, production of ethane
(C2H6) starts. At about 350 °C, production of ethylene (C2H4) begins. Acetylene (C2H2)
starts between 500 °C and 700 °C. In the past, the presence of only trace amounts of
acetylene (C2H2) was considered to indicate a temperature of at least 700 °C had
occurred. But recent discoveries have led to the conclusion that a thermal fault (hot spot)
of 500 °C can produce trace amounts (a few ppm). Larger amounts of acetylene can only
be produced above 700 °C by internal arcing. Notice that between 200 °C and 300 °C, the
production of methane exceeds hydrogen. Starting about 275 °C and on up, the
production of ethane exceeds methane. At about 450°C, hydrogen production exceeds all
others until about 750 °C to 800 °C; then more acetylene is produced. It should be noted
that small amounts of H2, CH4, and CO are produced by normal aging. Thermal
60
decomposition of oil-impregnated cellulose produces CO, CO2, H2, CH4, and O2.
Decomposition of cellulose insulation begins at only about 100 °C or less. Therefore,
operation of transformers at not more than 90 °C is imperative. Faults will produce
internal ―hot spots‖ of far higher temperatures than these, and the resultant gases show up
in the DGA.
Figure 4.1: Gas generation chart [3]
4.4 SOLUBILITY OF GASES IN TRANSFORMER OIL
The solubilities of the fault gases in Transformer oil as well as their temperature
dependence are also important factors for consideration in fault gas analyses. It should be
noted that there are almost two orders of magnitude difference between the least soluble
(H2) and the most soluble (C2H2) gas. The majority of gases that are indicative of faults
are also those that are in general the more soluble in the oil. When the rates of gas
generation are being followed it is important to take into account the solubility of these
gases as a function of temperature (Fig. 4.2). Over a temperature range of 0-80oC some
gases increases in their solubility upto 79% while others decrease their solubility upto
66%.
61
Figure 4.2: Relative solubility (Y-Axis) as function of temperature OC (X-Axis)[112]
Oil may oxidize with the formation of small quantities of CO and CO2, which can
accumulate over long periods of time into more substantial amounts.
4.5 DECOMPOSITION OF CELLULOSE INSULATION
The polymeric chains of solid cellulose insulation (paper, pressboard, wood blocks)
contain a large number of an hydro glucose rings, and weak C-O molecular bonds and
glycosides bonds which are thermally less stable than the hydrocarbon bonds in oil, and
which decompose at lower temperatures. Significant rates of polymer chain scission occur
at temperature higher than 105°C, with complete decomposition and carbonization above
300°C. Mostly CO, CO2 as well as H2O are formed, in much larger quantities than by
oxidation of oil at the same temperature, together with minor amounts of hydrocarbon
gases and furanic compounds. The latter can be analyzed according to IEC 61198, and
used to complement DGA interpretation and confirm whether or not cellulosic insulation
is involved in a fault. CO and CO2 formation increases not only with temperature but also
with the oxygen content of oil and the moisture content of paper.
4.6 OTHER SOURCES OF GASES IN TRANSFORMERS
Gases may be generated in some cases not as a result of faults in the equipment but
through rusting or other chemical reactions involving steel, uncoated surfaces or
protective paints.
Hydrogen may be produced by reaction of steel with water, as long as oxygen is available
from the oil near by. Large quantities of hydrogen have thus been reported in some
transformers that had never been energized. Hydrogen may also be formed by reaction of
free water with special coatings on metal surfaces, or by catalytic reaction of some types
of stainless steel with oil, in particular oil containing dissolved oxygen at elevated
62
temperatures. Hydrogen may also be formed in new stainless steel, absorbed during its
manufacturing process, or produced by welding and released slowly into the oil.
Hydrogen may also be formed by the decomposition of thin oil film between overheated
core laminates at temperatures of 140°C and above. Gases may also be produced by
exposure of oil to sunlight or may be formed during repair of the equipment. Internal
transformer paints, such as alkyd resins and modified polyurethane‘s containing fatty
acids in their formulation, may also form gases.
These occurrences, however are very unusual, and can be detected by performing DGA
analyses on new equipment, which has never been energized, and by material
compatibility tests. The presence of hydrogen with the total absence of other hydrocarbon
gases, for example may be indication of such a problem.
In-service transformers always have some fault gases dissolved in their oil. Only when
these levels exceed some threshold value, a fault is suspected. Several recommended safe
values have been published. Limits as per IEC 60599 in power transformers are listed in
Table 4.1 and Table 4.2 for permissible range on ageing of transformers.
Table 4.1: Recommended safe value of various gases in [µl/L(ppm)]
Table 4.2: Permissible concentration of dissolved gases on ageing
Gas H2 CH4 C2H2 C2H4 C2H6 CO CO2
Violations Limits
Max.(ppm)
100 120 35 50 65 350 2500
Typical Ratio of Gas
Increased(ppm/day)
<5 <2 <0.1 <2 <2 <50 <200
Gas Less than 4 years in
service (ppm)
4-10 years in
service (ppm)
More than 10 years in
service (ppm)
Hydrogen 100-150 200-300 200-300
Methane 50-70 100-150 200-300
Acetylene 10-30 30-50 50-80
Ethylene 100-150 150-200 200-400
Ethane 30-50 100-150 150-300
Carbon Monoxide 200-300 400-500 550-700
Carbon Dioxide 3000-3500 3500-600 6000-12000
63
A list of key gases and their related and key faults are shown in Table 4.3. This has been
discussed in detail in IEEE Std. C-57, 104TM
-2008, 1991 ―IEEE Guide for the
Interpretation of Gases Generated in Oil-Immersed Transformers‖ [41, 45].
Table 4.3: Key gases generated by particular fault
4.7 DISSOLVED GAS ANALYSIS
Testing of the insulating oil is one of the most common tests used to evaluate the
condition of transformers in service. The Dissolve Gas Analysis (DGA) test is one of the
important diagnostic techniques for fault diagnosis in power transformers. Insulating oils
under abnormal electrical or thermal stresses break down to liberate small quantities of
gases. The composition of these gases is dependent upon the type of fault. By means of
dissolved gas analysis, it is possible to distinguish faults such as partial discharge
(corona), over-heating, and arcing in a great variety of oil filled equipment. A number of
samples must be taken over a period of time to discern trends and to determine the
severity and progression of incipient faults. The gases in oil tests commonly evaluate the
concentration of hydrogen, methane, acetylene, ethylene, ethane, carbon monoxide,
carbon dioxide, nitrogen, and oxygen.
The relative ratios and the amount of gas detected in the sample are used to detect
problems with the insulation structure.
Information from the analysis of gasses dissolved in insulating oil is one of the most
valuable tools in evaluating the health of a transformer and has become an integral part of
preventive maintenance programs. Data from DGA can provide:
i. Advanced warning of developing faults
ii. Monitoring the rate of fault development
iii. Confirm the presence of faults
iv. Means for conveniently scheduling repairs
v. Monitoring of condition during overload
Material Condition Key Gases
Oil Overheated >110 0C Methane (CH4), Ethane (C2H6),
Ethylene (C2H4), Organic Acids
Oil Electric Stress (Partial Discharge
and Arcing to 1000 0C)
Hydrogen (H2), Acetylene (C2H2),
Waxes and Water
Cellulose Overheated >150 0C Carbon Monoxide (CO), Carbon
Oxide (CO2), Water (H2O)
64
4.8 FAULTS IN PAPER AND FAULTS IN TRANSFORMER OIL
Faults in paper are generally considered as more serious than faults in oil, because paper
is often locat5d in areas of high electric field (in the windings, or as voltage barriers), and
the destruction of paper insulation may lead to short circuits or severe arcing. Faults in
paper, fortunately, are much less frequent than faults in oil (typically, in 10 % of cases
only), however, because of the more serious consequences, their detection by DGA or
other means is of great interest.
A popular way of detecting faults in paper by DGA is by looking at the CO2 to CO ratio.
Values < 3 are a good indication of faults in paper of a temperature > 200 to 300°C
(including arcing), where paper degrades very rapidly or even carbonizes. However, there
is always a large background of CO and CO2 in oil (except in the first years of operation
of the transformers) so that caution should be exercised when interpreting the value of
this ratio. Using increment values of CO and CO2 over the last analysis are preferable, but
the uncertainty on the incremented ratio is high and should be calculated to determine its
reliability.
Values of the CO2 to CO ratio > 10 are also an indication of thermal faults in paper at
temperatures < 150 °C, but such temperatures have only a long term aging effect on paper
and on the reduction of transformer life, which can be more precisely evaluated by furans
formation, (when regular kraft, not thermally-upgraded paper, is used).
A sharp increase in the formation of furans may in some cases be a confirmation of faults
in paper at temperatures > 250 °C.
4.9 SAMPLING OF LIQUID INSULATION AND GAS EXTRACTION FROM
OIL
DGA has become very renowned as an effective method that aims at predicting the life
span of transformer. The quality and accuracy of DGA depends on right performance of
the method. The first part consist of appropriate oil sampling, the second part consist of
extraction of dissolved gases and third part for qualitative and quantitative analysis of
extracted gas by gas chromatography or spectroscopy. This portion describes oil sampling
procedure to be adopted for DGA analysis and the precautions to be taken during oil
sampling. Influence of light, air bubbles and temperature on storage of oil sample is
examined.
4.9.1 Sampling Procedures
The validity of the test results is dependent upon being certain that the oil sample is truly
representative of the oil in the equipment. Thus, it is essential that the sample container of
65
whatever type be free of contaminants, that the point of sampling within the equipment
will produce a representative sample, and that sampling taps and connecting tubing be
purged before the sample is taken. Ordinarily, samples for physical tests can be taken in
1-L (1-ft) glass bottles, plastic containers with screw top, Teflon-lined containers, or
stainless steel cylinders. Glass bottles may either be cork or glass stopper or fitted with
screw caps having cork or aluminum liners (inserts). Corks should be of good quality. DO
NOT use rubber stoppers. Gas-in-oil samples may be taken in 150 or 1,000-mL stainless
steel cylinders the 1,000-mL cylinders provide enough samples for both physical and gas-
in-oil tests. Steel containers, except Teflon lined, should be rinsed in chlorothene or
another comparable nonflammable solvent, dried, and washed in strong soapsuds. Then
they should be thoroughly rinsed with water, dried in an oven at 105°C (220°F) for
several hours, and removed from the oven. They should be corked while still warm. As
the bottles cool, dipping the necks in wax should seal them, then; the bottles should be
stored for future use.
These bottles should be opened only when the bottle temperature and the ambient
temperature are the same or nearly so. Because of the importance of properly cleaned and
treated containers, laboratory-approved containers should be used. Oil samples in glass or
other clear containers should be kept out of sunlight. Sunlight will break down materials
in the oil and affect the test results.
4.9.2 Taking Gas Samples for Gas Chromatograph Analysis
This procedure has been developed to maintain uniformity of all gas samples taken in the
field while maintaining the principal objective of obtaining a representative sample that is
free of all contaminants. Special stainless steel or glass cylinders are to be used for
collecting samples of gas for use with the gas chromatograph. These containers are NOT
to be used for any other purpose.
If a laboratory performs the required testing, they normally clean the container to
eliminate all contaminants. Special care should be taken to avoid getting moisture in the
sample since moisture can damage the analyzing equipment.
4.10 INFLUENCE OF STORAGE OF OIL SAMPLES ON DGA
Oil sample containers should be free from contamination and air tight sealed. Sample
container should be fully filled up and kept without air. Adequate sample storage is of
vital importance to obtain accurate results. Some experiments were conducted in order to
investigate the influence of factors such as air bubbles, light and temperature on stored oil
samples [113, 114]. Samples were stored for certain periods and gas in oil analysis was
66
conducted .The oil sample is introduced into a container and is steadily agitated and re-
circulated at constant temperature in order to establish the equilibrium between headspace
gases and oil. Attention was focused on the concentration of hydrogen as a critical
component.
4.10.1 Effect of Air Bubbles
Air bubbles of different volumes were introduced into the sample and then stored in
darkness for 4 days. Other samples without air bubbles were also stored in darkness as a
reference for comparison. It was observed that concentration of H2, CO, CH4 tends to
decrease as the volume of air bubble increases see Fig. 4.3. Especially the concentration
of H2 was significantly affected by a decrease of approx. 35% when air volume is more
than 1% of the sample volume. Bigger air bubbles allow larger diffusion of the less
soluble components in oil; hence there will be lower concentration of dissolved gases in
oil resulting in erroneous DGA result.
Figure 4.3: Influence of air bubbles of different volume on DGA results [113, 114]
4.10.2 Effect of Temperature
Generally samples should be stored at ambient temperature and in darkness. However
during sample transportation to laboratories, it can be exposed to high temperatures. The
test was carried at 90º C in darkness without air, Concentration remained almost
invariable and therefore storage at relatively high temperature did not cause significant
variation see Fig.4.4.
Sometimes Stray gassing is observed i.e. increased production of hydrogen and in some
cases also ethane and methane because of oil interactions with metallic parts such as
copper when they are heated at relatively low temperatures. Stray gassing can be
mistaken for a fictitious fault and DGA analysis can lead to wrong interpretation.
67
Figure 4.4: Stray gassing effect at 90 0C [113, 114]
4.10.3 Effect of Light
The tests were done with samples prepared with air-saturated oil of new and aged
condition. Result shows that concentration of H2 and CO2 were most affected. In case of
new oil, the concentration of CO reduces 12% when stored in darkness, while H2
concentration does not suffer any change since its concentration is less than 5 ppm.
It was seen that under the influence of sunlight the concentration of H2 increases up to
24% while there is no change in CO2 see Fig. 4.5 and Fig. 4.6. In case of darkness H2
concentration increases approx. 10% due to further reactions occurring during storage. No
significant deviation was seen in other components with respect to reference values.
Figure 4.5: Influence of sunlight and darkness on concentration of H2 and CO2 on New
oil[113, 114]
68
Figure 4.6: Influence of sunlight and darkness on concentration of H2 and CO2 on Aged
oil[113, 114]
.4.11 TAKING OIL SAMPLES FOR DGA
For in-service liquid filled electric apparatus, sampling of the liquid dielectric provides a
method to determine the condition of the solid and liquid insulation as well as the
operating condition of the apparatus without opening or de-energizing the apparatus.
Sampling provides a means to check the condition of oil in storage whether it be new or
used, to determine if it complies with specifications. Sampling can also help to determine:
i. If accidental mixing of different dielectric liquids has taken place;
ii. if the method of transportation contaminated the dielectric liquid; and
iii. if the handling equipment to transfer the dielectric liquid contaminated the product.
4.12 TECHNIQUES FOR SAMPLING INSULATING LIQUIDS
There are a number of practices that clearly define the proper way to retrieve samples
from electrical apparatus or storage containers. These practices have been developed over
many years and have incorporated the expertise of many individuals
i. Doble Reference Book on Insulating Liquids and Gases
ii. ASTM D 923: Standard Practice for Sampling Electrical Insulating Liquids
iii. ASTM D 3613: Standard Practice for Sampling Electrical Insulating Oils for Gas
Analysis and Determination of Water Content
iv. IEC 60475: Method of Sampling Liquid Dielectrics
v. IEC 60567: Guide for the Sampling of Gases and of Oil from Oil-filled Electrical
Equipment and for the Analysis of Free and Dissolved Gases.
Sampling procedures and lab handling are usually areas that cause the most problems in
getting an accurate DGA. There are times when atmospheric gases, moisture, or hydrogen
take a sudden leap from one DGA to the next. As has been mentioned, at these times, one
should immediately take another sample to confirm DGA values. It is, of course, possible
69
that the transformer has developed an atmospheric leak, or that a fault has suddenly
occurred inside. More often, the sample has not been taken properly, or it has been
contaminated with atmospheric gases or mishandled in other ways. The sample must be
protected from all contamination, including atmospheric exposure.
Do not take samples from the small sample ports located on the side of the large sample
(drain) valves. These ports are too small to adequately flush the large valve and pipe
nipple connected to the tank; in addition, air can be drawn past the threads and
contaminate the sample. Fluid in the valve and pipe nipple remains dormant during
operation and can be contaminated with moisture, microscopic stem packing particles,
and other particles. The volume of oil in this location can also be contaminated with
gases, especially hydrogen. Hydrogen is one of the easiest gasses to form. With hot sun
on the side of the transformer tank where the sample valve is located, high ambient
temperature, high oil temperature, and captured oil in the sample valve and extension,
hydrogen formed will stay in this area until a sample is drawn.
The large sample (drain) valve can also be contaminated with hydrogen by galvanic
action of dissimilar metals. Sampling valves are usually brass and a brass pipe plug
should be installed when the valve is not being used. If a galvanized or black iron pipe is
installed in a brass valve, the dissimilar metals produce a thermocouple effect, and
circulating currents are produced. As a result, hydrogen is generated in the void between
the plug and valve gate. If the valve is not flushed very thoroughly the DGA will show
high hydrogen.
Oil should not be sampled for DGA purposes when the transformer is at or below
freezing temperature. Test values which are affected by water (such as dielectric strength,
power factor, and dissolved moisture content) will be inaccurate.
This is typically not a problem with conservator transformers. If the transformer is
nitrogen blanketed, look at the pressure/vacuum gauge. If the pressure is positive, go
ahead and take the sample. If the pressure is negative, a vacuum exists at the top of the
transformer. If there is a vacuum at the bottom, air will be pulled in when the sampling
valve is opened. Wait until the pressure gage reads positive before sampling. Pulling in a
volume of air could be disastrous if the transformer is energized.
If negative pressure (vacuum) is not too high, the weight of oil (head) will make positive
pressure at the sample valve, and it will be safe to take a sample. Oil head is about 2.9
feet (2 feet 10.8 inches) of oil per pounds per square inch (psi).
70
If it is important to take the sample even with a vacuum showing at the top, proceed as
described in standards.
Use the sample tubing and adaptors described below to adapt the large sample valve to
1/8 inch tygon tubing. Fill a length (2 to 3 feet) of tygon tubing with new transformer oil
(no air bubbles) and attach one end to the pipe plug and the other end to the small valve.
Open the large sample (drain) valve a small amount and very slowly crack open the small
valve. If oil in the tygon tubing moves toward the transformer, shut off the valves
immediately. Do not allow air to be pulled into the transformer. If oil moves toward the
transformer, there is a vacuum at the sample valve. Wait until the pressure is positive
before taking the DGA sample. If oil is pushed out of the tygon tubing into the waste
container, there is a positive pressure and it is safe to proceed with DGA sampling. Shut
off the valves and configure the tubing and valves to take the sample per the instructions
in the standards.
4.13 GAS EXTRACTION FROM SAMPLE
The key gases are extracted by a degassing technique [110], for which often a so called
Toepler-pump-system is used, Fig. 4.7. The extraction is efficient and gives good results
for a subsequent analysis. This vacuum pump contains usually high amounts of mercury,
thus being a risk for the operating personnel and the environment.
Figure 4.7: Toepler pump for laboratory use [110]
71
Alternative procedures like the so called headspace extraction, where the amount of
dissolved gases is identified from a dissolution of the gases in the liquid into a defined gas
in the headspace of a phinole lead to somewhat different results and is as explained in the
standard IEC 60567 more imprecise.
In this contribution therefore another system for DGA usable even at the transformer-
operating site is introduced. This system consists of a mercury free gas extraction device
combined with a portable gas chromatograph (GC), which identifies the type and amount
of gases after the extraction from the oil sample.
The GC device contains a capillary column, in which the different gases are separated in a
carrier gas flow, for which in most cases Argon (Ar) is used. After passing the ripping-up
column the gases are injected in a Thermal Conductivity Detector (TCD) or a Flame
Ionization Detector (FID), each one sensitive for a special fraction of the detectable gases.
4.14 NEW DGA SYSTEM
With the new DGA system the analysis is done in 2 steps with high selectivity and
accuracy fully compliable with laboratory procedure.
a) Degassing of the taken oil sample through vacuum extraction (according to IEC
60567).
b) Dissolved gas analysis by gas chromatography with TCD (thermal conductivity
detector), FID (Flame Ionization detector) and Methanizer.
Both steps are synchronized by a micro-controller, which guarantees a fast and reliable
measurement routine.
4.15 VACUUM DEGASSING
A new invented vacuum degassing method (Patent DE 102 52 652 4) degases the oil
sample in multiple periods. After the extraction by a 4 step vacuum pump the gases are
enclosed in a vacuum chamber where they are separated. The gases are evacuated from
the vacuum chamber with a vacuum pump; therefore a very high sensitivity can be
reached. The extraction principle is schematically shown in Fig. 4.8 and a photo of the
device is given in Fig. 4.9. Inlet and outlet of the oil sample as well as the switching valve
are microprocessor controlled and a linear pressure sensor measures online the total gas
content (TGC), which is an important parameter concerning bubbling problems.
The injection of the gas samples into the gas chromatograph is made via a fixed mounted
and heated transfer pipe but also an additional port for a withdrawal of gas samples is
installed.
72
Figure 4.8: Schematic view of the gas extraction unit
Figure 4.9: Photo of the gas extraction unit [110]
4.16 MOBILE GAS CHROMATOGRAPHY
The analysis of the dissolved gases is done after the automatic or manual injection in an
automatic sample loop (6 port valve) of the gas chromatograph. The gas sample is
separated in an inert gas stream into 11components. This guarantees a high selectivity
with no Interference with other gas signals.
A capillary column (type Carboxen 1010, length 30m x ID 0.32 mm) is used for the
separation of the different compounds, whereas the detection is carried out with a TCD
(Thermal Conductivity Detector) and FID (Flame Ionization Detector). The compounds
CO and CO2 are converted with a methanizer for detection in the FID.
73
This procedure guarantees a very high sensitivity and the determination of all mentioned
key gases including Propane and Propene.
Due to the small size of the GC, which is shown in Fig. 4.10, the device is easy portable
and therefore also suitable for analyses on-site e.g. from free gas in the Buchholz-relay.
This is often very important in cases in which the Buchholz-relay responds after a failure
thus the transformer is protected against expanded failures.
In these cases the free gas accumulated in the relay can be analyzed on the spot, thus a
first statement on the severity of the fault becomes possible. Therefore a two injection
port is installed, thus it is also possible to monitor the Buchholz gases automatically. As
soon as the relay releases, the measurement routine is interrupted and the relay gases are
analyzed accordingly.
Figure 4.10: Mobile GC [110]
4.17 ONLINE DGA SYSTEM
The introduced systems can be used as an online DGA system as shown in Fig. 4.11
which is then controlled by a microprocessor. A serial interface (RS 232) allows
transferring the data to a remote computer. The intervals of the analysis cycle can be
freely defined by the operator (number of analyses per day or hour).
The system has 8 additional analog inputs (0-1 V) for the measurement of other
transformer parameters, like oil temperature, winding temperature, ambient conditions
etc. The connection of these external sensors is realized via relay contacts, which can be
specified by the customer as NC (normally closed) or NO (normally opened).
74
Figure 4.11: Online DGA system [110]
Beside the display of the results in ppm all other operating conditions are registered
online and stored in a database together with all other parameters. The graphical display
of the results is in real time on the monitor, in a table or/and as a test report in a print file.
The data can also be transmitted to a remote computer or a control center. All relevant
data are stored in a measurement file. The results are stored with other measurement
parameters in the history (log file). The history starts with the day of the commissioning.
The trend is shown in a diagram (hour / day / week / month / year) and the operator can
adjust the interval. The integrated expert system will analyze even smallest deviations in
the gas parameters. A plausibility test ensures the accuracy and reproducibility of the test
results.
Furthermore different alarm settings are possible, thus the following incidents lead to an
alarm;
i. Monitoring of all gas parameters (11 gases), excess of one parameter will activate an
alarm and data will be send to a remote control system,
ii. Trend analysis for predictive maintenance for each gas compound if the rise exceeds
beyond predefined threshold levels (considering the temperature of the transformer),
iii. Interruption of power supply,
iv. Status report of the system with automatic self diagnostic function, and
v. Export of all data and alert to a remote control system.
All threshold values can be individually selected for each gas component. The trend
analysis starts with the day of installation and is stored as an unbroken chain of evidence.
The trend will show any irregularity above 0.5 ppm or for organic components 0.1 ppm
on the FID channel. The degree of degradation of the insulation of the transformer can be
75
displayed with a comparison of the different parameters versus the time, which is an
important criterion for prophylactic control of possible defects in the transformer.
A system for DGA was introduced containing a new degassing technique and a mobile
GC, thus these devices can be combined to an online DGA transformer monitoring
system, which could enhance the reliability of transformers significantly.
Uncertainties related to oil sampling can be avoided, when this system is installed online,
thus the condition and gassing history of the transformer could be continuously
monitored.
Comparison tests with a Toepler-pump system and a high precision laboratory GC have
shown a high preciseness and good reproducibility of the system.
4.18 GASES ANALYZED BY DGA
The main gases formed by decomposition of oil and paper are summarized in Table 4.4
below. These gases dissolve in oil or accumulate above it and are analyzed by DGA.
DGA is the technique for detecting and monitoring faults in electrical equipment. The
combustible Concentration Limits differ from country to country, continent to continent
and transformer to transformer. It is not practical to set concentration limits because of
the many variations involved. The Gas Concentrations in the oil depend upon:
a) The volume of oil involved (dilution factors)
b) The age of the transformer (new or old)
c) The type of transformer;
Generator or Transmission
Sealed or free breathing
Construction of Tap changer
Table 4.4: Dissolved gases in transformer oil
Sl. No. Name Symbol
1 Nitrogen N2
2 Oxygen O2
2 Hydrogen H2
4 Carbon Monoxide CO
5 Carbon Dioxide CO2
6 Methane CH4
76
7 Ethane C2H6
8 Ethylene C2H4
9 Acetylene C2H2
Some laboratories also report the contents of C3 and C4 hydrocarbon gases formed.
4.19 FAULTS DETECTION BY DGA
Seven basic types of faults detectable by DGA have been defined:
4.19.1 Partial Discharges (PD) of the corona-type (PD)
Partial discharge is localized dielectric break down of a small portion of liquid or solid
electrical insulation system under high voltage stress, which does not bridge the space
between two conductors. Whereas corona is usually revealed by a relatively steady glow
of brush discharge in air, PD in solid systems is not visible.
Typical examples: discharges in gas bubbles or voids trapped in paper, as a result of poor
drying or poor oil-impregnation.
4.19.2 Discharges of Low Energy (D1)
Paper insulation in the transformers get carbonized under high electric stresses. This is the
way of conducting discharge through the oil between two conducting parts. Therefore,
carbonized paper is trailed under overloaded conditions, as in the process of modifying
the carbon particles in oil form.
Typical examples: Partial discharges of the sparking-type, including carbonized punctures
in paper. Low energy arcing, including surface tracking of paper and carbon particles in
oil.
4.19.3 Discharges of High Energy (D2)
Powerful discharges take place and the paper the damage by excessive carbon recovery.
Very large discharges caused the metal part to melt, thus excessive carbon recovery is
found in the transformer oil. In this fault, flow through the defective regions disables the
transformers.
Typical examples: High energy arcing, flashovers and short circuits with power follow-
through, resulting in extensive damage to paper, large formation of carbon particles in oil,
metal fusion, tripping of the equipment or gas alarms.
4.19.4 Thermal Faults of temperatures < 300°C (T1)
The continuous transformer overloaded conditions in case of emergency becomes the
reason for thermal faults. Under the temperature range less than 3000C, the insulation
paper is brown. Faults T1 are evidenced by paper turning brownish.
77
i. Brown (> 200°C)
ii. Black or Carbonized ( > 300°C)
Typical examples: Overloading, Blocked oil ducts.
4.19.5 Thermal Faults of temperatures between 300 °C and 700°C (T2)
For T2 faults, the temperature range is 300-700 0C. T2 fault is the cause of circulating
currents between the metallic parts in the transformer. Faults T2 are evidenced by:
i. Carbonization of paper
ii. Formation of carbon particles in oil.
Typical examples: Defective contacts, Defective welds, Circulating currents.
4.19.6 Thermal Faults of temperatures > 700 °C (T3)
T3 fault in the transformer happens above 700 0C. This fault is the cause of large
circulating currents in tank and core. Faults T3 are evidenced by:
i. Extensive formation of carbon particles in oil.
ii. Metal Coloration (800°C) or metal fusion (> 1000°C)
Typical examples: Large circulating currents in tank and core, Short circuits in
laminations.
4.20 STANDARDS AND CODES FOR DGA INTERPRETATIONS ADOPTED IN
INDIA
The following international and national standard/ interpretation guides of dissolved gas
analysis for fault diagnosis of power transformers are widely adopted in India and across
the world:
i. IEEE Std. C57-104TM
-2008, “IEEE Guide for the Interpretation of Gases Generated
in Oil- Immersed Transformers,” September 2008.
ii. IEC Publication 60599, “Mineral Oil Impregnated Equipments in Service- Guide to
the Interpretation, of Dissolved and Free Gases Analysis,” March 1999.
iii. IS 10593: 1992(Reaffirmed 2006) & IS-9434: (1992), ―Indian Standard Method of
Evaluating the Analysis of Gases in Oil-Filled Electrical Equipment in Service‟‟ (First
Revision), 1995.
iv. ASTM D 3612- Determination of Individual Dissolved Gases ―Laboratory methods for
analyzing the gas blanket and the gases extracted from oil”.
v. ANSI- American National Standards Institute. www.ansi.org/
vi. FIST (Facilities Instructions Standards and Techniques Vol.-3-31), ―Transformer
Diagnostic‖, US Department of the Interior Bureau of Reclamation, June 2003.
78
vii. CIGRE Working Group 12.18-2002, ―Life Management of Transformers-Guidelines
for Life Management Techniques of Power Transformers,‖ 2002.
4.21 OTHER TRANSFORMER CHEMISTRY SERVICES DURING
INTERPRETATION
Various other evaluations and compensations are given due considerations for proper
fault interpretations from the history of transformer. Some of those are as:
i. Key gases: CSUS values (Age compensated)
ii. BS 5800/IEC 599 ratios (providing the Total Combustible Gases present are above
300 ppm)
iii. Rogers Ratio‘s
iv. Trend (Production rates of gases) Morgan-Schaffer Tables
v. Total Combustible Gas Production Rates TDCG
vi. Total Combustible Gas Westinghouse Guidelines
vii. Age of transformer.
viii. History of transformer (Repaired, degasses, etc).
Analysing insulating oil taken from transformers is a unique way of identifying problems
occurring within a transformer.
By identifying and quantifying the gases found in transformer oil, the condition of the
transformer can be monitored.
If faults are found to be occurring, outages can be planned and the fault can be rectified
before major damage can occur.
The interpretation of transformer oil gas analysis is still an art and not an exact science.
The interpretation should be left to a specialist and his advice and recommendations
should be followed. Samples should be taken regularly and records are kept by utilities
with transformers. Some of the typical values of gas interpretations are shown in below
Tables 4.5 to Table 4.10 around the world [111].
Table 4.5 issued by California State University of US for total combustible gases for their
individual normal and abnormal ranges of gases. This table only shows five possible
faults given as below:
Table 4.5: California State University Sacremento guidelines for combustible gas
GAS NORMAL ABNORMAL INTERPRETATION
HYDROGEN(H2) <150 ppm >1000 ppm Arcing corona
CARBON MONOXIDE(CO) <500 ppm >1000 ppm Severe Overloading
79
CARBON DIOXIDE(CO2) <10000 ppm >15000 ppm Severe Overloading
METHANE(CH4) <25 ppm >80 ppm Sparking
ETHANE(C2H6) <10 ppm >35 ppm Local overheating
ETHYLENE(C2H4) <20 ppm >100 ppm Severe Overheating
ECYTYLENE(C2H2) <15 ppm >70 ppm Arcing
NITROGEN (N2) 1-10% NA -
OXYGEN (O2) 0.03% >0.5% Combustibles
Table 4.6 again issued in addition to above table to show the comparison between three
different followed standards and particularly to provide compensation for the age of
transformer to the existing gas levels.
Table 4.6 California State University Sacremento guidelines for combustible gas
GAS Dornenburg/
Stritt.
IEEE Bureau of
Reclamation(US)
Age
Compensated
HYDROGEN(H2) 200 100 500 20n+50
CARBON
MONOXIDE(CO)
500 350-
720
750 25+500-
110n+710
METHANE(CH4) 50 120 125 20n+50
ETHANE(C2H6) 35 65 75 20n+50
ETHYLENE(C2H4) 80 50 175 20n+50
ECYTYLENE(C2H2) 5 35 7 5n+10
TDCG(total above) 6000 2500 10000 100n+1500
n=years in
service
Table 4.7 is guidelines of three standards to show the different gas levels for generator
transformers and transmission transformers for interpretation of faults to present the
violation limits.
Table 4.7: CEGB/ANSI/IEEE Guide for gas concentration limits in ppm V/V
GAS GENERATOR
TRANSFORMER
TRANSMISSION
TRANSFORMER
HYDROGEN(H2) 240 100
80
CARBON
MONOXIDE(CO)
580 350
METHANE(CH4) 160 120
ETHANE(C2H6) 115 65
ETHYLENE(C2H4) 190 30
ECYTELENE(C2H2) 11 35
Table 4.8 shows the limits of the individual gases used in Canada, Switzerland and
Finland as standards for assessing the faults.
Table 4.8: Other International gas concentration limits in ppm V/V
GAS HYDRO
QUEBEC
CANADA
BBC
SWITZERLAND
OY
STROMBERG
FINLAND
HYDROGEN(H2) 250 200 100
CARBON
MONOXIDE(CO)
850 1000 500
METHANE(CH4) 33 50 100
ETHANE(C2H6) 15 15 150
ETHYLENE(C2H4) 40 60 100
ECYTELENE(C2H2) 25 15 30
Table 4.9 provides the individual gas limits for different range of transformer ratings
issued by SERC, Japan regulating authority, which is adopted by most their utilities for
transformer fault assessments.
Table 4.9: SERC – Japan limiting values in ppm V/V
GAS TRANSFORMERS
>275KV & - >10MVA
TRANSFORMERS
>275KV & - <10MVA
TRANSFORMERS
> 500KV
H2 400 400 300
CO 300 300 200
CH4 150 200 100
C2H6 150 150 50
C2H4 200 300 100
TCG 700 1000 400
81
Table 4.10 issued by EDC, France showing separate ranges of individual gas limits for
generator and transmission transformers without on-load tap changers.
Table 4.10: EDF – France Generator/Transmission transformers without OLTCs
GAS GENERATOR
TRANSFORMER
TRANSMISSION
TRANSFORMER
HYDROGEN(H2) 33 130
CARBON MONOXIDE(CO) 770 1000
METHANE(CH4) 44 130
ETHANE(C2H6) 33 150
ETHYLENE(C2H4) 11 44
ECYTELENE(C2H2) 0.4 0.4
4.22 DIFFERENT DGA METHODS
If DGA values are above typical concentration values and/or rates of increase, an actual
fault in the transformer is probable, and diagnostic methods may be used for its
identification.
The main diagnostic methods used are:
i. The IEEE methods (Key gases methods, Rogers‘s and Doernenburg‘s)
ii. The IEC ratio codes
iii. The Duval Triangle
iv. The CIGRE standards
Out of above methods, following are used in thesis.
4.23 KEY GASES METHOD (C-57-104TM
, 2008)
The principle of the Key Gas method [41] is based on the quantity of fault gases released
from the insulating oil when a fault occurs which in turn increase the temperature in the
power transformer.
4.23.1 Evaluation of Possible Fault Type by the Key Gas Method
Previous discussions of the dependence on temperature of the types of oil and cellulose
decomposition gases provides the basis for the qualitative determination of fault types
from the gases that are typical, or predominant, at various temperatures. These significant
gases and proportions are called ―Key Gases‖. Fig 4.12 to Fig.4.15 indicate, these ―Key
Gases‖ and relative proportions for the four general fault types. The presence of the fault
gases depends on the temperature or energy that will break the link or relation of the
82
insulating oil chemical structure. This method uses the individual gas rather than the
calculation of gas ratios for detecting fault.
Figure 4.12: Key gases (Over Heating, Oil)
Figure 4.13: Key gases (Over Heating, Paper)
Figure 4.14: Key gases (Partial Discharge)
Figure 4.15 Key gases (Arcing)
83
4.24 EVALUATION OF TYPES OF POSSIBLE FAULT BY ANALYSIS OF THE
SEPARATE COMBUSTIBLE GASES GENERATED
The use of gas ratios to indicate a single possible fault type is an empirical process based
upon the experience of each individual investigator in correlating the gas analyses of
many units with the fault type subsequently assigned as the cause for disturbance or
failure when the unit was examined. This process was attributed to Dernenburg‘s and
subsequently confirmed by Roger‘s on European systems, from which the bulk of the
diagnostic correlation is obtained. U.S. investigators have applied the European rules to
units on U.S. systems with varying degrees of success; however, a U.S. database of
comparable size to the European reports does not exist.
The diagnostic theories based upon the thermal degradation principles oil and paper
decompositions employ an array of ratios of certain key combustible gases as the fault
type indicators. These five ratios are:
i. Ratio 1 (R1) = CH4/H2
ii. Ratio 2 (R2) = C2H2/C2H4
iii. Ratio 3 (R3) = C2H2/CH4
iv. Ratio 4 (R4) = C2H6/C2H2
v. Ratio 5 (R5) = C2H4/C2H6
The first ratio method (Dornenburg‘s given below) utilizes Ratios 1, 2, 3, and 4. This
procedure requires significant levels of the gases to be present in order for the diagnosis
to be valid.
The second method (Rogers given below) utilizes Ratios 1, 2, and 5. The Roger‘s method
does not depend on specific gas concentrations to exist in the transformer for the
diagnosis to be valid. However, it suggests that the method be used only when the normal
limits of the individual gases have been exceeded.
4.25 THE DORNENBURG’S RATIO METHOD
This method started out as only two ratios:
CH4 /H2 & C2H2/C2H4
The ratios were plotted on a log-log scale. The areas corresponded to thermal
deterioration, arcing and partial discharge. By two ratio method too many faults missed
and then went to four ratios.
Ratio 1 (R1) = CH4/H2
Ratio 2 (R2) = C2H2/C2H4
Ratio 3 (R3) = C2H2/CH4
84
Ratio 4 (R4) = C2H6/C2H2
Used to determine three types of general faults;
i. Thermal faults
ii. Electrical Faults, low intensity discharges
iii. Electrical Faults, high intensity arcing
The Dornenburg‘s ratio method suggests the existence of three general fault types as
discussed. The method utilizes gas concentrations from which Ratios 1, 2, 3, and 4 are
calculated. According to IEEE Standard C-57.104TM
-1991, the step-by-step procedure to
diagnose faults using Dornenburg‘s ratio method as follows:
The values for these gases are first compared to special concentration limits in Table 4.11
(see Steps 2, 3, and 4 in Fig. 4.16)—to ascertain whether there really is a problem with
the unit and then whether there is sufficient generation of each gas for the ratio analysis to
be applicable. Then the ratios in the order Ratio 1, Ratio 2, Ratio 3, and Ratio 4 are
compared to limiting values, providing a suggested fault diagnosis as given in Table 4.12
and. Table 4.11 gives the limiting values for ratios of gases dissolved in the oil and gases
obtained from the transformer gas space or gas relay.
Table 4.11: Concentration limits of dissolved gas
Key Gases Concentrations limits
[µL/L(ppm)]
H2 100
CH4 120
C2H6 65
C2H4 50
C2H2 1
CO 350
The flow chart in Fig. 4.16 illustrates the step-by-step application of the Dornenburg‘s
ratio method for gases extracted from the transformer oil only. Exactly the same
procedure is followed for gases obtained from the gas space or gas relays, except the
limiting values for the ratios will be those appropriate for gas space.
Descriptions of the steps indicated in flow-chart are as follows:
Step 1; Gas concentrations are obtained by extracting the gases and separating them by
chromatograph
85
Step 2; If at least one of the gas concentrations [in microliters/liter, (ppm)] for H2, CH4,
C2H2, and C2H4 exceeds twice the values for concentration limit (see Table 4.11) and one
of the other two gases exceeds the values for limit L1, the unit is considered faulty;
proceed to Step 3 to determine validity of the ratio procedure.
Step 3; Determining validity of ratio procedure: If at least one of the gases in each ratio
R1, R2, R3, or R4 exceeds concentration limits, the ratio procedure is valid; otherwise,
the ratios are not significant, and the unit should be re-sampled and investigated by
alternate procedures.
Step 4; Assuming that the ratio analysis is valid, each successive ratio is compared to the
values obtained from Table 4.12. in the order R1, R2, R3, and R4.
Step 5; If all succeeding ratios for a specific fault type fall within the given limit values,
the suggested diagnosis is valid.
i. Criteria for application - a fault exists
a) One Gas > 2 times the minimum level
b) At least one of two other gases > minimum level
ii. Determine Validity, concentration limits norm test
One gas in each ratio > minimum
iii. Compare ratios to Fault Diagnosis Table
iv. All fall within one condition-valid diagnosis
Table 4.12: Dornenburg‘s ratios for key gases
Suggested Fault
Diagnosis
Ratio1 (R1)
CH4/H2
Ratio2 (R2)
C2H2/C2H4
Ratio3 (R3)
C2H2/CH4
Ratio4 (R4)
C2H6/C2H2
Oil Gas
Space
Oil Gas
Space
Oil Gas
Space
Oil Gas
Space
1.Thermal decomposition >1.0 >0.1 <0.75 <1.0 <0.3 <0.1 >0.4 >0.2
2.Partial discharge (low-
intensity PD)
<0.1 <0.01 Not significant <0.3 <0.1 >0.4 >0.2
3.Arcing (high-intensity
PD)
>0.1- -
<1.0
>0.01--
<0.1
>0.75 >1.0 >0.3 >0.1 <0.4 <0.2
86
Figure 4.16: Dornenburg‘s ratio method flow-chart
4.26 THE MODIFIED ROGER’S RATIO METHOD
Roger‘s originally developed four ratios CH4/H2 (R1), C2H6/CH4, C2H4/C2H6 (R5) and
C2H2/C2H4 (R2). The Roger‘s ratio method follows the same general procedure as the
Dornenburg method, using only three ratios (R1, R2, and R5) are used. This method,
shown in the step-by-step flow chart (see Fig. 4.17), is also based on the thermal
degradation principles. The validity of this method is based on correlation of the results of
a much larger number of failure investigations with the gas analysis for each case. But, as
with the Dornenburg method, the Roger‘s ratios can give ratios that do not fit into the
diagnostic codes. Therefore, other analytical methods by TCDG and Key Gas ratio should
be considered, as well as other options. These ratios, according to Rogers, are applicable
to both gases taken from the gas space (or relay) and gases extracted from the oil. The
fault types (cases) given in Table 4.13 have been chosen by combining some cases from
the number of fault types originally suggested by Roger.
Table 4.13: Roger‘s ratios for key gases
Case Suggested Fault Diagnosis Ratio1 (R1)
CH4/H2
Ratio2 (R2)
C2H2/C2H4
Ratio5 (R5)
C2H2/CH4
0 Unit Normal >0.1 to <1.0 <0.1 <1.0
1 Low-energy density arcing—PD@
<0.1 <0.1 <1.0
2 Arcing--high-energy discharge 0.1 to 1.0 0.1 to 3.0 >3.0
3 Low temperature thermal >0.1 to <1.0 <0.1 0.1 to 3.0
4 Thermal <7000C >0.1 <0.1 0.1 to 3.0
5 Thermal >7000C >0.1 <0.1 >3.0
@There will be a tendency for the ratios R2 and R5 to increase to a ratio above 3.0 as the
discharge develops in intensity.
87
Figure 4.17: Modified Roger‘s ratio method Flow-chart
4.27 DUVAL TRIANGLE METHOD
The Duval Triangle was first developed in 1974. It uses three hydrocarbon gases only
(CH4, C2H4 and C2H2). These three gases correspond to the increasing levels of energy
necessary to generate gases in transformers in service. The Triangle method is indicated
in Fig. 4.18.
Figure 4.18: Duval Triangle 1
88
Figure 4.19: Duval triangle indicating regions of different faults
In addition to the six zones of individual faults as mentioned (PD, D1, D2, T1, T2 or T3),
an intermediate zone DT has been attributed to mixtures of electrical and thermal faults in
the transformer. Colour codes for Duval Triangle shown in Fig. 4.19 are used for an easy
interpretation. C2H2 and C2H4 are used in all other interpretation methods to represent
high energy faults (such as arcs) and high temperature faults. H2 is preferred in several of
these methods to represent very low energy faults such as PDs, where it is produced in
large quantities. CH4, however, is also representative of such faults and always formed in
addition to H2 in these faults, in smaller but still large enough amounts to be quantified.
CH4 has been chosen for the Duval triangle because it not only allows to identify these
faults, but provides better overall diagnosis results for all the other types of faults than
when using H2.
The following are the some more methods used by some of the utilities in India to cross
verify the faults in transformers and many other countries subjected to their weather and
tropical conditions. A few are discussed as:
4.28 IEC 60599
This method identifies six different fault types listed below. This interpretation method
adopted in most of the part of the world. Almost IEEE and IS standards followed in India
are extracted from these interpretation methods.
a) PD: Partial Discharge
b) D1: Discharge of low energy
c) D2: Discharge of high energy
d) T1: Thermal fault, t <300°C
e) T2: Thermal fault, 300°C < t < 700 °C
f) T3: Thermal fault, t > 700 °C
89
This method uses a combination of ratios (based on Roger‘s Ratios), gas concentrations
and rates of gas increase between the consecutive sample testings.
4.28.1 IEC 60599 Ratio-Fault Diagnosis
Fault diagnosis becomes the combined outcome of gas ratios R1, R2 and R5 along with
the limits of the ratios to represent one of the fault, out of the six faults. This is shown in
Table 4.14.
Table 4.14: IEC 60599 ratio-fault diagnosis
R2
C2H2/C2H4
R1
CH4/H2
R5
C2H4/C2H6
Faults
NS <0.1 <0.2 PD
>1 0.1-0.5 >1 D1-Low Energy
0.6-2.5 0.1-1 >2 D2-High Energy
NS >1 (NS) <1 T1- <3000C
<0.1 >1 1-4 T2- >3000C - <700
0C
<0.2 >1 >4 Thermal (T3)>7000C
NS- Not Significant regardless of values.
4.29.2 Rates of Gas Increase
The second procedure to divide the faults depends upon the percentage of gas increase
between the recommended periodicity during the fault monitoring of the transformer.
i. If gas increases >10% per month above typical levels = active fault.
ii. If gas increases >50% per week or evolving faults of higher energy = serious faults
4.29.3 IEC 60599 Typical Gas Levels
The Table 4.15 is given by IEC- 60599 standard to follow strictly the safe gas levels for
evaluation of faults in different types of the transformers.
Table 4.15: IEC 60599 typical gas levels
H2 CO CH4 C2H6 C2H4 C2H2 CO2
IEC 60599
Typical Range
60-150 540-900 40-110 50-90 60-280 3-50 5100-
13000
Communicating
OLTC
75-150 400-850 35-130 50-70 110-250 80-270 5300-
12000
Note in IEC 60599: Typical values are higher in sealed transformers than free breathing
transformers.
90
4.30 NOMOGRAPH METHOD
The logarithmic nomograph method was developed by J. O. Church. This method
combines the fault gas ratio concept with the Key Gas threshold value in order to improve
the accuracy of fault diagnosis. It was intended to provide both a graphic presentation of
fault-gas data and the means to interpret its significance. The nomograph consists of a
series of vertical logarithmic scales representing the concentrations of the individual
gases as shown in Fig. 4.20.
With this method, straight lines are drawn between adjacent scales to connect the points
representing the values of the individual gas concentration. The slopes of these lines are
the diagnostic criteria for determining the type of fault. The key at the bottom of the chart
between the two axes indicates the fault type for the two axes. A visual comparison of the
slopes of the line segments with the keys given at the bottom of the Nomograph is all
needed to identify the type of fault. The position of the lines relative to the concentration
scales provides a means of assessing the severity of the fault.
Figure 4.20: Logarithmic nomograph [112]
Each vertical scale has a threshold value labeled with an arrow. For the slope of a line to
be considered significant, at least one of the two tie-points should lie above a threshold
value. If neither tie-point lies above a threshold value then the fault indication of that
slope is not considered significant.
4.31 DENKYOKEN METHOD
A special committee on ―Conservation and Council of oil-insulated components by
diagnosis of gas in oil‖ in Japan developed the Denkyyon Method. This is the diagnostic
method recommended by Electrical Cooperative Research Association (ECRA). This
method is based on the amount of H2, CH4, C2H6, C2H4 and C2H2 gases in ppm, the most
abundant gas is taken as 100% and the all other gases are then impressed with reference
91
to the value for a given specimen data. Thirteen different patents are designated for cause
of failure in oil-insulated power apparatus.
4.32 CIGRE’S METHOD
The method is use the key gas ratios as well as gas concentration for faults diagnosis. The
five key gas ratios those are considered are C2H2/C2H6, H2/CH4, C2H4/C2H6, C2H2/H2, and
CO/CO2. The key gas concentrations are C2H2, H2, sum of carbon hydrides, CO and CO2.
The ratio ranges and concentration limits are suggested. There is no combined application
of both these methods in case all the ratios and concentration are below the limits, a
transformer is considered to be healthy. The advantage of this method is that two or more
faults can be detected.
4.33 NBR7274 METHOD
This Brazilian standard is based on IEC 599/78 and the five gases considered H2, CH4,
C2H6, C2H4 and C2H2 in ppm. It employs the three ratios C2H2/C2H6, CH4/H2 and
C2H4/C2H6 based on the three ratios. The faults are classified in eight types including
normal ageing.
4.34 IS 10593:2006 Method
The Indian Standard is identical to the IEC 60599/1999 and was adopted by Bureau of
Indian Standards. On the recommendations of Fluids for Electrochemical Applications
Sectional Committee and approved by the Electrochemical Division Council. In this
adopted standard reference appears to certain International Standards for which can be
summarized into six types:
i. The methods have been developed based on experience and judgments of the human
experts only.
ii. Due to the in-completeness of the possible ratio combination and uncertainty of the
validity of the defined ranges of key gas ratios, there is high degree of inconsistency
and ambiguity.
iii. When these schemes are applied some times the decision is dependent only on the
expert.
iv. Concurrently within the power transformer multiple faults occur, all these methods
are unable to detect new or unknown faults.
v. The ratio methods do not cover all ranges of data.
92
4.35 COMPARATIVE ANALYSIS OF DIFFERENT DGA METHODS (KEY GAS,
DORNENBURG’S, ROGER’S RATIO & DUVAL TRIANGLE METHOD)
To understand the accuracy and simplicity among the different fault interpretation
methods, two case studies have been presented and shown in Appendix-A. It shows the
advantages of Duval triangle method over other convention methods of fault
interpretations. Table 4.16 reveals that Duval triangle method provides least unresolved
diagnosis and diagnosis of power transformers..
Table 4.16: Relative Performance of Different Methods
4.36 SUMMARY
This chapter concludes with the mechanism of decomposition of oil and the formation of
many gases in the transformer oil at different temperature ranges. It also shows how the
solubility of the gases changes with temperature in the oil? Various sampling
standards/procedures are listed. Safe and error free sampling procedures and gas
extraction process for the transformer and the oil samples are also explained. Small
descriptions of the systems and apparatus for DGA are included. Some of the new
degassing techniques and mobile GC for on-line monitoring are introduced. These
devices can be combined to an off-line/online DGA fault diagnosis/monitoring systems,
which can enhance the reliability of transformers significantly.
A list of gases produced in the oil during degradation is given in the beginning of the
chapter. The faults detectable by DGA in the transformers as per IEEE/IEC standards are
briefly described. Since gases limit interpretation is counted to be an art, the gases
limiting tables suggested in different standards, countries and organizations subjected to
their local environmental conditions. IEEE/IEC standards adopted in India are largely
explained in this chapter. Brief introduction of some others are also given. From the case
study conducted on fault analysis on some given transformer reports, the study concludes
that Duval triangle method analysis have low uncertainty regarding fault interpretation
compare to other conventional methods.
DGA Method % Unresolved diagnosis % Wrong diagnosis % Total
Key gases 0 58 58
Roger‘s 33 5 38
Doernenburg‘s 26 3 29
IEC 15 8 23
Duval triangle 0 4 4
93
CHAPTER 5
FAULT DIAGNOSIS WITH DUVAL TRIANGLE
5.1 INTRODUCTION
In this chapter Interpretation of DGA in Power Transformers is presented, with special
emphasis and use on the Duval Triangle method. This method uses three hydrocarbon
gases only (CH4, C
2H
4 and C
2H
2). These three gases correspond to the increasing levels of
energy necessary to generate gases in transformers in service. In addition to the six zones
of individual faults an intermediate zone DT has been attributed to mixtures of electrical
and thermal faults in the transformer.
5.2 FAULT DIAGNOSIS BY DUVAL TRIANGLE METHOD
The main gases formed as a result of electrical and thermal faults in transformers and
evaluated by DGA are H2, CH4, C2H2, C2H4, C2H2, CO, and CO2, whose relative
concentrations depend on the fault type. A general rule, based on thermodynamic
considerations, is that the degree of chemical un-saturation of the gases formed is related
to the energy density of the fault. Acetylene (C2H2) is thus mainly associated with arcing,
where temperatures reach several thousand degrees, ethylene (CH4,) with hot spots
between 150 °C and 1000 °C, and hydrogen (H2) with the cold gas plasma of corona
discharges. Although mixtures of all gases, including the other saturated hydrocarbons
(CH4, C2H6), are usually obtained in most cases of faults. Their relative proportions have
been correlated, through empirical observations and laboratory simulations, with the
various types of fault commonly encountered in transformers in service. The most widely
used tool for that purpose is the IEC/IEEE ratio method. One drawback of this method in
its present form is that a significant number of DGA results in service fall outside the
proposed codes, and cannot be diagnosed. Other methods that overcome this limitation
have therefore been developed, shown in Fig. 5.1 based on a triangular representation and
Hydro-Quebec experience using nomographs.
Figure 5.1: Duval triangle
94
a. 1. High-energy arcing (I>20 In)
b. 2. Low- energy arcing. Tracking
c. 3. Corona discharges
d. 4. Hpt spots, T < 200 0C
e. 5. Hot spots, 200 < T < 400 0
C
f. 6. Hot spots, T > 400 0C
To apply Duval triangle, thrre gses are changed to their percentages from the sum of these
gases. These three percentages can be represented on the each side of the equiviletral
triangle to show the graphical presentation for fault analysis. The triangular coordinates
are represented as given below:
% CH4 = for, x = [ CH4] in ppm (5.1)
% C2H4 = for, y = [ C2H4] in ppm (5.2)
% C2H2 = for, z = [ C2H2 ] in ppm (5.3)
5.3 IMPROVED VERSION OF DUVAL TRIANGLE 1 METHOD
Several methods of interpretation of DGA in transformers in service are provided in the
IEEE Guide C-57.104TM
[41], IEC Standard 60599 [45], as well as in published reviews
on the subject. The Duval Triangle method is described in the IEC Standard and in these
published reviews, however, users sometimes are not quite at ease with the use of
triangular coordinates.
The performance of the improved Duval Triangle with CH4
might be related to the fact
that H2
diffuses much more rapidly than the hydrocarbon gases from the oil through
gaskets and even metal welds. Therefore, gas ratios using H2
are probably more affected
by the loss of this gas than those using hydrocarbons gases only, which have much lower
and comparable diffusion rates.
The three sides of the Triangle are expressed in triangular coordinates (X, Y, Z)
representing the relative proportions of CH4,
C2H
4 and C
2H
2, from 0% to 100% for each
gas.
In order to display a DGA result in the Duval Triangle 1 method, the following procedure
must be adopted. One must start with the concentrations of the three gases, (CH4) = A,
(C2H
4) = B and (C
2H
2) = C, in ppm.
i. Calculate the sum of these three values: (CH4
+ C2H
4 + C
2H
2) = S, in ppm.
95
ii. Calculate the relative proportion of the three gases, in %:
X = % CH4
= 100 (A/S), Y = % C2H
4 = 100 (B/S), Z = % C
2H
2 = 100 (C/S).
X, Y and Z are necessarily between 0 and 100%, and (X+Y+Z) should always = 100 %.
iii. Plotting X, Y and Z in the Triangle, provides only one point in the Duval Triangle.
For example, if the DGA results are A = B = C = 100 ppm, X = Y = Z = 33.3%, which
corresponds to only one point in the centre of the Triangle, as indicated in Figure 5.2. The
zone in which the (X, Y, Z) point falls in the Duval Triangle allows to identify the fault
responsible for the DGA results. Plotting the (X, Y, Z) point in the Triangle can also be
done manually, preferably using a triangular graphical paper such as in Fig. 5.2 for better
precision.
Figure 5.2: Duval triangle 1 graphical plot
Zone boundaries in the Triangle [65-68] have been deduced empirically from a large
number of cases of faults visually inspected in transformers worldwide over the last 60
years by CIGRE and other resources in Fig. 5.3. Well documented and reliable new cases
of faults inspected in service may be used to confirm or re-adjust slightly these
boundaries.
96
Figure 5.3: Actual cases of faults visually inspected in different transformers
5.4 THE INFLUENCE OF LABORATORY ACCURACY ON FAULT DIAGNOSIS
The accuracy of DGA diagnosis, whatever the diagnosis method used, depends greatly on
the accuracy and reliability of the DGA results coming from the laboratory. Note that, by
convention among chemists, accuracy is represented by the difference with actual value
(the analytical error in %), so that higher (better) accuracies are represented by a smaller
number in %.
A few laboratories worldwide provide very accurate results, with an accuracy higher (or
error lower) than ±5% at routine gas concentration levels (typically, above 10 ppm for
hydrocarbon gases). Some others are known to provide very inaccurate results (±50%).
In-between, the average accuracy of laboratories worldwide has been evaluated by
CIGRE TF11 as ~ ±15% at routine levels. The average accuracy worsens rapidly to ~
35% at lower concentration levels (between 2 and 10 ppm for hydrocarbon gases), and
even more so (to 100% and more) as concentrations approach analytical detection limits.
This is illustrated in Fig. 5.4, where the diagnosis uncertainty corresponding to the
various DGA cases.
97
Figure 5.4: Uncertainty on diagnoses for many cases
The coloured polygons from the Fig. 5.4 represents the 10% uncertainty around the fault
in the each zone boundaries. This range of uncertainty from the laboratory results are
given in Table 5.1.
Table 5.1: Uncertainty on laboratory results
Sl. No. Fault CH4 C2H4 C2H2
1 PD 99 1 0
9.9 0.1 0
2 D1 38 12 50
3.8 1.2 5
3 D2 15 50 35
1.5 5 3.5
4 T2 69 30 1
6.9 3 0.1
5 T3 20 75 5
2 7.5 0.5
The more inaccurate the laboratory results, the larger the uncertainty on the diagnosis, as
illustrated in Fig. 5.5.
Figure 5.5: Diagnosis uncertainties to laboratory analytical accuracy of +- 15,30,50&75%
When a polygon crosses two or more zones, a wrong or uncertain diagnosis may result.
This may have serious consequences for the equipment if for example an arcing problem
is mistakenly diagnosed as a less severe thermal fault. In order to get good reliable
98
diagnoses, laboratory accuracy should below ±10%. Between ±10% and ±40%, diagnoses
will likely become more and more uncertain, and above ±40% they are totally
meaningless. DGA users are therefore strongly recommended to verify the accuracy of
their laboratories, using samples of gas-in-oil national/international standards.
DGA users should also always look at inconsistencies in the DGA results, for instance
values going up and down within short periods of time for no explainable reason. These
are often an indication of a gross laboratory or sampling error rather than just inaccurate
results.
5.4.1 Influence of Factors Affecting Transformer Reliability
Typical values for hydrocarbons (except C2H
2) are markedly higher in power
transformers of the shell-type and in shunt reactors than in the mostly core-type
transformers, possibly because they operate at higher temperatures. Typical values in
instrument transformers are much lower than in power transformers. Typical values are
higher in the early years of the transformers, suggesting that some unstable chemical
bonds in the paper or oil insulation are broken in the early years, then the remaining ones
are more stable afterwards. Typical values are also slightly higher for faults in oil than in
paper. But typical values are not dependent on oil volume, suggesting that smaller
amounts of gases (and smaller faults) are formed in smaller equipment.
5.4.2 Calculation of Own Typical Values
Since typical values are influenced by such factors as transformer age and type and
loading practices, each individual network is encouraged to calculate the typical values
corresponding to its own transformer population. This can be done easily by listing DGA
results by increasing order of values, for each of the fault gases (e.g., H2). The value
corresponding to 90 % of the cumulative number of DGA analyses is the 90 % typical
value. For example, 90 % of H2
values in the transformer population of the network are
below this typical value, and 10 % (the upper percentile) are above. This can be done for
both concentrations values and rates of gas increase (RGI). By default, if typical values
cannot be calculated, for example because of an insufficient DGA data bank, the typical
gas concentration levels and rates of increase reported in various countries by CIGRE and
the IEC may be used as a rough approximation Table 5.2 and Table 5.3.
Table 5.2: Ranges of 90% typical values for power transformer in ppm
C2H2 H2 CH4 C2H4 C2H6 CO CO2
All Transformers 50- 30- 60- 20-90 400-600 3800-
99
150 130 280 1400
No OLTCs 2-20
Communicating
OLTC
60-280
Table 5.3: Ranges of 90% typical rate of gas increase for power transformer in ppm/year
C2H2 H2 CH4 C2H4 C2H6 CO CO2
All Transformers 35-132 10-120 32-145 5-90 260-
1060
1700-10000
No OLTCs 2-20
Communicating
OLTCs
21-37
5.4.3 Adjustment of Pre-failure and Alarm Values
Pre-failure concentration values and pre-failure rates of gas increase can be obtained by
calculating the probability of having a failure-related event (PFS, in %) in a transformer
in service, as a function of gas concentration level in oil. This is done by calculating the
following ratio, for each individual gas, at different concentrations. Thus number of DGA
analyses followed by an event such as tripping, tank rupture, fire or explosion, divided by
the total number of analyses.
In Fig. 5.6, the PFS value is indicated as a function of the concentration of C2H
2, in power
transformers without communicating OLTCs [70]. It can be seen that even at low
concentration values (near the 90 % typical value of 5 ppm), the PFS is not zero but
around 12 %. In such cases, a fault probably developed in the transformer very rapidly
after the DGA analysis, without advanced warning. Above a value of around 350 ppm,
there is an inflexion point in the curve above which the PFS increases rapidly.
PFS, in %
Figure 5.6: Probability of having a failure-related event(PFS, In %) as a function of C2H2
concentration in service of in ppm and norm in %
This corresponds approximately to the 99 % typical value and to 1 % of DGA analyses,
which is not far from the annual failure rate of transformers (0.3 %). This value has been
100
defined as the pre-failure gas concentration value (PFGC). The PFGC values observed for
the other gases are indicated in Table 5.4.
Table 5.4: Pre-failure gas concentration values at CIGRE for core- type transformers
H2 CH4 C2H6 C2H4 C2H2 CO
550-1320 340-460 750-1050 700-990 310-600 980-3000
By combining pre-failure values and actual rates of increase in service, one may have an
idea of how long it may take to reach failure (if rates do not accelerate), and plan
appropriate actions.
Alarm gas concentration values may be defined as the values corresponding to x times the
PFGC population. For example, if x = 2, the alarm value corresponds to the 98 % typical
value, or 170 ppm. Alarm values thus calculated for the other gases can be found in.
5.5 APPLICATION OF DUVAL TRIANGLE METHOD 1 (DTM 1)
There are two different procedures to use DTMs as follows;
1. By using total accumulated gas
2. By using total increase in gases between conjugative samples
For this thesis work, total accumulated gas based procedure for the DGA using DTM 1
has been adopted.
5.5.1 Graphical Fault Interpretation
Graphical DGA fault diagnosis by Duval triangle 1 is carried out as follows; Consider the
three side of triangle in triangular coordinates (x, y and z) representing the relative
proportion of CH4, C2H4 and C2H2, from 0% to 100% for each gas. Numerical boundary
zones for seven key faults are shown in Table 5.5. To find the faults graphically (manual),
first calculate the percentage of each gas as per above equations (5.1 - 5.3). Then draw the
lines % CH4 quantity parallel to C2H2 line, % C2H4 quantity parallel to CH4 line and
%C2H2 quantity parallel to CH4 as on the specially supplied graphical sheets shown in
Fig. 5.2. Key Faults along with boundary zones are presented of Duval Triangle method1
in Table 5.5.
Table 5.5 Numerical zone boundaries
No. Fault Limits of zones
1 PD 98 % CH4 2 D1 23 % C2H4 13 % C2H2
3 D2 23 % C2H4 13 % C2H2 40 % C2H4 29 % C2H2
4 DT 40% C2H4 50% C2H4 04% C2H2 13% C2H2 15% C2H2
5 T1 04% C2H2 20 % C2H4
6 T2 04% C2H2 10 % C2H4 50 % C2H4
7 T3 15% C2H2 50 % C2H4
101
Thus drawn intersection of all three lines would indicate the fault responsible for the
DGA results in the transformer. Such verification of faults by Duval triangle method
(manual) DGA has been done (fault reported transformers). These results were verified
with DGA interpretation for total dissolved combustible gases by other procedures used
by different utilities in INDIA.
Example: CH4= 56 ppm, C2H4= 55 ppm and C2H2= 43 ppm, graphically (manually)
calculated result D2 displayed in Fig. 5.7.
Figure 5.7: Graphical analyses on DTM 1
5.5.2 Simulated Fault Interpretation with DTM 1
Following steps are used for simulated fault interpretations;
Step 1; In this research work, firstly, polygon coordinates for the numerical zone
boundaries of seven key faults of DTM 1 have been generated in terms of percentages of
CH4, C2H4 and C2H2, from 0% to 100% respectively shown in Table 5.6.
Table 5.6: Triangular coordinates for DTM 1 zones
No. Area Points %CH4 %C2H4 %C2H2
1 PD PD1 98 2 00
2 PD2 100 00 00
3 PD3 98 00 2
4 D1 D11 0 0 100
5 D12 0 23 77
6 D13 64 23 13
7 D14 87 00 13
102
Step 2; A flow-chart for simulation of fault interpretations developed for DTM 1 on
MATLAB 7.6 is presented in Fig. 5.8. To define each polygon, the defined points are
converted to Cartesian coordinates for percentage of gases for types of fault. Same inputs,
which are used for graphical (Manually) fault interpretations in Step 1 analyzed by DTM
1simulation on MATLAB-7.10 and revealed the same result D2 represented in Fig. 5.9.
The DGA fault analysis on simulated DTM 1 for the available fault reports are thoroughly
carried out for all the samples (reports) on MATLAB 7.10 and cross verified. All the
8 D2
D21 00 23 77
9 D22 0 71 29
10 D23 31 40 29
11 D24 47 40 13
12 D25 64 23 13
13 DT DT1 00 71 29
14 DT2 00 85 15
15 DT3 35 50 15
16 DT4 46 50 4
17 DT5 96 00 4
18 DT6 87 00 13
19 DT7 47 40 13
20 DT8 31 40 29
21 T1 T11 76 20 4
22 T12 80 20 00
23 T13 98 2 00
24 T14 98 00 2
25 T15 96 00 4
26 T2 T21 46 50 4
27 T22 50 50 00
28 T23 80 20 00
29 T24 76 20 4
30 T3 T31 00 85 15
31 T32 00 100 00
32 T33 50 50 00
33 T34 35 50 15
103
comparative fault analysis between manual and software implementation of DTM 1 along
with the fault analysis with other diagnostic techniques (faults analyzed by the respective
authorities and utilities) are prepared and a comparative study carried out and a complete
Table 5.28.
1
Figure 5.8: Flow-chart of DTM 1
CH4>120
C2H4>50
C2H2>1
Take gas levels of
C2H2, C2H4, CH4
Define the boundaries of the faults
acc. to the Duval Triangle 1.
Calculate the percentage of
C2H2,C2H4,CH4 as p1,p2,p3
p1>=98
DUVAL TRIANGLE TEST IS NOT
APPLICABLE
No
Yes
THERMAL FAULT
> 700 0C -- T3
DISCHARGE OF LOW
ENERGY -- D1
THERMAL FAULT <
300 0C -- T1
THERMAL FAULT 300-
700 0C -- T2
PARTIAL DISCHARGE --
PD
DISCHRAGE OF
HIGH ENERGY -- D2
p2 20-50 p3<4
P3<15
p2<23 p3<13
p2<20
p2>=50
p2 23-40 p3 29-77
No
No
No
No
No
Yes
Yes
No
Yes Yes
Yes
Yes
Yes
Yes
Yes
No THERMAL AND ELECTRICAL
FAULT - DT
Start
Stop
Stop
Stop
Stop
Stop
Stop
Stop
Stop
104
By comparing the graphical and simulated DTM 1 based analysis results are obtained,
shown as Fig. 5.9.
Figure 5.9: Duval Triangle method 1 simulation analysis
Step 3; Two other conventional methods: Dornenburg‘s and Roger‘s diagnostic
techniques are developed. A generalized flow-charts for fault analysis through these two
diagnostic tools are given in Fig. 4.25 and Fig.4.26. All the comparative fault analysis
between manual and simulation implication of DTM 1 along with the fault analysis for
Dornenburg and Roger‘s diagnostic techniques (faults analyzed by the respective
authorities and utilities) are prepared and given in Annexure-. It is found that DTM1
fault diagnosis on power transformers gives better fault interpretation as compared to
Dornenburg‘s ratios and Roger‘s ratios fault interpretations.
5.6 DUVAL TRIANGLE METOD 2 FOR LOAD TAP CHANGERS OF THE
OIL TYPE
The normal operation of load tap changers (LTCs) of the oil type involves arcing in the
oil. The ―Duval Triangle 2 for LTCs of the oil type‖, taking this into account in a more
user-friendly manner is presented in Fig. 5.10. Zone boundaries in the Duval Triangle 2
for LTCs of the oil type are based on the basis of examples of inspected cases of faults
indicated in Fig. 5.11. DGA results and references for these inspected cases are given in
Table 5.7. However, the N zone may be smaller for some individual types or brands of
LTCs of the oil type, and its actual boundaries can be determined by using DGA results
from such individual units which have been newly installed or where normal operation is
ascertained. Fault diagnoses should not be attempted in LTCs if concentrations of C2H2
and C2H4 are < 10 ppm, because of possible contamination and uncertainty of laboratory
results.
105
Figure 5.10: The Duval triangle 2 for LTCs of Oil Type Transformers
Figure 5.11: Inspected Cases of Faults in the Duval Triangle 2 for OLTCs of Oil-type
Transformers
The identification of fault zones in Fig. 5.10 and Fig. 5.11 is indicated in Table 5.7. In the
second column of Table 5.7, the types of faults mentioned are: D1 = Discharges of low
energy; D2 = Discharges of high energy; T2 = Thermal faults of temperature 300 ºC < T
<700 ºC; T3 = Thermal faults of temperature > 700 ºC.
Table 5.7: Identification of fault zones in above OLTC figure
No. Zone Identification Recommended actions
1 N Normal operation
2 D1 Abnormal arcing D1(outside of N
zone)
Inspect the LTC for small sign
of arcing.
3 X1 Abnormal arcing D1 or thermal
fault in progress
Still area is under
investigations.
4 X3 Fault T3 or T2 in progress with Test or inspect LTC for sign of
106
light coking or increased resistance
of contacts. Or severe D2 arcing
light coking or resistance of
contacts, or of severe arcing.
5 T2 Severe thermal fault T2
(300<T<700oC), coking
Change the oil.
Inspect the LTC for coking of
contacts.
6 T3 Severe thermal fault T3
(T>700oC), heavy coking
Numerical zone boundaries in Fig. 5.10 and Fig. 5.11 are indicated in Table 5.8. The
numerical zone boundaries indicated in Table 5.8 for the N (normal operation) zone cover
all types of LTCs of the oil type and may be used by default.
Table 5.8: Numerical zone boundaries of Duval triangle method 2
No. Fault Limits of Zones (%)
N CH4= 19, C2H4=23, CH4=2, C2H4=6
D1 CH4= 19, C2H4=6, CH4=2, C2H4=23
X1 CH4= 19, C2H4=23
X3 C2H4=23, C2H2=15
T2 C2H4=23, C2H2=15, C2H4=50
T3 C2H4=50, C2H2=15
5.6.1 Implementation and Results from Duval triangle method 2 (DTM 2) for
OLTCs
Here, with the help of reconstructed numerical zone boundaries of faults for DTM 2 is
reconstructed and shown in Table 5.9. Further Table 5.10 shows the polygon coordinates
of DTM 2 fault ranges.
Table 5.9: Numerical zones of DTM 2
No. Fault Limit of zones
1 N 19 % CH4 23% C2H4 2% CH4 6% C2H4
2 D1 19 % C2H4 6 % C2H2 2 %CH4 23% C2H4
3 X1 19 % C2H2 23% C2H4
4 X3 23% C2H4 15% C2H2
5 T2 2% C2H4 15 % C2H2 502H4
6 T3 50 % C2H4 15 % C2H4
107
Table 5.10: Triangular coordinates for Duval triangle method 2 zones
No. Area Points %CH4 %C2H4 %C2H2
1 N N1 19 6 75
2 N2 19 23 58
3 N3 2 6 92
4 N4 2 23 75
5 D1 D11 19 0 81
6 D12 19 6 75
7 D13 2 6 92
8 D14 2 23 75
9 D2 D15 0 23 77
10 D16 0 0 100
11 X1 X11 19 0 81
12 X12 100 0 0
13 X13 77 23 0
14 X14 19 23 58
15 X3 X3.1 0 23 77
16 X3.2 62 23 15
17 X3.3 0 85 15
18
T2
T2.1 62 23 15
19 T2.2 77 23 0
20 T2.3 50 50 0
21 T2.4 35 50 15
22 T3 T31 35 50 15
23 T32 50 50 0
24 T33 0 100 0
25 T34 0 85 15
Based on numerical zone boundaries and developed polygon coordinates, a simulation on
MATLAB is developed. The result of one of the sample for input given data is represented
in Fig. 5.12.
Enter the input data--
108
Figure 5.12: Fault analysis by DTM 2 for OLTC
5.7 THE DUVAL TRIANGLES METHOD 3 FOR NON-MINERAL OILS
Mineral oils are the most widely used insulating fluid in electrical equipment. However,
other non-mineral oils are increasingly used because they are less flammable or more
environmentally friendly.
This has raised the question of whether new DGA interpretation methods are needed for
them, or if existing methods used for mineral oils can be adjusted for non-mineral oils.
Very few inspected cases of faults in equipment filled with non-mineral oils are available,
because these fluids have been used only for a small number of years and in a limited
quantity of equipment. To evaluate gas formation patterns in these fluids, therefore, these
tests are simulating faults in the laboratory, and performed in parallel on both mineral oils
and the new fluids.
A number of such test results have been published and are readjusted to form the new
zone boundaries for silicone oil, synthetic ester (Midel), and natural esters/ vegetable oils
(FR3 and BioTemp), using the Duval Triangle 1 for transformers DGA results were
obtained either by the laboratory gas extraction method of Head Space (and were
109
corrected for the different Ostwald coefficients of the non-mineral oil ), or by the Toepler
or Partial Degassing methods (and do not need significant corrections).
The differences in gas formation patterns observed by Duval triangle method 1 between
mineral and non-mineral oils can be used to adjust zone boundaries of the Duval Triangle
1 and to create a Duval Triangle 3 for each of the non-mineral oils, as indicated in Fig.
5.13.
The reason for the different gas patterns of mineral and non-mineral oils can only be
speculative at this stage and is still under investigations. These differences are probably
related to the differences in chemical structure of the oils. For instance, the carbon-
oxygen bonds in natural and synthetic esters (FR3, BioTemp and Midel), and the silicon-
oxygen bonds in silicones, may slightly affect the chemical strength of neighboring
carbon-carbon and carbon-hydrogen bonds, and might explain the different relative
amounts of hydrocarbon gases formed at the same levels of fault energy or temperature.
The two natural esters (vegetable/seed oils) examined based on saturated/unsaturated
oleic/fatty acids [13], and the synthetic ester on pentaerythritol. Non-mineral oils with
different chemical structures may result in different gas patterns and zone boundaries.
Figure 5.13: Duval Triangle 3 for Non-mineral oils with their zone boundaries in colours
110
Zone boundaries of the Duval Triangle method 3 for non-mineral oils (Silicone, Midel,
FR3 and BioTemp) are the same as those of the Duval triangle method 1 for mineral oils
except for the boundaries (in % C2H4) between zones D1/ D2, T1/ T2, and T2/ T3, as
indicated in Fig. 5.13 and Table 5.11. The boundary between faults D2 and DT may also
have been modified when actual cases of faults D2 (high-energy arcing) are reported in
transformers in service filled with these non-mineral oils.
Table 5.11: Zone Boundaries (% C2H4) for Duval triangle method 3 (non-mineral oils)
Boundary
between
zones
Mineral
oils
Silicone Midel BioTemp FR3
D1/D2 23 9 26 20 25
T1/T2 20 16 39 52 43
T2/T3 50 (46) (68) (82) 63
5.7.1 Implementation and results from Duval triangle methods 3 (DTMs 3) for Non-
mineral oils
Simulation for all four Duval triangle method 3 are as follows:
5.7.1.1 Silicone oil
With the help of its numerical zone boundaries as shown in Table 5.12, whose polygon
coordinates are created and shown in Table 5.13. Therefore, the simulation for DTM 3,
silicone oil is developed.
Table 5.12: Numerical zone boundaries for DTM 3 (Silicone oil)
No. Fault Limits of zones
1 PD 98 % CH4 2 D1 10 % C2H4 12 % C2H2
3 D2 10 % C2H4 12 % C2H2 40 % C2H4 28 % C2H2
4 DT 40% C2H4 46% C2H4 05% C2H2 12% C2H2 28% C2H2
5 T1 05% C2H2 16 % C2H4
6 T2 05% C2H2 16 % C2H4 46 % C2H4
7 T3 14% C2H2 46 % C2H4
Table 5.13: Triangular coordinates for DTM 3 zones (Silicone Oil)
No. Area Points CH4 C2H4 C2H2
1 PD PD1 98 2 0
2 PD2 100 0 0
3 PD3 98 2 0
111
4 D1 D11 0 0 100
5 D12 88 0 12
6 D13 78 10 12
7 D14 0 10 90
8 D2 D21 0 10 90
9 D22 78 10 12
10 D23 48 40 12
11 D24 32 40 28
12 D25 0 72 28
13 DT DT1 0 72 28
14 DT2 32 40 28
15 DT3 48 40 12
16 DT4 88 0 12
17 DT5 95 00 5
18 DT6 49 46 5
19 DT7 40 46 14
20 DT8 0 86 14
21 T1 T11 95 0 5
22 T12 98 0 2
23 T13 98 2 0
24 T14 84 16 0
25 T15 79 16 5
26 T2 T21 84 46 0
27 T22 79 16 5
28 T23 49 46 5
29 T24 54 46 0
30 T3 T31 54 46 0
31 T32 40 46 14
32 T33 0 86 14
33 T34 0 100 0
The fault analysis result for one of the sample of below given data is analyzed and
represented in Fig. 5.14. This shows type of fault as well as the graphical representation
of the fault in simulation.
112
Figure 5.14: Fault analysis results of DTM 3 for non-mineral oil (Silicon oil)
5.7.1.2 Midel oil
Similarly, numerical zone boundaries of DTM3, Midel oil are shown in Table 5.14 and
polygon coordinates are created for these zone boundaries are given in Table 5.15.
Table 5.14: Numerical zone boundaries for DTM 3 (Midel oil)
No. Fault Limits of zones
1 PD 98 % CH4 2 D1 28 % C2H4 12 % C2H2
3 D2 28 % C2H4 12 % C2H2 42 % C2H4 27 % C2H2
4 DT 40% C2H4 64% C2H4 05% C2H2 14% C2H2 12% C2H2
5 T1 05% C2H2 40% C2H4
6 T2 05% C2H2 40 % C2H4 69 % C2H4
7 T3 14% C2H2 69 % C2H4
Table 5.15: Triangular coordinates for DTM 3 zones (Midel Oil)
No. Area Points %CH4 %C2H4 %C2H2
1 PD PD1 98 2 0
2 PD2 100 0 0
3 PD3 98 2 0
4 D1 D11 0 0 100
5 D12 88 0 12
6 D13 60 28 12
7 D14 0 28 72
8 D2 D21 0 28 72
9 D22 60 28 12
10 D23 48 40 12
11 D24 32 40 28
113
12 D25 0 72 28
13 DT DT1 88 0 12
14 DT2 95 0 5
15 DT3 26 69 5
16 DT4 17 69 14
17 DT5 00 86 14
18 DT6 00 72 28
19 DT7 32 40 28
20 DT8 0 86 14
21 T1 T11 95 0 5
22 T12 98 0 2
23 T13 98 2 0
24 T14 60 40 0
25 T15 55 40 5
26 T2 T21 55 40 5
27 T22 60 40 00
28 T23 31 69 00
29 T24 26 69 14
30 T3 T31 31 69 0
31 T32 17 69 14
32 T33 0
86 14
33 T34 0 100 0
The three inputs from a sample are entered in the simulation environment. The fault
diagnosis results by DTM3 Midel oil are shown in words as sell as graphical analysis
shown in Fig. 5.15.
enter the value of Methane=23
enter the value of Ethylene=12
enter the value of Acetylene=65
"Duval Triangle Test is Applicable"
""""""DUVAL TRIANGLE RESULTS"""
DISCHARGE OF LOW ENERGY -- D1
114
Figure 5.15: Fault analysis results by DTM3 (Midel oil)
5.7.1.3 FR 3 oil
In a similar way, numerical zone boundaries of DTM3, FR3 natural ester oil is shown in
Table 5.16 whose polygon coordinates are developed and shown in Table 5.17.
Table 5.16: Numerical zone boundaries for DTM 3 (FR3 oil)
No. Fault Limits of zones
1 PD 99 % CH4
2 D1 25 % C2H4 13 % C2H2
3 D2 25 % C2H4 13 % C2H2 40 % C2H4 19 % C2H2
4 DT 40% C2H4 64% C2H4 05% C2H2 13% C2H2
29% C2H2
5 T1 05% C2H2 44% C2H4
6 T2 05% C2H2 44 % C2H4 64 % C2H4
7 T3 14% C2H2 65 % C2H4
Table 5.17: Triangular coordinates for DTM 3 zones (FR3 Oil)
No. Area Points CH4 C2H4 C2H2
1 PD PD1 99 0 1
2 PD2 100 0 0
3 PD3 99 1 0
4 D1 D11 0 0 100
5 D12 87 0 13
6 D13 62 25 13
7 D14 0 25 75
8 D2 D21 0 25 75
9 D22 62 25 13
10 D23 47 40 13
11 D24 41 40 29
12 D25 0 71 29
115
13 DT DT1 87 0 13
14 DT2 95 0 5
15 DT3 31 64 5
16 DT4 12 64 14
17 DT5 00 86 14
18 DT6 00 71 29
19 DT7 41 40 29
20 DT8 41 40 13
21 T1 T11 95 0 5
22 T12 99 0 1
23 T13 99 1 0
24 T14 56 44 0
25 T15 51 44 5
26 T2 T21 51 49 5
27 T22 56 44 00
28 T23 36 64 00
29 T24 31 64 5
30
T3
T31 36 64 0
31 T32 22 64 14
32 T33 0
86 14
33 T34 0 100 0
A sample of below given inputs is tested to find the type of fault though DTM3 FR3 oil.
The analysis results are shown in simulated DTM3 FR3 Fig. 5.16 and as written fault in
words.
enter the value of Methane=23
enter the value of Ethylene=25
enter the value of Acetylene=25
"Duval Triangle Test is Applicable"
""""""DUVAL TRIANGLE RESULTS"""
DISCHRAGE OF HIGH ENERGY -- D2
116
>>
Figure 5.16: Fault analysis results by DTM3 (FR3 oil)
5.7.1.4. Bio-Temp oil
Further, numerical zone boundaries of Bio-Temp oil are given in Table 5.18 and polygon
coordinates for the same boundaries of BioTemp oil are developed and shown in Table
5.19.
Table 5.18: Numerical zone boundaries for DTM 3 (BioTemp oil)
No. Fault Limits of zones
1 PD 98 % CH4 2 D1 10 % C2H4 12 % C2H2
3 D2 10 % C2H4 12 % C2H2 40 % C2H4 28 % C2H2
4 DT 40% C2H4 46% C2H4 05% C2H2 12% C2H2 28% C2H2
5 T1 05% C2H2 16 % C2H4
6 T2 05% C2H2 16 % C2H4 46 % C2H4
7 T3 14% C2H2 46 % C2H4
Table 5.19: Triangular coordinates for DTM 3 zones (BioTemp oil)
No. Area Points %CH4 %C2H4 %C2H2
1 PD PD1 98 0 2
2 PD2 100 0 0
3 PD3 98 2 0
4 D1 D11 0 0 100
5 D12 89 0 11
6 D13 68 21 11
7 D14 0 21 79
8 D2 D21 0 21 79
9 D22 68 21 11
10 D23 48 41 11
11 D24 31 41 28
117
12 D25 0 72 28
13 DT DT1 89 0 11
14 DT2 97 0 3
15 DT3 13 84 3
16 DT4 00 89 11
17 DT5 31 41 28
18 DT6 48 41 11
19 T1 T11 97 0 3
20 T12 98 0 2
21 T13 98 2 0
22 T14 47 53 0
23 T15 44 53 3
24 T2 T21 44 53 3
25 T22 47 53 00
26 T23 16 84 00
27 T24 13 84 3
28 T3 T31 0 84 16
29 T32 0 16 84
30 T33 0
100 00
Here a sample for fault diagnosis purposes tested for DTM3 BioTemp oil as below given
inputs. The fault diagnosis results from simulated MatLab environment in graphical
presentation shown in Fig, 5.17 and also in words.
enter the value of Methane=23
enter the value of Ethylene=25
enter the value of Acetylene=25
"Duval Triangle Test is Applicable"
""""""DUVAL TRIANGLE RESULTS"""
DISCHRAGE OF HIGH ENERGY -- D2
Figure 5.17 Fault analysis result by DTM 3 (BioTemp oil)
118
5.8 DUVAL TRIANGLE METHODS 4 AND 5 FOR LOW-TEMPERATURE
FAULTS IN TRANSFORMERS
Duval triangle method 1 for transformers has proved quite efficient to determine the
general type of fault occurring in transformers in service (PD, D1, D2, T1, T2, T3, DT). It
is based on the use of 3 gases (CH4, C2H4 and C2H2) representative of low-, high-, and
very high-energy or temperature faults, respectively [64]. When DGA results are close to
the boundary between zones PD and T1. However, manually it may be difficult to
distinguish between these two types of faults. Also, some types of oils on the market tend
to be ―stray gassing‖, i.e., to unexpectedly form gases at relatively low temperatures such
as 80 °C to 200 °C. Stray gases may appear in the PD, T1, or T2 zones, and therefore may
interfere with the identification of these faults in transformers.
The Duval Triangles 4 and 5 for low-temperature faults in transformers have been
developed to remove these uncertainties. They should be applied only to faults identified
by the Duval Triangle 1 as possible faults PD, T1, or T2, and should be viewed only as a
complement of information for the Duval Triangle 1. These should not be applied to
faults identified by the Duval Triangle 1 as faults D1, D2, or T3.
5.8.1 Duval Triangle Method 4 for Low-Energy Gases
The Duval Triangle method 4 (DTM 4) uses the 3 low-energy gases (H2, CH4, and C2H6)
for incipient fault diagnosis of transformers filled with mineral oils in the main tank.
DTM 4 is shown in Fig.5.18.
Figure 5.18: Duval triangle method 4 transformers filled with mineral oil
The identification of fault zones in DTM 4 is indicated in Table 5.20 and the numerical
zones boundaries (in %) in Table 5.21.
119
Table 5.20 Duval triangle method 4 fault zones identifications
No. Zones Identification
1 PD Corona Partial Discharge
2 S Stray gassing of mineral oil
3 C Hot spot with carbonization of paper (T > 300oC)
4 O Overheating (T <250oC)
Table 5.21 Duval triangle 4 numerical zone boundary
No. Fault Limits of Zones
1 PD CH4=2, C2H6=1, CH4=15
2 S C2H6=44, H2=9, C2H6=24, CH4=36, CH4=15,
C2H6=1, CH4=2
3 C CH4=36, C2H6=24
4 O H2=9, C2H6=24
5.8.2 Implementation and Results for Duval Triangle Method 4
From the fault zone limits of DTM4, the percentage numerical zone boundaries are
represented in Table 5.22 and DTM4‘s polygon coordinates are created, shown in Table
5.23.
Table 5.22: Numerical zone boundaries for DTM 4 (low intensity faults)
No. Fault Limits of Zones%
1 PD 2 CH4 C2H6 15 CH4
2 S 44 C2H6 19 CH4 24 C2H6 36 CH4 15 CH4 1 C2H6 2 CH4
3 C 36 CH4 24 C2H6
4 O 9 H2 24 C2H6
Table 5.23: Triangular coordinates for DTM 4 zones
No. Area Points % H2 CH4%
CH4
% C2H6
1 PD PD1 98 2 0
2 PD2 85 15 0
3 PD3 84 15 1
4 PD4 93 2 1
5 S S1 56 0 44
6 S2 100 0 0
7 S3 98 2 0
8 S4 97 2 1
9 S5 84 15 1
120
10 S6 64 36 0
11 S7 74 26 0
12 S8 42 36 24
13 S9 9 67 24
14 C C1 0 76 24
15 C2 40 36 24
16 C3 0 100 0
17 C4 0 76 24
18 O O1 0 0 100
19 O2 9 0 81
20 O3 9 69 24
21 O4 0 76 24
The inputs of a sample are entered to simulate the DTM4 in MATLAB environment and
the fault diagnosis in DTM4 simulated results are shown in Fig.5.19.
Enter the input data—
Figure 5.19: Fault analysis results by DTM 4 for low-intensity faults
5.8.3 Duval Triangle Method 5 (DTM5) for Low-Temperature Faults
It is seen that gases formed by corona partial discharges PD, and low-temperature faults
in oil are relatively similar. This can be explained, however, by the fact that temperature
121
in such PDs, occurring in the gas phase and also known as cold plasma discharges, is
indeed quite low compared to sparking partial discharges of the D1 type.
The Duval triangle method 5 for low-temperature faults in transformers can be used to
confirm fault attributions which are still uncertain after using the Duval triangle method
4. It uses the 3 so-called temperature gases (C2H4, CH4, and C2H6) and is presented in
Fig.5.20. It is based on the inspected cases of transformer faults and stray gassing
laboratory test results. References for these inspected cases are the same as in low-energy
gases for T3 faults.
Figure 5.20: Duval triangle method 5 for transformers filled with mineral oil (Low-
Temperature Gases)
The identification of fault zones of DTM5 is indicated in Table 5.24 and the numerical
zones boundaries in Table 5.25.
Table 5.24: Duval triangle method 5 fault zones identifications
No. Fault Identification
1 PD Corona Partial Discharge
2 S Stray gassing of mineral oil
3 C Hot spot with carbonization of paper (T > 300oC)
4 O Overheating (T <250oC)
5 T3 Thermal faults of very high temperatures(T > 700oC)
Table 5.25: Duval triangle method 5 numerical zone boundaries
No. Fault Limits of Zones
1 PD C2H4=1, C2H6=15,
122
2 S C2H6=53, C2H4=10, C2H6=15,
3 C C2H4=36, C2H6=30, C2H4=35,
4 O C2H4=10, C2H6=53 C2H6=15, C2H4=1,
5 T3 C2H4=34,
Faults occurring in zone O are only evidenced by paper turning brown and are often
difficult to find by visual inspection of the oil-impregnated paper. Cases through Duval
triangle method 4 & 5 in are also located in zone O, and have been attributed to stray
gassing. This suggests that temperature in these cases possibly was > 200°C. New stray
gassing tests at 250 to 300 °C would be necessary to confirm this possibility.
Stray gassing of non-mineral oils (Midel, FR3, and BioTemp), based on laboratory
results, occurs in the Duval triangle method 4 in zone S and in the unnamed zone between
zones S and O. It occurs in the Duval triangle method 5 in zone O and in the unnamed
zone between zones O and C.
Only a few cases of abnormal gassing in transformers in service filled with FR3 oil have
been available so far. They occur mostly in zones O and S of the Duval triangle 4 and in
zone O of the Duval triangle 5.
5.8.4 Implementation and Results for Duval Triangle Method 5 (DTM 5)
Similar analysis is carried out for DTM 5. Its percentage numerical zone boundaries are
shown in Table 5.26, whose polygon coordinates are created and shown in Table 527.
Table 5.26: Numerical zone boundaries for DTM 5
No. Faults Limits of Zones
1 PD 1% C2H4 15% C2H6
2 S 53% C2H6 10% C2H4 13% C2H6
3 C 10% C2H4 30% C2H6 36 C2H4
4 O 10% C2H4 53% C2H6 15% C2H6
5 T3 36% C2H4
Table 5.27 Triangular coordinates for DTM 5 zones
No. Area Points % CH4 % C2H4 % C2H6
1 PD PD1 85 0 15
2 PD2 99 0 1
3 PD3 98 1 1
4 PD4 84 15 1
5 S S1 47 0 53
6 S2 85 0 15
123
7 S3 75 10 15
8 S4 37 10 53
9 O O1 0 0 100
10 O2 47 0 53
11 O3 37 10 53
12 O4 0 10 90
13 O5 85 0 15
14 O6 84 15 1
15 O7 75 10 15
16 O8 85 10 0
17 O9 100 0 0
18 O10 99 0 1
19 O11 98 1 1
20 C C1 60 10 30
21 C2 34 36 30
22 C3 64 36 0
23 C4 90 10 0
24 T3
T31 0 37 63
25 T32 64 36 0
26 T33 0 100 0
Simulated result from DTM5 for a given sample input in fault interpretation of power
transformer is shown in Fig. 5.21.
Figure 5.21: Fault analysis results by DTM 5 for low temperature fault
124
Duval triangle method 1 to Duval triangle method 5 can be applied as DGA for fault
diagnosis of power transformers, whether oil in main tank, oil in OLTCs, different types
of synthetic oils and also separately for Low-energy faults. Duval triangle methods can
provide complete solutions. A generalized flow- chart for complete solution is shown in
Fig. 5.22.
Figure 5.22: Generalized flow-chart for complete solution by Duval triangle methods
SUMULATE DTM 5
Start
TYPE OF OIL? SIMULATE DTM 3
CHECKING FOR
WHICH OIL? SIMULATE DTM 2
AMBIGUIS
FAULTS?
SIMULATE DTM 4
SIMULATE DTM 1
AMBIGUIS
FAULTS? SIMULATE DTM 5
SIMULATE DTM 4
Stop Stop Stop Stop
OLTC
MAIN TANK
SYNTHETIC
MINERAL
LOW INTENSITY
HIGH INTENSITY
LOW INTENSITY
HIGH INTENSITY
125
The following conclusions are drawn from the DTMs:
Duval triangle fault interpretation methods can be used graphically in a very easy
way and the evaluation of faults is also easy.
Available eight Duval triangles are capable of fault diagnosis separately; such as
main transformer, online load tap-changers (OLTCs), faults with low energy,
faults with low temperatures and synthetic oils.
Simulated fault analysis results in DTM1 compared with manual fault diagnosis of power
transformers for the reported faults are shown in Table 5.28 by different conventional
fault interpretations method used across the world. Same simulated results are compared
with Roger‘s ratio and Dornenburg‘s ratio fault interpretations techniques to emphasize
the advantages of Duval triangle method fault interpretation are given in Table 5.29.
These fault analyses tables have been are prepared from the fault reports presented in
Appendix-B, collected from different authorities and organizations.
126
Table 5.28: Comparative analysis between simulated and graphical fault interpretations in Duval triangle 1
S.No. Transformer Rating GASSES
TOTAL %CH4 %C2H4 %C2H2 DGA Fault Diagnosis/ Comments DTM 1 RESULTS
MVA KV CH4 C2H4 C2H2 CEBG IEC, CIGRE IEEE/Reco. Comnts SIMULATION MANUAL
1 100 220/66 187 614 ND 801 23.35 76.65 ND C2H4 High Hot spot T3 T3
2 20 11/0.4 10 23 11 44 22.73 52.27 25.00 C2H2 High Overloading DT DT
3 9989 66-33/114 6852 3137 ND 9989 68.60 31.40 ND C2H2 High Old OIL T2 T2
4 100 220/66 254 313 417 984 25.81 31.81 42.38 C2H2 High Hot spot D2 D2
5 100 220/66 83 221 3 307 27.04 71.99 0.98 C2H4 High Hot spot T3 T3
6 315 400/200 55 72 0 127 43.31 56.69 0.00 300-700 °C >700°C T3 T3
7 315 400/220 79 89 0
1
68 47.02 52.98 0.00 300-700 °C <700°C
T3 T2
8 250 NG 547 530 nd 1077 50.79 49.21 ND 300-700 °C Cir. C T2 T2
9 25 NG 29 21 47 97 29.90 21.65 48.45 AD D1 D2
10 50 NG 748 2587 10 3345 22.36 77.34 0.30 AD T3 T3
11 50 400 108 50 0 158 68.35 31.65 0.00 Cir. C 300-700 °C Cir. C C2H4 High T2 T1
12 200 400/21 80 120 0 200 40.00 60.00 0.00 Cir. C >700°C Cir. C C2H4 High T3 T3
13 250 16/420 40 10 0 50 80.00 20.00 0.00
200-
300°C NF
200-300°C Ind. Gases D1 T1
14 200 400/21 90 125 0 215 41.86 58.14 0.00 Cir. C >700°C Cir. C C2H4 High T3 T3
15 250 16/420 30 0 0 30 100.00 0.00 0.00 <150°C 150-300°C 150°C C2H6 High PD PD
16 50 420 25 10 0 35 71.43 28.57 0.00 PD NF N SG T2 T1
17 63 420 105 15 0 120 87.50 12.50 0.00 PD NF N IG D1 T1
18 200 400/21 2 2 0 4 50.00 50.00 0.00 PD NF N SG T3 T2
19 200 420/21 1 1 0 2 50.00 50.00 0.00 PD NF N IG T3 T2
20 200 400/21 20 7 0 27 74.07 25.93 0.00 S OK NF N SG T2 T2
21 200 420/133 1 6 0 7 14.29 85.71 0.00 PD NF N NA T3 T3
22 200 420/33 31 10 0 41 75.61 24.39 0.00 S OK NF N NA T2 T2
23 240 21/420 4 1 0 5 80.00 20.00 0.00 S OK NF N NA T1 T2
127
24 80 132/11 0 0 0 0 nd nd nd PD NF N DT
25 200 21/1420 20 3 0 23 86.96 13.04 0.00 SOK NF N NA T1 T1
26 200 420/21 5 1 0 6 83.33 16.67 0.00 SOK NF N NA T1 T1
27 200 420/21 2 2 0 4 50.00 50.00 0.00 PD NF N IG T2 T2
28 80 132/11 10 1 0 11 90.91 9.09 0.00 SOK NF N NA T1 T1
29 75 400/345 8 5 0 13 61.54 38.46 0.00 SOK NF T2 T2
30 50 33/6.9 25 39 0 64 39.06 60.94 0.00 Cir. C 300-700 °C Cir. C C2H4 High T3 T3
31 25 11/6.65 3 0 0 3 100.00 0.00 0.00 SOK NF PD PD
32 25 11/6.6 15 0 0 15 100.00 0.00 0.00 SOK NF PD PD
33 50 420 5 1 0 6 83.33 16.67 0.00 DZD NF DZD B T1 T1
34 13 33/6.9 6 0 0 6 100.00 0.00 0.00 SOK NF N PD PD
35 32 33/6.9 8 3 0 11 72.73 27.27 0.00 SOK NF N T2 T2
36 50 33/6.9 17 33 0 50 34.00 66.00 0.00 MF <150°C MF T3 T3
37 16 16/6.9 26 2 0 28 92.86 7.14 0.00
200-
300°C NF 200-300°C T1 T1
38 25 220/6.9 5 13 0 18 27.78 72.22 0.00 CO <150°C CO T3 T3
39 13 33/6.9 5 10 0 15 33.33 66.67 0.00 SOK NF N T3 T3
40 16 15/6.9 3 0 0 3 100.00 0.00 0.00 PD NF N PD PD
41 25 16/6.9 26 2 0 28 92.86 7.14 0.00
200-
300°C NF 200-300°C T1 T1
42 63 145 85 3 0 88 96.59 3.41 0.00 SOK NF N T1 T1
43 13 33/6.9 7 0 0 7 100.00 0.00 0.00 SOK NF N PD PD
44 25 11/6.6 8 18 0 26 30.77 69.23 0.00 SOK NF N T3 T3
45 75 NG 14 0 0 14 100.00 0.00 0.00
45 75 NG 11 0 0 11 100.00 0.00 0.00
45 75 NG 1 0 0 1 100.00 0.00 0.00
45 75 NG 1 0 0 1 100.00 0.00 0.00
45 75 NG 3 0 0 3 100.00 0.00 0.00
128
45 75 NG 4 0 0 4 100.00 0.00 0.00
45 75 NG 5 0 0 5 100.00 0.00 0.00
45 75 NG 14 0 0 14 100.00 0.00 0.00 PD PD,D1,T1 N H2, CO2 PD PD
46 200 NG 160 35 0 195 82.05 17.95 0.00
47 16 NG 35 2 0 37 94.59 5.41 0.00
47 16 NG 32 1 0 33 96.97 3.03 0.00
47 16 NG 30 2 0 32 93.75 6.25 0.00
47 16 NG 30 1 0 31 96.77 3.23 0.00
47 16 NG 20 13 0 33 60.61 39.39 0.00
47 16 NG 10 5 0 15 66.67 33.33 0.00
47 16 NG 3 5 0 8 37.50 62.50 0.00
47 16 NG 56 55 43 154 36.36 35.71 27.92 N D1,D2 SRQ D2 D2
48 63 NG 395 475 0 870 45.40 54.60 0.00
48 63 NG 413 432 0 845 48.88 51.12 0.00
48 63 NG 403 412 0 815 49.45 50.55 0.00
48 63 NG 393 401 0 794 49.50 50.50 0.00
48 63 NG 494 361 0 855 57.78 42.22 0.00
48 63 NG 680 416 0 1096 62.04 37.96 0.00
48 63 NG 672 405 0 1077 62.40 37.60 0.00
48 63 NG 663 420 0 1083 61.22 38.78 0.00 >700°C T3,>700°C SQR >700°C T2 T2
49 50 NG 490 451 15 956 51.26 47.18 1.57
49 50 NG 185 510 18 713 25.95 71.53 2.52
49 50 NG 2100 1408 26 3534 59.42 39.84 0.74
49 50 NG 3010 1946 34 4990 60.32 39.00 0.68
49 50 NG 3586 2280 38 5904 60.74 38.62 0.64
49 50 NG 51 65 0 116 43.97 56.03 0.00
49 50 NG 36 110 3 149 24.16 73.83 2.01
129
9 50 NG 92 120 4 216 42.59 55.56 1.85 >700°C D2,T3 N H2, C2H4 T3 T3
50 16 132/33 2 8 0 10 20.00 80.00 0.00 T OH T3 T3
51 10 33/11 3 10 0 13 23.08 76.92 0.00 T N T3 T3
52 10 33/11 2 5 0 7 28.57 71.43 0.00 T N T3 T3
53 16 132/33 4 63 2 69 5.80 91.30 2.90 T OH T3 T3
54 16 132/33 4 9 0 13 30.77 69.23 0.00 T OH T3 T3
55 6.3 33/11 3 17 0 20 15.00 85.00 0.00 T OH T3 T3
56 6.3 33/11 2 8 0 10 20.00 80.00 0.00 T N T3 T3
57 16 132/33 45 110 147 302 14.90 36.42 48.68 T AD D2 D2
58 16 132/33 52 106 214 372 13.98 28.49 57.53 T AD D2 D2
59 16 132/33 1 18 0 19 5.26 94.74 0.00 T OH T3 T3
60 16/20 132/33 17 17 0 34 50.00 50.00 0.00 T OH T3 T3
61 16/20 132/33 1 12 0 13 7.69 92.31 0.00 T N T3 T3
62 16 132/33 3 64 42 109 2.75 58.72 38.53 OH OH DT D2
63 6.3 33/11 2 28 0 30 6.67 93.33 0.00 OH OH T3 T3
64 20 66/11 47 72 196 315 14.92 22.86 62.22 Cir. C& AD C2H4,C2H2 DT D2
65 20 66/11 51 79 201 331 15.41 23.87 60.73 Cir. C& AD C2H4,C2H2 D2 D2
66 20 66/11 45 70 189 304 14.80 23.03 62.17 Cir. C& AD C2H4,C2H2 D2 D2
67 20 66/11 49 75 199 323 15.17 23.22 61.61 Cir. C& AD C2H4,C2H2 D2 D2
DUVAL TRIANGLE fault diagnosis satisfy that more then 95% interpretations are correct compare to other fault diagnosis
techniques, those; CEBG, IEC CIGRE,IEEE, Roger‘s and Dornenburg‘s ratios, etc techniques are yet used by different agencies.
130
Table 5.29: DTM1 simulated fault diagnosis compared with Roger‘s and Dornenburg‘s ratio fault interpretations
S.
No.
Transformer Rating GASSES
TOTAL H2 C2H6 CO
DGA Fault Diagnosis/ Comments DTM 1
Simulated
Roger’s Dornen-
burg’s
MVA KV CH4 C2H2 C2H4 CEBG
IEC,
CIGRE IEEE/Reco. Comnts
1 100 220/66 187 ND 614 801 ND
178 ND C2H4 High Hot spot T3 >700C >700C
2 20 11/0.4 10 11 23 44 ND
2 ND C2H2 High Overloading DT NI NI
3 10/12.5 66-33/114 6852 ND 3137 9989 2195
3137 ND C2H2 High Old OIL T2 >700C TF
4 100 220/66 254 417 313 984 328
17 ND C2H2 High Hot spot D2 He Arc DA
5 100 220/66 83 3 221 307 13
10 ND C2H4 High Hot spot T3 >700C TF
6 315 400/200 55 0 72 127 39
22 80 300-700 °C >700°C
T3 >700C NF
7 315 400/220 79 0 89 168 40
42 110 300-700 °C <700°C
T3 <700C NF
8 250 NG 547 nd 530 1077 18
217 ND 300-700 °C
Cir. C T2 <700C TF
9 25 NG 20 47 21 88 139
10 ND AD D1 NI DA
10 50 NG 748 10 2587 3345 198
182 ND AD T3 >700C TF
11 50 400 108 0 50 158 42
Cir. C 300-700 °C Cir. C
C2H4 High T2 NI NF
12 200 400/21 80 0 120 200 45
Cir. C >700°C Cir. C
C2H4 High T3 NI NF
13 250 16/420 40 0 10 50 52 48 150 200-
300°C NF
200-300°C Ind. Gases D1 NF NF
14 200 400/21 90 0 125 215 50
30 95 Cir. C
>700°C Cir. C
C2H4 High T3 >700C TF
15 250 16/420 30 0 0 30 25
15 110 <150°C 150-300°C 150°C
C2H6 High PD NI NF
16 50 420 25 0 10 35 20
65 190 PD NF N SG T2
NI NF
17 63 420 105 0 15 120 45
66 310 PD NF N IG D1
NI TF
18 200 400/21 2 0 2 4 30 9 182 PD NF N SG T3
PD NF
19 200 420/21 1 0 1 2 0 1 50 PD NF N IG T3
<700C NF
20 200 400/21 20 0 7 27 48
5 90 S OK NF N SG T2
LTO NF
21 200 420/133 1 0 6 7 30
2 185 PD NF N
NA T3 NI NF
22 200 420/33 31 0 10 41 35
19 185 S OK NF N
NA T2 NF NF
131
23 240 21/420 4 0 1 5 35
5 232 S OK NF N
NA T1 NF NF
24 80 132/11 0 0 0 0 5 ND ND PD NF N DT NI NF
25 200 21/1420 20 0 3 23 45
3 230 SOK NF N NA T1 NF NF
26 200 420/21 5 0 1 6 10 1 90 SOK NF N NA T1 NI NF
27 200 420/21 2 0 2 4 36
3 47 PD NF N IG T2 PD NF
28 80 132/11 10 0 1 11 39 1 230 SOK NF N NA T1 NF NF
29 75 400/345 8 0 5 13 12
40 130 SOK NF T2 NF NF
30 50 33/6.9 25 0 39 64 16 19 70 Cir. C 300-700 °C Cir. C C2H4 High T3 <700C NF
31 25 11/6.65 3 0 0 3 22
0 36 SOK NF PD NI NF
32 25 11/6.6 15 0 0 15 33 0 160 SOK NF PD NI NF
33 50 420 5 0 1 6 83
0 105 DZD NF DZD B T1 NI NF
34 13 33/6.9 6 0 0 6 13 1 190 SOK NF N PD NI NF
35 32 33/6.9 8 0 3 11 22
1 210 SOK NF N T2 LTO NF
36 50 33/6.9 17 0 33 50 32 22 65 MF <150°C MF T3 LTO NF
37 16 16/6.9 26 0 2 28 36
45 192
200-
300°C NF 200-300°C T1 NF NF
38 25 220/6.9 5 0 13 18 30 10 380 CO <150°C CO T3 LTO NF
39 13 33/6.9 5 0 10 15 16
2 95 SOK NF N T3 NI NF
40 16 15/6.9 3 0 0 3 40 1 16 PD NF N PD NI NF
41 25 16/6.9 26 0 2 28 36
45 192
200-
300°C NF 200-300°C T1 NF NF
42 63 145 85 0 3 88 150
10 310 SOK NF N T1 NI NF
43 13 33/6.9 7 0 0 7 14
0 126 SOK NF N PD NI NF
44 25 11/6.6 8 0 18 26 22
3 179 SOK NF N T3 NI NF
45 75 NG 14 0 0 14
45 75 NG 11 0 0 11
45 75 NG 1 0 0 1
132
45 75 NG 1 0 0 1
45 75 NG 3 0 0 3
45 75 NG 4 0 0 4
45 75 NG 5 0 0 5
45 75 NG 14 0 0 14 41
0 47 PD PD,D1,T1 N H2, CO2 PD NI NF
46 200 NG 160 0 35 195 210 340
47 16 NG 35 0 2 37
47 16 NG 32 0 1 33
47 16 NG 30 0 2 32
47 16 NG 30 0 1 31
47 16 NG 20 0 13 33
47 16 NG 10 0 5 15
47 16 NG 3 0 5 8
47 16 NG 56 43 55 154 16 2 201 N D1,D2 SRQ D2 NI NI
48 63 NG 395 0 475 870
48 63 NG 413 0 432 845
48 63 NG 403 0 412 815
48 63 NG 393 0 401 794
48 63 NG 494 0 361 855
48 63 NG 680 0 416 1096
48 63 NG 672 0 405 1077
48 63 NG 663 0 420 1083 127 54 12 >700°C T3,>700°C SQR >700°C T2 >700C TF
49 50 NG 490 15 451 956
49 50 NG 185 18 510 713
49 50 NG 2100 26 1408 3534
133
49 50 NG 3010 34 1946 4990
49 50 NG 3586 38 2280 5904
49 50 NG 51 0 65 116
49 50 NG 36 3 110 149
49 50 NG 92 4 120 216 115
18 90 >700°C
D2,T3 N H2, C2H4 T3 NI NI
50 16 132/33 2 0 8 10 18 3 158 T OH T3 LTO NF
51 10 33/11 3 0 10 13 4
2 186 T N T3 NI NF
52 10 33/11 2 0 5 7 4 6 173 T N T3 NF NF
53 16 132/33 4 2 63 69 3.5
16 51 T OH T3 >700C NF
54 16 132/33 4 0 9 13 11 189 7 T OH T3 NF NI
55 6.3 33/11 3 0 17 20 4
6 25 T OH T3 LTO NF
56 6.3 33/11 2 0 8 10 4 4 144 T N T3 LTO NF
57 16 132/33 45 147 110 302 41
11 297 T AD D2 NI NI
58 16 132/33 52 214 106 372 103 7 275 T AD D2 He Arc DA
59 16 132/33 1 0 18 19 29
25 231 T OH T3 PD NF
60 16/20 132/33 17 0 17 34 60 10 405 T OH T3 LTO NF
61 16/20 132/33 1 0 12 13 1
16 262 T N T3 NF NF
62 16 132/33 3 42 64 109 2 31 290 OH OH DT NI NF
63 6.3 33/11 2 0 28 30 2
20 497 OH OH T3 LTO NF
64 20 66/11 47 196 72 315 4 Cir. C& AD C2H4,C2H2 DT NI NF
65 20 66/11 51 201 79 331 41
Cir. C& AD C2H4,C2H2 D2 NF NF
66 20 66/11 45 189 70 304 Cir. C& AD C2H4,C2H2 D2 NF NF
67 20 66/11 49 199 75 323
Cir. C& AD C2H4,C2H2 D2 NF NF
134
5.9 SUMMARY
In this chapter graphical and simulations of different Duval Triangle methods for DGA
provides the following results:
1. Duval Triangle Methods (DTMs) for DGA fault interpretations in power transformers
is very simple (with three gases only) and consuming less time.
2. These methods have satisfied the fault diagnosis (both graphical (manual) and
simulation implementation) more than 95% accurate than any other method of
diagnostics.
3. Traces of one of the three gases can provide the quick fault diagnosis to a little
experienced worker on the power transformers.
4. These methods always provide a diagnosis, with a very low percentage of wrong
diagnosis.
5. Duval triangle representation also allows to follow graphically in a very easy way and
the evaluation of faults with time visually.
6. From all above analysis and their developed simulations for all Duval Triangle
method 1 to Duval Triangle method5 respectively has proved that this technique can
provide full solutions in fault analysis of power transformers by DGA.
135
CHAPTER 6
APPLICATION OF ARTIFICIAL INTELLIGENCE IN DUVAL TRIANGLE
METHODS
6.1 INTRODUCTION
Certain amount of uncertainty in the data concerning dissolved gas analysis due to
generation, sampling, and chromatography analysis exist and thus variation in gas
interpretation affects the variation of the gases by the utilities. Due to the diverse gas
content of the insulating oil of transformers, many artificial intelligence (AI) techniques
may be used such as expert systems, support vector machines (SVM), fuzzy logic (FL),
artificial neural networks (ANN), genetic algorithm (GA) and even novel cerebellar
model articulation controller based method for off-line and on-line monitoring.
6.2 ARTIFICIAL INTELLIGENCE TECHNIQUES
Artificial intelligence (AI) is the intelligence of the machines and the branch of computer
science that aims to create it. AI can be said as the study of intelligent agent, where an
intelligent agent is a system that perceives its environment and takes action that
maximizes its chances of success. This was founded on claim that a central property of
human, intelligence can be precisely described that it can be simulated by a machine.
Therefore, artificial intelligence is a science based upon the human cognitive process. AI
has been the subject optimism, but also has suffered the setbacks. Nevertheless, today,
AI has become the part of the technology industry, providing the solutions to most
difficult problems in computer science.
6.3 FUZZY LOGIC TECHNIQUE (FLT)
Fuzzy logic is the theory of fuzzy sets, sets that calibrates roughness. Fuzzy is not logic
that is fuzzy, but a logic that is used to describe fuzzyness. Fuzzy logic is based on the
idea that all things admit of degrees, such as temperature, height, speed, beauty, distance,
etc, which can be seen of sliding scale. Thus fuzzy logic is a branch of fuzzy set theory,
which deals (similar to logic systems do) with the representation and inference form of
knowledge. The fuzzy analysis consists of three parts: fuzzification, fuzzy inference and
defuzzification. Fuzzification is the process of transforming crisp input values into grades
of membership for linguistic terms of fuzzy sets. The membership function is used to
associate a grade to each linguistic term. A chosen fuzzy inference system (FIS) is
responsible for drawing conclusions from the knowledge-based fuzzy rule set of if-then
linguistic statements. Fault types are first listed and then form the fuzzy rule set for the
136
diagnosis system. Defuzzification then converts the fuzzy output values back into crisp
output actions see Fig. 6.1.
Figure 6.1 Basic fuzzy logic control
A fuzzy set F of a universe of discourse X = 0, 1 is defined as a mapping, µF (x) : X →
[0, α], by which each x is assigned a number in the range [0, α], indicating the extent to
which x has the attribute F. If x is say ‗small‘ a value of fuzzy variables from the range
of ‗0 to ∞‘, then µsmall (x) € [0, α] is called a ‗membership function‘. A membership
function is normalized (ie., α = 1), thus µF (x) : X → [0, 1], fuzzy logic is called ‗normal‘.
The normalization of fuzzy set is expressed by
sup µ (x) = 1 (6.1)
x € X
Normalization of a set of numbers is achieved by dividing each number of the set by the
largest one, the supermum. Where, X be a time-invariant set of objects x. A fuzzy set is
X will be expressed by a set of ordered pairs:
= [(x, µ (x))| x € X ] (6.2)
Where, µ is the membership function that maps X to the membership space M (ie. 0 to
1) and µ (x) is the grade of membership (degree of truth) x in .
Tomsovic et al made use of fuzzy information approach to integrating different
transformer DGA diagnostic methods [51], which has shown that fussy relation are
combined with a faults tree to provide the best analysis. In this DGA has been
implemented with satisfying results. It could also be desirable to suggest further
measurements which could help clarify the diagnosis in marginal cases. Also in this paper
current implementations of the non-DGA methods were tried since these techniques are
gaining the experiences. Huang et al designed robust transformer DGA fault diagnosis
system with the help of evolutionary FL [52, 53]. Using IEC/IEEE criteria as reference,
an evolutionary based fuzzy system was developed to identify the incipient faults in the
Knowledge
Base
Fuzzyfication
Interface
Defuzzification
Interface
Decision
Making Logic
137
transformers. In this research the conventional DGA methods were improved in
accordance with the record of the practical examination of suspected transformers, an
evolutionary based optimization algorithm was employed to further modify the fuzzy if-
then rules and simultaneous adjust the membership function of fuzzy subsets. Yang et al
[54, 55] used adaptive fuzzy diagnostic system to enhance the fault diagnosis capabilities
for DGA of power transformers. Using practical gas records and associated fault causes, a
fuzzy reasoning algorithm was presented to establish a preliminary fuzzy diagnosis
system. The proposed scheme has enhanced the tuning of the fuzzy membership functions
a well as rule-pruning scheme for more compact diagnosis system. Zhang et al applied
fuzzy equivalent matrix (FEM) for fault diagnosis of oil-immersed insulation [56].
Compared with conventional methods, it is not based on inferring rules, but abstracting
fault information from the original data unsupervised to the recognition of DGA data for
power transformers. Subsequently, [57] which applied fuzzy data processing for DGA of
power transformers, fuzzy clustering analysis (FCA) technique introduced by the same
researcher to identify the fault patterns, its clustering centers are dynamic, which can
classify and recombine different samples successfully. Denghua et al [58] used a new
fuzzy information optimization processing for monitoring the transformers, used different
diagnostic methods and information as mathematics theory to establish the membership
functions of different data for oil-filled transformers. Q. Su [59, 60] utilized fuzzy logic
tool for transformer fault diagnosis. Since conventional interpretation methods of DGA
such as IEC codes can detect faults in many cases, where more than one fault exists in a
transformer. Thus, this technique is very useful tool for diagnosis and maintenance
planning of power transformers.
Hooshmand et al [61] have applied fuzzy systems in DGA fault diagnosis using IEC,
ASTM and CEGB standards and categorized the results into two groups: crisp and fuzzy
set. One of the problems in fault diagnosis of transformer based on dissolved gas is lack
of matching the result of fault diagnosis of different standards with the real world. The
result of the different standards is analyzed using fuzzy and the result is compared with
the empirical test. The comparison between the suggested method and existing methods
indicate the capability of the suggested method in on-line fault diagnosis of the
transformers. In addition, in some cases the existing standards are not able to diagnose the
fault. The presented method has the potential of diagnosing the fault. The used the
information of three transformers to the show the capability of the suggested method in
138
diagnosing the fault. The results validate the capability of the presented method in fault
diagnosis of the transformer.
Flores et al [62] applied type-2 fuzzy logic as novel algorithm in fault diagnosis of power
transformers through dissolved gas analysis. Afiqah et al. [63] proposed two DGA
methods which are Rogers Ratio and IEC Ratio to interpret the DGA results. However,
there are situations of errors and misleading results occurring due to borderline and
multiple faults. Fuzzy logic is implemented here as an improved DGA interpretation
method that provides higher reliability and precision of fault diagnosis. Aragon-Patil at al
[64] has shown improvements of interpretation of DGA in power transformers for IEC
code ratios. Sun et al [65] utilized in research with fuzzy model hierarchy classification in
his study for fault diagnosis of power transformers through DGA.
6.4 FUZZY LOGIC APPROACH IN DTM1
The fuzzy logic approach used for fault diagnosis of power transformer with inputs and
an output is depicted in Fig. 6.2.
Figure 6.2: Fuzzy logic approach in DTMs
The following steps are used in Duval triangle method for fault diagnosis of power
transformers:
Step 1; Firstly, polygon coordinates for the numerical zone boundaries of seven key
faults of Duval Triangle method 1 are used in terms of percentages of CH4, C2H4 and
C2H2, from 0% to 100% respectively as mentioned in Table 5.6.
Step 2; In this diagnostic each crisp value of gas percentages (CH4, C2H4, C2H2) from in
Table 5.6, which are changed to triangular fuzzy-membership function ranges (ie.
139
A.......M, A1......I1 and A2......H2 respectively) shown in Table 6.1. Fuzzy-membership
function for input gases obtained in MATLAB fuzzy-box tool is illustrated in Fig. 6.3(a)
to Fig. 6.3(c).
Table 6.1: Membership function ranges for three gas inputs of DTM1
No. CH4 C2H4 C2H2
1 A 0-30 A1 0-1 A2 0-1
2 B 25-34 B1 1-15 B2 1-3
3 C 34-35 C1 10-22 C2 3-10
4 D 35-46 D1 22-32 D2 8-14
5 E 46-49 E1 30-42 E2 14-25
6 F 49-60 F1 40-55 F2 25-40
7 G 55-75 G1 55-75 G2 70-85
8 H 70-79 H1 70-87 H2 80-100
9 I 79-86 I1 85-100
10 J 85-95
11 K 90-97
12 L 97-99
13 M 99-100
Figure 6.3(a): %CH4 membership-function
140
Figure 6.3(b): %C2H4 member functions
Figure 6.3(c): % C2H2 member functions
Step 3: Fuzzy outputs- transformer faults; In this diagnostic, each crisp value for
transformer faults are further rearranged, shown in Table 6.2 and changed to output
trapezium fuzzy-membership function into 33 faults. Fuzzy-membership function for
faults by DTM1 are obtained and illustrated in Fig. 6.4.
Table 6.2: Transformer fault outputs DTM 1
No. Area Points Faults
1 PD PD1 1
PD2 2
PD3 3
2 D1 D11 4
D12 5
D13 6
D14 7
141
3 D2 D21 8
D22 9
D23 10
D24 11
D25 12
4 DT DT1 13
DT2 14
DT3 15
DT4 16
DT5 17
DT6 18
DT7 19
DT8 20
5 T1 T11 21
T12 22
T13 23
T14 24
T15 25
6 T2 T21 26
T22 27
T23 28
T24 29
7 T3 T31 30
T32 31
T33 32
T34 33
142
Figure 6.4: Membership function of DTM 1 faults
Step 4; Fuzzy-rule base; The set of fuzzy inputs of three gases assigned with crisp
values (% three gases) with their respective membership function form the integral part of
fuzzy logic analysis. The fuzzy rule set (If-then linguistic statements) with AND operator
for minimum and OR operator for maximum fault conditions is then used form
‗judgment‘ on the fuzzy inputs derived from the three gas percentages, whose sum is
always 100%. For example;
All such thirty three fuzzy rules just derived for mapping the fault types and screenshot of
these rules is shown Table 6.3 in DTM1. Although these faults are defined strictly for the
percentages of the zone boundaries of seven faults in Duval triangle method 1
coordinates.
Table 6.3: Fuzzy Rules for DTM 1
Rules % Gas input Membership Function Output
CH4 C2H4 C2H2
1 L A1 B2 PD1
2 M A1 A2 PD2
3 L B1 A2 PD3
4 A A1 H2 D11
5 A D1 G2 D12
6 G D1 D2 D13
7 J A1 D2 D14
8 A D1 G2 D21
9 A H1 F2 D22
10 B E1 F2 D23
11 E E1 D2 D24
143
12 G D1 D2 D25
13 A G1 F2 DT1
14 A H1 E2 DT2
15 C F1 E2 DT3
16 D F1 C2 DT4
17 K A1 C2 DT5
18 J A1 D2 DT6
19 E E1 D2 DT7
20 B E1 F2 DT8
21 H C1 C2 T11
22 I C1 A2 T12
23 L B1 A2 T13
24 L A1 B2 T14
25 K A1 C2 T15
26 D F1 C2 T21
27 F F1 A2 T22
28 I C1 A2 T23
29 H C1 C2 T24
30 A H1 E2 T31
31 A H1 A2 T32
32 F F1 A2 T33
33 C F1 E2 T34
Step 5; Fuzzy Inference System (FIS), FIS involves the operation between input fuzzy
sets as shown in Fig. 6.5, known as ‗Mamdani‘ type. This derives output fuzzy sets
‗judging‘ all the possible fuzzy rules by finding the membership for the fault types as
represented by thirty three fuzzy output rules. The solution is reached by weighted
average of the fuzzy inputs. The spikes in the figure denoting probabilities between 0 and
1 are the outputs from each fuzzy rule, which denotes the fault type. Thus each rule is a
row of the plots and each column is variable.
144
Figure 6.5: FIS analysis of fuzzy rules for DTM1
6.5 FUZZY LOGIC APPROACH FOR DTM 2
Fuzzy logic approach for fault diagnosis in oil-filled OLTCs with the help of DTM2, the
similar steps are followed as the fault diagnosis by DTM1 for the incipient faults in the
power transformers is adopted. The approach as follows:
Input ranges: The ranges of three inputs gases in DTM2 for fault diagnosis in OLTCs of
power transformers are given in Table 6.4.
Table 6.4: DTM 2 ranges of input gases
CH4 C2H4 C2H2
A1 1-20 A2 6-22 A3 58-92
C1 20-100 C2 22-85 C3 78-100
D1 0-62 D2 22-50 D3 0-81
E1 35-76 E2 50-100 E3 15-80
F1 0-50
F3 0-15
The DTM2 triangular fuzzy-membership function ranges are arranged as (ie. A1.......H1,
A2......F2 and A3......H3 respectively) for the three input gas percentages of (CH4, C2H4,
C2H2). These are shown in Fig. 6.6 (a) – Fig. 6.6 (c) for corresponding three gases.
145
Figure 6.6 (a): DTM 2 Membership functions of CH4
Figure 6.6 (b) DTM 2 Membership functions of C2H4
Figure 6.6 (c): DTM 2 Membership functions of C2H2
Output of DTM 2; Ranges for OLTC‘s six fault outputs are given in Table 6.5 and their
fuzzy membership function for the same output ranges are shown in Fig. 6.7.
Table 6.5: Types of faults in OLTCs
No. Faults Ranges
1 N 0-10
146
2 D1 10-30
3 X1 20-40
4 X3 40-50
5 T2 50-70
6 T3 60-70
Figure 6.7: OLTC‘s faults output ranges
Fuzzy rules; Fuzzy rules for the fault diagnosis in oil filled OLTC tank for the given
input and output combinations are shown in Table 6.6.
Table 6.6: Fuzzy rules for OLTC faults
Rules % Gas input Membership Function Output
CH4 C2H4 C2H2
1 A1 A2 A3 N
2 A1 B2 B3 D1
3 C1 B2 D3 X1
4 D1 C2 E3 X3
5 E1 D2 F3 T2
6 F1 F2 F3 T3
Fuzzy fault output in FIS analysis in DTM2 for given rules are shown in Fig. 6.8.
147
Figure 6.8: FIS analysis of fuzzy rules for DTM2
6.6 FUZZY LOGIC APPROACH FOR DTM 3
For fuzzy logic approach for DTM3 in fault diagnosis for synthetic oils will involve the
same procedure like the steps followed for fault diagnosis for DTM1 and DTM2. There
are four types of synthetic oil (Silicone, Midel, FR3 and BioTemp oil) are available with
different fault zones, and number of faults. Therefore, separate four DTM3 are used for
four types of synthetic oils. These are as follows:
6.6.1 Fuzzy Logic Approach for DTM3, Silicone oil
Input ranges: Inputs to the fuzzy logic DTM3 are arranged in the different ranges and
shown in Table 6.7. The fuzzy membership functions of same three input gases are shown
in Fig. 6.9 (a) to Fig. 6.9 (c).
Table 6.7: Input Ranges for DTM3, Silicone oil
CH4 C2H4 C2H2
A1 0-25 A2 0-1 A3 0-1
B1 25-35 B2 1-3 B3 1-3
C1 35-45 C2 5-15 C3 3-7
D1 45-49 D2 15-25 D3 8-13
E1 48-50 E2 35-45 E3 12-28
F1 50-60 F2 45-55 F3 20-30
G1 70-80 G2 70-80 G3 80-95
H1 75-80 H2 80-90 H3 95-100
I1 80-88 I2 90-100
J1 85-90
148
K1 90-97
L1 97-99
M1 99-100
Figure 6.9(a): Membership functions for CH4 DTM3, Silicone oil
Figure 6.9(b): Membership functions for C2H4 DTM3, Silicone oil
Figure 6.9(c): Membership function for C2H2 DTM3, Silicone oil
149
DTM3 faults in Silicone oil: Output membership function for DTM3 Silicone oil are
further subdivided, shown in Table 6.8 and fuzzy membership functions from fuzzy tool-
box are shown in Fig. 6.10
Table 6.8: Fault output membership functions of DTM3, Silicone oil
No. Area Points Faults
1 PD PD1 1
PD2 2
PD3 3
2 D1 D11 4
D12 5
D13 6
D14 7
3 D2 D21 8
D22 9
D23 10
D24 11
D25 12
4 DT DT1 13
DT2 14
DT3 15
DT4 16
DT5 17
DT6 18
DT7 19
DT8 20
5 T1 T11 21
T12 22
T13 23
T14 24
T15 25
6 T2 T21 26
150
T22 27
T23 28
T24 29
7 T3 T31 30
T32 31
T33 32
T34 33
Figure 6.10: Output membership functions of silicone oil
Fuzzy rules; Fuzzy rules for fault diagnosis through DTM3, silicone oil for the above
mentioned inputs and output membership functions are stated in Table 6.9.
Table 6.9: Fuzzy Rules for DTM3, Silicone oil
Rules % Gas input Membership Function Output
CH4 C2H4 C2H2
1 L1 B2 A3 PD1
2 M1 A2 A3 PD2
3 L1 B2 A3 PD3
4 A1 A2 H3 D11
5 J1 A2 D3 D12
6 A1 C2 G3 D13
7 A1 C2 G3 D14
8 G1 C2 D3 D21
151
9 B1 E2 F3 D22
10 D1 E2 D3 D23
11 G1 C2 D3 D24
12 A1 G2 F3 D25
13 A1 G2 F3 DT1
14 B1 E2 F3 DT2
15 D1 E2 D3 DT3
16 J1 A2 B3 DT4
17 K1 A2 C3 DT5
18 E1 A2 C3 DT6
19 C1 F2 E3 DT7
20 A1 H2 E3 DT8
21 K1 A2 C3 T11
22 L1 A2 B3 T12
23 L1 B2 A3 T13
24 I1 D2 A3 T14
25 H1 B2 C3 T15
26 I1 F2 A3 T21
27 H1 D2 A3 T22
28 E1 F2 C3 T23
29 F1 F2 A3 T24
30 F1 F2 A3 T31
31 C1 F2 E3 T32
32 A1 H2 E3 T33
33 A1 I2 A3 T34
Fuzzy fault output in FIS analysis in DTM3, silicone oil for given rules are shown in Fig.
6.11.
152
Figure 6.11: FIS analysis of fuzzy rules for DTM3, Silicone oil
6.6.2 Fuzzy Logic Approach for DTM3, Midel Oil
Input ranges; The ranges of input for DTM3 Midel oil and their fuzzy membership are
given in Table 6.10 and Fig. 6.12 (a) – 6.12 (c) for three input gases in fuzzy logic
approach.
Table 6.10: Ranges of input for DTM3, Midel oil
CH4 C2H4 C2H2
A 0-10 A 0-1 A 0-1
B 10-20 B 1-5 B 1-3
C 20-30 C 20-30 C 3-8
D 28-32 D 35-45 D 5-15
E 30-35 E 65-70 E 12-15
F 40-48 F 70-75 F 20-30
G 50-60 G 80-90 G 65-75
H 58-63 H 90-100 H 90-100
I 80-90
J 90-95
K 95-98
L 98-100
153
Figure 6.12 (a): Fuzzy membership function of CH4 for DTM3, Midel oil
Figure 6.12 (b): Fuzzy membership function of C2H4 for DTM3, Midel oil
Figure 6.12 (c): Fuzzy membership function of C2H2 for DTM3 Midel oil
DTM3 fault output; There are thirty three fault outputs for DTM3, Midel oil shows in
Table 6.11, and fuzzy membership function listed in Fig. 6.13.
154
Table 6.11: Output ranges for DTM3, Midel oil
No. Area Points faults
1 PD PD1 1
PD2 2
PD3 3
2 D1 D11 4
D12 5
D13 6
D14 7
3 D2 D21 8
D22 9
D23 10
D24 11
D25 12
4 DT DT1 13
DT2 14
DT3 15
DT4 16
DT5 17
DT6 18
DT7 19
DT8 20
5 T1 T11 21
T12 22
T13 23
T14 24
T15 25
6 T2 T21 26
T22 27
T23 28
T24 29
155
7 T3 T31 30
T32 31
T33 32
T34 33
Figure 6.13: Output membership functions for DTM3, Midel oil
Fuzzy rules; There are thirty three output fuzzy rules are framed for DTM3, Midel oil
shown in Table 6.11.
Table 6.12: Fuzzy rules for DTM3, Midel oil
Rules % Gas input Membership Function Output
CH4 C2H4 C2H2
1 K B A PD1
2 L A A PD2
3 K B A PD3
4 A A H D11
5 I A D D12
6 H C D D13
7 A C G D14
8 A C G D21
9 H C D D22
10 F D D D23
11 E D F D24
12 A E F D25
156
13 I A C DT1
14 K A C DT2
15 C E C DT3
16 B E E DT4
17 A G E DT5
18 A E F DT6
19 E D F DT7
20 A G E DT8
21 J A C T11
22 L A B T12
23 L B A T13
24 H D A T14
25 G D C T15
26 G D C T21
27 H D A T22
28 D E A T23
29 C E E T24
30 D E A T31
31 B E E T32
32 A G E T33
33 A H A T34
Fuzzy fault output in FIS analysis in DTM3, midel oil for given rules are shown in Fig.
6.14.
157
Figure 6.14: FIS analysis of fuzzy rules for DTM3, Midel oil
6.6.3 Fuzzy Logic Approach for DTM3, FR3 Oil
Fuzzy logic approach for DTM3, FR3 oil in DGA fault diagnosis of power transformers
as follows:
Input ranges; Three input ranges for DTM3, FR3 oil are shown in Table 6.13.
Table 6.13: Input ranges for DTM3, FR3 oil
CH4 C2H4 C2H2
A 0-10 A 0-1 A 0-1
B 10-20 B 0-20 B 0-3
C 20-30 C 20-30 C 3-10
D 25-35 D 30-40 D 7-15
E 30-45 E 40-45 E 15-20
F 40-48 F 45-50 F 25-35
G 45-55 G 55-65 G 70-80
H 50-55 H 65-75 H 90-100
I 55-60 I 80-90
J 60-70 J 90-100
K 80-90
L 85-95
M 95-100
N 98-100
158
Fuzzy membership functions for above stated inputs of DTM3 FR3 oil are shown in the
Fig.6.15 (a) – Fig. 6. 15 (c).
Figure 6.15 (a): Fuzzy membership functions for DTM3, FR3 oil
Figure 6.15 (b): Fuzzy membership functions for DTM3, FR3 oil
Figure 6.15 (c): Fuzzy membership functions for DTM3, FR3 oil
159
Fault outputs; Thirty three outputs are assigned for fault diagnosis of power transformer
by DTM3, FR3 oil are given in Table 6.14.
Table 6.14: Fault output for DTM3, FR3 oil
No. Area Points Faults
1 PD PD1 1
PD2 2
PD3 3
2 D1 D11 4
D12 5
D13 6
D14 7
3 D2 D21 8
D22 9
D23 10
D24 11
D25 12
4 DT DT1 13
DT2 14
DT3 15
DT4 16
DT5 17
DT6 18
DT7 19
DT8 20
5 T1 T11 21
T12 22
T13 23
T14 24
T15 25
6 T2 T21 26
T22 27
160
T23 28
T24 29
7 T3 T31 30
T32 31
T33 32
T34 33
Output membership functions for DTM3, FR3 oil in fuzzy environment are given in Fig.
6.16.
Figure 6.16: Output membership functions for DTM3 FR3 oil
Fuzzy rules; Fuzzy rules for DTM3, FR3 oil fault diagnosis approach are shown in Table
6.15.
Table 6.15: Fuzzy rules for DTM3, FR3 oil
Rules % Gas input Membership Function Output
CH4 C2H4 C2H2
1 L1 B2 A3 PD1
2 M1 A2 A3 PD2
3 L1 B2 A3 PD3
4 A1 A2 H3 D11
5 J1 A2 D3 D12
6 A1 C2 G3 D13
7 A1 C2 G3 D14
161
8 G1 C2 D3 D21
9 B1 E2 F3 D22
10 D1 E2 D3 D23
11 G1 C2 D3 D24
12 A1 G2 F3 D25
13 A1 G2 F3 DT1
14 B1 E2 F3 DT2
15 D1 E2 D3 DT3
16 J1 A2 B3 DT4
17 A1 I2 E3 DT5
18 A1 H2 F3 DT6
19 F1 D2 F3 DT7
20 F1 D2 D3 DT8
21 L1 A2 C3 T11
22 N1 A2 B3 T12
23 N1 B2 A3 T13
24 I1 E2 A3 T14
25 H1 E2 C3 T15
26 H1 F2 C3 T21
27 I1 E2 A3 T22
28 E1 G2 A3 T23
29 D1 G2 C3 T24
30 E1 G2 A3 T31
31 C1 G2 E3 T32
32 A1 I2 E3 T33
33 A1 J2 A3 T34
Fuzzy fault output in FIS analysis in DTM3, FR3 oil for given rules are shown in Fig.
6.17.
162
Figure 6.17: FIS analysis of fuzzy rules for DTM3, FR3 oil
6.6.4 Fuzzy Logic Approach for DTM 3, Bio-Temp Oil
Similar way, a fuzzy logic approach is also applied for DTM3, BioTemp oil in DGA fault
diagnosis of power transformers as:
Input ranges; Three gas inputs are arranged in different ranges for DTM3, BioTemp oil
filled transformers for fault diagnosis given in Table 6.16.
Table 6.16: Input ranges for DTM3, BioTemp oil
CH4 C2H4 C2H2
A 0-10 A 0-1 A 0-1
B 4-15 B 1-3 B 1-3
C 15-25 C 10-20 C 2-5
D 25-35 D 20-30 D 5-15
E 35-45 E 34-45 E 10-20
F 45-50 F 50-60 F 25-35
G 48-56 G 70-80 G 70-80
H 60-70 H 75-85 H 80-90
I 80-90 I 85-95 I 90-100
J 90-97 J 90-100
K 99-100
163
Triangular fuzzy membership function ranges for three gas inputs of DTM3, BioTemp oil
are shown in Fig. 6.18 (a) – Fig. 6.18 (c).
Figure 6.18 (a): Input membership function CH4 in DTM3, BioTemp oil
Figure 6.18 (b): Input membership function C2H4 in DTM3, Biotemp oil
Figure 6.18 (c): Input membership function of C2H2 in DTM3, BioTemp oil
Fault outputs; In DTM3, BioTemp oil outputs assigned for fault diagnosis of power
transformer are given in Table 6.17.
Table 6.17: Fault output for DTM3 BioTemp
No. Area Points faults
1 PD PD1 1
164
PD2 2
PD3 3
2 D1 D11 4
D12 5
D13 6
D14 7
3 D2 D21 8
D22 9
D23 10
D24 11
D25 12
4 DT DT1 13
DT2 14
DT3 15
DT4 16
DT5 17
DT6 18
5 T1 T11 19
T12 20
T13 21
T14 22
T15 23
6 T2 T21 24
T22 25
T23 26
T24 27
7 T3 T31 28
T32 29
T33 30
Fuzzy membership functions for fault output variables from DTM3, BioTemp oil are
shown in Fig. 6.19.
165
Figure 6.19: Fuzzy outputs of DTM3, BioTemp oil
Fuzzy rules; Fuzzy rules for assigned thirty outputs in DTM3, BioTemp oil fault
diagnosis of power transformer are given in Table 6.18.
Table 6.18: Fuzzy rules for DTM3, BioTemp oil
Rules % Gas input Membership Function Output
CH4 C2H4 C2H2
1 K A B PD1
2 K A A PD2
3 K B A PD3
4 A A I D11
5 I A D D12
6 H D D D13
7 A D G D14
8 A D G D21
9 H D D D22
10 G E D D23
11 D E F D24
12 A G F D25
13 I A D DT1
14 J A C DT2
15 B H C DT3
16 B I D DT4
17 D E F DT5
18 G E D DT6
166
19 J A C T11
20 K A B T12
21 K B A T13
22 F F A T14
23 E F C T15
24 E F C T21
25 F F A T22
26 C H A T23
27 B H C T24
28 A H E T31
29 A C H T32
30 A J A T33
Fuzzy fault output in FIS analysis in DTM3, BioTemp oil for given rules are shown in
Fig. 6.20.
Figure 6.20: FIS analysis of fuzzy rules for DTM3, BioTemp oil
167
6.7 FUZZY LIGIC APPROACH FOR DTM4, LOW ENERGY FAULTS
DTM4 is used for low intensity faults over and above the fault diagnosis with DTM1. The
procedure is similar to fault diagnosis to other DTMs. This follows as:
Input ranges; DTM4 is used low energy faults in the main transformer tank, input gas
ranges (H2, CH4, C2H6) are given in Table 6.19.
Table 6.19: Input ranges for DTM4
H2 CH4 C2H6
A 0-5 A1 0-1 A2 0-0.5
B 5-15 B1 1-3 B2 0.5-5
C 30-45 C1 10-20 C2 20-30
D 40-55 D1 20-30 D2 40-55
E 54-62 E1 30-40 E2 75-88
F 62-72 F1 55-68 F2 90-100
G 72-82 G1 68-75
H 82-86 H1 75-85
I 84-88 I1 90-100
J 90-95
K 90-98
L 97-100
Fuzzy membership functions for inputs in DTM4 used for low energy faults occurring in
power transformers are shown in Fig. 6.21(a) – Fig. 6.21(c).
Figure 6.21 (a): Fuzzy membership functions of H2 for DTM4
168
Figure 6.21 (b): Fuzzy membership functions of CH4for DTM4
Figure 6.21 (c): Fuzzy membership functions of C2H6 for DTM4
Fault outputs; Faults output ranges for DTM4 are described and shown in Table 6.20.
Table 6.20: DTM4 output ranges
No. Area Points Faults
1 PD PD1 1
PD2 2
PD3 3
PD4 4
2 O O1 5
O2 6
O3 7
O4 8
3 S S1 9
S2 10
169
S3 11
S4 12
S5 13
S6 14
S7 15
S9 16
S10 17
4 C C1 18
C2 19
C3 20
C4 21
Fuzzy rules; Fuzzy rules for the input and output ranges for DTM 4 for fault diagnosis
are given in Table 6.21.
Table 6.21: Fuzzy rules for DTM4
Rules % Gas input Membership Function Output
H2 CH4 C2H6
1 L B1 A2 PD1
2 I C1 A2 PD2
3 H C1 B2 PD3
4 J B1 B2 PD4
5 E A1 D2 O1
6 L A1 A2 O2
7 L B1 A2 O3
8 K B1 B2 O4
9 H C1 B2 S1
10 F E1 A2 S2
11 G D1 A2 S3
12 D E1 C2 S4
13 B F1 C2 S5
14 A F1 C2 S6
15 C E1 C2 S7
170
16 D E1 A2 S8
17 B F1 A2 S9
18 A A1 F2 C1
19 B A1 E2 C2
20 B G1 C2 C3
Fuzzy fault output in FIS analysis in DTM4 for given rules are shown in Fig. 6.22.
Figure 6.22: FIS analysis of fuzzy rules for DTM4
6.8 FUZZY LIGIC APPROACH FOR DTM5, LOW TEMPERATURE FAULTS
DTM5 is used as low -temperature faults in the main transformer tank. This fault
diagnosis method is used along with DTM1 to ascertain fault in low intensity electrical
discharges in transformers. Rest of the procedure is similar to other DTMs as follows:
Input ranges; Ranges of three inputs for DTM5 used for fault in power transformers are
arranged as given in Table 6.22.
Table 6.22: Input ranges of DTM5
CH4 C2H4 C2H6
A1 0-5 A2 0-1 A3 0-0.5
B1 5-15 B2 1-3 B3 0.5-5
171
C1 30-45 C2 10-20 C3 20-30
D1 45-55 D2 20-30 D3 40-55
E1 54-62 E2 30-40 E3 75-88
F1 62-72 F2 55-68 F3 90-100
G1 72-82 G2 68-75
H1 82-86 H2 75-85
I1 84-88 I2 90-100
J1 90-95
K1 90-98
L1 97-100
M1 98-100
Fuzzy membership functions for the three inputs of DTM5 are shown in Fig. 6.23 (a) –
Fig. 6.23 (c).
Figure 6.23 (a): Fuzzy membership functions of CH4 in DTM5
Figure 6.23 (b): Fuzzy membership functions of C2H4 in DTM5
172
Figure 6.23 (c): Fuzzy membership functions of C2H6 in DTM5
Fault outputs; Fault outputs in DTM5 are rearranged to represent faults are given in
Table 6.23.
Table 6.23: Fault outputs in DTM5
No. Area Points Faults
1 PD PD1 1
PD2 2
PD3 3
PD4 4
2 S S1 5
S2 6
S3 7
S4 8
3 O O1 9
O2 10
O3 11
O4 12
O5 13
O6 14
O7 15
O9 16
O10 17
O11 18
173
4 C C1 19
C2 20
C3 21
C4 22
5 T3 T31 23
T32 24
T33 25
DTM5 outputs as fuzzy membership functions are given in Fig. 6.24.
Figure 6.24: Fuzzy output membership function for DTM5
Fuzzy rules; Fuzzy rules in DTM5 for fault diagnosis for input output combinations are
given in Table 6. 24.
Table 6.24: Fuzzy rules for DTM5
Rules % Gas input Membership Function Output
CH4 C2H4 C2H2
1 A1 A2 C3 PD1
2 I1 A2 B3 PD2
3 K1 B2 B3 PD3
4 H1 D2 B3 PD4
5 D1 A2 E3 S1
6 I1 A2 C3 S2
7 G1 C2 CE S3
8 C1 C2 E3 S4
174
9 A1 A2 H3 O1
10 D1 A2 E3 O2
11 C1 C2 E3 O3
12 A1 C2 G3 O4
13 I1 A2 C3 O5
14 H1 D2 B3 O6
15 G1 C2 C3 O7
16 I1 C2 A3 O8
17 L1 A2 A3 O9
18 L1 A2 B3 O10
19 K1 B2 B3 O11
20 E1 C2 D3 C1
21 B1 E2 D3 C2
22 F1 E2 A3 C3
23 J1 C2 A3 C4
24 A1 F2 F3 T31
25 F1 E2 A3 T33
26 A1 G2 A3 T32
Fuzzy fault output in FIS analysis in DTM5 for given rules are shown in Fig. 6.25.
Figure 6.25: FIS analysis of fuzzy rules for DTM5
175
6.9 ARTIFICIAL NEURAL NETWORK TECHNIQUES (ANNT)
Artificial Neural Networks is a massively parallel distributed processor as shown in Fig.
6.26, having a natural tendency to acquire sufficient experimental Knowledge and making
it available for use.
Figure 6.26: Two-layered Feedforward Nueral Network
A neuron model is the most basic information processing unit in a neural network.
Depending on complexity they are organized in three or more layers: the input layer, the
output layer and one or many hidden layers. Each neuron model receives input signals,
which are multiplied by synaptic weights. An activation function transforms these signals
into an output signal to the next neuron model and so on. Generalization ability and
learning stability are the two main issues for multilayer feed-forward neural network
(FNN). A back-propagation algorithm is a well suitable method for training the multilayer
feed-forward ANN, which has good ability of pattern recognition. Hencefore , a well
suitable Feed-forward Back-propagation Two-layer artificial neural network has been
applied in Duval Triangle 1 method of fault diagnosis in power transformers in this study.
A sigmoid activation transfer function is applied because of its two advantages: firstly, it
is highly nonlinear, and secondly, it has good performance when working with back-
propagation learning algorithms. Typically, for the conventional back-propagation (BP)
training algorithm , the mean square error (MSE) is used. The mean square error is
defined as follows:
(6.3)
where, Ep is the error for the pth pattern vector, ypj is the desired value of the output node
and opj is the actual output of the jth output node. The aim of learning is to minimize the
instantaneous square error of the output signal by modifying the weights and bias the bias
176
term can be regarded as weights and iteratively alters in the same manner as the other
weights. The ‗sigmoid‘ function used is as follows:
Opj = f(netpj) = 1/ (1 + e-netpj -qj
) (6.4)
Here,
netpj (6.5)
where, the sum is taken over the output Opj, if all nodes in layer ( L-1).
Among these AI techniques, the ANN is widely designed to diagnose transformer faults.
An important advantage of fault diagnosis based on the ANN is that it can learn directly
from the training samples and update its knowledge when necessary. The high nonlinear
mapping capability of neurons always provides a comparable and superior performance
over a fuzzy system solution. However, although the computational complexity of the
ANN is not too high, especially in the fault diagnostic process, it involves certain
problems including slow convergence, oscillation and so on, all of which must be
resolved before it is practically applied.
An Artificial Neural Network included selection of inputs, outputs, network topology and
weighted connection of node. Input feature-selection constitutes an essential first step.
This is chosen very carefully so that the input features will correctly reflects the
characteristics of the problem. Another major task of the ANN design is to choose
network topology. This is done experimentally through a repeated process to optimize the
number of hidden layers and nodes according to training performance and prediction
accuracy.
For many years there was no theoretically sound algorithm for training multi-layer ANN,
and therefore, the applications of ANN were severely limited. The invention of Back-
propagation algorithm has played a vital role in the resurgence of interest in ANN. Back-
propagation is a systematic method for training multi-layer ANN. It has a strong
mathematical foundation.
A two-layer feed-forward network is shown in Fig. 6.25. It consists of number of neurons
connected by links divided into two layers. A set of inputs is applied from outside or from
previous layer. Each of these is multiplied by a corresponding weight w. The sum of the
weighted inputs and the bias .b. forms the input .n. to the transfer/activation function ―F‖.
Neurons may use any differentiable, monotonic increasing transfer functions to generate
their outputs. Back-propagation networks often use the log-sigmoid and tan-sigmoid
transfer functions.
177
Zhang et al [66] proposed ANN approach to transformer fault diagnosis with DGA and
paper insulation degradation. In the current study, five key gases H2, CH4, C2H6, C2H4,
and C2H2 are chosen as input features. Since overheating, partial discharge and arcing are
the three major fault types in power transformers; hence there will be four output patterns
to be identified including the normal condition. A two step ANN method is used to detect
fault with and without paper cellulose, and obtained the good accuracy. Y. M. Tu et al
[67] utilizes improved ANN for many methods and criteria of DGA diagnosis to produce
the comparative study. The ANN will extract essential characteristics from the input data
without physical model to improve the accuracy of diagnostic. The diagnosing accuracy
of ANN will be improved further by using multi-subdivided ANNs and biological nerves.
Yanming et al [68] applied BP-ANN in the collected DGA of seven gases data from
transformer faults with improved algorithm is used in which the adjustment of the
weights is adaptive. Zheng et al [69] tested a synthetic diagnostic method for faults of oil
immersed power transformers and suggested many of its advantages. Thang et al [70] a
comparison of conventional DGA interpretations methods such CIGRE‘s, IEC, Roger‘s,
Dornenburg‘s ratios are discussed. Finally, a new approach includes simple statistical
analysis on DGA record and followed by high level data-mining using self – organizing
map algorithm to prove alternate for incipient fault diagnosis and monitoring applications.
Guardalo et al [71] presented a comparative study of neural network efficiency for
detection of incipient faults in transformers. ANN was trained for five DGA diagnostic
methods ie Roger‘s, Modified Roger,s, Dornenburg,s, IEC and CSUS, the study shown
that NN rate of successful fault evaluation is in the range of 87-100%. Patel et al [72]
introduces ANN based transformer fault diagnosis to show the application of adaptive BP
learning algorithm, which drastically reduces the number of epochs required to increase
the accuracy of the diagnosis compared to conventional BP algorithm. Dhlamini et al
presented parallel ANN [73] to successfully diagnose the condition of bushing through
DGA as per IEEE C-57-104 along with other methods and revealed that ANN produces
the accuracy of 99%.
Yan-jing et al [74] improved the back propagation (BP)-based artificial neural networks
(ANN) can identify complicated relationships among dissolved gas contents in
transformer oil and corresponding fault types, using the highly nonlinear mapping nature
of the neural nets. An efficient BP-ALM (BP with Adaptive Learning Rate and
Momentum coefficient) algorithm is proposed to reduce the training time and avoid being
trapped into local minima, where the learning rate and the momentum coefficient are
178
altered at iterations. We developed a system of transformer fault diagnosis based on
Dissolved Gases Analysis (DGA) with a BP-ALM algorithm. Training patterns were
selected from the results of a Refined Three-Ratio method (RTR). Test results show that
the system has a better ability of quick learning and global convergence than other
methods and a superior performance in fault diagnosis compared to convectional BP-
based neural networks and RTR.
Moreira et al [75] developed power transformer diagnosis system on two neural networks.
This uses gaseous chromatography and physical-chemical analysis. Results obtained
provided high accuracy rate of correct diagnosis compared with commonly used
traditional methods in addition proposed system presented itself as an alternative to
automation of the analysis process. Xiang-Hao at al [76] proposed artificial immune
network (ANIC) algorithm mimics the adaptive learning and defense mechanism to
respond to fault sample of the transformers, which has shown remarkable diagnosis
accuracy, multiple incipient faults can be classified effectively.
6.10 ANN APPROACH FOR DTM 1
Following steps are used in DTM 1 for fault diagnosis of power transformers with
artificial neural network technique:
Step 1; Polygon coordinates for the numerical zone boundaries of seven key faults of
Duval Triangle 1 are already stated in Table 5.6 and the same input percentages of
numerical zones are being used to create strings of three gas ratios (CH4, C2H4 and C2H2).
Step 2; Input and output patterns A complete ANN includes selection of inputs,
outputs, network topology and weighted connections of nodes. In this thesis the input
pattern are obtained from the DGA fault diagnosis reports of the faulty transformers
collected from organizations/utilities across INDIA. For each input pattern there exists an
output pattern which describes the fault types for a given diagnosis criterion. Both input
and output pattern formulate an ANN training set. Since Duval triangle 1 needs only three
combustible gases methane (CH4), ethylene (C2H4) and acetylene (C2H2), therefore, for
this study, input and output patterns for fault diagnosis in power transformers are as
follows:
[Input] = [x] = [%CH4 %C2H4 %C2H2] T
[0utput] = [O] = [PD D1 D2 DT T1 T2 T3] T
Output represents faults and the number assigned for each is as follows:
[O] = [1 2 3 4 5 6 7] T
179
A brief parameter details for this thesis work of DGA fault diagnosis in power
transformers through ANN based Duval triangle method 1 approach are shown in Table
6.25.
Table 6.25: Parameters for ANN Duval Triangle method 1
Input [%CH4 %C2H6 %C2H2]T
Output Faults [1 2 3 4 5 6 7]T
Maximum Input
Value
100%, 100%, 100%
Minimum Input
Value
0.1%, 0.1%, 0.1%
Number of Samples 1236
Epochs 53
Training Function Levenberg-Marquardt
(trainlm) Performance
Functions
Mean Square Errors (MSE)
Data Division Random (dividerand)
Network Types Feed-forward Back-
propagation Number of Neurons 2
Properties of Layer 1
Number of Neurons 10
Transfer Function TANSIG
Properties of layer 2
Transfer Function PURELIN
Training Parameters As desired
Epochs 1000
Time Infinite
Goal 0
Step 3; Neural Network configurations and training For this thesis two-layer feed-
forward network used. It is consist of number of neurons connected by links divided into
two layers. ANN training is the process by which synoptic weights in the network are
adjusted according to signal received. These signals are the input and output vectors
defined previously. Here BP network with tan-sigmoid (TANSIG) transfer function has
been used. Also ANN training purposes Levenberg-Marquardt (trainlm) is used. A
complete configuration and training of two-layer ANN in MATLAB environment is
shown in Fig. 6.27.
180
Figure 6.27: Two-layer BP-ANN for DTM1
Figure 6.26 shows that for training purposes of ANN, there are three fault gases as input
to the BP network, N = 3. Types of faults are chosen as network outputs, thus M = 7.
Therefore, number of nodes in the input and output layers are determined. The number of
hidden neural nodes is decided accordingly be 10. Then raw data out of 1236 samples are
used for training the patterns to train this neural network with the desired error goal set to
0.0. After successful training, the weight matrices and biases are stored as files. The final
number of hidden nodes is confirmed to be 11 and the 3-11-7 ‗network topology‘ is fixed.
Step 4; ANN result outputs A two-layer BP-ANN with 53 epochs is designed and
trained by different training patterns obtained from different power transformers‘ fault
diagnosis reports having capacities from 6.3-250 MVA, 11/0.4-420/21 KV collected from
various organizations/utilities in INDIA. From the MATLAB environment, it is seen that
from Fig. 6.28 that best validation performance is 0.49576 at epoch 47 out of 53 epochs.
181
Figure 6.28: Validation performance chart
Similarly, from training states, the Mu (the momentum) for the weights and biases is set
at 1, and also the gradient (local minima) is 0.0051068 can be observed from Fig. 6.29
(a). Though output goal is set to be 0 and other input /output targets for this study are
given in Fig. 6.29 (b).
Figure 6.29 (a) Mu and Gradient plot Figure 6.29 (b) Target plots
Taking all above details, for this thesis work a string of 1236 input samples between the
range of 0-100% in the variations of 0.1%, 0.2%, 0.25% and 0.5% as per requirements
were developed for three input vectors. This string of sample vectors is used for feed-
forward BP-ANN Duval triangle 1 diagnostic method.
6.11 ANN APPROACH FOR DTM 2
Considering Fig. 5.10 of Duval triangle 2 for OLTCs filled with transformer oil, the
procedure is as follows;
182
Step 1; Polygon coordinates for the numerical zone boundaries of six key faults of Duval
triangle 2 are listed in Table 5.7 and the same input percentages of numerical zones are
being used to create strings of three gas ratios (CH4, C2H4 and C2H2).
Step 2; Input and output patterns ANN technique used for DTM2 is similar to DTM1.
Which needs three combustible gases methane (CH4), ethylene (C2H4) and acetylene
(C2H2), therefore, for this thesis work, input and output patterns of DTM2 for fault
diagnosis in power transformers are as follows:
[Input] = [x] = [%CH4 %C2H4 %C2H2] T
[0utput] = [O] = [N D1 X1 X3 T2 T3] T
Output represents faults and the number assigned for each is as follows:
[O] = [1 2 3 4 5 6 ] T
A brief parameter details for this thesis on DGA for fault diagnosis in power transformers
through ANN based Duval triangle 2 approach are shown in Table 6.26.
Table 6.26: Parameters for ANN Duval triangle 2
Input [%CH4 %C2H6 %C2H2]T
Output Faults [1 2 3 4 5 6]T
Maximum Input
Value
100%, 100%, 100%
Minimum Input
Value
0.1%, 0.1%, 0.1%
Number of Samples 840
Epochs 175
Training Function Levenberg-Marquardt
(trainlm) Performance
Functions
Mean Square Errors (MSE)
Data Division Random (dividerand)
Network Types Feed-forward Back-
propagation Number of Neurons 2
Properties of Layer 1
Number of Neurons 10
Transfer Function TANSIG
Properties of layer 2
Transfer Function PURELIN
Training Parameters As desired
Epochs 1000
Time Infinite
Goal 0
Step 3; Neural Network configurations and training For this thesis two-layer feed-
forward network similar to DTM1 is used and complete configuration and training of
two-layer ANN in MATLAB environment is shown in Fig. 6.30.
183
Figure 6.30: Two-layer BP-ANN for DTM2
There are 840 samples as faults string are used for training the neural network in DTM2.
The network topology is fixed to be 3-11-6 in DTM2.
Step 4; ANN result outputs A two-layer BP-ANN with 175 epochs is designed and
trained with different training patterns. From Fig. 6.31, it is noticed that best validation
performance is 0.41116 at epoch 167 out of 175 epochs.
Figure 6.31: Validation performance chart
The gradient (local minima) is 0.029756, Mu 0.1 and validation checks 06 can be
observed from Fig. 6.32 (a) and output goal is set to be 0 and other input /output targets
are given in Fig. 6.32 (b).
184
Figure 6.32 (a): Mu and Gradient plot Figure 6.32 (b): Target plots
For DTM2 fault analysis purposes, a string of 840 fictitious input samples between the
range of 0-100% in the variations of 0.1%, 0.2%, 0.25% and 0.5% are used to develop
three input vectors.
6.12 ANN APPROACH FOR DTM3
Since there are four Duval triangles are available for four synthetic oils (silocone, midel,
FR3 and biotemp oils). For this thesis work each DTM3 is considered separately and
analysed for fictitious input samples because synthetic oils are not popular transformer
oils in India.
6.12.1 ANN Approach for DTM3, Silicone Oil
From Fig. 5.13 of Duval triangle 3 for non-mineral silicone oil, the ANN approach is as
follows;
Step 1; Polygon coordinates for the numerical zone boundaries of seven key faults of
Duval triangle 3, silicone are listed in Table 5.10 and the same input percentages of
numerical zones are being used to create strings of three gas ratios (CH4, C2H4 and C2H2).
Step 2; Input and output patterns ANN technique used for DTM3 is similar to DTM1
and DTM2. Which needs three combustible gases methane (CH4), ethylene (C2H4) and
acetylene (C2H2), therefore, for this thesis work, input and output patterns of DTM3 are
equal to DTM1 for fault diagnosis in power transformers are as follows:
[Input] = [x] = [%CH4 %C2H4 %C2H2] T
[0utput] = [O] = [PD D1 D2 DT T1 T2 T3] T
Output represents faults and the number assigned for each is as follows:
[O] = [1 2 3 4 5 6 7] T
185
A brief parameter details for this thesis work of DGA fault diagnosis in power
transformers through ANN based Duval triangle 3, silicone oil approach are shown in
Table 6.27.
Table 6.27: Parameters for ANN Duval triangle 3, silicone oil
Input [%CH4 %C2H6 %C2H2]T
Output Faults [1 2 3 4 5 6 7]T
Maximum Input
Value
100%, 100%, 100%
Minimum Input
Value
0.1%, 0.1%, 0.1%
Number of Samples 1640
Epochs 116
Training Function Levenberg-Marquardt
(trainlm) Performance
Functions
Mean Square Errors (MSE)
Data Division Random (dividerand)
Network Types Feed-forward Back-
propagation Number of Neurons 2
Properties of Layer 1
Number of Neurons 10
Transfer Function TANSIG
Properties of layer 2
Transfer Function PURELIN
Training Parameters As desired
Epochs 1000
Time Infinite
Goal 0
Step 3; Neural Network configurations and training For this thesis two-layer feed-
forward network similar to DTM1 is used and complete configuration and training of
two-layer ANN in MATLAB environment is shown in Fig. 6.33.
186
Figure 6.33: Two-layer BP-ANN for DTM3, silicone oil
There are 1640 samples as faults string are used for training the neural network in DTM3,
silicone oil. The network topology is fixed to be 3-11-7 in DTM3, silicone oil.
Step 4; ANN result outputs A two-layer BP-ANN with 116 epochs is designed and
trained with different training patterns. From Fig. 6.34, it is noticed that best validation
performance is 0.65835 at epoch 110 out of 116 epochs.
Figure 6.34: Validation performance chart
The gradient (local minima) is 0.52905, Mu 0.001 and validation checks 06 can be
observed from Fig. 6.35 (a) and output goal is set to be 0 and other input /output targets
are given in Fig. 6.35 (b).
187
Figure 6.35 (a): Mu and Gradient plot Figure 6.35 (b): Target plots
For DTM3, silicone oil fault analysis purposes, a string of 1640 fictitious input samples
between the range of 0-100% in the variations of 0.1%, 0.2%, 0.25% and 0.5% are used
to develop three input vectors.
6.12.2 ANN Approach for DTM3, Midel Oil
From Fig. 5.13 of Duval triangle 3 for non-mineral midel oil, the procedure for ANN
approach is as follows;
Step 1; Polygon coordinates for the numerical zone boundaries of seven key faults of
Duval triangle 3, midel oil are listed in Table 5.15 and the same input percentages of
numerical zones are being used to create strings of three gas ratios (CH4, C2H4 and C2H2).
Step 2; Input and output patterns ANN technique used for DTM3 is similar to DTM1
and DTM2. Which needs three combustible gases methane (CH4), ethylene (C2H4) and
acetylene (C2H2), therefore, for this thesis work, input and output patterns of DTM3 are
equal to DTM1 for fault diagnosis in power transformers are as follows:
[Input] = [x] = [%CH4 %C2H4 %C2H2] T
[0utput] = [O] = [PD D1 D2 DT T1 T2 T3] T
Output represents faults and the number assigned for each is as follows:
[O] = [1 2 3 4 5 6 7] T
A brief parameter details for this thesis work of DGA fault diagnosis in power
transformers through ANN based Duval triangle 3, midel oil approach is shown in Table
6.28.
Table 6.28: Parameters for ANN DTM 3, Midel oil
Input [%CH4 %C2H6 %C2H2]T
Output Faults [1 2 3 4 5 6 7]T
188
Maximum Input
Value
100%, 100%, 100%
Minimum Input
Value
0.1%, 0.1%, 0.1%
Number of Samples 1630
Epochs 78
Training Function Levenberg-Marquardt
(trainlm) Performance
Functions
Mean Square Errors (MSE)
Data Division Random (dividerand)
Network Types Feed-forward Back-
propagation Number of Neurons 2
Properties of Layer 1
Number of Neurons 10
Transfer Function TANSIG
Properties of layer 2
Transfer Function PURELIN
Training Parameters As desired
Epochs 1000
Time Infinite
Goal 0 Step 3; Neural Network configurations and training For this thesis two-layer feed-
forward network similar to DTM1 and DTM2 are used and complete configuration and
training of two-layer ANN in MATLAB environment is shown in Fig. 6.36.
Figure 6.36: Two-layer BP-ANN for DTM3, silicone oil
189
There are 1630 samples as faults string are used for training the neural network in DTM3,
silicone oil. The network topology is fixed to be 3-11-7 in DTM3, midel oil.
Step 4; ANN result outputs A two-layer BP-ANN with 78 epochs is designed and
trained with different training patterns. From Fig. 6.37, it is noticed that best validation
performance is 0.53649 at epoch 72 out of 78 epochs.
Figure 6.37: Validation performance chart
The gradient (local minima) is 0.0231165, Mu 0.1 and validation checks 06 can be
observed from Fig. 6.38 (a) and output goal is set to be 0 and other input /output targets
are given in Fig. 6.38 (b).
Figure 6.38 (a): Mu and Gradient plot Figure 6.38 (b): Target plots
For DTM3, silicone oil fault analysis purposes, a string of 1630 fictitious input samples
between the range of 0-100% in the variations of 0.1%, 0.2%, 0.25% and 0.5% are used
to develop three input vectors.
190
6.12.3 ANN Approach for DTM3, FR3 Oil
From Fig. 5.13 of Duval triangle 3 for non-mineral FR3 oil, the procedure for ANN
approach is as follows;
Step 1; Polygon coordinates for the numerical zone boundaries of seven key faults of
Duval triangle 3, FR3 oil are listed in Table 5.17 and the same input percentages of
numerical zones are being used to create strings of three gas ratios (CH4, C2H4 and C2H2).
Step 2; Input and output patterns ANN technique used for DTM3 is similar to DTM1
and DTM2. Which needs three combustible gases methane (CH4), ethylene (C2H4) and
acetylene (C2H2), therefore, for this thesis work, input and output patterns of DTM3 are
equal to DTM1 for fault diagnosis in power transformers are as follows:
[Input] = [x] = [%CH4 %C2H4 %C2H2] T
[0utput] = [O] = [PD D1 D2 DT T1 T2 T3] T
Output represents faults and the number assigned for each is as follows:
[O] = [1 2 3 4 5 6 7] T
A brief parameter details for this thesis work of DGA fault diagnosis in power
transformers through ANN based Duval triangle 3, FR3 oil approach is shown in Table
6.29.
Table 6.29: Parameters for ANN Duval triangle method 3, FR3 oil
Input [%CH4 %C2H6 %C2H2]T
Output Faults [1 2 3 4 5 6 7]T
Maximum Input
Value
100%, 100%, 100%
Minimum Input
Value
0.1%, 0.1%, 0.1%
Number of Samples 1625
Epochs 120
Training Function Levenberg-Marquardt
(trainlm) Performance
Functions
Mean Square Errors (MSE)
Data Division Random (dividerand)
Network Types Feed-forward Back-
propagation Number of Neurons 2
Properties of Layer 1
Number of Neurons 10
Transfer Function TANSIG
Properties of layer 2
Transfer Function PURELIN
Training Parameters As desired
Epochs 1000
Time Infinite
Goal 0
191
Step 3; Neural Network configurations and training For this thesis two-layer feed-
forward network similar to DTM1 and DTM2 are used and complete configuration and
training of two-layer ANN in MATLAB environment is shown in Fig. 6.39.
Figure 6.39: Two-layer BP-ANN for DTM3, FR3 oil
There are 1625 samples as faults string are used for training the neural network in DTM3,
silicone oil. The network topology is fixed to be 3-11-7 in DTM3, FR3 oil.
Step 4; ANN result outputs A two-layer BP-ANN with 120 epochs is designed and
trained with different training patterns. From Fig. 6.40, it is noticed that best validation
performance is 0.3374 at epoch 114 out of 120 epochs.
192
Figure 6.40: Validation performance chart
The gradient (local minima) is 0.191162, Mu 0.1 and validation checks 06 can be
observed from Fig. 6.41 (a) and output goal is set to be 0 and other input /output targets
are given in Fig. 6.41 (b).
Figure 6.41 (a): Mu and Gradient plot Figure 6.41 (b): Target plots
For DTM3, FR3 oil fault analysis purposes, a string of 1625 fictitious input samples
between the range of 0-100% in the variations of 0.1%, 0.2%, 0.25% and 0.5% are used
to develop three input vectors.
6.12.4 ANN Approach for DTM3, BioTemp Oil
From Fig. 5.13 of Duval triangle 3 for non-mineral BioTemp oil, the procedure for ANN
approach is as follows;
Step 1; Polygon coordinates for the numerical zone boundaries of seven key faults of
Duval triangle 3, FR3 oil are listed in Table 5.19 and the same input percentages of
numerical zones are being used to create strings of three gas ratios (CH4, C2H4 and C2H2).
193
Step 2; Input and output patterns ANN technique used for DTM3 is similar to DTM1
and DTM2. Which needs three combustible gases methane (CH4), ethylene (C2H4) and
acetylene (C2H2), therefore, for this thesis work, input and output patterns of DTM3 are
equal to DTM1 for fault diagnosis in power transformers are as follows:
[Input] = [x] = [%CH4 %C2H4 %C2H2] T
[0utput] = [O] = [PD D1 D2 DT T1 T2 T3] T
Output represents faults and the number assigned for each is as follows:
[O] = [1 2 3 4 5 6 7] T
A brief parameter details for this thesis work of DGA fault diagnosis in power
transformers through ANN based Duval triangle 3, biotemp oil approach is shown in
Table 6.30.
Table 6.30: Parameters for ANN Duval triangle method 3, BioTemp oil
Input [%CH4 %C2H6 %C2H2]T
Output Faults [1 2 3 4 5 6 7]T
Maximum Input
Value
100%, 100%, 100%
Minimum Input
Value
0.1%, 0.1%, 0.1%
Number of Samples 1625
Epochs 343
Training Function Levenberg-Marquardt
(trainlm) Performance
Functions
Mean Square Errors (MSE)
Data Division Random (dividerand)
Network Types Feed-forward Back-
propagation Number of Neurons 2
Properties of Layer 1
Number of Neurons 10
Transfer Function TANSIG
Properties of layer 2
Transfer Function PURELIN
Training Parameters As desired
Epochs 1000
Time Infinite
Goal 0 Step 3; Neural Network configurations and training For this thesis two-layer feed-
forward network similar to DTM1 and DTM2 are used and complete configuration and
training of two-layer ANN in MATLAB environment is shown in Fig. 6.42.
194
Figure 6.42: Two-layer BP-ANN for DTM3, BioTemp oil
There are 1625 samples as faults string are used for training the neural network in DTM3,
silicone oil. The network topology is fixed to be 3-11-7 in DTM3, biotemp oil.
Step 4; ANN result outputs A two-layer BP-ANN with 343 epochs is designed and
trained with different training patterns. From Fig. 6.43, it is noticed that best validation
performance is 0.33681 at epoch 337 out of 343 epochs.
195
Figure 6.43: Validation performance chart
The gradient (local minima) is 0.11522, Mu 0.1 and validation checks 06 can be observed
from Fig. 6.44 (a) and output goal is set to be 0 and other input /output targets are given
in Fig. 6.44 (b).
Figure 6.44 (a): Mu and Gradient plot Figure 6.44 (b): Target plots
For DTM3, biotemp oil fault analysis purposes, a string of 1630 fictitious input samples
between the range of 0-100% in the variations of 0.1%, 0.2%, 0.25% and 0.5% are used
to develop three input vectors.
6.13 ANN APPROACH FOR DTM4
DTM 4 is used for low-energy fault diagnosis in power transformers in addition to
DTM1. Three gases used in DTM 4 are H2, CH4 and C2H6. As shown in Fig. 5.18, Duval
triangle 4 has four types of faults for mineral oil in main tank, the procedure for ANN
approach is as follows;
196
Step 1; Polygon coordinates for the numerical zone boundaries of seven key faults of
Duval triangle 3, FR3 oil are listed in Table 5.26 and the same input percentages of
numerical zones are being used to create strings of three gas ratios (CH4, C2H4 and C2H2).
Step 2; Input and output patterns ANN technique used for DTM4 is similar to other
DTMs. Which needs three combustible gases hydrogen (H2), methane (CH4) and ethylene
(C2H6), therefore, for this thesis work, input and output patterns of DTM4 are equal to
DTM1 for fault diagnosis in power transformers are as follows:
[Input] = [x] = [%H2 %CH4 %C2H6] T
[0utput] = [O] = [PD S C O] T
Output represents faults and the number assigned for each is as follows:
[O] = [1 2 3 4] T
A brief parameter details for this thesis work of DGA fault diagnosis in power
transformers through ANN based Duval triangle 4 approach is shown in Table 6.31.
Table 6.31: Parameters for ANN Duval triangle method 4
Input [%CH2 %CH4 %C2H6]T
Output Faults [1 2 3 4]T
Maximum Input
Value
100%, 100%, 100%
Minimum Input
Value
0.1%, 0.1%, 0.1%
Number of Samples 1236
Epochs 61
Training Function Levenberg-Marquardt
(trainlm) Performance
Functions
Mean Square Errors (MSE)
Data Division Random (dividerand)
Network Types Feed-forward Back-
propagation Number of Neurons 2
Properties of Layer 1
Number of Neurons 10
Transfer Function TANSIG
Properties of layer 2
Transfer Function PURELIN
Training Parameters As desired
Epochs 1000
Time Infinite
Goal 0 Step 3; Neural Network configurations and training For this thesis two-layer feed-
forward network similar to DTM1 and DTM2 are used and complete configuration and
training of two-layer ANN in MATLAB environment is shown in Fig. 6.45.
197
Figure 6.45: Two-layer BP-ANN for DTM4
There are 1636 samples as faults string are used for training the neural network in DTM4.
The network topology is fixed to be 3-11-4 in DTM4.
Step 4; ANN result outputs A two-layer BP-ANN with 61 epochs is designed and
trained with different training patterns. From Fig. 6.46, it is noticed that best validation
performance is 0.10028 at epoch 55 out of 61 epochs.
Figure 6.46: Validation performance chart
198
The gradient (local minima) is 0.0031822, Mu 0.1 and validation checks 06 can be
observed from Fig. 6.46 (a) and output goal is set to be 0 and other input /output targets
are given in Fig. 6.46 (b).
Figure 6.46 (a): Mu and Gradient plot Figure 6.46 (b): Target plots
6.14 ANN APPROACH FOR DTM5
DTM 5 is used for low-temperature fault diagnosis in power transformers in addition to
DTM1. Three gases used in DTM 5 are CH4 C2H4 and C2H6. As shown in Fig. 5.20,
Duval triangle has five types of faults for mineral oil in main tank, the procedure for
ANN approach is as follows;
Step 1; Polygon coordinates for the numerical zone boundaries of seven key faults of
Duval triangle 5 are listed in Table 5.27 and the same input percentages of numerical
zones are being used to create strings of three gas ratios.
Step 2; Input and output patterns ANN technique used for DTM5 is similar to other
DTMs. Which needs three combustible gases methane (CH4), ethane (C2H4) and ethylene
(C2H6), therefore, for this thesis work, input and output patterns of DTM4 are equal to
DTM1 for fault diagnosis in power transformers are as follows:
[Input] = [x] = [%CH4 %C2H4 %C2H6] T
[0utput] = [O] = [PD S C O T3] T
Output represents faults and the number assigned for each is as follows:
[O] = [1 2 3 4] T
A brief parameter details for this thesis work of DGA fault diagnosis in power
transformers through ANN based Duval triangle 5 approach is shown in Table 6.32.
Table 6.32: Parameters for ANN Duval triangle 5
Input [%CH2 %CH4 %C2H6]T
Output Faults [1 2 3 4]T
Maximum Input
Value
100%, 100%, 100%
199
Minimum Input
Value
0.1%, 0.1%, 0.1%
Number of Samples 1184
Epochs 42
Training Function Levenberg-Marquardt
(trainlm) Performance
Functions
Mean Square Errors (MSE)
Data Division Random (dividerand)
Network Types Feed-forward Back-
propagation Number of Neurons 2
Properties of Layer 1
Number of Neurons 10
Transfer Function TANSIG
Properties of layer 2
Transfer Function PURELIN
Training Parameters As desired
Epochs 1000
Time Infinite
Goal 0
Step 3; Neural Network configurations and training For this thesis two-layer feed-
forward network similar to DTM1, DTM2, DTM3s and DTM4 are used. Complete
configuration and training of two-layer ANN in MATLAB environment is shown in Fig.
6.48.
Figure 6.48: Two-layer BP-ANN for DTM5
200
There are 1184 samples as faults string are used for training the neural network in DTM4.
The network topology is fixed to be 3-11-4 in DTM4.
Step 4; ANN result outputs A two-layer BP-ANN with 42 epochs is designed and
trained with different training patterns. From Fig. 6.49, it is noticed that best validation
performance is 61.0869 at epoch 36 out of 42 epochs.
Figure 6.49: Validation performance chart
The gradient (local minima) is 2.548, Mu 10 and validation checks 06 can be observed
from Fig. 6.50 (a) and output goal is set to be 0 and other input /output targets are given
in Fig. 6.50 (b).
Figure 6.50 (a): Mu and Gradient plot Figure 6.50 (b): Target plots
Selected samples of DGA fault reports (APPENDIX-B) by other conventional diagnostic
methods are used for neural network training in different groups. Their vector input
ranges were rounded-up to the first digit of decimal (0.50) to match the variations in
increasing/decreasing trend. Many unknown new input vectors were also used for neural
testing. Neural network does not know these vectors. The result table presented for
201
known faults which is cross verified with other software schemes with this new method
shows that neural network efficiency for this set of input vector is in the range of 90-
100%, which is quite acceptable.
Results using FL and ANN approaches and fault diagnosis reports available from various
organizations gives the following conclusions:
Percentage of gases based rules in FL gives flexibility in Duval triangle methods
to get accurate analysis. Results obtained from fuzzy logic approach on DTM1 are
compared with simulation results shown in Table 6.33. It is found that FLA gives better
results as compared to the conventional method results e.g. for 250 and 315 MVA
transformers conventional methods provides the range of temperatures while FLA
provides particular fault.
Table 6.33: Comparative results of DTM1 simulation and Fuzzy logic DTM1
Sr.
No
Transformer
Ratings
GASES
(ppm)
GASES (%)
DGA Fault Diagnosis/ Comments
DTM 1
RESULTS
MVA KV CH4 C2H4 C2H2 CH4 C2H4 C2H2 CEBG
IEC,
CIGRE
IEEE/R
eco.
Comnt
s Fuzzy logic
1 6.3 33/11 2 28 0 6.67 93.33 0.00
OH OH T32
2 16 132/33 3 64 42 2.75 58.72 38.53
OH OH D22
3 16/20 132/33 17 17 0 50.00 50.00 0.00
T OH T33
4 200 420/21 1 1 0 50.00 50.00 0.00 PD NF N IG T22
5 250 16/420 40 10 0 80.00 20.00 0.00
200-
300°C NF
200-
300°C
Ind.
Gases T12
6 315 400/220 79 89 0 47.02 52.98 0.00
300-700
°C <700°C
T21
Superior learning capabilities of ANN gives better and quicker analysis as
compared to other fault diagnosis methods. Results obtained from ANN approach on
DTM1 compared with simulation results are shown in Table 6.34.
Table 6.34: Comparison of results simulation and ANN
S.No.
Transformer Rating GASES
(ppm)
GASES (%)
DGA Fault Diagnosis/ Comments
DUVAL
RESULTS
MVA KV CH4 C2H4 C2H2 CH4 C2H4 C2H2 CEBG
IEC,
CIGRE IEEE/Reco. Comnts ANN
1 10/12.5 66-33/114 6852 3137 ND 68.50 31.50 ND
C2H2 High Old OIL OH T2
2 100 220/66 187 614 ND 23.50 76.50 ND
C2H4 High Hot spot
T3
3 250 16/420 40 10 0 80.00 20.00 0.00
200-
300°C NF
200-300°C
Ind.
Gases D1
202
4 200 420/21 1 1 0 50.00 50.00 0.00 PD NF N IG T2
5 315 400/200 55 72 0 43.50 56.50 0.00
300-700 °C >700°C OH T3
Further, it is also found that ANNA gives better results as compared to the conventional
method results e.g. for 100 and 315 MVA transformers conventional methods provides
the range of temperatures while ANNA provides particular fault in less time.
Simulated results using fuzzy logic and artificial neural network approach on Duval
triangle method1 are compared with fault diagnosis reports as shown in Table 6.35.
Table 6.35: Comparison of FL and ANN results
Sr.
No.
Transf0rmers
Ratin
gs
GASES
(ppm)
GASES (%)
DGA Fault Diagnosis/ Comments
DTM 1
RESULTS
MVA KV CH4 C2H4 C2H2 CH4 C2H4 C2H2 CEBG
IEC,
CIGRE
IEEE/
Reco. Comnts. ANN
Fuzzy
logic
1. 6.3 33/11 2 28 0 6.67 93.33 0.00
OH OH T3 T32
2. 10/12.5
66-
33/114
685
2 3137 ND 68.50 31.50 ND
C2H2
High
Old
OIL OH T2 T2
3. 16/20 132/33 17 17 0 50.00 50.00 0.00
T OH T3 T33
4. 100 220/66 187 614 ND 23.50 76.50 ND
C2H4
High
Hot
spot
T3 T3
5. 200 420/21 1 1 0 50.00 50.00 0.00 PD NF N IG T2 T22
6 315
400/22
0 79 89 0 47.02 52.98 0.00
300-700
°C <700°C
T2 T21
It is observed that ANNA provides quick and accurate results once trained properly. This
capability of ANNA will be very useful for complete fault diagnosis of power
transformers.
6.15 SUMMARY
Simulated results with fuzzy logic and artificial neural network approaches on Duval
triangle 1 are compared with fault diagnosis report results. It is found that fuzzy logic and
artificial neural network approaches will give better results than actual fault diagnosis
report results. Further it is it is observed that ANN provides fast and accurate results as
compared to fuzzy logic result.
203
CHAPTER 7
CONCLUSIONS AND FUTURE SCOPE OF WORK
7.1 CONCLUSIONS
This chapter includes the conclusions and the suggestions for future work. Major
conclusions are as follows;
Critical specifications of transformer oil can be performed for fault
diagnosis/condition monitoring (periodic test) as pre-warnings of the faults (if any) after
thermal degradation of transformer oil test analysis.
Type of fault and reasons for the faults are correlated to confirm the faults from
the test reports of critical specifications of oil.
Most of the solutions are regeneration of oil such as filtering, removal of water
contents and sludge removal.
After comparison of different fault diagnosis methods of power transformers, it is
found that Duval Triangle Methods (DTMs) gives better results.
DTMs have satisfied the fault diagnosis for more than 95% accuracy compared to
the other fault diagnosis methods.
Traces of one of the three gases can be helpful for early fault indication in the
power transformers.
Application of FL and ANN in DTMs can provide higher accuracy, low
uncertainty and quicker analysis in fault diagnosis of power transformers.
It has been found that ANN based DTMs can provide full solutions in fault
diagnosis of power transformers.
7.2 FUTURE SCOPE OF WORK
This thesis gives following suggestions for future scope of work.
Optimization techniques may be attempted for improvement of accuracy in fault
diagnosis of power transformers.
Intelligent techniques may be applied to reduce the uncertainties among faults.
204
REFERENCES
[1] Xiang Zhang, Gockenbach E., ―Asset-management of transformers based on condition
monitoring and standard diagnosis,‖ IEEE Electrical Insulation Magazine, vol.24,
no. 4, pp. 26-40, 2008.
[2] CIGRE Working Group 12.05, ―An international survey on failures in large power
transformers in service,‖ Electra, no. 88, pp. 21-47, 1983.
[3] J. P. Gibeault , J. K. Kirkup, ―Early detection and continuous monitoring of dissolved
key fault gases in transformers and shunt reactors,‖ in Proc. Elect. Electronics Insul.
Conf., Elect. Manufact. Coil Winding Conf., pp. 285–293, Sep. 1995.
[4] B. Pahlavanpour, A. Wilson, ―Analysis of transformer oil for transformer condition
monitoring,‖ in IEE Colloq. Engineer. Rev. Liquid Insul., pp. 1/1–1/5, Jan. 1997.
[5] B. Sparling, J. Aubin, ―Field experience with multigas on-line monitoring of power
transformers,”in Proc. IEEE Transmission Distrib. Conf., vol. 2, pp. 895–900, April
1999.
[6] B. Sparling, ―Transformer monitoring and diagnostics,‖ Proceeding of IEEE Power
Engineer. Society Winter Power Meeting, vol.2, New York, pp. 978–980, 1999.
[7] J. Lapworth, ―A novel approach (scoring system) for integrating dissolved gas
analysis results into a life management system,‖ in Conf. Record of IEEE Int. Symp.
Elect. Insul., pp. 137–144, April 2002.
[8] J. Sabau, L. Silberg, Paul Vaillancourt, ―The impact of oil decay on gassing and
reliability of aging power transformers,‖ in Annual Reprt Conf. Elect. Insul. Dielect.
Phenomena, pp. 408–411, Oct. 2002.
[9] A. Shahsiah, R. C. Degeneff, ―A new dynamic model for propagation of characteristic
gases in transformer oil-cellulose structures due to temperature variations,‖ in Annual
Report Conf. Elect. Insul. Dielect. Phenomena, pp. 269–272, Oct. 2005.
[10] J. Sabau, R. Stokhuyzen, ―The side effects of gassing in transmission power
transformers,‖ in Annual Report Conf. Elect. Insul .Dielect. Phenomena, vol. 1, pp.
264–267, Oct. 2000.
[11] R. Sans, John., K. Muge Bilgin, J. Joseph, ―Large -scale survey of furanic
compounds in operating transformers and life‖, Proc. of 1998 IEEE international
symposium on electrical insulation, Arlington, USA, 543-553, June 7-10, 1998.
205
[12] R. Blue, D. Uttamchandani, O. Farish, ―A novel solid state material for
furfuraldehyde detection,‖ IEEE transactions on Dielectrics and electrical
insulation, vol. 4, no.3, 341-343, 1997.
[13] R. B. Harrison, ―Detection, determination and identification of furfuraldehyde in
hydrocarbon oil,‖ Analyst, 88, 644-646, 1963.
[14] Y. Han, Y. H. Song ―Condition monitoring techniques for electrical equipments-A
literature survey,‖ IEEE Transactions on Power delivery, Vol.17, pp.4-13, 2003.
[15] M. Kohtoh, S. Kaneko, S. Okabe, T. Amimoto, ―Ageing effect on electrical
characteristics of insulating oil in field transformer,‖ IEEE Transactions on
Dielectrics and Electrical Insulation, vol.166, pp.1698-1706, 2009.
[16] M. Kohtoh, Genyo, S. Okabe, ―Transformer Insulating oil characteristic changes
observed using accelerated degradation in consideration of field transformer
conditions,‖ IEEE Transactions on Dielectrics and Electrical Insulation, Vol.18,
pp.808-818, 2010.
[17] Syed Islam, Abus Siada, Lai Pin Sin, ―Remnant Life estimation of Power
transformer using oil UV-Vis Spectral response,‖ IEEE International conference on
Power systems, pp.15-18, March 2009.
[86] IS-6792 - Methods for Determination of Electric strength of insulating oil.
[19] ASTM D-3828- Standard Test Methods for determining Flash Point, Fire point for
oil.
[20] BS- 2000 Part 1-Methods of Determination of acidity, neutralization value.
[21] ASTM D-2162-06, IP 70 Standard Test Methods and protection rating for
determining Viscosity of oil.
[22] IEC 60247- Insulating Liquids-Measurement.
[23] Liao Ruijin, Sun Caixin, Chen Weigen, Wang Caisheng, ―On-line detection of gases
dissolved in transformer oil and the faults diagnosis‖, in Proceedings of the
International Symposium on Electrical Insulating Materials, Chongqing Univ.,
Chongqing, China, pp. 771-775, 1998.
[24] T. McGrail and A. Wilson, ―On-line gas sensors,‖ in IEE Colloq. Condition
Monitoring Large Machines Power Transform., pp. 1/1–1/4, June 1997.
[25] J. Samat, ―Interpretation of DGA data and its use in the detection of malfunctions
[transformers],‖ in IEE Colloq. Assessment Degradation Transformer Insul. Syst.,
pp. 3/1–3/4, Dec 1991.
206
[26] S. J. Ferrito, ―A comparative study of dissolved gas analysis techniques: the vacuum
extraction method versus the direct injection method,‖ IEEE Trans. Power Delivery,
vol. 5, no. 1, pp. 20–25, 2002.
[27] Y. Inoue, K. Suganuma, ―Development of oil-dissolved hydrogen gas detector for
diagnosis of transformers,‖ IEEE Trans. Power Delivery, vol. 5, no. 1, pp. 226–232,
1990.
[28] M. Duval, F. Langdeau, P. Gervais, G. Belanger, ―Influence of paper insulation on
acceptable gas-in-oil levels in transformers,‖ in Annual Report Conference on
Electrical Insulation and Dielectric Phenomena, pp. 358–362, 1989.
[29] M. Duval, ―Dissolved gas analysis: It can save your transformer,‖ IEEE Electrical
Insulation Magazine, vol. 5, no. 6, pp. 22–27, 1989.
[30] M. Duval, F. Langdeau, P. Gervais, G. Belanger, ―Acceptable gas-in-oil levels in
generation and transmission power transformers,‖ Annual Report Conference on
Electrical insulation and Dielectric Phenomena, pp. 325–330, 1990.
[31] C. Lamarre, M. Duval, J.-P. Crine, ―On the influence of metal deactivators on some
properties of transformer oils,‖ in Conference Record 10th
International Conference
on Conduction and Breakdown in Dielectric Liquids, pp. 510–513, 1990.
[32] M. Duval, H. Dupas, F. Langdeau, P. Gervais, G. Belanger, ―Acceptable gas-in-oil
levels in instrument transformers (current Vs voltage equipment),‖ Annual Report
Conference on Electrical Insulation and Dielectric Phenomena, pp. 506–511, 1991.
[33] M. Duval, A. dePablo, ―Interpretation of gas-in-oil analysis using new IEC
publication 60599 and IEC TC 10 databases,‖ IEEE Electr. Insul. Mag., vol. 17, no.
2, pp. 31–41, 2001.
[34] M. Duval, ―A review of faults detectable by gas-in-oil analysis in transformers,‖
IEEE Electrical Insulation Magazine, vol. 18, no. 3, pp. 8–17, 2002.
[35] M. Duval, ―New techniques for dissolved gas-in-oil analysis,‖ IEEE Elect. Insul.
Mag., vol. 19, no. 2, pp. 6–15, 2003.
[36] M. Duval, J. Dukarm, ―Improving the reliability of transformer gas-in-oil diagnosis,‖
IEEE Electrical Insulation Magazine, vol. 21, no. 4, pp. 21–27, 2005.
[37] Michel Duval, ―Dissolved gas analysis and the Duval triangle,‖.
(www.avo.co.nz/Technical papers/2006 –Conference_Duval.pdf.)
[38] M. Duval, ―Calculation of DGA limit values and sampling intervals in transformers
in service,‖ IEEE Electrical Insulation Magazine, vol. 24, no. 5, pp.7–13, 2008.
207
[39] M. Duval, ―The Duval triangle for load tap changers, non-mineral oils and low
temperature faults in transformers,‖ IEEE Electrical Insulation Magazine, vol. 24,
no. 6, pp. 22–29, 2008.
[40] CIGRE Working Group 12.18-2002, ―Life Management of Transformers-Guidelines
for Life Management Techniques of Power Transformers,‖ 2002.
[41] IEEE Std. C57-104TM
-2008, “IEEE Guide for the Interpretation of Gases Generated
in Oil- Immersed Transformers,” September 2008.
[42] IEC Publication 60599, 1978, “Interpretation of the analysis of gases in
transformers and other oil-filled electrical equipment in service”.
[43] IEC Publication 567-1992, “Guide for the Sampling of Gases and Oil from Oil-
Filled Electrical Equipment and for the Analysis of Free and Dissolved Gases‖.
[44] IEC Publication 60599, “Mineral Oil Impregnated Equipments in Service- Guide to
the Interpretation, of Dissolved and Free Gases Analysis,” March 1999.
[45] IEC Publication 60599, “Mineral Oil Impregnated Equipments in Service- Guide to
the Interpretation, of Dissolved and Free Gases Analysis,” 2003.
[46] IS 10593: 1992(Reaffirmed 2006) & IS-9434: (1992), ―Indian Standard Method of
Evaluating the Analysis of Gases in Oil-Filled Electrical Equipment in Service‟‟
(First Revision), 1995.
[47] ASTM D 3612- Determination of Individual Dissolved Gases, ―Laboratory methods
for analyzing the gas blanket and the gases extracted from oil”.
[48] FIST (Facilities Instructions Standards and Techniques Vol.-3-31), ―Transformer
Diagnostic‖, US Department of the Interior Bureau of Reclamation, June 2003.
[49] ANSI- American National Standards Institute. www.ansi.org/
[50] R. R. Roger, ―IEEE and IEC codes to interpret incipient faults in transformers, using
gas in oil analysis,‖ IEEE Trans. Electr. Insul., vol. 13, no. 5, pp. 349–354, 1978.
[51] K. Tomsovic, M. Tapper, T. Ianvarsson, ―A fuzzy information approach to
integrating different transformer diagnostic method,‖ IEEE Trans. Power Delivery,
vol. 8, no.3, pp. 1638-1646, 1993.
[52] H.Y. Chang, Y.H. Tzer, ―An evolutionary computation based fuzzy fault diagnosis
system for a power transformer,‖ in Proceeding of the Asian Fuzzy System
Symposium. „Soft Computing Intelligent System Information Process, pp. 218–223,
1996.
208
[53] H. Y. Chang, Y. H. Tzer, ―Design of robust transformer fault diagnosis system using
evolutionary fuzzy logic,‖ in Proceedings of the IEEE International Symposium on
Circuits and Systems „Connecting the World‟, vol. 1, pp. 613–616, 1996.
[54] H. Y. Chang, Y. H. Tzer, ―Developing a new transformer fault diagnosis system
through evolutionary fuzzy logic,‖ IEEE Trans. Power Delivery, vol. 12, no. 2, pp.
761–767, 1997.
[55] Y. H. Tzer, L. C. Chou, ―Adaptive fuzzy diagnosis system for dissolved gas analysis
of power transformers,‖ IEEE Trans. Power Delivery, vol. 14, no. 4, pp. 1342–1350,
1999.
[56] G. Zhang, K. Yasuoka, ―Application of fuzzy equivalent matrix for fault diagnosis
of oil-immersed insulation,‖ in Proceedings of the IEEE 13th
International
Conference on Dielectric Liquids, pp. 400–403, 1999.
[57] Z. Guanjun, L. Y. Shing, ―Application of fuzzy data processing for fault diagnosis of
power transformers,‖ in Proceedings of the Eleventh International Symposium of
High Voltage Engineering, vol. 5, pp. 160–163, 1999.
[58] M. Denghua, ―A new fuzzy information optimization processing technique for
monitoring the transformer,‖ in Proceedings of the Eighth International. Conference
of Dielectric Materials, Measurements and. Applications, pp. 192–195, 2000.
[59] Q. Su, ―A fuzzy logic tool for transformer fault diagnosis,‖ IEEE Trans. Power
Delivery, vol. 14, no. 2, pp 265-268, 2000.
[60] Q. Su, C. Mi, ―A fuzzy dissolved gas analysis method for the diagnosis of multiple
incipient faults in a transformer,‖ IEEE Trans. Power Systems, vol. 15, no. 2, pp.
593–598, 2000.
[61] R.Hooshmand, M Banejad, ―Application of fuzzy logic in fault diagnosis in
transformers using dissolved gas based on different standards,‖ in Proceedings of the
World Academy of Science, Engineering and Technology, Conference 2006, pp. 157-
161, 2006.
[62] W. Flores, E Mombello, ― A Novel algorithm for the diagnostics of power
transformers using Type-2 fuzzy logic systems,‖ presented at the IEEE Conference
Supported by German Academy Exchange Service (DAAI), German, 2008.
[63] R. N. Rafiquah, I. Musirin, D. Johri, M. M. Othman, T. K. A. Rahman, Zoyhman, ―
Fuzzy logic application in DGA method to classify faults type in power transformer,‖
in Proceedings of the Selected Topic in Power Systems and Remote Sensing
(CEPES), university technology MARA, Shah Alam, Malaysia, pp 83-88, 2009.
209
[64] J. Aragon-Patil, M. Fischer, S. Tenbohlen, ―Improvement of interpretation of
dissolved gas analysis for power transformer,‖ presented at the Institute of Power
Transmission and High Voltage Technology, Pfaffenwaldring -47, 70569 Stuttgart,
Germany, 2002.
[65] C. X. Sun, J. F. Guo, R. J. Liao, L. Du., W. G. Chen, ― Study on fault diagnostic
method of DGA for transformer oil with fuzzy model hierarchy classification,‖ in
Proceedings of the Chinese Society of Electrical Engineering, College of Electrical
Engineering, Chongqing Univ., Chongqing, China, vol. 21, no. 2, pp. 37-41, 2001.
[66] Y. Zhang, X. Ding, ―An artificial neural network approach to transformer fault
diagnosis,‖ IEEE Trans. Power Delivery, vol. 11, no. 3, pp. 1836–1841, 1996.
[67] Y. M. Tu, J. M. Huang, J. M. Huang, N. Gao, Z. S. Yan, ― Transformer insulation
diagnosis based on ANN analysis,‖ in Proceedings of the 5th
International
Conference on Properties and Applications of Dielectric Materials, Seoul, Korea, pp
263-266, 1997.
[68] T. Yanming, Q. Zheng, ―DGA based insulation diagnosis of power transformer via
ANN,‖ in Proceedings of the 6th International Conference of Properties,
Applications of Dielectric Materials, vol. 1, pp. 133–136, 2000.
[69] Q. Zheng, Y. Minzhong, Y. Zhang, ―Synthetic diagnostic method for insulation fault
of oil-immersed power transformer,‖ in Proceedings of the 6th International
Conference on Properties and Applications of Dielectric Materials, vol. 5, pp. 872–
875, 2000.
[70] K. F. Thang. R. K. Aggarwal, D. G. Esp, A. J. McGrail, ―Statistical and neural
network analysis of dissolved gases in power transformers,‖ in Proceedings of the
Dielectric Materials, Measurements and Applications Conference, vol. 473, pp. 324-
329, 2000.
[71] J. L. Guardado. J. L. Naredo, ―A comparative study of neural network efficiency in
power transformers diagnosis using dissolved gas analysis,‖ IEEE Trans. Power
Delivery, vol. 16, no. 4, pp. 643–647, 2001.
[72] N. K. Patel, R. K. Khubchandani, ―ANN based transformer fault diagnosis,”
Institution of Engineers (I) Electrical Engineering Journal, volume 85, pp. 60-65,
August 2004.
[73] S. M. Dhlamini, T. Marwala, ―Bushing diagnostics uses an ensemble of parallel
neural networks,‖ in Proceedings of the International Symposium on Electrical
Insulating Materials, vol. 1, pp. 289–292, 2005.
210
[74] S. Yan-jing, Z. Shen, M. Chang-xin, L. Jing-meng, ―Improved BP neural network for
transformer fault diagnosis,‖ Journal of China University of Mining & Technology,
vol.17, no.1, pp. 138-142, 2007.
[75] M. P. Mreira, Leonardo T. B. Santos, Marley M.B.R. Vettasco, ―Power transformers
diagnosis using neural networks,‖ in Proceedings of the IEEE International
Conference on Neural Networks, Orlando, USA, Aug. 12-17, 2007.
[76] X. Hao, S. Cai-Xin, C. Wei-Gen, D. Lin, L. Yu-Xiang, ―Artificial immune network
classification algorithm for fault diagnosis of power transformers,‖ in Proceedings
of the Conference on Automation of Electric Power Systems, vol. 30, no. 6, pp. 57-
60, 2006.
[77] L. W. Song, C.P. Hung, ―CMAC-based fault diagnosis of power transformers,‖ in
Proceedings of the International Joint Conference on Neural Networks, vol. 1, pp.
986–991, 2002.
[78] B. Song, Y. Ping, ―Study on the fault diagnosis of transformer based on the grey
relational analysis,‖ in Proceedings of the International Conference Power System
Technology, vol. 4, pp. 2231–2234, 2002.
[79] L. Chia-Hung, W. Chien-Hsien, H. Ping-Zan, ―Grey clustering analysis for incipient
fault diagnosis in oil-immersed transformers,‖ Journal of Expert Systems with
Applications, vol. 36, 1371–1379, 2009.
[80] W. Zang, L. Yang, Y. LI, F. Liu, C. Ma, ―The condition estimate of power
transformer based on rough set theory‖, in Proceedings of the 20th
International
Conference on Electricity Distribution (CIRED), Prague, pp. 8-11, 2009.
[81] H. Yann-Chang, S. Huo-Ching, H. Kun-Yuan, L. Yi-Shi, ―Fault diagnosis of power
transformer using rough set theory,” presented at the Department of Electrical
Engineering, Heng Shiu Univ. Kaohsiung, Taiwan, 2009.
[82] Z. J. Richardson, J.Fitch, W.H.Tang, J.Y.Goulermas, Q. H. Wu, ―A probabilistic
classifier for transformer dissolved gas analysis with a particle swarm optimizer,‖
IEEE Trans. Power Delivery, vol. 23, no. 2, pp.751-759, 2008.
[83] W. Chen, C. Pan, Y. Yun, Y. Liu, ―Wavelet networks in power transformers
diagnosis using dissolved gas analysis,‖ IEEE Trans. Power Delivery, vol. 24, no. 1,
pp. 187-194, 2009.
[84] Yaan-Chang Huang, Chao-Ming Huang, ―Evolving wavelet networks for power
transformer condition monitoring,‖ IEEE Trans. IEEE Trans. Power Delivery, vol.
17, no. 2, pp. 412-416, 2002.
211
[85] Y. H. Tzer, H. Y. Chang, ―Intelligent decision support for diagnosis of incipient
transformer faults using self-organizing polynomial networks,‖ in Proceedings of the
20th
International Conference on Power Industry and Computer Applications, pp.
60–66, 1997.
[86] K. F. Thang. R. K. Aggarwal, A. J. McGrail, D. G. Esp, ―Application of self-
organizing map algorithm for analysis and interpretation of dissolved gases in power
transformers,‖ in Proceedings of the IEEE Power Engineering Society, Summer
Meeting, vol. 3, pp. 1881–1886, 2001.
[87] D. Ming, M. Yuan-Yuan, X. Chang-Xiang, Y. Zhang, ―Fault diagnosis model for
power transformer based on support vector machine and dissolved gas analysis,‖ in
Proceedings of the Chinese Society of Electrical. Engineering, Xi'an Jiaotong Univ.,
China, vol. 23, no. 7, pp. 88-92, 2003.
[88] X. Cheol Jeong, K. Jong-Wook, ― A Pattern-Based fault classification algorithm for
distribution transformers,‖ IEEE Trans. Power Delivery, vol. 20, no. 4, pp. 2483-
2490, 2005.
[89] Y. Huang, ―A new data mining approach to dissolved gas analysis of oil-insulated
power apparatus,‖ IEEE Trans. Power Delivery, vol. 18, no. 4, pp. 1257–1261, 2003.
[90] Z. Limin, L. Zheng, ―Power transformer fault diagnosis based on extension theory,‖
in Proceedings of the Eighth International Conference on Electrical Machines and
Systems, vol. 3, pp. 1763–1766, 2005.
[91] W. Yongqiang, L. Fangcheng, ―The fault diagnosis method for electrical equipment
based on Bayesian network,‖ in Proceedings of the Eighth International Conference
on Electrical Machines and Systems, vol. 3, pp. 2259–2261, Sep. 2005.
[92] X. Hao, S. Cai-Xin, L. Rui-Jin, D. Lin, “ Study of Kernel-based possibilistic and
dissolved gas analysis for fault diagnosis of power transformer,‖ in Proceedings of
the Chinese Society of Electrical Engineering , vol. 25, no. 20, pp. 162-166, 2005.
[93] A. Singh, P. Verma, ―A Review of intelligent diagnostic methods for condition
assessment of insulation system in power transformers,‖ in Proceedings of the IEEE
International Conference on Condition Monitoring and Diagnosis, Beijing, China,
April 21-24, 2008.
[94] D. Bhalla, R. K. Bansal, H. O. Gupta, ―Application of artificial intelligence
techniques for dissolved gas analysis of transformers –A Review,‖ World Academy
of Science, Engineering and Technology, vol. 62, pp. 221-229, 2010.
212
[95] D.V.S.S., Siva Sarma, G.N.S. Kalyani, ―Application of AI techniques for non-
destructive evaluation of power transformer using DGA,‖ WASET, vol.4, no. 4,
pp.45-52, 2008.
[96] S. Souahlia, K. Bacha, A. Chaari, ―Artificial intelligence tools aided decision for
power transformer faults diagnosis,‖ International Journal of Computer applications,
vol.38, no.3, pp 1-7, 2012
[97] S. Qacdi, S.Seyedtabaii, ―Improvement in power transformer intelligent dissolved
gas analysis method,‖ WASET, vol.6, pp-1144-1147, 2012.
[98] G. Duke, Predictive maintenance a case study in infrared thermography,‖ Electrical
Maintenance, 11-12, 1998.
[99] J. Singh, Y. R. Sood, R. K. Jarial, ―Condition monitoring of power transformers -
Bibliography Survey,‖ IEEE Electrical Insulation Magazine, vol. 24, no. 3, pp. 11-
24, 2008.
[100] J. Singh, Y. R. Sood, ―Novel method for detection of transformer winding faults
using Sweep Frequency Response Analysis,‖ IEEE Power Engineer. Society General
Meeting, Tampa, FL USA, pp. 1–9, June, 2007.
[101] Bale Garcia, Juan Carlos Burgos, Angel Matias Alosno, ―Transformer tank
vibration modeling as a method of detecting winding deformations-Part I:
experimental verification,‖ IEEE Trans. Power Delivery, vol.21, no. 1, pp. 157-164,
2006.
[102] Bale Garcia, Juan Carlos Burgos, Angel Matias Alosno, ―Transformer tank
vibration modeling as a method of detecting winding deformations-Part II:
experimental verification,‖ IEEE Trans. Power Delivery, vol.21, no. 1, pp. 164-169,
2006.
[103] Imad-U-Khan, Zhongdong Wang, Cotton I., Northcote S, ―Dissolved gas
analysis of alternative fluids for power transformers,‖ IEEE Elect. Insul. Mag., vol.
23, no. 5, pp. 5-14, 2007.
[104] W. J. Chatterton and J. L. Goudie, ―An update on silicone transformer fluid:
specification, manufacturing, maintenance and end of life options,‖ in Conf. Record
2000 IEEE Int. Symp. Elect. Insul., pp. 412–416, April 2000.
[105] Kisao Kan, ― Proposal for an improvement in transformer diagnosis using dissolved
gas analysis (DGA),‖ IEEE Electrical insulation Magazine, vol. 11, no. 6, pp. 15-21,
1995.
213
[106] S. A. Ward, ―Evaluating transformer condition using DGA oil analysis,‖ in Annual
Rep. Conf. on Elect. Insul. Dielect. Phenomena, pp. 463–468, 2003.
[107] Selwyn Palmer, ―Detection of faults in new and old transformers by dissolved-gas
analysis,‖ IEE Power engineering Journal, pp. 52-54, 1988.
[108] V. G. Arakelian, ―Effective diagnosis for oil-filled equipment,‖ IEEE Elect. Insul.
Mag., vol. 18, no. 6, pp 26-38, Nov./Dec. 2002.
[109] L. Pettersson, ―Estimation of the remaining service life of power transformers and
their insulation‖, Electra, no.133, 65-71, 1990.
[110] P. W. C. Radigk, W. Sorgatz, M. Halm, V. Wassenberg, ‖Comparison of Different
DGA (Dissolved Gas Analysis) Methods for Condition Assessment of Power
Transformers,‖.
[111] I.A.R GRAY Transformer Chemistry Services: ―A Guide of fault Analysis‖, 2007.
[112] Joseph B Digiorgio, ―NTT (Northern Technology & Testing) Technical Bulletin,
Dissolved Gas‖, pp. 7-19, 2008
[113] E. Bräsel, U. Sasum, J. Aragon-Patil, S. Tenbohlen, ―Improved monitoring of
dissolved transformer gases on the basis of a natural internal standard (NIS),‖ CIGRE
23, pp. 1-8, BRUGGE 2007 (www.uni.stuttgart.de/...cigre2007).
[114] S. Tenbohlen, J. Aragon-Patil, M. Fischer, M. Schäfer, Z. D. Wang, I. Höhlein
Atanasova, ―Investigation on sampling, measurement and interpretation of gas-in-oil
analysis for power transformers,‖ D1-24, CIGRE 2008, pp. 1-8
(www.uni.stuttgart.de/...cigre2008).
[115] J. Srivastava, ―Prevention of catastrophic failure of transformer by fault finding
with the help of dissolved gas analysis,‖ Ph. D. thesis, AAU, 2010.
214
APPENDIX-A
CASE STUDY 1: 650 MVA, 400/220 KV, Transformer
DGA result [115] for gas concentration on 11th
November 2001 shown in Table A-1.
Table A-1: Gas concentrations
Analysis:
Routine DGA during six month period, the above mentioned analysis showed no
significant changes in fault gas concentrations. However the decision was made to
increase the DGA testing frequency and to install an online hydrogen and water monitor
on the suspect transformer.
Monitoring equipments were installed in May, 2002 and decision was made to degas the
transformer in order to reset the dissolved gas reference to zero.
Hydrogen generation rate from 30th
May to 26th
September 2002 are given in Table A-2.
Table A-2: Gas generating rate
Time Period H2 Generating Rate
(ppm/day)
May 3 to May 20 <1.0
May 30 to June 15 3.8
June 16 to June 28 5.0
June 29 to Aug. 12 2.8
Aug. 13 to Aug. 25 4.1
Aug. 26 to Sept. 12 1.4
Sept. 13 to Sept. 26 12.1
Dissolved Gases Gas Concentration (ppm)
Hydrogen 392
Methane 594
Ethane 260
Ethylene 260
Acetylene N/D (Not Defined)
Carbon monoxide 736
Carbon dioxide 5900
215
From the above data, it can be observed that the hydrogen-generating rate significantly
increased on 13th
September, thus indicating a sudden change in gas generating
conditions.
On site DGA performed during the period are shown in Table A-3.
Table A-3: Gas concentrations
Dissolved Gases May 30, 2002 July 8, 2002 Sept. 17 , 2002
Hydrogen N/D 170 419
Methane N/D 372 865
Ethane N/D 184 505
Ethylene N/D 179 528
Acetylene N/D 35 30
Carbon monoxide 4 160 262
Carbon dioxide 25 1360 2726
Though large increase in fault gas concentration was observed, percent changes between
tests for both absolute concentrations and key ratios remained more or less constant.
H2 generating rates between 27th
September and 29th
September, 2002 are listed in Table
A-4.
Table A-4: H2 generating rate
Time Period H2 Concentration
(ppm)
Max. H2 Generating
Rate (ppm/hour)
Sept.27,2002,5:52 to Sept.28,2002,5:53 1367 100
Sept.28,2002,8:53 to Sept.28,2002,23:42 5006 300
Sept.29,2002,11:43 to Sept.29,2002,23:42 9494 392
Analysis by Ratio method: (Modified Roger’s Ratio-Fault Diagnosis) from the Table
A-5.
Table A-5: Roger‘s Ratio fault diagnosis
Dissolved Gases July 8, 2002 Sept. 17 , 2002
Hydrogen 170 419
Methane 372 865
Ethane 184 505
216
Ethylene 179 528
Acetylene 35 30
Carbon monoxide 160 262
Carbon dioxide 1360 2726
On 17/9/02 the results are:
R2= C2H2/C2H4 = 60/528 = 0.1136, R2 (0.1-3.0)
R1= CH4/H2 = 865/419 = 2.0644, R1 (>1.0)
R5=C2H4/C2H6 = 528/505=1.0455, R5 (>1.0)
After analyzing the data contained in the above tables, a diagnostic of thermal fault of
medium range temperature was established from 8th
July to 17th
September.
DGA results 30th
September 2002. Fault analysis by Dornenburg‘s ratio of Table A-6.
Table A-6: Dornenburg‘s ratio fault diagnosis
Dissolved Gases Concentration (ppm)
Hydrogen 8150
Methane 11000
Ethane 2460
Ethylene 17960
Acetylene 570
Carbon monoxide 351
Carbon dioxide 2620
R2= C2H2/C2H4 = 570/17960 = 0.0317, R2 (< 0.1)
R1= CH4/H2 = 11000/8150 = 1.3496, R1 (>1.0)
R5=C2H4/C2H6= 17960/2460 =7.3008, R5 (>3.0)
Test results of 30th
September data then led to high temperature (> 700°C) thermal fault
diagnostic.
Analysis by Duval Triangle; on the basis of gas concentrations on 17th
September, 2002.
Concentration of CH4 = 865
Concentration of C2H4 = 528
Concentration of C2H2 = 30
Sum of (CH4 + C2H4 + C2H2) = 865+528+30= 1423
% CH4 = 100 (865/1423) = 60.78
% C2H4 = 100 (528/1423) = 37.10
217
% C2H2 = 100 (30/1423) = 2.108
Plotting the coordinates in the triangle, a thermal fault (300 °C-700 °C) is located.
On 30/9/02:
Concentration of CH4 = 11000
Concentration of C2H4 = 17960
Concentration of C2H2 =570
Sum of (CH4 + C2H4 + C2H2) = 11000+17960+570= 29530
% CH4 = 100 (11000/29530) = 37.25
% C2H4 = 100 (17960/29530) = 60.81
% C2H2 = 100 (570/29530) = 1.93
Plotting the coordinates in the triangle a thermal fault >700 °C is located.
At some stage between 17th
September and 27th
September additional energy, possibly
due to circulating currents, increased significantly the temperature of metal and oil at the
fault location. This led to accelerated production of hydrogen, methane, ethylene and a
small relative amount of acetylene, which are indicative of overheating of oil above 700
°C.
From these diagnostics one can suspect that a source of high electrical resistance was
initially present in transformer (e.g. a bad connection). This high resistance created a
resistive hot spot of increasing resistance, thus the observed increase in hydrogen,
methane, ethane and ethylene. The transformer was taken out of service; it was found that
a high voltage lead was burned and ready to fail.
CASE STUDY 2- 105 MVA, 400/220/33 KV, 1Φ, Autotransformer
The transformer was commissioned on 18.7.97 showed fault gases in its DGA results,
which are tabulated in Table A-7 below:
Table A-7: Gas concentration
Dissolved Gases July28,1997 Aug.14, 1997 Aug.24, 1997
Hydrogen 70 181 319
Methane 09 24 38
Ethane 2 2 2
Ethylene 11 23 33
Acetylene 32 104 134
Carbon monoxide 25 39 61
218
Carbon dioxide 205 260 332
Analysis by Ratio method: (Modified Roger’s Ratio-Fault Diagnosis)
On August 24, 1997:
R2= C2H2/C2H4 = 134/33 =4.06, R2 >3
R1= CH4/H2 = 38/319 = 0.12, R1 >0.1 & R1 < 1.0
R5=C2H4/C2H6 = 33/2 =16.5, R5 > 3
DGA result indicates definite arcing inside the transformer as the acetylene increased
from 32 to 104 ppm with in fortnight. The IEC 60599 analyses interpreted the fault as
arcing with discharge of low energy due to continuous sparking in oil between bad
connection of different potential or to floating potential. Roger‘s ratio method also
analyses as low energy fault with continuous sparking in floating potential.
The first internal inspection by manufacturer did not conclusively reveal the source of the
arcing fault within transformer. The transformer was recharged and fault gases started
rising in the subsequent monthly DGA samples as shown below:
Analysis by Ratio method: (Modified Roger’s Ratio, Fault Diagnostic)
On 24th
April, 1998:
R2 = C2H2/C2H4 = 106/24 =4.416, R2 >3
R1 = CH4/H2 = 33/450 = 0.068, R1 < 1.0
R5 = C2H4/C2H6 = 24/2 = 12.0, R5 > 3
Test results by Roger‘s Ratio method did not conclusively reveal the type of fault as the
DGA results were falling outside the ratio codes. But rise of acetylene interpreted the
fault as arcing with discharge of low energy.
Analysis by Duval Triangle; on the basis of gas concentrations on 24th
August 1998.
Concentration of CH4 = 33
Concentration of C2H4 = 24
Concentration of C2H2 = 106
Sum of (CH4 + C2H4 + C2H2) = 33+24+106= 163
% CH4 = 100 (33/163) = 20.24
% C2H4 = 100 (24/163) = 14.72
% C2H2 = 100 (106/163) = 65.03
Plotting the coordinates in the Duval triangle, the fault is located in the area of low energy
discharge (small arcing). Hence Duval triangle method help to diagnose the fault as it is a
closed system rather than open one.
219
Observing the fault gas pattern from Fig. A-8, it was interpreted that fault does not
originate from the winding. The fault/arcing appears to be originated from any metallic
joint within the transformer. During the 2nd internal inspection the wall shunts clamping
structure was found loose with arcing marks on the earth strips and the boss. No other
abnormality was noticed inside the transformer.
Table A-8: Gas concentrations
Dissolved Gases 10th
March 1998 4th
April 1998 13th
April 1998
Hydrogen 11 317 450
Methane 4 30 33
Ethane 1 2 2
Ethylene 7 24 24
Acetylene 8 121 106
Carbon monoxide 25 75 91
Carbon dioxide 109 313 275
220
APPENDIX-B
Case 1: 100MVA, 220/66kV Power Transformer:
Gases Sept. 2004 Jun. 2005 Jul. 2005 As per
IEC
C2 H4
High
As IEEE
Std…
HOT
SPOT
Methane 18 63 187
Ethane 15 48 178
Ethylene 18 194 614
Acetylene ND ND ND
Carbon Dioxide 155 582 2034
Carbon mono-oxide ND ND ND
Hydrogen ND ND ND
Total gas content (TGC) 432 543 765
Case 2: 20MVA, 11/0.4kV Power Transformer:
Gases s-1 s-2 s-3 s-4 s-5 s-6
AS per
IEC
C2H2 High
As per
IEEE Std..
Overloading
fault
Methane 3 5 4 9 9 10
Ethane 1 ND ND 2 2 2
Ethylene 5 7 9 14 18 23
Acetylene ND ND 3 5 9 11
Carbon Dioxide 340 409 439 1419 603 733
Carbon mono-oxide ND ND ND ND ND ND
Hydrogen ND ND ND ND ND ND
Total gas content (TGC) 61 765 588 876 743 765
Case 3: 10/12.5MVA, 66-33/114kV Power Transformer:
Gases 12.09.05 23.09.05 17.10.05 27.10.05 25.11.05 29.11.05
As per Methane 2779 2738 1222 34 392 6852
221
Ethane 1330 1245 1333 30 299 3137 IEC
C2H4
High
IEEE
As per
OLD
OIL
Ethylene 3359 3102 2887 55 879 8797
Acetylene 11 13 15 ND ND ND
Carbon
Dioxide
1713 3243 3820 97 508 1342
Carbon
mono-
oxide
ND ND ND ND ND ND
Hydrogen 733 620 392 ND 3791 2195
Total gas
content
(TGC)
699 720 943 477 943 765
Date of
Filtration
Jun - 03 Jun-03 Jun-03 23.10.05 23.10.05 23.10.05
Case 4: 100MVA, 220/66kV Power Transformer:
Gases 2/15/05 5/13/05 6/28/05 06/07/05 10/07/05
As per
IEC
C2H2
High
As per
IEEE
Std.
HOT
SPOT
Methane 38 102 10 2 254
Ethane 20 60 47 3 17
Ethylene 36 108 141 7 313
Acetylene ND ND 137 5 417
Carbon Dioxide 2382 4399 9109 252 365
Carbon mono-oxide ND ND ND ND ND
Hydrogen ND ND 930 ND 328
Total gas content (TGC) 605 820 699 432 521
Date of Filtration 1988/-- 1988/-- 1988/-- 1988/--
06.07.05
1988/--
06.07.05
Case 5: 100MVA, 220/66kV Power Transformer:
Gases 11/05/05 14/06/05 30/06/05 10/07/05 05/08/05
As per Methane 5 36 51 86 83
222
Ethane ND 6 7 11 10 IEC
C2H4
High
As per
IEEE
HOT
SPOT
Ethylene 17 106 145 213 221
Acetylene ND ND 4 9 3
Carbon Dioxide 899 499 1193 1214 1819
Carbon mono-
oxide
ND ND ND ND ND
Hydrogen ND ND 8 9 13
Total gas content
(TGC)
743 92 543 654 61
Date of Filtration 2002/-- 2002/--
25/5/05
2002/--
25/3/05
2002/--
25/3/05
2002/--
25/3/05
Case 6: 315MVA, 400/200kV Power Transformer:
Parameter Measured Value
As per IEC
300-700 oC
As per IEEE
Above 700 oC
Total Gas content 6.85
Nitrogen(N2) 4.90
Oxygen(O2) 1.11
Hydrogen()H2 39
Methane(CH4) 55
Ethylene(C2H4) 72
Ethane(C2H6) 22
Acetylene(C2H2) 0.0
Carbon mono-oxide(CO) 80
Carbon Dioxide(CO2) 2328
Case 7: 315MVA, 400/220kV Power Transformer:
Parameter Measured Value
Total Gas content 9.40
223
Nitrogen(N2) 7.02 As per IEC
300-700 oC
As per IEEE
Below 700 oC
Oxygen(O2) 1.51
Hydrogen(H2) 40
Methane(CH4) 79
Ethylene(C52H4) 89
Ethane(C2H6) 42
Acetylene(C2H2) 0.0
Carbon mono-oxide(CO) 110
Carbon Dioxide(CO2) 3802
Case 8: 250MVA Generator Transformer:
Gas 18/1/01 19/4/01 10/7/01
IEC
300-700 oC
As per
IEEE Std.
Circulating
Current
TGC 9.52 9.52 10.87
Methane 547 631 362
Ethane 199 386 217
Ethylene 530 924 512
Acetylene ND ND ND
Hydrogen 18 81 54
Carbon Dioxide 2905 3449 4155
Case 9: 25MVA Transformer:
Gas Contents
As per IEC
NO FAULT
As per IEEE
Arc Discharge
Methane 20
Ethane 10
Ethylene 21
Acetylene 47
Hydrogen 139
224
Case 10: 50MVA Transformer:
Gas 9/5/05 2/9/05
AS per IEC
NO FAULT
As per IEEE Std.
Arc Discharge
Methane 748 772
Ethane 238 182
Ethylene 2587 1936
Acetylene 10 1
Hydrogen 198 149
Case 11: 50MVA, 400kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible
Gas
438 CEGB METHOD: Circulating current-bad
contact.
IEC, CIGRE STANDARDS: Thermal fault
of medium temperature range 300-700 oC.
Local overheating of core due to
concentration of flux. Increasing hot spot
temperatures varying from small hot pots in
core, shorting links in core, overheating of
copper due to eddy currents bad
contacts/joints upto core & tank circulating
currents.
ANSI/IEEE STANDARDS: Circulating
currents in windings.
REMARKS: DGA show C2H4 level being
marginal wrt IEEE norms (50). C2H4,
however has not shown increase wrt earlier
sample. Next sample is due after 3 months
for further monitoring.
Hydrogen (H2) 42
Methane(CH4) 108
Ethylene(C2H4) 50
Ethane(C2H6) 37
Acetylene(C2H2) 0
Carbon mono-
oxide(CO)
196
Carbon Dioxide(CO2) 2343
Nitrogen (N2 %) 4.20
Oxygen (O2 %) 0.00
CH4/H2 2.57
C2H6/CH4 0.34
C2H4/C2H6 1.35
C2H2/C2H4 0.00
Case 12: 200MVA, 400/21kV, Power Transformer (2005):
225
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 400 CEGB METHOD: Circulating current-bad
contact.
IEC, CIGRE STANDARDS: Thermal fault of
high temperature range >700 oC. Local
overheating of core due to concentration of flux.
Increasing hot spot temperatures varying from
small hot pots in core, shorting links in core,
overheating of copper due to eddy currents bad
contacts/joints upto core & tank circulating
currents.
ANSI/IEEE STANDARDS: Circulating currents
in windings.
REMARKS: DGA show C2H4 level is exceeding
IEEE norms (50). C2H4, however has not shown
increase wrt earlier sample. Next sample is due
after 1 month for further monitoring.
Hydrogen (H2) 45
Methane(CH4) 80
Ethylene(C2H4) 120
Ethane(C2H6) 35
Acetylene(C2H2) 0
Carbon mono-
oxide(CO)
90
Carbon Dioxide(CO2) 950
Nitrogen (N2 %) 4.30
Oxygen (O2 %) 0.00
CH4/H2 1.78
C2H6/CH4 0.44
C2H4/C2H6 3.43
C2H2/C2H4 0.00
Case 13: 250MVA, 16/420kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 310 CEGB METHOD: Overheating 200-300 oC..
IEC, CIGRE STANDARDS: No fault
normal ageing.
ANSI/IEEE STANDARDS: Overheating
200-300 oC.
REMARKS: DGA show individual gases
within IEEE norms (50). Next sample is
due after 3 months for further monitoring.
Hydrogen (H2) 52
Methane(CH4) 40
Ethylene(C2H4) 10
Ethane(C2H6) 48
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 150
Carbon Dioxide(CO2) 1593
Nitrogen (N2 %) 4.30
226
Oxygen (O2 %) 0.00
CH4/H2 0.77
C2H6/CH4 1.20
C2H4/C2H6 0.21
C2H2/C2H4 0.00
Case 14: 200MVA, 400/21kV, Power Transformer(2005):
Parameter Measured
Value(ppm)
Faults
Total
Combustible
Gas
425 CEGB METHOD: Circulating current-bad contact.
IEC, CIGRE STANDARDS: Thermal fault of high
temperature range >700 oC. Local overheating of core
due to concentration of flux. Increasing hot spot
temperatures varying from small hot pots in core,
shorting links in core, overheating of copper due to
eddy currents bad contacts/joints upto core & tank
circulating currents.
ANSI/IEEE STANDARDS: Circulating currents core
& tank; over loaded joints.
REMARKS: DGA show C2H4 level is exceeding IEEE
norms (50). C2H4, however has not shown increase
wrt earlier sample. Next sample is due after 1 month
for further monitoring.
Hydrogen (H2) 50
Methane(CH4) 90
Ethylene(C2H4) 125
Ethane(C2H6) 30
Acetylene(C2H2) 0
Carbon mono-
oxide(CO)
95
Carbon
Dioxide(CO2)
960
Nitrogen (N2 %) 9.30
Oxygen (O2 %) 0.009
CH4/H2 1.80
C2H6/CH4 0.33
C2H4/C2H6 4.17
C2H2/C2H4 0.00
227
Case 15: 250MVA, 16/420 kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 185 CEGB METHOD: Overheating below 150 oC.
IEC, CIGRE STANDARDS: Thermal fault of low
temperature range 150-300 oC. Increasing hot spot
temperatures varying from small hot pots in core,
shorting links in core, overheating of copper due to
eddy currents bad contacts/joints upto core & tank
circulating currents.
ANSI/IEEE STANDARDS: Slight overheating to
150 oC.
REMARKS: Sample taken after oil filtration and
recharging: present DGA results may be kept as
benchmark for further monitoring. Next sample is
due after 1 month.
Hydrogen (H2) 25
Methane(CH4) 30
Ethylene(C2H4) 0
Ethane(C2H6) 15
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 110
Carbon Dioxide(CO2) 560
Nitrogen (N2 %) 4.40
Oxygen (O2 %) 0.00
CH4/H2 1.20
C2H6/CH4 0.50
C2H4/C2H6 0.00
C2H2/C2H4 ****
Case 16: 50MVA, 420 kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total
Combustible
Gas
340 CEGB METHOD: Overheating 150-200 oC.
IEC, CIGRE STANDARDS: Thermal fault of low
temperature range 150-300 oC. Local overheating
of core due to concentration of flux. Increasing hot
spot temperatures varying from small hot pots in
core, shorting links in core, overheating of copper
due to eddy currents bad contacts/joints upto core
& tank circulating currents.
ANSI/IEEE STANDARDS: Slight overheating
150-200 oC.
Hydrogen (H2) 20
Methane(CH4) 25
Ethylene(C2H4) 10
Ethane(C2H6) 65
Acetylene(C2H2) 0
228
Carbon mono-
oxide(CO)
190 REMARKS: DGA results show C2H6 level being
marginal wrt IEEE norms.. Next sample is due
after 3 months for further monitoring. Carbon
Dioxide(CO2)
2310
Nitrogen (N2 %) 4.20
Oxygen (O2 %) 0.00
CH4/H2 1.25
C2H6/CH4 2.60
C2H4/C2H6 0.15
C2H2/C2H4 0.00
Case 17: 63MVA, 420 kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 556 CEGB METHOD: Overheating below 150 0C.
IEC, CIGRE STANDARDS: Thermal fault of
low temperature range 150-300 oC. Local
overheating of core due to concentration of flux.
Increasing hot spot temperatures varying from
small hot pots in core, shorting links in core,
overheating of copper due to eddy currents bad
contacts/joints upto core & tank circulating
currents.
ANSI/IEEE STANDARDS: Slight overheating
to 150 oC.
REMARKS: DGA results show C2H6 level
exceeding IEEE norms. Next sample is due after
1 month.
Hydrogen (H2) 45
Methane(CH4) 105
Ethylene(C2H4) 15
Ethane(C2H6) 66
Acetylene(C2H2) 0
Carbon mono-
oxide(CO)
310
Carbon Dioxide(CO2) 2490
Nitrogen (N2 %) 7.20
Oxygen (O2 %) 0.00
CH4/H2 2.33
C2H6/CH4 0.63
C2H4/C2H6 0.23
C2H2/C2H4 0
229
Case 18: 200 MVA, 400/21kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 225 CEGB METHOD: Partial Discharge.
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results show gases well
within IEEE norms. Next sample is due after
3 months.
Hydrogen (H2) 30
Methane(CH4) 2
Ethylene(C2H4) 2
Ethane(C2H6) 9
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 182
Carbon Dioxide(CO2) 704
Nitrogen (N2 %) 4.40
Oxygen (O2 %) 0.00
CH4/H2 0.07
C2H6/CH4 4.50
C2H4/C2H6 0.22
C2H2/C2H4 0
Case 19: 200 MVA, 420/21kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 87 CEGB METHOD: Partial Discharge.
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results show individual
gases well within IEEE norms. Next sample
is due after 3 months for further monitoring.
Hydrogen (H2) 34
Methane(CH4) 1
Ethylene(C2H4) 1
Ethane(C2H6) 1
Acetylene(C2H2) 0
Carbon mono-
oxide(CO)
50
230
Carbon Dioxide(CO2) 1358
Nitrogen (N2 %) 4,30
Oxygen (O2 %) 0.00
CH4/H2 0.03
C2H6/CH4 1.00
C2H4/C2H6 1.00
C2H2/C2H4 0.00
Case 20: 200 MVA, 400/21kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 187 CEGB METHOD: Sample is o.k.
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results show gases well
within IEEE norms. Next sample is due
after 3 months for further monitoring.
Hydrogen (H2) 48
Methane(CH4) 20
Ethylene(C2H4) 7
Ethane(C2H6) 5
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 90
Carbon Dioxide(CO2) 1336
Nitrogen (N2 %) 4.30
Oxygen (O2 %) 0.00
CH4/H2 0.42
C2H6/CH4 4.30
C2H4/C2H6 1.40
C2H2/C2H4 0
Case 21: 200 MVA, 420/133kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
231
Total Combustible Gas 226 CEGB METHOD: Partial Discharge.
IEC, CIGRE STANDARDS: No fault-Normal
Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results do not show any
abnormality. Next sample is due after 6 months.
Hydrogen (H2) 30
Methane(CH4) 1
Ethylene(C2H4) 6
Ethane(C2H6) 2
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 185
Carbon Dioxide(CO2) 790
Nitrogen (N2 %) 4.40
Oxygen (O2 %) 0.00
CH4/H2 0.03
C2H6/CH4 4.40
C2H4/C2H6 3.0
C2H2/C2H4 0
Case 22: 200 MVA, 420/33kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 310 CEGB METHOD: sample is o.k.
IEC, CIGRE STANDARDS: No fault-Normal
Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results do not show any
abnormality. Next sample is due after 6 months.
Hydrogen (H2) 35
Methane(CH4) 31
Ethylene(C2H4) 10
Ethane(C2H6) 19
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 185
Carbon Dioxide(CO2) 180
Nitrogen (N2 %) 4.30
Oxygen (O2 %) 0.00
232
CH4/H2 0.89
C2H6/CH4 0.61
C2H4/C2H6 0.53
C2H2/C2H4 0
Case 23: 240 MVA, 21/420kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 279 CEGB METHOD: sample is o.k...
IEC, CIGRE STANDARDS: No fault-Normal
Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results do not show any
abnormality. Next sample is due after 2months.
Hydrogen (H2) 35
Methane(CH4) 4
Ethylene(C2H4) 1
Ethane(C2H6) 5
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 232
Carbon Dioxide(CO2) 1296
Nitrogen (N2 %) 4.30
Oxygen (O2 %) 0.00
CH4/H2 0.11
C2H6/CH4 1.25
C2H4/C2H6 0.20
C2H2/C2H4 0
Case24: 80 MVA, 132/11kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 5 CEGB METHOD: Partial Discharge.
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
ANSI/IEEE STANDARDS: Normal
Hydrogen (H2) 5
Methane(CH4) 0
Ethylene(C2H4) 0
233
Ethane(C2H6) 0 REMARKS: sample is taken after oil
filtration/prior to first charging: BDV,
Moisture, & Tan-Delta are as per IS-1866
requirements. SP. resistance value is on
the lower side wrt requirements ( 6E21
min.). Sample is due immediately after
charging.
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 0
Carbon Dioxide(CO2) 130
Nitrogen (N2 %) 4.40
Oxygen (O2 %) 0.00
CH4/H2 0.03
C2H6/CH4 ****
C2H4/C2H6 ****
C2H2/C2H4 *****
Case25: 200 MVA, 21/1420kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 301 CEGB METHOD: sample is o.k..
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results do not show any
abnormality. Next sample is due after 3
months.
Hydrogen (H2) 45
Methane(CH4) 20
Ethylene(C2H4) 3
Ethane(C2H6) 3
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 230
Carbon Dioxide(CO2) 910
Nitrogen (N2 %) 4.40
Oxygen (O2 %) 0.00
CH4/H2 0.44
C2H6/CH4 0.15
C2H4/C2H6 1.0
C2H2/C2H4 0
234
Case26: 200 MVA, 420/21kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 107 CEGB METHOD: Sample is o.k..
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results do not show
any abnormality. Next sample is due after
3 months.
Hydrogen (H2) 10
Methane(CH4) 5
Ethylene(C2H4) 1
Ethane(C2H6) 1
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 90
Carbon Dioxide(CO2) 820
Nitrogen (N2 %) 4.40
Oxygen (O2 %) 0.00
CH4/H2 0.50
C2H6/CH4 0.20
C2H4/C2H6 1.00
C2H2/C2H4 0
Case 27: 200 MVA, 420/21kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 90 CEGB METHOD: Partial Discharge.
IEC, CIGRE STANDARDS: No fault-Normal
Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results show individual gases
well within IEEE norms. Next sample is due after
3 months.
Hydrogen (H2) 36
Methane(CH4) 2
Ethylene(C2H4) 2
Ethane(C2H6) 3
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 47
235
Carbon Dioxide(CO2) 1017
Nitrogen (N2 %) 4.30
Oxygen (O2 %) 0.00
CH4/H2 0.06
C2H6/CH4 1.50
C2H4/C2H6 0.67
C2H2/C2H4 0
Case 28: 80 MVA, 132/11kV, Power Transformer (2005):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 271 CEGB METHOD: sample is o.k.
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
ANSI/IEEE STANDARDS: Normal
REMARKS: DGA results do not show any
abnormality. Next sample is due after 6
months.
Hydrogen (H2) 29
Methane(CH4) 10
Ethylene(C2H4) 1
Ethane(C2H6) 1
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 230
Carbon Dioxide(CO2) 1510
Nitrogen (N2 %) 4.30
Oxygen (O2 %) 0.00
CH4/H2 0,34
C2H6/CH4 0.10
C2H4/C2H6 1.00
C2H2/C2H4 0
Case29: 75 MVA,400/345kV, Power Transformer (2006):
Parameter Measured
Value(ppm)
Faults
236
Total Combustible Gas 200 CEGB METHOD: sample is o.k.
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
Hydrogen (H2) 12
Methane(CH4) 8
Ethylene(C2H4) 5
Ethane(C2H6) 40
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 130
Carbon Dioxide(CO2) 1030
Nitrogen (N2 %) 4.61
Oxygen (O2 %) 1.75
CH4/H2 0,67
C2H6/CH4 5.00
C2H4/C2H6 0.13
C2H2/C2H4 0
Case30: 50 MVA,33/6.9kV, Power Transformer (2006):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 171 CEGB METHOD: Circulating current-bad
contact
IEC, CIGRE STANDARDS: Thermal fault of
medium range 300-700 0C. Local overheating
of core due to concentration of flux. Increasing
hot spot temperature varying from small hot
pots in core, shorting links in , overheating of
copper due to eddy currents bad contacts/joints
upto core & tank circulating currents.
ANSI/IEEE STANDARDS: Circulating
currents in windings.
REMARKS: DGA show C2H4 level being
marginal wrt IEEE norms (50). Next sample is
due after 3 months for further monitoring.
Hydrogen (H2) 16
Methane(CH4) 25
Ethylene(C2H4) 39
Ethane(C2H6) 19
Acetylene(C2H2) 0
Carbon mono-
oxide(CO)
70
Carbon Dioxide(CO2) 810
Nitrogen (N2 %) 00
Oxygen (O2 %) 00
237
CH4/H2 1.56
C2H6/CH4 0.76
C2H4/C2H6 2.05
C2H2/C2H4 0
Case31: 25 MVA,11/6.65kV, Power Transformer (2006):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 61 CEGB METHOD: sample is o.k.
IEC, CIGRE STANDARDS: No fault-
Normal Ageing.
Hydrogen (H2) 22
Methane(CH4) 3
Ethylene(C2H4) 0
Ethane(C2H6) 0
Acetylene(C2H2) 0
Carbon mono-
oxide(CO)
36
Carbon Dioxide(CO2) 606
Nitrogen (N2 %) 4.70
Oxygen (O2 %) 2.70
CH4/H2 0.14
C2H6/CH4 00
C2H4/C2H6 ***
C2H2/C2H4 ***
Case32:25 MVA,11/6.6kV, Power Transformer (2006):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 208 CEGB METHOD: sample is o.k.
IEC, CIGRE STANDARDS: No Hydrogen (H2) 33
238
Methane(CH4) 15 fault-Normal Ageing.
Ethylene(C2H4) 0
Ethane(C2H6) 0
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 160
Carbon Dioxide(CO2) 1760
Nitrogen (N2 %) 4.62
Oxygen (O2 %) 2.66
CH4/H2 0.45
C2H6/CH4 00
C2H4/C2H6 ***
C2H2/C2H4 ***
Case33: 50MVA,420 kV, Power Transformer (2006):
Parameter Measured Value(ppm) Faults
Total Combustible
Gas
195 CEGB METHOD: Denominator
zero- Diagnosis
IEC, CIGRE STANDARDS: No
fault-Normal
ANSI/IEEE STANDARD :
Denominator zero- Diagnosis
Bypassed
IS:1866(old) NORMS
VOILATION:BDV,
MOISTURE
Hydrogen (H2) 83
Methane(CH4) 5
Ethylene(C2H4) 1
Ethane(C2H6) 0
Acetylene(C2H2) 0
Carbon mono-
oxide(CO)
105
Carbon
Dioxide(CO2)
860
Nitrogen (N2 %) 2.70
Oxygen (O2 %) 1.66
CH4/H2 0.06
C2H6/CH4 00
239
C2H4/C2H6 ***
C2H2/C2H4 0
Case34: 13 MVA,33/6.9kV, Power Transformer (2006):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 210 CEGB METHOD: sample is o.k.
IEC, CIGRE STANDARDS: No
fault-Normal Ageing.
ANSI/IEEE STANDARD:
NORMAL
Hydrogen (H2) 13
Methane(CH4) 6
Ethylene(C2H4) 0
Ethane(C2H6) 1
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 190
Carbon Dioxide(CO2) 1960
Nitrogen (N2 %) 0
Oxygen (O2 %) 0
CH4/H2 0.43
C2H6/CH4 0.17
C2H4/C2H6 0
C2H2/C2H4 ****
Case 35: 50MVA, 33/6.9 kV, Power Transformer (2008):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 171 CEGB METHOD: Mixed fault
IEC, CIGRE STANDARDS:
Thermal fault of low temperature
less than 150 oC
- General insulated conductor
overheating
ANSI/IEEE STANDARDS: Mixed
Hydrogen (H2) 32
Methane(CH4) 17
Ethylene(C2H4) 33
Ethane(C2H6) 22
240
Acetylene(C2H2) 0 fault
Carbon mono-oxide(CO) 65
Carbon Dioxide(CO2) 560
Nitrogen (N2 %) 3.40
Oxygen (O2 %) 1.00
CH4/H2 0.53
C2H6/CH4 1.29
C2H4/C2H6 1.5
C2H2/C2H4 0.00
Case 36: 16MVA, 16/6.9 kV, Power Transformer (2008):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 311 CEGB METHOD: over heating200-
300 oC
IEC, CIGRE STANDARDS: No fault
–Normal ageing.
ANSI/IEEE STANDARDS: Over
heating 200-300 oC
Hydrogen (H2) 36
Methane(CH4) 26
Ethylene(C2H4) 2
Ethane(C2H6) 45
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 192
Carbon Dioxide(CO2) 1510
Nitrogen (N2 %) 4.62
Oxygen (O2 %) 2.69
CH4/H2 0.72
C2H6/CH4 1.73
C2H4/C2H6 0.07
C2H2/C2H4 0
241
Case 37: 25MVA, 220/6.9 kV, Power Transformer (2008):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 448 CEGB METHOD: Conductor
overheating
IEC, CIGRE STANDARDS: Thermal
fault of low temperature less than 150 oC
- General insulated conductor
overheating
- ANSI/IEEE STANDARDS:
General conductor
overheating
Dissolved gas norms violation :co,co2
IS:1866(old) Norms Voilation :
Specific resistance, TAN DELTA
Hydrogen (H2) 30
Methane(CH4) 5
Ethylene(C2H4) 13
Ethane(C2H6) 10
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 380
Carbon Dioxide(CO2) 2610
Nitrogen (N2 %) 0
Oxygen (O2 %) 0
CH4/H2 0.17
C2H6/CH4 2
C2H4/C2H6 1.3
C2H2/C2H4 0.00
242
Case 38: 13MVA, 33/6.9 kV, Power Transformer (2008):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 129 CEGB METHOD: sample OK
IEC, CIGRE STANDARDS: No fault
–Normal ageing.
ANSI/IEEE STANDARDS: Normal
CELLULOSE INVOLVEMENT: YES
Hydrogen (H2) 16
Methane(CH4) 5
Ethylene(C2H4) 10
Ethane(C2H6) 2
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 95
Carbon Dioxide(CO2) 1956
Nitrogen (N2 %) 0
Oxygen (O2 %) 0
CH4/H2 0.31
C2H6/CH4 0.40
C2H4/C2H6 5
C2H2/C2H4 0
Case 39: 16MVA, 15/6.9 kV, Power Transformer (2008):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 104 CEGB METHOD: Partial Discharge
IEC, CIGRE STANDARDS: NO fault-
normal Ageing
ANSI/IEEE STANDARDS: normal
CELLULOSE INVOLVEMENT: YES
Hydrogen (H2) 40
Methane(CH4) 3
Ethylene(C2H4) 0
Ethane(C2H6) 1
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 60
243
Carbon Dioxide(CO2) 1430
Nitrogen (N2 %) 4.68
Oxygen (O2 %) 2.63
CH4/H2 0.08
C2H6/CH4 0.33
C2H4/C2H6 0
C2H2/C2H4 *****
Case 40: 16MVA, 16/6.9 kV, Power Transformer (2008):
Parameter Measured
Value(ppm)
Faults
Total Combustible Gas 311 CEGB METHOD: over heating200-300 oC
IEC, CIGRE STANDARDS: No fault –
Normal ageing.
ANSI/IEEE STANDARDS: Over heating
200-300 oC
Hydrogen (H2) 36
Methane(CH4) 26
Ethylene(C2H4) 2
Ethane(C2H6) 45
Acetylene(C2H2) 0
Carbon mono-oxide(CO) 192
Carbon Dioxide(CO2) 1510
Nitrogen (N2 %) 4.62
Oxygen (O2 %) 2.69
CH4/H2 0.72
C2H6/CH4 1.73
C2H4/C2H6 0.04
C2H2/C2H4 0
244
PUBLICATION FROM THESIS WORK
Published Papers
[1] Sukhbir Singh, Dheeraj Joshi, M. N. Bandyopadhyay, ― Software Implementation of
Duval Triangle Technique for DGA in Power Transformers‖, International Journal of
Electrical Engineering, vol. 4, number 5, pp. 529-540, 2011.
[2] Sukhbir Singh, M. N. Bandyopadhyay, ―Dissolved Gas Analysis Technique for
Incipient fault Diagnosis in Power Transformers: A Bibliographic Survey‖, IEEE
Electrical Insulation Magazine, vol. 26, no. 6, pp.41-46, Non-Dec 2010.
[3] Sukhbir Singh, M. N. Bandyopadhyay, ―Duval Triangle: A Noble Technique for DGA
in Power Transformers‖, International Journal of Electrical and Power Engineering,
vol. 4 no. 3, pp. 193-197, 2010.
[4] Sukhbir Singh, Dheeraj Joshi, M. N. Bandyopadhyay, ―Dissolved gas analysis for
power transformers‖, Electrical India, pp. 112-119, March-2012.
[5] Sukhbir Singh, ―A Survey on Testing and Residual Life Assessment of Power
Transformer‖, National Seminar on Power for All by 2012 and Challenges, IEC
Gautam Budh Nagar (Greater Noida) , 11-12th
Jan.2009.
[6] Sukhbir Singh, ―Modern Trends on Power Transformer Fault Diagnosis‖,
Management and Technology Journal, Agra, vol. I, pp. 21-27., Dec. 2008.
[7] Sukhbir Singh‘ ―Comparative Study on Fault Diagnosis on Power Transformers‖,
International Conference on Transformers TRAFOSEM-2008, ITMA, New Delhi, 11-
12th
Nov.2008.
[8] Sukhbir Singh, ―New Trends on Power Transformer Fault Diagnosis‖, National
Conference on Recent Advances in Electrical Engineering RAEE-2008, NIT
Hamirpur (HP), Dec. 26-27, 2008.
Accepted Papers
[1] Sukhbir Singh, Dheeraj Joshi, M. N. Bandyopadhyay, ―Review on computer
applications for fault diagnosis on power transformers‖, Journal of Electrical
Engineering, 2011.
[1] Sukhbir Singh, Dheeraj Joshi, M. N. Bandyopadhyay, ―Feed-forward BP ANN Duval
triangle method for DGA fault analysis of power transformers‖, IEEE Electrical
Insulation Magazine, 2012.
245
Communicated Papers
[1] Sukhbir Singh, Dheeraj Joshi, M. N. Bandyopadhyay, ―New fuzzy logic Duval
triangle techniques for fault diagnosis in power transformers‖, WSEAS Trans. on
Power Systems, 2011.
[2] Sukhbir Singh, Dheeraj Joshi, M. N. Bandyopadhyay, ― A review on computational
intelligence for fault diagnosis on power transformers,‖ IEEE Computational
Intelligence Magazine, 2011.
[3] Sukhbir Singh, Dheeraj Joshi, M. N. Bandyopadhyay, ―Feed-forward back-
propagation ANN based Duval triangle method for fault diagnosis in power
transformers‖, International Journal of Electrical Power Engineering, 2011.