INDIAN INSTITUTE OF TECHNOLOGY, KHARAGPUR 721302, DECEMBER 27-29, 2002 389

Abstract-- This paper presents some preliminary results of

investigations being carried out for accounting the consumption of electrical energy in the rural sector of the state of Karnataka, in the absence of complete metering. In this pilot study extensive measurements have been carried out to separate line losses from energy consumed and to validate methods for computing line losses. The study indicates that the loss formula prescribed by the Rural Electrification Corporation (REC) which is used all over the country may need to be modified to take into account the preliminarily wide variabilities in the supply conditions.

Index Terms— Energy Accounting, Distribution losses, Loss formula, Rural radial systems.

I INTRODUCTION

ransmission and Distribution (T&D) losses in India are known to be very high, but no dependable figures are

available for them. In states like Karnataka, Andhra Pradesh, Haryana etc. where agricultural loads are very high, the energy consumed by IP sets are not metered. A substantial part of the electrical energy thus remains unaccounted. In Karnataka, for example, only about 1/3rd of the electrical energy is metered and billed. Out of the remaining 2/3rd of the energy, there is no dependable account of the amount of energy consumed by the IP sets and the amount lost in distribution at 11 kV and 430 volts levels. Thus the loss figures presented by State Electricity Boards (SEB), for years, are considered by many to be at best poor estimates or at worst, deliberately fudged. Transmission losses are relatively low and can be measured with reasonable accuracy. Distribution losses at 11 kV and 430 volt levels on the other hand, are very high and in the absence of meters, particularly at the rural sector, cannot be measured. These losses are some times estimated, using formulae which are not validated using network data which are not authenticated. The LT network (430 volts) drawings are particularly inaccurate, with unauthorized loads that are not known. As a result, there are no means to separate technical and commercial losses. The SEBs are often accused of playing down technical losses and guestimate the commercial loss or theft. A commonly quoted amount for theft of electricity in India is Rs 25,000 crores/year. At an average of Rs 2.5 per unit,

D.P. Sen Gupta, Indraneel sen and P.S. Nagendra Rao, are with Indian Institute of Science, Bangalore. Vinod Havanagi is with Karnataka Power Transmission Corporation Limited.

this would amount to 100 billion units of electricity, which is nearly 25% of the total electricity generated in the country. No one really knows how this figure was arrived at, particularly in the absence of metering. Unfortunately such figures tend to be passed around without validation and picked up by many analysts.

Rural feeders are often long, with 70-90 transformer centres providing power to the villages along the way. Out of a total of 3700, 11 kV feeders in Karnataka State, more than 1000 feeders are long and overloaded (at times having to carry more than 200 A when it is designed to carry about 100A) As a result, copper losses are very high and voltage regulations are extremely poor. In the absence of meters at receiving ends, it is not possible to measure the distribution losses. Hardly any of the 1.3 million IP sets, consuming almost 1/3rd of the total electrical energy in the state, is fitted with meters. As a result, KPTCL (Karnataka Power Transmission Corporation Limited) ends up making estimates of T&D losses using approximate methods, having no means to verify the results. The estimated losses are calculated for a year and no information of the variation of losses with seasonal changes of loads is available. There also appears to be a lot of uncertainty and controversy about the losses in the LT networks, which are once again computed, using approximated equations and incomplete network and load data.

It was therefore decided to setup a joint group with participation from the Indian Institute of Science and KPTCL, to study various issues connected with the distribution systems. In this context the following issues were also considered in great detail

a. Examine the validity of the current methods in use for computing losses in the distribution network.

b. Propose modified or new methods of computing line losses with reasonable accuracy without requiring extensive metering.

This paper briefly describes some of the important outcomes of this effort. This paper is more concerned with presenting data collected which is unique and not readily available rather than analyzing the same.

II METHODOLOGY

Four rural feeders (11kV) selected across four geographically disparate areas have been extensively metered at the transformer centres where the voltage is stepped down for LT supply at 430 volts to villages, primarily for IP set operation. Fig 1 provides the thumbnail

D.P. Sen Gupta, Indraneel Sen, P.S. Nagendra Rao & Vinod Havanagi

Energy Accounting in the Rural Sector in Karnataka State

T

NATIONAL POWER SYSTEMS CONFERENCE, NPSC 2002 390

sketches of the four feeders chosen for the study. All the transformers on the selected feeders are fitted with trivector electronic meters on the LT side. These meters record voltage, current, active and reactive power every half an hour and the data are stored for over a month. These data are collected with the help of a MRI (meter reading instrument) and download into a computer. Fig 2 provides a typical set of Load curves obtained from one of the feeders for one month.

The sum of the energy readings obtained from all the meters when subtracted from the sending end energy for a particular period, provides the loss in the 11 kV feeder and the transformation loss. Since it is not feasible to carry out similar studies on the 3700, 11 kV feeder in Karnataka State, with an average of about 75 transformers centres each, it was considered useful to develop a method of computing losses on an 11 kV feeder, based on measurements at the sending end. All 11-kv substations in Karnataka State have been fitted with trivector meters. Unfortunately these metes are hardly put to any use except metering energy which is noted down and entered into a log book. As has been stated earlier there is hardly any energy accounting beyond the 11-kV substations. The following steps have been taken to obtain a reasonable estimate of the technical losses and the average pumpset consumption:

1. Measure 11-kV losses and develop a reliable computational method to estimate losses based on data collated from the electronic meter at the sending end at 11-kV substation.

2. Compute LT line losses and validate them against measurements

3. From 1 and 2 and the loss information the sending end measurements calculate average pumpset consumption.

A. The Standard kmkVA Method For Computing HT

Line Losses The standard method of computing annual loss in 11 kV

feeders is given in equation 1.

Annual Energy Loss= 2

2

2105.0

DFLLF

LDFRLP

××

× Units/Year (1)

where, P = ∑ kVA of all the transformers

R = Resistance of the conductor/unit length L = Length of the conductor

LDF = Load Distribution Factor = ∑

×kmkVA

LP

* In this paper unless specified otherwise ∑ stands for

summation over all segments

DF( Diversity Factor) = kVAinLoadPeak

P

LLF the Loss Load Factor, defined as 0.2 LF + 0.8 LF2 (2)

or 0.3LF + 0.7 LF2

where, LF(Load Factor) = 8760×LoadPeakoutsentEnergy

Needless to say, the formula is approximate and was developed when computers or even calculators had not been developed. This method continues to be used by the State Electricity Boards (SEBs) without validating the results by actual measurements.

It needs to be mentioned in this context that this standard method does not take into account two crucial factors

1. The power supply to the rural sector in most states is highly irregular. In Karnataka state, for example, 3 phase power is supplied for 6-8 hours, single phase power is supplied for 8-10 hours and there is no supply at all for the remaining period of the day [Fig 3]. This feature significantly affects the Loss Load Factor (LLF).

2. During the maximum of March and April, irrigation requirement reaches its peak and the conductors often carry twice the load they are designed to carry. As a result the conductors are overheated and resistance of the overloaded sections of the conductors increases substantially. This factor is hardly ever considered in calculating annual loss as depicted in (1), where R is taken as a constant.

It is essential that the above effects are incorporated in calculating line losses. An important feature in calculating line losses is to compute monthly losses rather than annual losses, especially in rural feeders where the load during premonsoon peak season is nearly 3 times that of the off-peak months. This clearly implies that line losses during peak seasons would be very high as compared with those during off peak seasons. In a similar way the voltage profiles during peak and off peak seasons are significantly different. In such a situation computing annual average values may be misleading. The monthly variations in losses, measured and computed using different methods are shown in Fig 4. It is evident that losses computed, using standard technique described in this section (equation 1) consistently gives lower values for the line losses. Equation 1 while calculating monthly loss, (after suitable modification), considers monthly peaks. B. kmkVA Method with modified LLF The Loss Load Factor essentially relates peak power loss with average energy losses that take place in the conductors. Since the load curves in the rural feeders are highly non typical, standard relationships as in (1) can hardly be used in computing line losses. Single- phasing and Load shedding make the relations more untenable. In the present context, as the values of Load despatched are recorded every half an hour by trivector meters, a more justifiable method may be used for calculating LLF. The proposed method for obtaining the LLF for any period of time T would be

LLF(new) =

INDIAN INSTITUTE OF TECHNOLOGY, KHARAGPUR 721302, DECEMBER 27-29, 2002 391

2

2

)(.

)(

TduringobservedPeakLoadinTperiodshourhalfofNo

hourhalfeachinLoadT

×

∑

--- (3)

The use of equation 3 for computing LLF significantly improves the results which tally reasonably well with measured values (fig 4) C. An Alternative LDF

A careful analysis of the loss formula (eqn 1) shows that there is one other empirical factor in use. The factor LDF in (1) is expected to account for the variation of implied uniform load distribution in the derivation.

Our investigations show that the assumption of Uniform Load Distribution (UDL) needs to be examined carefully. It is obvious that for any system given a load and LDF, it is possible to have many different load distributions, each one giving rise to a different loss. The range of such variation of loss (varying from 1/LDF to 1/LDF2) is quite significant. The new LDF that we propose is

∑

=kmkVA

LPLDFnew 2

2

where P is the total connected capacity in KVA and kVA2km is calculated using an equivalent conductor of uniform resistance

The results obtained with this formula are also presented in Fig 4.

D. Effect of temperature on resistance

As has been stated earlier, losses increase with temperature rise and during peak loads, correction for this becomes necessary. Extensive studies have indicated that on an average about 10-12% increase in the loss needs to be incorporated, to account for the change of resistance while the conductors carry large currents. Table1 provides the losses measured and computed for one year by various method briefly described above.

E. Loss Graph or Chart

It is useful if a loss chart is prepared, using Load Flow or any other technique listed above, so that for every possible load, the loss is calculated and stored in the computer. The grid details are already fed into the computer in the 11-kV station. The Load despatched from the station every half an hour is known. Fig 5, shows, the “Loss Curve”, in which the

loss is plotted against the sending end power, precomputed and stored in the computer. This is valid for a certain topology and connected kVA and needs to be updated as they change. In the absence of metering, it provides a useful and quick way to estimate losses in the 11-kV feeders. F. Voltage Profile An estimate of voltage regulation is essential for any system design and systems improvement exercise. The Electricity Boards prescribe a stringent 9% voltage regulation whereas, in reality, the voltage regulation at peak load often reaches 50%. Fig 6, shows the voltage recorded at 4 TCs of Bukkasagara feeder for the month of march 2002. It may be seen that the voltage sharply drops beyond the 17th transformer of the feeder which has about 70 transformers. With such voltages, it is almost impossible to run IP sets. Unless provided with suitable protective devices, motors of IP sets are often burnt. It is a common practice to have motors rewound so that they may be operated from low voltage supply. It is also a reasonably common practice to use converters so that IP sets may be operated out of single phase supply. Load curves (Fig 2) clearly reflect such practices.

III LT LOSSES

Data of LT feeders supplying the IP sets were collected from all transformer centres on the selected feeders. Ordinary energy meters have been installed on the pumpsets of a few selected feeders. The Line losses on the LT feeders are computed as below. The phase currents recorded by the meters every half an hour help to estimate the losses as

∑ ×++F

RIII cba1)( 222 (4)

where, R is the total resistance of each phase of the feeder and the summation is carried over the time horizon of interest F is a factor to account for the load distribution and is computed as follows

( )∑

∑×

=)( 2

2

lhpLmotorsallofhp

F (5)

The calculated values should match the measured values obtained as below : LTpis EEE =− (6)

where Es = Sending end energy from transformer secondary piE = Total energy consumed by the pumpsets which

forms the bulk of energy used as other loads are negligible

NATIONAL POWER SYSTEMS CONFERENCE, NPSC 2002 392

This method of line loss measurement worked in some cases and the results tallied with the computed values. In a number of cases it failed because the meters installed on the IP sets were bypassed. Fig 7 shows HT (11 kV) and estimated LT (430 V) losses in Bukkasagara feeder as they vary from month to month. Fig 8 shows the monthly variation of IP set consumptions in the four regions selected for the study. A. HT And LT Line Losses The computations and measurements carried out showed that contrary to popular belief, the HT losses in rural feeders with a large number of transformer centres, are on an average larger than average LT line losses. Fig 7 provides a graphical representation of this. The fact that currents are stepped up by a factor of 25 in a 11000/433 V transformer, does not necessarily imply that LT losses should be larger. Much depends on how many LT feeders radiate from the transformers and how many transformers are actually there, branching out the total HT current. Consider that there are n TCs, and on an average there are two feeders on the secondary of the transformers. Assuming uniform load distribution, the same resistance/length for HT and LT and an LT/HT length ratio of 4 we have

3

42

2531 22 RIR

nI

=×

×

25≈∴ n (7)

This would imply that the LT line losses will be approximately the same as HT losses (with the assumptions stated above) if there are 25 transformer centres (TC). When the number of TCs exceed 25, which normally is the case in rural feeders, HT losses could easily exceed LT losses.

IV CONCLUSIONS The study carried out is possibly the first of this kind where extensive measurements and computations have been carried out to a. separate HT and LT losses in rural feeders and estimate average pumpset consumption. b. check the correctness of formulae for computing HT and LT losses that have been and continue to be used by the Electricity Boards without establishing their validity. c. develop equations that make it possible to measure hourly losses in HT lines based on parameters measured from the 11-kV sending end substation. d. provide guidelines to Boards/Utilities to decide on a priority list for systems improvement based on the values of line losses and voltage regulation estimated from the 11 kV station. The studies help to establish that the 11-kV and 430 volt line losses are very large, and call for immediate rectification. V ACKNOWLEDGEMENT This project is supported by the Karnataka Power Transmission Corporation Limited.

INDIAN INSTITUTE OF TECHNOLOGY, KHARAGPUR 721302, DECEMBER 27-29, 2002 393

NATIONAL POWER SYSTEMS CONFERENCE, NPSC 2002 394