adaption of the current load model to consider residential ...fig. 3 shows the lumped p-consumption...
TRANSCRIPT
Adaption of the Current Load Model to Consider
Residential Customers Having Turned to LED
Lighting
Daniel-Leon Schultis
Albana Ilo
Institute of Energy Systems and Electrical Drives
TU Wien
Vienna, Austria
Abstract—Recent measurements show that the substitution of
General Incandescent Lamps by the high efficiency LEDs
changes the behavior of the modern residential customers.
They consume less active power and behave capacitive in the
evening. Actually, also the extended load models, which
consider the load composition and power factors of individual
load subcategories, describe an inductive behavior of the load
during the whole day. This paper presents an adapted load
model that considers residential customers having turned to
LED lighting. The current load models are adapted to reflect
the estimated load composition and the recent measurements.
The corresponding ZIP-coefficients are calculated and
provided for download in a public data repository. They change
during the day and reach high values when the reactive power
consumption of the residential customer changes the sign. In
this case, the capacitive and inductive power contributions of
the single load devices within the residential customer plant
compensate each other.
Index Terms—Distribution grids, load modelling, power flow
analysis, residential customers, ZIP model.
I. INTRODUCTION
The goal of this paper is to review the current state-of-the-art of the modelling of the reactive power part of residential loads for power flow analysis; and to derive an adapted load model that considers residential customers having turned to LED lighting. The distribution system engineer’s decisions concerning the integration of Distributed Energy Resources rely on the results of power flow studies. For performing distribution grid analysis, models are developed for all pertinent system components including Distributed Generation, distribution feeders, loads, etc. Much attention has been paid to model the active power part of the load devices connected to distribution grids by categorizing customers in residential, commercial, etc., and by developing ZIP models for different load categories and subcategories [1,2]. The reactive power part is derived from the active one using constant power factor and ZIP models [1,3]. During the day, the residential load have always shown a slightly inductive behavior [4].
Recent measurements on residential customer plants have shown that their reactive power behavior varies during the
day [5]. The development of high efficiency LEDs [6] and their increasing use is changing the reactive power behavior of the customers in the evening.
This paper presents an adapted load model that considers residential customers having turned to LED lighting. Section II describes the used methodology. The current state-of-art of load modelling of residential customers is discussed in Section III. Section IV describes the adapted load model, while its comparison with the current model is given in Section V. Conclusions are presented in Section VI.
II. METHODOLOGY
Fig. 1 shows the methodology used to derive the adapted load model of residential customers.
Figure 1. Methodology used to derive the adapted load model.
The estimated load composition and recent active (P) and reactive power (Q) measurements of modern residential customers are used to adapt the current load model.
For both load models, i.e. the current and the adapted one, the reactive power is calculated in two different ways: a) simplified with a fixed power factor of 0.95 inductive; and b) extended by considering the load composition and power factors of individual load subcategories.
III. CURRENT LOAD MODELS
A. Load Classification
All types of load devices within residential customer plants are divided into four general load categories [1,2]: Switch-Mode Power Supply (SMPS) loads, motors, resistive loads, and lighting, Table I. SMPS loads may possess active
(aPFC), passive (pPFC) or no Power Factor Correction (noPFC). Motors are split in directly connected Single-Phase Induction Motors (SPIM) and Single-phase Adjustable Speed Drives (SASD). SPIMs are sub-categorized in Inductor-Run (IR) motors with Constant or Quadratic Torque (CT or QT) and Capacitor-Run (CR) motors with CT. SASDs are split in low and high Power (lowP and highP) drives each with CT or QT. Lighting devices are sub-categorized in General Incandescent Lamps (GIL), Compact Fluorescent Lamps (CFL) and Light-Emitting Diodes (LED).
B. Load Profile
Fig. 2 shows the load profiles of the current load model for various load categories and sub-categories [2].
Figure 2. Load profiles of the current load model for various load
categories and sub-categories.
The lumped active power consumption ,
lpd
nom tP of the
customers for nominal grid voltage and time-point t is determined by
, ·lpd lpd peak
nom t tP f P (1)
where lpd
tf is the customers’ lumped load profile factor for
time-point t as shown in Fig. 2; and peak
P is their peak winter active power demand.
In many cases, the lumped reactive power consumption of residential customers is determined by using a fixed power factor of 0.95 inductive [3], as in
,
, , tan(arccos(0.95))·lpd fixPF lpd
nom t nom tQ P , (2)
where ,
,
lpd fixPF
nom tQ is the lumped reactive power consumption
with fixed power factor for nominal grid voltage and time-point t.
The load profile is decomposed into different load sub-categories, allowing to calculate a more accurate value for the
reactive power [2]. The active ,
i
nom tP and reactive power
,
i
nom tQ
consumption of load sub-category i for nominal grid voltage and time-point t are given by
, ·i i peak
nom t tP f P (3a)
and,
for ind. cos
for cap. cos
· ·tan ,
· ·tan ,
i
i
i peak i
i t
nom t i peak i
t
f PQ
f P
(3b)
where i
tf is the load profile factor of load sub-category i and
for time-point t. Normalizing (3) on peak winter demand yields
,
, ,/
NORM i i peak
nom t nom tP P P (4a)
and,
, ,/
NORM i i peak
nom t nom tQ Q P (4b)
Adding up the power consumption of each load category yields the lumped power consumption, as in
, ,
lpd i
nom t nom tiP P
(5a)
and, ,
lpd i
nom t nom tiQ Q
(5b)
TABLE I. MODEL DATA OF LOAD SUBCATEGORIES.
Category
Subcategory Active power ZIP
coefficients
Reactive power ZIP
coefficients
Power
factor Ref.
Share of category
Estimated
in [7]
Adapted
load
model
i name ,Z i
PC
,I i
PC ,P i
PC ,Z i
QC ,I i
QC ,P i
QC cosi
is [%]
,A is [%]
SMPS
1 aPFC 0.00 0.00 1.00 - - - 1 [3,7] 27.00 27.00
2 pPFC 0.00 0.00 1.00 0.45 -1.44 1.99 0.970 ind. [3,7] 27.00 27.00
3 noPFC 0.00 0.00 1.00 3.63 -9.88 7.25 0.994 cap. [3,7] 46.00 46.00
Moto
rs
Directly
connected
SPIM
4 IR_QT 0.10 0.10 0.80 1.40 -0.90 0.50 0.620 ind. [2,3,7] 35.00
29.48
5 IR_CT 0.63 -1.20 1.57 1.40 -0.90 0.50 0.620 ind. [1,7] 5.52
6 CR_CT 0.50 -0.62 1.12 1.54 -1.43 0.89 0.900 ind. [2,3,7] 25.00 25.00
SA
SD
lowP 7 QT -0.27 0.76 0.51 3.67 -10.31 7.64 0.990 cap. [1,7]
20.00 9.70
8 CT 0.40 -0.89 1.49 3.32 -10.50 8.18 0.990 cap. [1,7] 10.30
highP 9 QT -0.27 0.76 0.51 0.54 -1.65 2.11 0.896 ind. [1,7]
20.00 9.70
10 CT 0.40 -0.89 1.49 1.54 -3.95 3.41 0.896 ind. [1,7] 10.30
Resistive 11 - 1.00 0.00 0.00 - - - 1 [2,3] 100.00 100.00
Lighting
12 GIL 0.43 0.69 -0.12 - - - 1 [2,3] 0.00 0.00
13 CFL -0.01 0.96 0.05 -0.10 0.73 0.37 0.910 cap. [3,7] 100.00 0.00
14 LED 0.69 0.92 -0.61 1.84 -0.91 0.07 0.480 cap. [6,8] - 100.00
Fig. 3 shows the lumped P-consumption of the current load model for nominal grid voltage according to (1); and furthermore, its lumped Q-consumption with fixed power factor calculated according to (2), and with variable power factor calculated according to (5b).
Figure 3. Lumped P- and Q-consumption of the current simplified and
extended-load-model for nominal grid voltage.
In both cases, the current load model shows an inductive behavior for the entire time horizon of 24 hours. Both curves are generally similar. Therefore, for simplicity, the reactive power of the load is calculated using constant power factor.
C. ZIP Model
For ,
0i
nom tP and
,0
i
nom tQ , active and reactive power
voltage dependency of each load subcategory i is considered
by using a ZIP model [2] according to
2
, , ,
,· ·
i i Z i NORM I i NORM P i
t nom t P t P t PP P C U C U C (6a)
2
, , ,
,· ·
i i Z i NORM I i NORM P i
t nom t Q t Q t QQ Q C U C U C (6b)
with /NORM
t t nomU U U (6c)
where ,Z i
PC ,
,I i
PC ,
,P i
PC and
,Z i
QC ,
,I i
QC ,
,P i
QC are P- and Q-
ZIP coefficients of load subcategory i; i
tP and
i
tQ are the
actual P- and Q-consumption of load subcategory i for time-
point t; t
U is the actual- and nom
U the nominal grid voltage.
The ZIP coefficients fulfil
, , , , , ,
1,Z i I i P i Z i I i P i
P P P Q Q QC C C C C C i (7)
For ,
0i
nom tP and
,0
i
nom tQ , it is 0
i
tP and 0
i
tQ ,
respectively. The addition of the actual power consumptions
of each load subcategory gives the actual lumped power
consumption for time-point t, as in
lpd i
t tiP P
(8a)
lpd i
t tiQ Q
(8b)
For ,
0lpd
nom tP and
,0
lpd
nom tQ , a time-varying lumped ZIP
model is calculated by considering the ZIP coefficients of the
individual load subcategories and thei
tf , as in
2
, , , ,· ·
lpd lpd Z NORM I NORM P
t nom t P t t P t t P tP P C U C U C (9a)
2
, , , ,· ·
lpd lpd Z NORM I NORM P
t nom t Q t t Q t t Q tQ Q C U C U C (9b)
where ,
Z
P tC ,
,
I
P tC ,
,
P
P tC and
,
Z
Q tC ,
,
I
Q tC ,
,
P
Q tC are lumped P-
and Q-ZIP coefficients for time-point t that fulfil
, , , , , ,
1,Z I P Z I P
P t P t P t Q t Q t Q tC C C C C C t (10)
The lumped ZIP coefficients are determined by
, , ,
, , ,·
Z NORM i Z i NORM i
P t nom t P nom ti iC P C P
(11a)
, , ,
, , ,·
I NORM i I i NORM i
P t nom t P nom ti iC P C P
(11b)
, , ,
, , ,·
P NORM i P i NORM i
P t nom t P nom ti iC P C P
(11c)
, , ,
, , ,·
Z NORM i Z i NORM i
Q t nom t Q nom ti iC Q C Q
(11d)
, , ,
, , ,·
I NORM i I i NORM i
Q t nom t Q nom ti iC Q C Q
(11e)
and , , ,
, , ,·
P NORM i P i NORM i
Q t nom t Q nom ti iC Q C Q
(11f)
For ,
0lpd
nom tP and
,0
lpd
nom tQ , it is 0
lpd
tP and 0
lpd
tQ ,
respectively. Fig. 4 shows the lumped ZIP coefficients of the
current extended-load-model according to (11).
Figure 4. Lumped ZIP coefficients of the current extended-load-model.
Between 0:00 and 18:00, P-consumption of the current
extended-load-model mainly behaves as a combination of
two almost equal load parts with constant Z and P. During
the remaining time, also the constant I load part becomes
perceptible. The Q-consumption behaves like a ZIP-load
during the whole time horizon. The ZIP coefficient values
lie within the interval (-1.5, 1.5).
IV. ADAPTED EXTENDED-LOAD-MODEL
A. Estimated Load Composition
In this section are used (1) to (11) with the additional superscripts “C” for the current and “A” for the adapted load
model. Reference [7] estimates the share i
s of each load
subcategory of the corresponding category for residential customers in 2020, Table I. The SMPS load with noPFC have an estimated share of 46%, while that with pPFC and aPFC respectively 27%. In the motor load category, the directly connected IR SPIM has an estimated share of 35%; the directly connected CR SPIM 25%, the low and high power SASD respectively 20%. In the lighting sector is estimated a share of 100% CFL. The estimated shares of IR SPIM, low and high power SASD are allocated to the corresponding load subcategories using
,4 ,4 ,4 ,50.35·
A C C C
r r rs P P P (12a)
,5 ,5 ,4 ,50.35·
A C C C
r r rs P P P (12b)
,7 ,9 ,4 ,4 ,5 ,60.2·
A A C C C C
r r r rs s P P P P (12c)
,8 ,10 ,5 ,6 ,4 ,5 ,60.2·
A A C C C C C
r r r r rs s P P P P P (12d)
where ,A i
s is for the adapted load model the share of load
subcategory i of the corresponding category, and ,C i
rP is for
the current load model the installed rating of load subcategory i, estimated as in
, ,
,max
C i C i
r nom tt
P P (13)
For the lighting load category is assumed a share of 100%
LED instead of CFL, since LED has not been considered in
[7], but is a more efficient technology than CFL with lower
operational costs [9], much higher average lifetime [10] and
rapidly increasing market share [11]. Compared to CFL,
LED has the advantage of no environmental pollution since
it is free of toxic substances such as mercury [10], and it
possesses improvements concerning light intensity,
brightness, color control, and reliability [12].
The shares,A is of all load subcategories are shown in
Table I. The load profile factors of the adapted load model’s
subcategories are calculated as in
ft
A,i = sA,i · ∑ f
t
C,i3i=1 , ∀ i ∈ (1,3), (14a)
ft
A,i = sA,i · ∑ f
t
C,i10i=4 , ∀ i ∈ (4,10), (14b)
,11 ,11 ,11
·A A C
t tf s f (14c)
In the residential lighting sector, GIL and CFL are
currently in the process of being replaced by LED. The
efficiency increase drastically lowers the number of lamps
required to supply a certain light demand. However, due to
rebound effect [13] it is assumed that the modern customer
increases his light demand by 30%, as considered in
,12 ,13
0,A A
t tf f t (15a)
,14 ,12 ,131.3· · ·
A C GIL LED C CFL LED
t t tf f f (15b)
where 2.6%GIL
, 11%CFL
, and 20%LED
are
efficiency of GIL, CFL and LED lighting, respectively [9].
B. Recent measurements
Recent measurements have shown a capacitive behavior
of a residential customer plant with LED and CFL lighting
and an annual energy consumption of 3000 kWh during
periods of low active power consumption, i.e. at nighttime
[5].
C. Load Profile
The load profiles of the adapted load model for various
load categories and sub-categories are shown in Fig. 5. The
corresponding load profile factors are calculated according
to (14) and (15).
Figure 5. Load profiles of the adapted load model for various load
categories and sub-categories.
The resulting lumped P-consumption of the adapted load
model for nominal grid voltage according to (1) is shown in
Fig. 6; and furthermore, there are shown its lumped Q-
consumption with fixed power factor according to (2), and
with variable power factor according to (5b).
Figure 6. Lumped P- and Q-consumption of the adapted simplified and
extended-load-model for nominal grid voltage
The adapted extended-load-model shows a capacitive
behavior between 21h and 24h, and an inductive behavior
for the remaining time horizon.
D. ZIP Model
Fig. 7 shows the lumped ZIP coefficients according to
(11), and the normalized lumped reactive power
consumption for the nominal grid voltage of the adapted
extended-load-model.
Figure 7. Lumped ZIP coefficients and reactive power consumption for
nominal grid voltage of the adapted extended-load-model.
The ZIP coefficients, especially ,
Z
Q tC and
,
I
Q tC , show
high peaks when reactive power consumption turns from
inductive to capacitive, and vica verse. This is because the
denominator in (11d)-(11f) is almost zero, but not the
reactive power consumption of single load devices. The
capacitive and inductive power contributions of the single
load devices within the residential customer plant
compensate each other.
The data of both, the current and the adapted extended-
load-model of residential customer plants (,
lpd
nom tP ,
,
lpd
nom tQ ,
,
Z
P tC
, ,
I
P tC ,
,
P
P tC ,
,
Z
Q tC ,
,
I
Q tC ,
,
P
Q tC ) are provided for download in a
public data repository [14]; the data is sampled into one
minute time-steps.
V. COMPARISON OF LOAD MODELS
Fig. 8 shows the lumped P- and Q-consumption of the
current and adapted extended-load-model for nominal grid
voltage.
Figure 8. Lumped P- and Q-consumption of the current and adapted
extended-load-model.
The substitution of GILs by high efficient LEDs reduces
the consumption of the active power in the evening up to
17.35% of peak
P . Between 0:00 and 18:00, the inductive Q-
consumption is reduced by 7.68% of peak
P , in average. From
18:00, the inductive Q-consumption steadily decreases, and
even changes to a capacitive one at about 20:45. The
maximum capacitive Q-consumption is about 4.17% of peak
P at 22:00.
VI. CONCLUSIONS
The calculation of the reactive power consumption of the
modern residential customers using the simplified-load-
model, i.e. fix power factor of 0.95 inductive, is inaccurate
compared to the extended-load-model, i.e. considering the
composition and power factors of individual load
subcategories. The use of the modern equipment, in
particular LEDs, reduces the active power consumption and
modifies significantly the reactive power behavior of the
residential customers. Also the inductive power
consumption is reduced, and it even changes to a capacitive
one in the evening. The adapted extended-load-model
considers the new behavior of modern residential customers.
The corresponding ZIP-coefficients are calculated; they
change during the day and reach high values when the
reactive power consumption of the residential customer
changes the sign. In this case, the capacitive and inductive
power contributions of the single load devices within the
residential customer plant compensate each other.
REFERENCES
[1] CIGRE Working Group C4.605, “Modelling and Aggregation of Loads in Flexible Power Networks,” Feb. 2014.
[2] A. J. Collin, I. Hernando-Gil, J. L. Acosta and S. Z. Djokic, "An 11
kV steady state residential aggregate load model. Part 1: Aggregation methodology," 2011 IEEE Trondheim PowerTech, Trondheim, 2011,
pp. 1-8.
[3] A. J. Collin, G. Tsagarakis, A. E. Kiprakis and S. McLaughlin, "Development of Low-Voltage Load Models for the Residential Load
Sector," IEEE Trans. Power Syst., vol. 29, no. 5, pp. 2180-2188, Sept. 2014.
[4] W. Peng, Y. Baghzouz and S. Haddad, "Local load power factor
correction by grid-interactive PV inverters," 2013 IEEE Grenoble Conference, Grenoble, 2013, pp. 1-6.
[5] C. Groiss, P. Zehetbauer, R. Schwalbe and C. Schirmer, "Reactive
power flow over system boundaries in the distribution grid," 25th International Conference on Electricity Distribution, Madrid, Spain,
2019, pp. 1-5.
[6] X. Xu, A. Collin, S. Z. Djokic, R. Langella, A. Testa and J. Drapela, "Experimental evaluation and classification of LED lamps for typical
residential applications," 2017 IEEE PES Innovative Smart Grid
Technologies Conference Europe, Torino, 2017, pp. 1-6. [7] C. Cresswell, “Steady State Load Models for Power System
Analysis,” Ph.D. dissertation, University of Edinburgh, May 2009.
[8] A. Arif, Z. Wang, J. Wang, B. Mather, H. Bashualdo and D. Zhao, "Load Modeling—A Review," IEEE Trans. Smart Grid, vol. 9, no. 6,
pp. 5986-5999, Nov. 2018.
[9] N. Khan and N. Abas, “Comparative study of energy saving light sources,” Renewable and Sustainable Energy Reviews, vol. 15, no. 1,
pp. 296-09, Jan. 2011.
[10] S. Di Mauro and A. Raciti, "Analysis and comparison of CFLs and LED lamps," 2014 AEIT Annual Conference - From Research to
Industry: The Need for a More Effective Technology Transfer (AEIT),
Trieste, 2014, pp. 1-6. [11] Goldman Sachs. n.d. "Estimated LED penetration of the global
lighting market from 2010 to 2020," Statista, Accessed June 19, 2019.
Available from https://www.statista.com/statistics/246030/estimated-led-penetration-of-the-global-lighting-market/.
[12] C.-F. Huang, Y.-F. Su, S.-Y. Yang, C.-L. Hsu, N.-C. Chen and K.-N.
Chiang, "Quantum efficiency investigation at high current density of Ultra-High-Brightness LEDs," 13th InterSociety Conference on
Thermal and Thermomechanical Phenomena in Electronic Systems,
San Diego, CA, 2012, pp. 303-307. [13] L. A. Greening, D. L. Greene and C. Difiglio, "Energy efficiency and
consumption – the rebound effect – a survey," Energy Policy, vol. 28,
no. 6-7, pp. 389-401, June 2000. [14] D.-L. Schultis, “Daily load profiles and ZIP models of current and new
residential customers,” Mendeley Data, v1, 2019,
http://dx.doi.org/10.17632/7gp7dpvw6b.1.