Distribution system voltage regulation

and var compensation for different static

load models
N. Mithulananthan, M. M. A. Salama, C. A. Canizares and J. Reeve
Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, Canada
E-mail: m.salama@ece.uwaterloo.ca
Abstract Voltage regulation computations for distribution systems are strongly dependent on power flow solutions.
The classical constant power load model is typically used in power flow studies of transmission or distribution
systems; however, the actual load of a distribution system cannot just be modeled using constant power models,
requiring the use of constant current, constant impedance, exponential or a mixture of all these load models to
accurately represent the load. This paper presents a study of voltage regulation of a distribution system using different
static load models. The effect of shunt capacitor compensation is also studied and illustrated in this paper for
systems with different static load models.
Keywords distribution systems; power flow; static load models; voltage regulation

Introduction
Voltage regulation is an important subject in electrical distribution engineering.
It is the utilities responsibility to keep the customer voltage within specified
tolerances. The performance of a distribution system and quality of the service
provided are not only measured in terms of frequency of interruption but in
the maintenance of satisfactory voltage levels at the customers premises.
According to Gonen,1 a high steady-state voltage can reduce light bulb life and
reduce the life of electronic devices. On the other hand, a low steady-state
voltage leads to low illumination levels, shirking of television pictures, slow
heating of heating devices, motor starting problems, and overheating in motors.
However, most equipment and appliances operate satisfactorily over some
reasonable range of voltages; hence, certain tolerances are allowable at the
customers end. Thus, it is common practice among utilities to stay within
preferred voltage levels and ranges of variations for satisfactory operation of
apparatus as set by various standards such as ANSI (American National
Standard Institution). For example, power acceptability curves given by IEEE
(IEEE orange book, IEEE standard 446) and FIPS (United States Federal
Information Processing Standard) indicate that steady-state voltage regulations
should be within +6% to 13% for satisfactory operation of various electrical
devices.2
Voltage regulation calculations depend on the power flow solutions of a
system. Most of the electrical loads of a power system are connected to lowvoltage/medium-voltage distribution systems rather than to a high-voltage
transmission system. The loads connected to the distribution system are certainly voltage dependent; thus, these types of load characteristics should be
International Journal of Electrical Engineering Education 37/4

Voltage regulation for different static load models

385

considered in load flow studies to get accurate results and to avoid costly
errors in the analysis of the system. For example, in voltage regulation improvement studies, possible under- or over-compensation can be avoided if more
accurate results of load flow solutions are available, as demonstrated in this
paper. However, most conventional load flows use a constant power load
model, which assumes that active and reactive powers are independent of
voltage changes. In reality, constant power load models are highly questionable
in distribution systems, as most nodes are not voltage controlled; therefore, it
is very important to consider better load models in these types of load flow
problems.
In this paper, distribution system voltage regulation and the effect of shunt
capacitor compensation on this regulation for different static load models are
studied. The paper is organized as follows: the next section briefly reviews
different types of static load models. Power flow equations and a MATLABbased solution technique, as well as the definition of voltage regulation and
methods to improve it are then discussed. Details of the distribution test system
used in this paper follow, together with a discussion of some interesting simulation results. Finally, major contributions of this paper are highlighted.
Static load models
In power flow studies, the common practice is to represent the composite load
characteristic as seen from power delivery points. In transmission system load
flows, loads can be represented by using constant power load models, as
voltages are typically regulated by various control devices at the delivery points.
In distribution systems, voltages vary widely along system feeders as there are
fewer voltage control devices; therefore, the V I characteristics of load are
more important in distribution system load flow studies.35
Load models are traditionally classified into two broad categories: static
models and dynamic models. Dynamic load models are not important in load
flow studies. Static load models, on the other hand, are relevant to load flow
studies as these express active and reactive steady state powers as functions of
the bus voltages (at a given fixed frequency). These are typically categorized as
follows:6
Constant impedance load model (constant Z). A static load model where the
power varies with the square of the voltage magnitude. It is also referred to as
constant admittance load model.
Constant current load model (constant I). A static load model where the power
varies directly with voltage magnitude.
Constant power load model (constant P). A static load model where the power
does not vary with changes in voltage magnitude. It is also known as constant
MVA load model.
International Journal of Electrical Engineering Education 37/4

Mithulananthan et al.

386

Exponential load model. A static load model that represents the power
relationship to voltage as an exponential equation in the following way:

AB

AB

V a
V b
,
Q=Q
P=P
0 V
0 V
0
0
where P and Q stand for the real and reactive powers consumed at a reference
0
0
voltage V . The exponents a and b depend on the type of load that is being
0
represented, e.g., for constant power load models a=b=0, for constant current
load models a=b=1 and for constant impedance load models a=b=2.
Table 1 shows typical values of a and b for several types of loads encountered
in power systems;6 it is interesting to note that none of these loads has a zero
exponent.
Polynomial load model. A static load model that represents the power-voltage
relationship as a polynomial equation of voltage magnitude. It is usually
referred to as the ZIP model, as it is made up of three different load models:
constant impedance (Z), constant current (I ) and constant power (P). The real
and reactive power characteristics of the ZIP load model are given by

C AB AB D
C AB AB D

V 2
V
+c
P=P a
+b
P
0 P V
P V
0
0
V 2
V
Q=Q a
+c
+b
0 Q V
Q
Q V
0
0
where a +b +c =a +b +c =1, and P and Q are the real and reactive
P
P
P
Q
Q
Q
0
0
power consumed at a reference voltage V . In this paper, three types of static
0
load models, i.e., constant power, constant current and constant impedance,
are considered to demonstrate their effect on voltage regulation calculations in
distribution systems. The studies presented in this paper can be readily extended
to other load models as well.
Power flow equations and voltage regulation
Power flow studies are of great importance in planning and designing future
expansions of power systems. The main information obtained from power flow
TABLE 1. A sample of fractional load exponents
Load component

Incandescent lamps
Room air conditioner
Furnace fan
Battery charge
Compact fluorescent lamps
Fluorescent lamps

1.54
0.50
0.08
2.59
0.951.03
2.07

2.50
1.60
4.06
0.310.46
3.21

International Journal of Electrical Engineering Education 37/4

Voltage regulation for different static load models

387

studies is the magnitude and angle of the phasor voltage at each node, and the
real and reactive power flowing in each line. With this information, the voltage
regulation of any feeder in the system can be easily computed.
Power flow
In general, if a node in a power system is considered (Fig. 1), the following
equations can be readily written by considering the real and reactive power
balance:
P =P P ,
Q =Q Q
k
gk
Lk
k
gk
Lk
where P and Q are the real and reactive power generated at node k; P
g
gk
L
and reactive power loads at node k, which could bek
and Q kare the real
Lk
constant or a function of the bus voltage magnitude; and P and Q are the
k
k
real and reactive power injected into the system,7 i.e.,
N
P =V Y V cos(d d h )
k
k
kl l
k
l
kl
l=1
N
Q =V Y V sin(d d h )
k
k
kl l
k
l
kl
l=1
where V and V are bus voltage magnitudes at nodes k and l, together with
k
l
their respective phase angles d and d ; Y and h are the magnitude and angle
k
l kl
kl
of the kl entry in the Y-bus matrix; and N is the total number of nodes in
the system.
A set of non-linear equations can be established for the system by considering
the power balance and injected power in terms of system parameters. Thus,
these equations can then be solved by using any non-linear equation solving
technique (e.g. NewtonRaphson). In this paper, these equations are solved,
for different static load models, using a MATLAB routine (fsolve) based on
least square optimization techniques; the main steps of this program are as
follows:
1. Read the data and form the Y-bus matrix of the distribution system.
Pgk, Qgk

Pk,Qk
To other Nodes

Load PLk, QLk

Fig. 1 Power balance at node k.

International Journal of Electrical Engineering Education 37/4

388

Mithulananthan et al.

2. Generate the symbolic set of nonlinear power flow equations by assuming

that the substation node is a PV bus while all other nodes are load buses.
3. Assume an initial guess (all voltages equal to the substation voltage and
all angles equal to zero).
4. Solve the set of nonlinear equations with the fsolve command in
MATLAB.
Voltage regulation
Once a load flow solution is obtained, the voltage regulation of any feeder can
be calculated as follows:1
|V ||V |
r 1100
V Reg.= s
|V |
r
where V is sending-end voltage and V is receiving-end voltage. In distribution
s
r
systems, this regulation may be typically improved by using one or more of
the following techniques:
(a) Increasing primary voltage.
( b) Activating voltage regulating equipment at the substations bus such as
capacitors or LTCs.
(c) Balancing of loads on primary feeders.
(d) Increased size of feeder conductor.
(e) Transferring loads to new feeders.
(f ) Installing new substations and primary feeders.
(g) Installing shunt capacitors or SVCs on primary feeders.
The most economical way of improving voltage profiles along a feeder, and
thus voltage regulation and overall system performance, is by using shunt
capacitors. Hence, the effect of shunt capacitors on voltage regulation with
different static load models is studied in this paper.
Simulation results
The topology of many distribution systems is like a tree with several lateral
and sub-lateral branches; the root of the tree is the feeding node or feeding
substation. This kind of general structure of distribution systems (a radial type)
has been fully exploited to develop some efficient load flow methods in the
past.8,9 In the present work, a similar type of system is used to analyze the
effect of various voltage dependent load models in distribution system voltage
regulation.
The distribution system used in this paper is depicted in Fig. 2. It is a
balanced three-phase radial system that consists of 30 nodes and 29 segments;
there are three sub-feeders and one main feeder. It is assumed that all the loads
are fed from the substation located at node 1. The loads belonging to one
segment are assumed to be placed at the end of each segment. The system data
was extracted from Ref. 10.
International Journal of Electrical Engineering Education 37/4

Voltage regulation for different static load models

389
16
15

30

14

29
SUB-1

SUB-3

13

1
SUB STATION

3
4
5
MAIN FEEDER

28

17

18

19

10

20

12

11

22

21

23

24

SUB-2
27

26

25

Fig. 2 Single-line diagram of the test distribution system.

The studies presented in this paper consist of two parts. The first part consists
of the load flow solutions and voltage regulation computations for the test
system with different static load models. The second part focuses on var
compensation by shunt capacitors with different static load models and its
effect on the system voltage regulation.
Voltage regulation
Initially, load flow solutions for the test distribution system with (Fig. 2)
constant power load models were obtained by using the developed MATLAB
program, and these results were validated by comparing them to those obtained
with a standard power flow program (PFLOW11).
Tables 2 and 3 compare the load flow solutions obtained for different static
TABLE 2 Voltage magnitude along the main feeder: classical load (cons P) vs. constant
current (cons. I)
Voltage (p.u.)
Node

Classical (cons. P)

Cons. I

Difference
(p.u.)

1
2
3
4
5
6
7
8
9
10
11
12

1.0000
0.9824
0.9670
0.9460
0.9279
0.9109
0.9008
0.8958
0.8903
0.8880
0.8866
0.8862

1.0000
0.9845
0.9711
0.9528
0.9372
0.9226
0.9136
0.9091
0.9043
0.9023
0.9010
0.9007

0.0000
0.0021
0.0041
0.0068
0.0093
0.0117
0.0128
0.0133
0.0140
0.0143
0.0144
0.0145

International Journal of Electrical Engineering Education 37/4

Mithulananthan et al.

390

TABLE 3 Voltage magnitude along the main feeder: classical load (cons P) vs. constant
impedance (cons. Z)
Voltage (p.u.)
Node

Classical (cons. P)

Cons. Z

Difference
(p.u.)

1
2
3
4
5
6
7
8
9
10
11
12

1.0000
0.9824
0.9670
0.9460
0.9279
0.9109
0.9008
0.8958
0.8903
0.8880
0.8866
0.8862

1.0000
0.9859
0.9737
0.9573
0.9434
0.9303
0.9220
0.9179
0.9135
0.9117
0.9105
0.9102

0.0000
0.0035
0.0067
0.0113
0.0155
0.0194
0.0212
0.0221
0.0232
0.0237
0.0239
0.0240

load models against solutions obtained using the classical (constant power)
load model. The voltage profiles along the main feeder for these cases are
shown on Fig. 3. As can be seen in Tables 2 and 3, different load models result
in different load flow solutions, as expected; the difference between the solutions
increases as one moves away from the substation or feeding point. These
differences in voltage may lead to significantly different design approaches in
1

Constant PQ
Constant Current
Constant Impedance
0.98

Voltage (pu.)

0.96

0.94

0.92

0.9

0.88

10

11

12

Nodes

Fig. 3 Voltage profiles along the main feeder: classical load (cons. P), constant current,
and constant impedance load models.
International Journal of Electrical Engineering Education 37/4

Voltage regulation for different static load models

391

a distribution system. Hence, proper load models should be considered in order

to obtain more accurate results.
Table 4 shows the distribution system voltage regulation calculated with
different static load models. The voltage regulation computed for constant
power load models is the highest of all loads models, as one would expect,
since power demand does not change with voltage variations, whereas for the
other models the power demand decreases as the voltage goes down.
Compensation
In order to see the effect of var compensation (shunt capacitors) on voltage
regulation, various simulations were performed for different sizes of shunt
capacitors. Shunt capacitors are installed on the main feeder at two-thirds of
its length from the substation (Golden Rule). Voltages and thus voltage regulation of the main feeder are calculated for the different load models and for
different shunt capacitor sizes. The results obtained from these calculations are
shown in Tables 5, 6, and 7, and the corresponding voltage profiles of the main
feeder are depicted in Figs. 4, 5 and 6. Figure 7 summarizes these results
by showing the voltage regulation of the main feeder for different shunt
capacitor sizes.
TABLE 4 Voltage regulation of the main feeder with diVerent load models
Load model

Voltage regulation (%)

Constant power
Constant current
Constant impedance

12.84
11.02
9.87

TABLE 5 Voltage magnitudes for diVerent compensation levels (cons. P)

Voltage (p.u.)
Node

5 Mvar

4 Mvar

3 Mvar

2 Mvar

1 Mvar

1
2
3
4
5
6
7
8
9
10
11
12

1.0000
0.9877
0.9772
0.9615
0.9484
0.9364
0.9267
0.9218
0.9164
0.9142
0.9129
0.9125

1.0000
0.9867
0.9754
0.9588
0.9447
0.9318
0.9220
0.9170
0.9117
0.9095
0.9081
0.9077

1.0000
0.9857
0.9734
0.9558
0.9408
0.9267
0.9170
0.9120
0.9066
0.9044
0.9030
0.9027

1.0000
0.9847
0.9714
0.9527
0.9367
0.9217
0.9118
0.9068
0.9014
0.8992
0.8978
0.8974

1.0000
0.9836
0.9692
0.9494
0.9324
0.9164
0.9064
0.9014
0.8959
0.8937
0.8923
0.8919

International Journal of Electrical Engineering Education 37/4

Mithulananthan et al.

392

TABLE 6 Voltage magnitudes for diVerent compensation levels (cons. I)

Voltage (p.u.)
Node

5 Mvar

4 Mvar

3 Mvar

2 Mvar

1 Mvar

1
2
3
4
5
6
7
8
9
10
11
12

1.0000
0.9893
0.9804
0.9668
0.9557
0.9456
0.9366
0.9321
0.9273
0.9253
0.9240
0.9237

1.0000
0.9885
0.9787
0.9644
0.9524
0.9414
0.9324
0.9279
0.9231
0.9211
0.9198
0.9195

1.0000
0.9876
0.9770
0.9617
0.9489
0.9370
0.9280
0.9235
0.9187
0.9167
0.9154
0.9151

1.0000
0.9866
0.9751
0.9589
0.9451
0.9324
0.9234
0.9189
0.9141
0.9121
0.9108
0.9104

1.0000
0.9856
0.9731
0.9560
0.9413
0.9276
0.9186
0.9141
0.9093
0.9073
0.9060
0.9056

TABLE 7 Voltage magnitudes for diVerent compensation levels (cons. Z)

Voltage (p.u.)
Node

5 Mvar

4 Mvar

3 Mvar

2 Mvar

1 Mvar

1
2
3
4
5
6
7
8
9
10
11
12

1.0000
0.9905
0.9825
0.9705
0.9606
0.9518
0.9434
0.9392
0.9347
0.9328
0.9317
0.9313

1.0000
0.9897
0.9810
0.9682
0.9575
0.9489
0.9395
0.9353
0.9308
0.9290
0.9278
0.9275

1.0000
0.9888
0.9793
0.9657
0.9543
0.9438
0.9354
0.9313
0.9268
0.9250
0.9238
0.9235

1.0000
0.9879
0.9776
0.9631
0.9508
0.9395
0.9312
0.9270
0.9226
0.9207
0.9196
0.9192

1.0000
0.9870
0.9757
0.9603
0.9472
0.9350
0.9267
0.9226
0.9181
0.9163
0.9152
0.9149

Observe that the system with constant power load model presents the highest
voltage regulation values followed by the constant current load model and
then the constant impedance load model, as expected. The voltage regulation
for all static load models appears to vary approximately linearly with the size
of the shunt capacitor; however, the slopes change with the load model, with
the effectiveness of the capacitor compensation being highest for the system
with constant power load models.
To illustrate the importance of selecting the proper load models in the
reactive power compensation problem, consider the following sample case: to
International Journal of Electrical Engineering Education 37/4

Voltage regulation for different static load models

393

5 MVar
4 MVar
3 MVar
2 MVar
1 MVar

0.98

Voltage (pu.)

0.96

0.94

0.92

0.9

10

11

12

Nodes

Fig. 4 Voltage profiles along the main feeder with constant power load models for
diVerent compensation levels.
1

5 MVar
4 MVar
3 MVar
2 MVar
1 MVar

0.99

0.98

0.97

Voltage (pu.)

0.96

0.95

0.94

0.93

0.92

0.91

0.9

10

11

12

Node

Fig. 5 Voltage profiles along the main feeder with constant current load models for
diVerent compensation levels.

achieve a voltage regulation of 10%, a 5 Mvar capacitor would be needed in

the case of constant power loads models, whereas only a 2 Mvar capacitor
would be required if the loads are modeled as constant current and no compensation would be required if constant impedance load models are used. Hence,
by properly modeling the load on a distribution system, over-design and the
associated extra costs can be avoided.
International Journal of Electrical Engineering Education 37/4

Mithulananthan et al.

394
1

5 MVar
4 MVar
3 MVar
2 MVar
1 MVar

0.99

0.98

0.97

Voltage (pu.)

0.96

0.95

0.94

0.93

0.92

0.91

0.9

10

11

12

Node

Fig. 6 Voltage profiles along the main feeder with constant impedance load models for
diVerent compensation levels.
13

Const. PQ
Const. Curr.
Const. Imp.
12

% Voltage Regulation

11

10

0.5

1.5

2
2.5
3
Capacitor Size (MVar)

3.5

4.5

Fig. 7 Voltage regulation for diVerent capacitor compensation levels.

Conclusions
This paper demonstrates how voltage regulation calculations in distribution
systems vary with different static load models. Systems with constant power
load models present high voltage drops along a feeder, and thus high voltage
regulation, followed by systems with constant current load models and then
by systems with constant impedance load models. Hence, it is important to
International Journal of Electrical Engineering Education 37/4

Voltage regulation for different static load models

395

choose the load models more suitable for a given system in order to obtain
accurate results.
Shunt capacitor compensation improves system voltage regulation for all
types of loads. However, different sizes of shunt capacitors are required for
different types of static load models to achieve proper voltage regulations. By
properly selecting the load models, efficient designs can be obtained for var
compensation in distribution systems.
References
1
2
3
4
5
6
7
8
9
10
11

T. Gonen, Electric Power Distribution System Engineering (McGraw Hill, New York, 1986).
G. T. Heydt, Electric Power Quality, 2nd edn (Stars in a Circle Publications, West LaFayette,
IN, 1991).
M. E. El-Hawary and L. G. Dias, Incorporation of load models in load-flow studies: form of
models effects, IEE Proc. C, 134(1) (1987), 2730.
P. S. R. Murty, Load modelling for power flow solution, J. Inst. Eng. (India), Part EL , 58(3)
(1977) 162165.
M. H. Haque, Load flow solution of distribution systems with voltage dependent load
models, Int. J. Electric Power System Res., 36 (1996), 151156.
T. Van Cutsem and C. Vournas, Voltage Stability of Electric Power Systems, Power Electronics
and Power System Series, Kluwer, 1998.
J. D. Glover and M. Sarma, Power System Analysis and Design, 2nd edn (PWS Publishing
Company, Boston, 1993).
C. G. Renato, New method for the analysis of distribution networks, IEEE T rans. Power
Delivery, 5(1) (1990), 391396.
D. Das, H. S. Nagi, and D. P. Kothari, Novel methods for solving radial distribution
networks, IEE Proc. Generation T ransmission and Distribution, 141(4) (1994).
M. M. A. Salama and A. Y. Chikhani, A simplified network approach to the var control
problem for radial distribution systems, IEEE T rans. Power Delivery, 8(3) (1993), 15291535.
C. A. Can izares et al., PFLOW: continuation and direct methods to locate fold bifurcations
in AC/DC/FACTS power systems, University of Waterloo, August 1998. Available at
http://www.power.uwaterloo.ca.

International Journal of Electrical Engineering Education 37/4