Modern Applied Science

May, 2009

An Optimal Load Shedding Approach for Distribution Networks

with DGs Considering Capacity Deficiency
Modelling of Bulked Power Supply
Ahmad Reza Malekpour
Department of Electrical & Electronic Eng.
School of Engineering, Shiraz University, Shiraz, I.R. of Iran
Tel: 98-711-230-3081

E-mail: [email protected]

Ali Reza Seifi (Corresponding author)

Department of Electrical & Electronic Eng.
School of Engineering, Shiraz University, Shiraz, I.R. of Iran
Tel: 98-711-230-3081

E-mail: [email protected]

Abstract
This paper discusses a genetic algorithm (GA) based optimal load shedding that can apply for electrical distribution
networks with and without dispersed generators (DG). Also, the proposed method has the ability for considering
constant and variable capacity deficiency caused by unscheduled outages in the bulked generation and transmission
system of bulked power supply. The genetic algorithm (GA) is employed to search for the optimal load shedding
strategy in distribution networks considering DGs in two cases of constant and variable modelling of bulked power
supply of distribution networks. Electrical power distribution systems have a radial network and unidirectional power
flows. With the advent of dispersed generations, the electrical distribution system has a locally looped network and
bidirectional power flows. Therefore, installed DG in the electrical distribution systems can cause operational problems
and impact on existing operational schemes. Introduction of DGs in electrical distribution systems has introduced many
new issues in operational and planning level. Load shedding as one of operational issue has no exempt. The objective is
to minimize the sum of curtailed load and also system losses within the frame-work of system operational and security
constraints. The proposed method is tested on a radial distribution system with 33 load points for more practical
applications.
Keywords: DG, Load shedding, Optimization, Capacity Deficiency Modelling
1. INTRODUCTION
The phenomenal growth in load demand both in developing and developed countries has emerged as a potential
challenge to the power system planners and operators. Projections show that the growth in load demand is always going
to be ahead of the growth in generation. Electric supply failures can have serious monetary impacts on the system
customers. During an emergency situation, system operators are required to make load shedding decisions based on
system security concerns, such as voltage, current, power and frequency constraints, to alleviate constraints and
maintain system stability.
The distribution systems are the final link of the interconnection between power systems and the consumers. If there is
the necessity to alleviate the load in order to guarantee the safety restrictions, usually the curtailment occurs in the
distribution system. An important aspect in the operation of these systems is that the load curtailment originates from
failures in the generation and transmission system. Distributed generation (DG) is normally defined as small generation
units (<10 MW) installed in distribution systems (Ackermann,T.; Andersson,G.; Soder, L. 2001). Distributed generation
is expected to play an increasing role in emerging electric power systems. Studies have predicted that DG will be a
significant percentage of all new generation going on line. It is predicted that they would have about 20% of new
generations being installed (W. El-Khattam, M. M. A. Salama, 2004). They use different types of resources and
technologies to serve energy to power systems. DG applications result in positive and negative side effects for both
utility and customers (Daly, P.A.; Morrison, J)( Dugan, R.C.; McDermott, T. E.). Different techniques have been
143

Vol. 3, No. 5

Modern Applied Science

proposed to solve the load shedding problem in distribution network. Aoki et al (K. Aoki, N. Nara, M. Itoh and T. Satoh
and H. Kuwabara, 1989)describes a load curtailment procedure as part of a service restoration algorithm considering a
violation vector with current capacity and voltage drop violations as components. A quantity called effective length of
remaining violations is defined in (K. Aoki, N. Nara, M. Itoh and T. Satoh and H. Kuwabara, 1989). Loads in the end
sections of the violating feeders that have the smallest value of this quantity are curtailed. Sarma et al (N.D.R. Sarma, S.
Ghosh, K.S. Prakasa Rao and M. Srinivas, 1994) consider load shedding in system with switch able capacitors and
on-load tap changers. For voltage drop violations, if the load point with violation does not have a switch able capacitor,
it will be shed. For current capacity violation at a component, a low priority load at a point beyond that component is
shed. However, the steps to be taken when several current capacity and voltage drop violations are simultaneously
present are not specified in this paper. Wang et al (P. Wang and R. Billinton, 2000) investigates the effect of
load-shedding procedures on distribution system reliability cost indices. Customer concerns regarding interruption costs
are incorporated in the load-shedding decision process when a bulk system deficiency occurs. Cost weight factors for
different feeder types, based on capacity and cost match, are used to determine the load-shedding priority among
feeders. In (Ding Xu and Adly Girgis, 2001), an optimal load shedding strategy for power system with multiple DGs is
presented and in this paper discritization and mathematical programming has been introduced. In (W.P. Luan, M.R.
Irving and J.S. Daniel, 2002), a genetic algorithm is employed to search for supply restoration and optimal load
shedding in distribution networks. In (P.S. Nagendra Rao and K.S. Papa Rao, 2003) line ampacity violations and
voltage drop violations at the load points are considering for load shedding in radial distribution systems. But, to our
knowledge, hardly anything has been reported in the literature on the problem of online load shedding with the
objective of minimizing system loss during generation deficiency conditions caused by unscheduled outages in
generation and transmission system for maneuver applications in case of emergency for distribution system. In (A. R.
Malekpour, A. R. Seifi, M. R. Hesamzadeh, 2006) authors applied a GA optimization method for load shedding in
distribution networks considering DG units.
This paper presents a new approach for solving the steady state load-shedding problem in distribution network during
generation deficiency conditions with DGs. The problem is formulated to minimize the sum of curtailed load and also
system losses. The problem is subjected to equality and inequality constraints. The formulated optimization problem is
solved using GAs technique (D.E. Goldberge, 1989). The method is tested on a radial distribution network with 33 load
points. The effects of GAs parameters and operators are studied. Results are reported and discussed.
2. MATHEMATICAL MODEL OF THE PROBLEM
Load shedding problem can be formulated as an optimization problem with the following objective function and
constraints:
Nb

(1)

Min ( Rk I k ) + W L PL
2

k =1

Such that:
N

(2)

Pgi Pdi Vi V j Yij cos (i j ij ) = 0

j =1

Qgi Qdi Vi VjYij sin (i j ij ) = 0

(3)

j =1

i=1 Nb

Vi min Vi Vi max

i=1 Nl

Pijmin Pij Pijmax

i=1NDG

Pgimin Pgi Pgimax

Qgi = Qg

if Qgimin Qgi Qgimax

Qgi = Qgimin if Qgi Qgimin

Qgi = Qgimax if Qgimax Qgi

Where:
Nb: Total number of branches
Rk: Resistance of kth branch
Ik: Absolute value of current of the kth branch
WL: Importance degree of customer L
PL: Demand active power of customer L
144

i=1 NDG

(4)
(5)
(6)
(7)

Modern Applied Science

May, 2009

Pgi

: Generating active power at bus i

Pgimin

: Minimum limit for generating active power at bus i

Pgimax : Maximum limit for generating active power at bus i

Pdi : Demand active power at bus i

Vi : Magnitude of voltage at bus i

Y ij : Magnitude of (i, j) element of YBus admittance matrix

ij

: Angle of (i, j) element of YBus admittance matrix

i : Angle of voltage at bus i

Q gi : Generating reactive power at bus i

Q di : Demand reactive power at bus i

V i min

: Minimum limit for magnitude of voltage at bus i

V i max : Maximum limit for magnitude of voltage at bus i

Pijmin : Minimum limit for active power of branch between buses i and j
Pijmax : Maximum limit for active power of branch between buses i and j
2

In the set of equations (1) through (7), R k I k is the Ohmic loss of the kth branch while k refers to kth branch of the
network. Equations (2) and (3) are well-known load flow equations. Security and operational constraints have been
formulated as (4) and (5). Where, (4) refers to voltage limits and (5) point at thermal limit of distribution lines of the
network. Equation set (7) refers to reactive limits of dispersed generators.
The steady state model of DG is used in this paper. This model is suitable for some kind of DGs such as gas turbine,
combustion engines and hydro generation. DGs are modelled as constant power factor units. Considering this point, the
bus connected to the DG can be modeled as PQ bus (Mardaneh M, Gharehpetian, G.B, 2004). The output and the ramp
rate are two constraints for this kind of DG. It must be pointed out that minimum output of some generation is an
important constraint because of the cogeneration. They must generate certain power to ensure the heat supply (Ding Xu
and Adly Girgis, 2001). These constraints can be written as set of equations 7 with N dg as the number of installed DG
in the system. Now the problem can be stated as minimization of the objective function (OBF) satisfying all system
constraints stated above. A GA software package was written for simulation of load shedding in electrical distribution
networks with and without DGs. This program initializes a random sample of individuals with different parameters to
be optimized using the genetic algorithm approach. The population size of 100 is found to be appropriate for our
problem. By tuning the GA parameters, the optimal performance was reached with one child per pair of parents.
Chromosome length is of length number of buses plus one fitness bit.
3. CASE STUDY
A radial distribution network with 33 load points is used to simulate the load shedding problem with dispersed
generation. The data of this test system is taken from (B.Venkatesh, R.Ranjan, 2003).
The system is a hypothetical 12.66 kV system. When there is a disturbance in the network, the system operator may
request the distribution utility or industrial customer to shed load to maintain the system integrity. Two cases are studied.
Case 1 occurred when DGs can not compensate decrease in power flows to the network. (Decrease in power is more
than total installed DGs). Case 2 occurred when DGs can compensate decrease in power flows to the network.
(Decrease in power is not more than total installed DGs).
3.1 Optimal load shedding considering constant capacity deficiency modelling of bulked power supply

Suppose now that the power flows to the network decreases to 1.5 per unit (the power flows to the network without DG
in normal case is 3.9128 per unit) and caused emergency case for loads. Using the proposed algorithm the resulting
network topology is shown in Fig 1.The active and reactive load powers before and after load shedding can be
compared in the network.
Table I shows the installation node and old and new operational power of DG's for the test system. DGs are installed in
heavy loaded node. Suppose now that the power flows to the network decreases to 1.5 pu. Using the proposed algorithm
in case 1 the resulting network topology is shown in Fig 2.
145

Vol. 3, No. 5

Modern Applied Science

Table shows objective function, summation of loads, loss and profile index (PI) of case study with/without DG.
3.2 Optimal load shedding considering variable capacity deficiency modeling of bulked power supply

For the proposed system, using the proposed formulation, the total supplied load decreases to 3.045 per unit by 0.02 per
unit decreasing step in case one and 2.985 to 1.505 per unit by 0.02 per unit decreasing step in case two.
For each case the result has been saved in a table and when the supplied power to the distribution system decreases in
case of an emergency state, the optimum load shedding can be loaded from the table and applied to the system by
system operator.
1) Load shedding is not necessary (case1)
Because the total maximum installed DGs is 0.793 per unit by decreasing the total supplied to 3.045 per unit by 0.02 per
unit decreasing step load shedding is not needed and DGs can compensate the decreased power. Figure 3 shows the
remained active load with and without DG in case one. Figure 4 and 5 shows the profile index (PI) and Percentage
loss (the ratio of total loss to total generated power) of the system with and without DG in case one.
2) Load shedding is necessary (case2)
In this case because DGs are at maximum injection power by decreasing the total supplied load from 2.985 to 1.505 per
unit by 0.02 per unit decreasing step load shedding is necessary and total decreased power can not be compensated by
DGs. Figure 6, 7, and 8 shows the remained active load, profile index (PI) and Percentage loss of the system with and
without DG considering variable capacity deficiency modeling of bulked power supply in case2.
4. ANALYSIS OF THE RESULTS
The proposed algorithm was successful in solving the optimization problem of optimal load shedding in distribution
networks with and without installed DGs considering constant and variable capacity deficiency modeling of bulked
power supply. When deficiency occurred the main objective would be maintaining much more load of the system. In
case1 this objective is satisfied by DGs. Also the proposed algorithm can improve the percentage loss in comparison the
cases without DG according to figures 3, 4, 5. But in contrast the voltage profile was worse than cases without DG.
These results were also shown in case2 and figures 6, 7, and 8.
5. Conclusion
The energy deficit originated from failures in the generation and transmission systems promotes load curtailments in the
distribution system. There are many policies or strategies that can be adopted to perform these load curtailments. This
paper proposed a GA-based methodology for finding optimum load shedding strategy for distribution networks with
and without installed DGs considering constant and variable capacity deficiency modeling of Bulked Power Supply
points of distribution networks. The model is based on minimization of total curtailed load based on their assigned
importance degree and system losses within the frame work of load shedding equality and inequality constraints. A test
system was used in order to apply the methodology and the results were presented for the various load shedding
alternatives.
REFERENCES
A. R. Malekpour, A. R. Seifi, M. R. Hesamzadeh, (2006). Considering Dispersed Generation in Optimal Load
Shedding for Distribution Networks, 14th Iranian Conference on Electrical Engineering, ICEE2006.
Ackermann,T.; Andersson,G.; Soder, L.; (2001). "Distributed Generation: A Definition", Electric Power Systems
Research, Vol 57, 2001, pp 195204.
B.Venkatesh, R.Ranjan, (2006). Optimal radial distribution system reconfiguration using fuzzy adoption of
evolutionary programming, Electric Power System Research, 25 (2003) 775-780.
D.E. Goldberge, (1989). Genetic Algorithms in Search, Optimization and Machine Learning, Addison-Wesely,
Reading, Ma.
Daly, P.A.; Morrison, J; "Understanding the Potential Benefits of Distributed Generation on Power Delivery Systems"
IEEE Power Engineering Society Summer Meeting pp. A2-1/A2-13
Ding Xu and Adly Girgis. (2001). Optimal Load Shedding Strategy in Power Systems with Distributed Generation,
IEEE Winter meeting, Power Engineering Society, 2001, V01.2, pp.788-792.
Dugan, R.C.; McDermott, T. E.; "Operating Conflicts for Distributed
Power Engineering Society Summer Meeting, ppA3-1/A3-6.

Generation on Distribution Systems", IEEE

K. Aoki, N. Nara, M. Itoh and T. Satoh and H. Kuwabara, (1989). A new algorithm for service restoration in
distribution systems IEEE PWRD, 4(3):1832-1839.

146

May, 2009

Modern Applied Science

Mardaneh M, Gharehpetian, G.B, (2004). Siting and sizing of DG units using GA and OPF based technique,
TENCON 2004. 2004 IEEE Region 10 Conference.
N.D.R. Sarma, S. Ghosh, K.S. Prakasa Rao and M. Srinivas, (1994). Real time service restoration in distribution
networks a practical approach IEEE PWRD, 9(4):2064-2070.
P. Wang and R. Billinton, (2000). Optimum load-shedding technique to reduce the total customer interruption cost in a
distribution system, IEE Proc. - Gener. Transm.Distrib., Vol. 147, No. 1, pp. 51-56, Jan. 2000.
P.S. Nagendra Rao and K.S. Papa Rao, (2003). An efficient load shedding algorithm for radial systems, TENCON
2003. IEEE Region
W. El-Khattam, M. M. A. Salama, (2004). "Distributed generation technologies, definitions and benefits", Electric
Power Systems Research, Vol 71, 2004, pp 119-128.
W.P. Luan, M.R. Irving and J.S. Daniel (2002). Genetic algorithm for supply restoration and optimal load shedding in
power system distribution networks, IEE Proc- Gener. Transm. Distrib, Vol 149, No. 2, March 2002.

TABLE 1. INSTALLED NODE WITH OPERATING POINT OF DISPERSED GENERATORS

Operating point of DG before

Operating point of DG after

Max active power

load shedding (kW/power factor)

load shedding (kW/power factor)

(kW/power factor)

50/0.8

83.3/0.8

193/0.8

100/0.9

161.23/0.9

175/0.9

25

200/0.9

284.63/0.9

300/0.9

30

100/1

163.04/1

175/1

Node #

TABLE 2. OBJECTIVE FUNCTION, TOTAL LOAD, LOSS, PI

Load
Index

Shedding
without DG

OBF

Load Shedding
with DG

6.189

4.12

1.464

2.211

0.8162

1.1806

Loss

0.023

0.0225

PI

0.019435

0.017425

load

load

147

Vol. 3, No. 5

Modern Applied Science

Figure 1. Optimal load shedding considering constant capacity deficiency modelling of bulked
power supply with out installed dispersed generators

148

May, 2009

Modern Applied Science

Figure 2. Optimal load shedding considering constant capacity deficiency modelling of bulked
power supply with installed dispersed generators

149

Vol. 3, No. 5

Modern Applied Science

Remained active power with and without DG

Remained active
power(p.u.)

4
3.5
Without DG
3

With DG

2.5

3.11

3.13

3.15

3.17

3.19

3.21

3.23

3.25

3.27

3.29

3.31

3.33

3.35

3.37

3.39

3.41

3.43

3.45

3.47

3.49

3.51

3.53

Supplied power to the system (p.u.)

Figure 3. Remained active load of distribution network considering variable capacity deficiency modelling of bulked
power supply with and without installed dispersed generators (case 1)

0.18
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0

PI without DG

3.
12
5

3.
16
5

3.
20
5

3.
24
5

3.
28
5

3.
32
5

3.
36
5

3.
40
5

3.
44
5

PI with DG

3.
48
5

3.
52
5

PI

PI with and without DG

Supplied power to the system (p.u.)

Figure 4. Profile index (PI) of distribution network considering variable capacity deficiency modeling of bulked power
supply with and without installed dispersed generators (case 1)
Percentage Loss with and without DG
0.05
Loss (%)

0.04
0.03

Loss (%) without DG

0.02

Loss (%) with DG

0.01

3.
04
5

3.
08
5

3.
12
5

3.
16
5

3.
20
5

3.
24
5

3.
28
5

3.
32
5

3.
36
5

3.
40
5

3.
44
5

3.
48
5

3.
52
5

Supplie d powe r to the syste m (p.u.)

Figure 5. Percentage loss of distribution network considering variable capacity deficiency modeling of bulked power
supply with and without installed dispersed generators (case 1)

150

May, 2009

Modern Applied Science

Remained active load (p.u.)

Remained active load with and without DG

4
3.5
3
2.5
2
1.5
1
0.5
0
2.
98
5
2.
90
5
2.
82
5
2.
74
5
2.
66
5
2.
58
5
2.
50
5
2.
42
5
2.
34
5
2.
26
5
2.
18
5
2.
10
5
2.
02
5
1.
94
5
1.
86
5
1.
78
5
1.
70
5
1.
62
5
1.
54
5

Without DG
With DG

Supplied power to the system (p.u.)

Figure 6. Remained active load of distribution network considering variable capacity deficiency modeling of bulked
power supply with and without installed dispersed generators (case 2)

PI with and without DG

0.12
0.1

PI

0.08
PI without DG
PI with DG

0.06
0.04
0.02

2.
98
5
2.
90
5
2.
82
5
2.
74
5
2.
66
5
2.
58
5
2.
50
5
2.
42
5
2.
34
5
2.
26
5
2.
18
5
2.
10
5
2.
02
5
1.
94
5
1.
86
5
1.
78
5
1.
70
5
1.
62
5
1.
54
5

Supplied power to the system (p.u.)

Figure 7. Profile index (PI) of distribution network considering variable capacity deficiency modeling of bulked power
supply with and without installed dispersed generators (case 2)

0.04
0.035
0.03
0.025
0.02
0.015
0.01
0.005
0
1.55

1.63

1.71

1.79

1.87

1.95

2.03

2.11

2.19

2.27

2.35

2.43

2.51

2.59

2.67

2.75

2.83

2.91

Loss (%) without DG

Loss (%) with DG

2.99

Loss(%)

Percentage Loss with and without DG

Supplied power to the system (p.u.)

Figure 8. Percentage loss of distribution network considering variable capacity deficiency modeling of bulked power
supply with and without installed dispersed generators (case 2)

151