You are on page 1of 5

2014 IEEE 8th International Power Engineering and Optimization Conference (PEOCO2014), Langkawi, The Jewel of Kedah,

Malaysia. 24-25 March 2014

Gravitational Search Algorithm Application in


Optimal Allocation and Sizing of
Multi Distributed Generation
1
M.R.H.M. Rasdi, 2I. Musirin, 3Z. Abdul Hamid, 4H. Che Mat Haris

Faculty of Electrical Engineering


Universiti Teknologi Mara,
Shah Alam, Selangor, Malaysia
1
rashdanhafiz88@gmail.com, 2ismailbm1@gmail.com, 3zulcromok086@gmail.com, 4irharizan@salam.uitm.edu.my

Abstract- Distribution system represents the final link between bulk generator into the distribution network. On the other hand,
power system and the consumers. Distributed generators or small- when operating at negative power factor, the generator will
scale energy sources are usually smaller than central energy sources consume reactive power. The power flow will be from the
and can produce electricity in the range of 10 kW to 10,000kW. It is
distribution network into the generator [9].
usually connected to the power system points close to the customer.
One of the problems faced when dealing with DG is to
Power loss in a distribution system consists of real and reactive
power loss. Real power loss is due to the transmission lines resistance determine the proper size. As the size of DG is increased, the
while reactive power loss is due to the inductive element, which total system loss will reduce up to a minimum point where if
allows the real power transfer in the distribution system. This paper the size of DG is further increase, the system loss will increase
presents the optimal allocation and sizing of multi-distributed [6]. Adnan in [6] verified that, location also affects the
generation (DG) in distribution system for minimum power loss. The reliability of DG. At different locations, DG would reduce
study involves the development of Gravitational Search Algorithm power loss differently. It can be said that DG is location-
(GSA) engine as an optimization technique. Comparison with dependent. The size of DG that will be allocated also need to
respect to Evolutionary Programming (EP) was conducted in order to
be equal or less to the summation of total systems load and
highlight its superiority. The developed algorithm has been validated
total systems loss [7]. This value will be the limit for the size
on 15-Bus IEEE Distribution System and the results revealed that
power loss is reduced. GSA outperformed EP in terms of achieving of DG to be allocated.
lower minimal power loss. The technique presented in this paper is to optimize the
total losses through the changes of loading condition at
Index Terms Gravitational Search Algorithm, Evolutionary selected buses using the Gravitational Search Algorithm
Programming, Allocation and sizing of DG (GSA). This technique was compared with Evolutionary
Programming (EP) to see which technique can offer better
I. INTRODUCTION results in minimizing power loss. The developed algorithm has
Renewable energy for electricity resource is gaining been tested on 15-Bus IEEE Distribution System. Study has
popularity in Malaysia. It can benefit the distribution system if revealed that the proposed algorithm has succeeded in
it is installed correctly. The main point of this research is to optimizing the power loss and offer better result compared to
have most minimum power loss. Power losses in the system EP.
mainly depend on the systems power flow [8].
Conventionally, distribution system was designed as a passive II. MECHANICS OF GRAVITATIONAL SEARCH ALGORITHM
network [7] and it has unidirectional power flow [12] purely Gravitational Search Algorithm (GSA) is an optimization
for delivering power [13]. By applying DG, the power flow algorithm based on the gravitation law. Based on Newtons
become active and it can either help reducing the loss or law of gravitation, each mass (of an object) attracts every
increasing it, depending on its size and location. Thus, DG is other mass with a gravitational force [1]. The force between
able to change the reactive power flow in the distribution the two masses can be formulated as follow:
network [9].

When operating at positive power factor, the generator  (1)

will produce reactive power. The power will flow from the

978-1-4799-2422-6/14/$31.00 2014 IEEE 364


2014 IEEE 8th International Power Engineering and Optimization Conference (PEOCO2014), Langkawi, The Jewel of Kedah,
Malaysia. 24-25 March 2014

where, F is the gravitational force, G is the gravitational



constant, M1 and M2 are the masses and R is the distance . (2)
between the two masses. This gravitational force will cause
global movement of masses toward the hheaviest masses. The number of iteration used in the algorithm affects this
Through GSA, agents (DG allocation and sizing) will be constant. Large numbers of iteratiion produce more accurate
considered as object with its own masses andd its performance optimization but consume more tiime for optimization. The
proposed number of iteration used for this paper is 100 since
(power loss) will be analyzed based on theeir masses. Each
the optimization is faster and proper optimized value
mass has its own acceleration rate. Heavy m mass move slower produced.
than the others and usually represent good soluution.
This algorithm comprises of 5 steps aand explained as 4. Calculate mass(M) and accelerattion(a) for each agent:
follow: Masses represent solutions. Heavy masses move slowly
1. Initialization of initial population: and has low acceleration rate. Thiis is the characteristic of a
An N number of agents are assigned. Foor this paper, we good solution/optimization value [1].
use 20 agents. This population is randomlyy generated [7].
Agents represent the positions of the mass. V Velocity of each (3)

agent is set to 0. The value of each agent is raandomly selected 

between 0 and 1. If the value is higher thann 1, the loss will (4)


increase instead of decrease since it w will exceed the 
summation of total systems load and total ssystems loss [7].
This decision based on the agents evaluationn at base case for (5)

both cases.
5. Update velocity(v) and position((x):
The velocity and position of eaach agent has changed and
updated. These data will be used again for another iteration
until ends criterion is met.

 (6)

(7)

III. RESULTS AND D DISCUSSION


The study has been conducted d on two cases. In Case 1,
the optimization was conducted on the sizing of DG while the
location is pre-determined arbitrarily. Two types of DG were
used which are photovoltaic (type 1) for DG1 and hydro for
DG2 (type 3) [2][5]. The selected location for case 1 is DG1
will be allocated at bus 11 and DG2 2 is allocated at bus 5. Case
2 is on the other hand is optimizattion of size of DG and the
locations to be allocated.
For Bus 6, the load flow will stop
s converging when load
applied is equal or greater than 10 MVAR.
M In Figure 2 for case
1, GSA provides less loss compared to EP. From Figure 3, in
case 2, GSA garph is a little bit incconsistant compared to EP.
However, more loss can be minim mized by applying optimal
location.
Figure 1: Flow Chart of GSA The load flow stop converging g for Bus 4 at 6.5 MVAR.
This time around, case 1 managees to minimize power loss
2. Evaluate fitness for each agent: better compared to case 2. Howeveer, it can be seen that GSA
Fitness is calculated using the load flow. T
This algorithm is
provides less loss compared to EP. For case 2 (Figure 5), GSA
combined with the load flow analysis to obtain the total
systems power loss [7]. produces lower loss than EP.
3. Update constant G, best and worst of the ppopulation:
After evaluating the fitness, the best and the worst fitness
will be updated including the constant G.

365
2014 IEEE 8th International Power Engineering and Optimization Conference (PEOCO2014), Langkawi, The Jewel of Kedah,
Malaysia. 24-25 March 2014

Figure 5: Optimal Location & Sizing (Loading at Bus 4)


Figure 2: Optimal Sizing (Loading at Bus 6)

Figure 3: Optimal Location & Sizing (Loading at Bus 6)


Figure 6: Optimal Sizing (Loading at Bus 13)

Figure 7: Optimal Location & Sizing (Loading at Bus 13)


Figure 4: Optimal Sizing (Loading at Bus 4)
Table I, Table II and Table III tabulate the optimal sizing
Divergence at Bus 13 occurs when the load is equal or of DG for Case1 using GSA. It is observed that the values of
greater than 2.5 MVAR. This time, case 2 significantly sizing are high. Table IV, Table V and Table VI tabulate the
reduces more losses as compared to case 1. It is also observed optimized location and size of DG performed using GSA.
that GSA provides much lower losses over EP. From all the From Table IV, Table V and Table VI, the locations of DG
six figures, it is discovered that all optimization results for all varied as the loading conditions are gradually increased. For
loading conditions at both cases do reduce power loss. load increment at Bus 6, the allocation is random and
Nevertheless, Case 2 demonstrates better results than case 1. inconsistent. Sometimes both DGs are placed at the same bus
GSA provides better loss minimization than EP especially location. For Table 5, the location is consistent during half
when optimal location is executed. way of the load increment and both DGs are located at the
same place. At loading increment for Bus 13, the allocations

366
2014 IEEE 8th International Power Engineering and Optimization Conference (PEOCO2014), Langkawi, The Jewel of Kedah,
Malaysia. 24-25 March 2014

have the same pattern as allocation for Bus 4 but with lesser Table VII, Table VIII and Table IX tabulate the optimal
load increment. sizing of DG for Case1 using EP. It is observed that the values
are consistently higher as compared to case 1 using GSA.
Table X, Table XI and Table XII shows the optimized location
TABLE I: OPTIMAL SIZING DURING LOAD INCREMENT AT BUS 6 and sizing of DG using EP. Same as case 1, using EP; the size
USING GSA
Qd6 Loss (MW) DG Sizing is almost consistent. The size of DG using EP is much higher
/ MVar Before After P1 P2 / Q2 than implementation using GSA.
/ MW MW / MVar For GSA, the size of DG for case 2 is much smaller as
2.0 0.2260 0.1120 0.4114 0.3290 0.4413
4.0 0.6607 0.3884 0.5024 0.3772 0.3990
compared to case 1. On the other hand, the size of DG for case
6.0 1.4860 0.8875 0.6425 0.4608 0.4104 2 is much bigger as compared to case 1 when EP was
8.0 3.0382 1.6896 0.8339 0.6052 0.6926 implemented. As mentioned in the introduction, DG is
location-dependent. The reason why EP is used, the sizing of
TABLE II: OPTIMAL SIZING DURING LOAD INCREMENT AT DG for case 2 is bigger which might be due to the location of
BUS 4 USING GSA
Qd4 / Loss (MW) DG Sizing DG is almost the same during load increment.
MVar P1 / P2 Q2 /
Before After
MW / MW MVar
2.0 0.2973 0.0533 0.3681 0.3870 0.4454 TABLE VII: OPTIMAL SIZING DURING LOAD INCREMENT
4.0 0.9957 0.1555 0.3946 0.4991 0.4935 AT BUS 6 USING EP
6.0 2.8894 0.3237 0.4376 0.6483 0.4657 Loss (MW) DG Sizing
Qd6 /
MVar P1 / P2 / Q2 /
Before After
TABLE III: OPTIMAL SIZING DURING LOAD INCREMENT AT MW MW MVar
BUS 13 USING GSA 2.0 0.2260 0.1820 0.7564 0.9910 0.3653
Qd13 / Loss (MW) DG Sizing 4.0 0.6607 0.4458 0.7564 0.9910 0.3653
MVar P1 / P2 / Q2 4.5 0.8229 0.5500 0.0291 0.0471 0.5257
Before After 6.0 1.4860 0.9825 0.0304 0.0478 0.5263
MW MW / MVar
0.5 0.1088 0.0241 0.3685 0.3135 0.3071 8.0 3.0382 1.8843 0.0312 0.0493 0.5260
1.0 0.2174 0.0554 0.3998 0.3134 0.4564
1.5 0.4013 0.1089 0.4363 0.3307 0.5579 TABLE VIII: OPTIMAL SIZING DURING LOAD INCREMENT
2.0 0.6978 0.1867 0.5310 0.3135 0.3601 AT BUS 4 USING EP
2.5 1.2043 0.2919 0.6362 0.3135 0.4495 Qd4 / Loss (MW) DG Sizing
MVar Before After P1 / P2 / Q2 /
MW MW MVar
TABLE IV: OPTIMAL LOCATION AND SIZING DURING LOAD
2.0 0.2973 0.1187 0.7564 0.9910 0.3653
INCREMENT AT BUS 6 USING GSA
4.0 0.9957 0.2072 0.0285 0.0471 0.5260
Qd6 / Loss (MW) DG Location DG Sizing
MVar before after DG1 DG2 P1 P2 Q2 6.0 2.8894 0.4086 0.0288 0.0494 0.5257
Bus Bus (MW) (MW) MVar
2.0 0.226 0.058 11 4 0.074 0.206 0.583 TABLE IX: OPTIMAL SIZING DURING LOAD INCREMENT
4.0 0.661 0.253 4 4 0.351 0.132 0.515 AT BUS 13 USING EP
6.0 1.486 0.632 5 11 0.113 0.106 0.695 Loss (MW) DG Sizing
Qd13 /
8.0 3.038 1.182 10 4 0.091 0.072 0.409 P1 / P2 / Q2 /
MVar Before After
MW MW MVar
0.5 0.1088 0.0976 0.7564 0.9910 0.3653
TABLE V: OPTIMAL LOCATION AND SIZING DURING LOAD
INCREMENT AT BUS 4 USING GSA 1.0 0.2174 0.1254 0.7564 0.9910 0.3653
1.5 0.4013 0.1731 0.7564 0.9910 0.3653
Loss (MW) DG Location DG Sizing
Qd4 / 2.0 0.6978 0.2436 0.7564 0.9910 0.3653
DG1 DG2 P1 / P2 Q2 /
MVar before after 2.5 1.2043 0.3482 0.0291 0.0471 0.5259
Bus Bus MW / MW MVar
2.0 0.297 0.064 8 8 0.245 0.056 0.463
4.0 0.996 0.313 8 8 0.193 0.104 0.387 TABLE X: OPTIMAL LOCATION AND SIZING DURING LOAD
6.0 2.889 0.737 8 8 0.482 0.090 0.684 INCREMENT AT BUS 6 USING EP
Qd6 / Loss (MW) DG Location DG Sizing
MVar before After DG1 DG2 P1 / P2 / Q2 /
TABLE VI: OPTIMAL LOCATION AND SIZING DURING LOAD
Bus Bus MW MW MVar
INCREMENT AT BUS 13 USING GSA
2.0 0.226 0.211 4 15 0.756 0.991 0.365
Loss (MW) DG Location DG Sizing
Qd13 / 4.0 0.661 0.474 4 15 0.756 0.991 0.365
DG1 DG2 P1 / P2 / Q2 /
MVar before After 6.0 1.486 0.963 14 12 0.653 0.414 0.700
Bus Bus MW MW MVar
1.5 0.401 0.109 6 6 0.376 0.007 0.537 8.0 3.038 1.855 14 12 0.649 0.410 0.696
2.0 0.698 0.191 6 6 0.300 0.007 0.492
2.5 1.204 0.308 7 7 0.393 0.134 0.218

367
2014 IEEE 8th International Power Engineering and Optimization Conference (PEOCO2014), Langkawi, The Jewel of Kedah,
Malaysia. 24-25 March 2014

TABLE XI: OPTIMAL LOCATION AND SIZING DURING LOAD


INCREMENT AT BUS 4 USING EP
REFERENCES
Loss (MW) DG Location DG Sizing [1] Esmat Rashedi, Hossein Nezamabadi-pour, Saeid Saryazdi, GSA: A
Qd4 / DG Gravitational Search Algorithm, Information Sciences, Volume 179, Issue
MVar Befor DG1 P1 / P2 / Q2 /
After 2 13, 13 June 2009, Pages 2232-2248, ISSN 0020-0255,
e Bus MW MW MVar
Bus 10.1016/j.ins.2009.03.004.
2.0 0.297 0.215 6 8 0.678 0.679 0.934 [2] Prommee, W.; Ongsakul, W., "Optimal multi-distributed generation
4.0 0.995 0.678 8 10 0.827 0.029 0.048 placement by adaptive weight particle swarm optimization," Control,
6.0 2.889 1.663 8 10 0.827 0.029 0.048 Automation and Systems, 2008. ICCAS 2008. International Conference
on, vol., no., pp.1663, 1668, 14-17 Oct. 2008.
[3] Tan, W.S.; Hassan, M.Y.; Majid, M.S.; Rahman, H.A., "Allocation and
TABLE XII: OPTIMAL LOCATION AND SIZING DURING LOAD sizing of DG using Cuckoo Search algorithm," Power and Energy
INCREMENT AT BUS 13 USING EP (PECon), 2012 IEEE International Conference on, vol., no., pp.133, 138,
Loss 2-5 Dec. 2012
DG Location DG Sizing [4] Zahra Moravej and Amir Akhlaghi, A novel approach based on cuckoo
Qd13 / (MW)
MVar DG1 DG2 P1 / P2 / Q2 / search for DG allocation in distribution network Electrical Power and
Before After Energy Systems, vol. 44, pp. 672-679, 2013.
Bus Bus MW MW MVar
0.5 0.108 0.104 6 8 0.678 0.679 0.9347 [5] Vinothkumar, K.; Selvan, M. P.; Srinath, S., "Impact of DG model and
load model on placement of multiple DGs in distribution
1.0 0.217 0.191 4 15 0.756 0.991 0.3653
system," Industrial and Information Systems (ICIIS), 2010 International
1.5 0.401 0.318 6 8 0.678 0.679 0.9347 Conference on, vol., no., pp.508, 513, July 29 2010-Aug. 1 2010
2.0 0.697 0.544 8 10 0.827 0.029 0.0475 [6] Anwar, A.; Pota, H.R., "Optimum allocation and sizing of DG unit for
2.5 1.204 0.895 8 10 0.827 0.029 0.0482 efficiency enhancement of distribution system," Power Engineering and
Optimization Conference (PEDCO) Melaka, Malaysia, 2012 Ieee
International, vol., no., pp.165, 170, 6-7 June 2012
[7] Mistry, K.; Bhavsar, V.; Roy, R., "GSA based optimal capacity and
V. CONCLUSION location determination of distributed generation in radial distribution
system for loss minimization," Environment and Electrical Engineering
This paper presents GSA based technique for power loss (EEEIC), 2012 11th International Conference on, vol., no., pp.513, 518,
minimization in power distribution system. The validation 18-25 May 201
process is performed on the IEEE 15-bus distribution system. [8] Salman, S.K., "The impact of embedded generation on voltage regulation
and losses of distribution networks," Embedded Generation on
The technique is tested with two cases, optimal sizing only Distribution Networks (Digest No. 1996/194), IEE Colloquium on the
and optimal location and sizing. These results were then Impact of, vol., no., pp.2/1,2/5, 15 Oct 1996
[9] Duong Quoc Hung; Mithulananthan, N., "An optimal operating strategy
compared with results of EP. From the result, it is found that of DG unit for power loss reduction in distribution systems," Industrial
GSA is better than EP in minimizing power loss. and Information Systems (ICIIS), 2012 7th IEEE International
Conference on, vol., no., pp.1, 6, 6-9 Aug. 2012
[10] Hany E. Farag, E.F. El-Saadany, Ramadan El Shatshat, Aboelsood
ACKNOWLEDGMENT Zidan, A generalized power flow analysis for distribution systems with
The authors would like to acknowledge The Research high penetration of distributed generation, Electric Power Systems
Research, Volume 81, Issue 7, July 2011, Pages 1499-1506, ISSN 0378-
Management Institute (RMI) of Universiti Teknologi MARA 7796, 10.1016/j.epsr.2011.03.001.
(UiTM) for the financial supports of this research. This [11] Kashem, M. A.; Le, A.D.T.; Negnevitsky, M.; Ledwich, G., "Distributed
research is supported by RMI under The Research Intensive generation for minimization of power losses in distribution systems,"
Faculty (RIF) with project code of 600-RMI/DANA 5/3/RIF Power Engineering Society General Meeting, 2006. IEEE , vol., no.,
pp.8 pp.,, 0-0 0
(372/2012). [12] Antonios G. Marinopoulos, Minas C. Alexiadis, Petros S. Dokopoulos,
Energy losses in a distribution line with distributed generation based on
stochastic power flow, Electric Power Systems Research, Volume 81,
Issue 10, October 2011, Pages 1986-1994, ISSN 0378-7796,
10.1016/j.epsr.2011.06.006.
[13] Anwar, A.; Pota, H.R., "Optimum capacity allocation of DG units based
on unbalanced three-phase optimal power flow," Power and Energy
Society General Meeting, 2012 IEEE, vol., no., pp.1, 8, 22-26 July 2012

368

You might also like