You are on page 1of 6

International Journal of Advanced Computer Science, Vol. 2, No. 10, Pp. 376-381, Oct., 2012.

Optimization of Distributed Generation Placement for Minimizing Power Losses and Voltage Profile Improvement Using Genetic Algorithm
Meysam Kalantari, Ahad Kazemi, & Mohammad Saleh Zakerinia
Manuscript
Received: 30,Nov., 2011 Revised: 2, May, 2012 Accepted: 23,Jun. 2012 Published: 15,Nov., 2012

Keywords
Genetic Algorithm, Backwardforward Sweep, Load Models, Power Loss

Abstract The optimal sizing and placement of distributed generation units (DGs) are becoming very attractive to researchers these days. In this paper, a genetic algorithm approach has been used for allocating DGs in distribution systems with mixed load model which includes residential loads, commercial loads, industrial loads and constant loads. In this method the multiobjective minimization is proposed using genetic algorithm. Power losses (active and reactive) and improvement of voltage profile are minimized regarding to system constraints (Power equality, line capacity limits and voltage drop limits). Power flow has been done by Backward-Forward Sweep method. The proposed model has been applied to 37-bus distribution system.

1. Introduction
In the restructured power systems in order to considering environmental issues and using green power, distributed generation units have been spread out in the power distribution systems. In the literature review there isn't unique definition of DGs and they have used different definitions such as; small power generation units compare to large centralized power plants, so that they can be connected to the grid in any point. Generation or saving power in the small scale, close to the load which provides possibility of selling and buying electricity with grid and in some cases generation of maximum energy efficiency is called DG. Also there are many other definitions that they are complementary. DGs are not developed only for commercial purposes but also they have capacity of improving power quality, environmental issues and other aspects like improvement of voltage profile, decrement of power loss, decrement of THD, etc. [1] Generally some of the advantages of using DGs are

This work was supported by the Iran University of Science and Technology. Meysam Kalantari, Iran University of Science and Technology (meysamkalan@gmail.com) Ahad Kazemi, Iran University of Science and Technology (kazem@iust.ac.ir) Mohammad Saleh Zakerinia, University of Tehran (mzakeri@gmail.com)

- preventing of investment in distribution and transmission infrastructure [2]. - Improvement of energy reliability and system security [3]. - Services in interruption mode and on time efficiency for customers [3]. - Lower capacity modules which follow load changes with the better accuracy. Some of disadvantages of using DGs are - High investment cost which causes less competition compare to conventional systems. - Operation and maintenance systems need new methods. - Decentralizing increases transaction costs. There are many approaches for placing DGs in distribution system, Promme et al. [4] is proposed the optimal placement of several DG units using Adaptive weight Particle Swarm Optimization (APSO), this method can control the velocity of particles and finally Basic Particle Swarm Optimization (BPSO) is compared to APSO in terms of power loss reduction. Authors of [5] in order to placement of DG units have used continuation of power flow analysis, determining buses with the most sensitive voltage buses to collapse voltage, influence of efficiency and effectiveness on voltage profile improvement, increment of power transmission capacity and maximum loading. Goswami et al. [6] have analyzed load voltage sensitivity considering load models of voltage dependent load, whereas this method has been done by genetic algorithm. Singh et al. [7] unlike other studies which dealt with the constant load models have studied on the effect of different load and sensitive to voltage and frequency and then they found the best location for DG units. Authors of [8] have used an analytical method for optimal DG allocation, this method is based on power flow for the radial network and calculate loss sensitivity factor and priority list, which causes reduction on the search space. In [2] are proposed a heuristic method for optimal sizing and placement of DG in order to reduce economic costs, like energy cost, investment, operational cost, loss cost and technical aspect such as energy loss and voltage level. In this paper, placement and sizing of DG unit is carried out using genetic algorithm and considering power loss reduction and voltage profile improvement. two third of power loss in electrical network is in the distribution system which is about 10 to 11 percent and in term of peak load is about 14 to 15 percent. So that power loss reduction causes

Meysam Kalantari et al.: Optimization of Distributed Generation Placement for Minimizing Power Losses and Voltage Profile Improvement Using Genetic Algorithm.

377

generation capacity and network capacity increment without investing on power generation sector. In distribution network, in different buses, voltage magnitude changes depends on different hours and causes voltage fluctuations which is harmful for network equipments. But presence of DG improves voltage profile.

which can be seen in the following formulas. + = 0 = 0 (Equ. 5) (Equ. 6)

2. LOAD MODEL AND PROBLEM FORMULATION


Active and reactive loads in buses are illustrated in the formulas (1), (2) as it said in the previous section in power flow analysis. Load models are usually assumed to be constant without considering voltage magnitude and system frequency, while in reality in this kind of loads active and reactive powers are depend on bus voltage and frequency of system. Load models are usually fall into constant loads, residential loads, commercial loads, industrial loads, etc. [6]
Pi = P0i Vi [1 + (f f0 )] = 0 [1 + ( 0 )]

Where , are active and reactive power at bus i, , are active and reactive loads at bus i, is active power injected by DG at bus i. B. Distribution Line Capacity Limits Power flow of lines should be less than maximum permitted power of line due to line thermal capacity. (,) (,) (Equ. 7)

Where (,) is MVA flow in the line connecting bus i and j,(,) is MVA capacity of line i and j. C. Voltage Drop Limit Bus voltages should be in the range of minimum and maximum voltage. < < (Equ. 8)

(Equ. 1) (Equ. 2)

Where and are active and reactive power at bus i, 0 and 0 are active and reactive operating point at bus i, is voltage at bus i, , active and reactive power exponents, , 0 are real and rated frequency and , are active and reactive power slopes. In this paper it is assumed that the frequency of system is constant [9], so that the formulas (1), (2) changes to formulas (3), (4). = 0 = 0

Where , maximum and minimum are allowable voltages at buses. The following indices are introduced for explanation of load models effect because of presence of DG units. 1) Active and Reactive Power Loss Indices [9] (LPI and LQI): The indices of active and reactive power loss are defined as below: = =

(Equ. 3) (Equ. 4)

(Equ. 9) (Equ. 10)

If = = 0 presented load model becomes constant load model and with modification of , different load models are obtained, values of , for different loads can be seen in table 1 [10].
TABLE 1 LOAD TYPE AND EXPONENT VALUES

Where and are the total active and reactive power losses of the distribution generation in presence of DG, and are the total active and reactive system power losses without DG in the distribution system. 2) Voltage Profile Index (VPI) [9]: One of the benefits of correct selection of location and size of DG is improvement of voltage profile. This index penalizes higher voltage deviations from the nominal (1 = 1). The closer this index to zero is the better performance of the network. VPI is calculated with the following formula.
= =2 ( | 1 || | ) | | 1

Load Type Constant Industrial Residential Commercial

0 0.18 0.92 1.51

0 6 4.04 3.4

In this paper load model is assumed to be mixed model which includes residential, commercial, industrial and constant. In power system and during power flow, some of constraints must be satisfied. These constraints are comes as follow: A. Active and reactive power equality Total generation power equals to demanding power,
International Journal Publishers Group (IJPG)

(Equ. 11)

The multi-objective function in order to find the size and placement of DG is a combination of active and reactive power indices and voltage profile index [9]. As it can be seen in formula (12) weight coefficients regarding to the network circumstances assign to indices,

378

International Journal of Advanced Computer Science, Vol. 2, No. 10, Pp. 376-381, Oct., 2012.

which are between 0 and 1. = (1 . + 2 . + 3 . ) Where


3 =1

(Equ. 12)

= 1 , [0,1]

The main purpose of this paper is minimizing the value of FF, which means reduction of active and reactive power losses and voltage profile improvement. Weights are defined based on cost. This cost is according to the saving energy on power loss and profile voltage improvement. Table 2 shows weight values for indices. In this table active power loss is a substantial weight (0.55), voltage profile is 0.2 which is due to effects of its power quality. Multi-objective function is minimized subject to operational constraints which are said before, so that it satisfies electrical requirements of the network.
TABLE 2 INDICES WEIGHTS

If the selected pair of size and location is optimal or close to optimal is selected for DG penetration level. The selected objective function considers all of technical constraints and if one of them exceeds the limits, this solution is rejected. Finally, genetic algorithm compares the new produced population with the old and new solutions, and chooses the best one. When the maximum irritation of the algorithm is reached or the difference between objective function values is smaller than the expected accuracy the algorithm is stopped, then the best solution is chose as a final solution. The proposed algorithm can be seen in Fig. 1.

Indices PLI QLI VPI

0.55 0.25 0.2

3. GENETIC ALGORITHM IMPLEMENTETION


In this paper, the optimization problem is solved with the genetic algorithm (GA) which can find the best optimal solution for DG location and sizing based on fitness function (FF). GA is a programming technique which uses genetic evolution as a solving pattern of the problem [11]. The problem to be solved is an input and the solutions are encoded in the solution space. Fitness function evaluates each solution that is selected most of them randomly. Each solution is defined as a chromosome; each chromosome is a two string solution. The first one indicated the placement of DG in power system and the second one shows its size. In GA first the initial population is generated randomly, and then objective function gives solutions based on this population. Some of solutions are not satisfied the system constraints, these solutions are called infeasible solutions. Infeasible solutions are eliminated, then using mutation and crossover operators the new population is generated, by repeating this procedure the best solution is obtained. In this method the initial population is defined in a way that all of them are possible solution. For each solution a pair of size and location (bus) which is produced by the designer with technical and economic justification.

Fig.1Genetic Algorithm Flowchart

4. CASE STUDY AND SIMULATION RESULTS


DG placement using genetic algorithm method with a multi-objective function has been done on the 37-bus test system, which can be seen in figure. 2 [10]. Total active load of system is 5084.26 KW, and total reactive load is 2547.32 KW. Compensation of system is not done well, and total losses are about 8% of total load. In this method, the placement of only one DG is carried out and it is assumed that DG size changes from 0 to 0.63. DG with size of zero, which means there is no DG in the system. DG is working with power factor of unit.

International Journal Publishers Group (IJPG)

Meysam Kalantari et al.: Optimization of Distributed Generation Placement for Minimizing Power Losses and Voltage Profile Improvement Using Genetic Algorithm.

379

Figure 3.illustrates voltage different between two cases of without having DG and having a DG with size of 0.63 p.u. and, improvement of voltage profile is noticeable in all of buses.
Table 4 Power Loss and Voltage Profile With and Without DG DG Location Capacity Active Power Loss (p.u) Reactive Power Loss (p.u) Voltage Profile Without DG 0.1587 0.1055 0.0765 Bus 14 0.63 0.1106 0.0729 0.0638

Fig. 2 38-bus tested system

As it is presented in the Fig. 1. There are two important parts. The first one is related to power flow, which is done by backward-forward sweep. In backward sweep, currents are calculated from the last branches and come to the first branches, and in forward sweep, voltages are calculated from the first buses to the last buses respectively. In the second part optimization with genetic algorithm is done by initializing, crossover and mutation. And then, if the constraints are satisfied optimization has been done. In this section two cases of one and two DGs allocation are studied separately. A. With One DG Units The proposed method has been run for several standard systems in distributed system, so that in these systems a DG is placed in different positions with different sizes. As it is said before, load model in this system is a mixed load model. The best DG locations with capacity of 0.63 and indices are illustrated in table 3.
TABLE 3 INDICES COMPARISON IN PRESENCE OF DG

Table 5 also shows four optimal results, in all of obtained results the size of DG is 0.63. In this case, with considering costs these results may not be repeated. As it is shown in table 3 when there is no DG in system, PL and QL indices are not proper values and buses voltage average is 0.9536 and fitness function is a significant value of 1, when DG with size of 0.63 is placed in bus 14, as it is shown in Fig. 3. Voltage profile is improved and PL, QL indices compare to the previous state have been improved. Average voltage has got to 0.9640 from 0.9536 and fitness function is reduced to 05689. B. With Two DG Units With putting two DGs with maximum capacity of 0.3 p.u. in the system the placement algorithm is ran, and the results for the best cases are shown in Table 6.
TABLE 5 ANOTHER OPTIMAL RESULTS

DG Location 15 16 13 12

PL 0.70063 0.70132 0.70242 0.70884

QL 0.69485 0.69693 0.6983 0.70665

VPI 0.06393 0.06356 0.06397 0.0638

FF 0.57184 0.57267 0.5737 0.5793

DG Location capacity PL QL VPI FF

14 0.63 0.697 0.6912 0.06376 0.5689

15 0.63 0.7006 0.6948 0.06393 0.57185

Losses of active and reactive power having DG have been shown in Table 4. It can be seen that active and reactive power losses and voltage profile has been decreased significantly. But, it is necessary to say that criterion for DG allocation is multi-objective function FF. Active power loss without DG is 0.1587, with adding a DG with size of 0.63 this value decreases to 0.1106. Also, reactive power loss decreases from 0.1055 to 0.0729.
International Journal Publishers Group (IJPG)

This table illustrates when a DG in bus 14 and the other DG is put in bus 31, the best results are obtained. Whereas, it may not result in the best indices, anyway, criterion is multi-objective function and this value should be minimized. Another point is that the sum of active power of two DGs is 0.6, while, in the previous case there is a DG with capacity of 0.63 in the network. With comparing the best results of two cases, it is concluded that with adding more DGs the better results have obtained, while the total capacity is decreased.

380

International Journal of Advanced Computer Science, Vol. 2, No. 10, Pp. 376-381, Oct., 2012.

Fig 3. Voltage profile comparison with DG in size of 0.63 and without DG

For instance, in case of two DGs the value of FF is 0.5533, while in presence of one DG this value is 0.5689. Figure 4 shows voltage profile without DG and with a DG with capacity of 0.63 p.u., also having two DGs with capacity of 0.3 p.u. It can be seen in buses in range of 7-18, having one DG is appropriate, but in the terminative buses having two DGs show the appropriate voltage profile. Another point in this paper is that, running placement algorithm with genetic algorithm leads to the same results with the same capacity, while if the cost index is added to multi-objective function and weighted coefficients differs as well, these results will change for sure and the size of DGs will not be the maximum size.
TABLE 6 RESULTS IN PRESCENCE OF TWO DG UNITS DG Location capacity PL QL VPI FF 14-31 0.3 , 0.3 0.6799 0.6721 0.0579 0.5535 15-32 0.3 , 0.3 0.6811 0.6722 0.0573 0.5541

Shows the difference between voltage profile without DG and with DG size of 0.63 and with two DG units. Indices comparison in this case with the earlier cases shows more improvement in indices and the most reduction in fitness function. Indices and fitness function show buses 12, 13, 14, 15 and 16 when the size of DG is 0.63 are the best buses for placing DG according to genetic algorithm. This value changes to 0.9657 from 0.9536, this change shows the improvement of voltage profile. This index is not stated in fitness function, but it is a good idea to use this index instead of VPI in future works. Totally, the best solution of GA for the proposed system is the selection of DG with size of 0.63 and in the bus 14; this is obvious that size increasing of DG leads to improvement of indices and fitness function of system. If other indices like operation and maintenance cost and investment cost are considered, the solution may change. If the load model changes, for example if there is only one type of load in the system like residential load, DG location in system will change. Since the proposed load model in this paper is a mixed load, the results of single load model have not considered. It can be stated that with change of load model, there is no change in the size of DG but the place of DG will change.

5. Conclusions
In this paper an exhaustive analysis using genetic algorithm which includes mixed load model, for multiobjective optimization of distributed generation units in distribution system has been carried out. The proposed objective function is defined as a weighted sum of VPI, PL and QL indices. They show voltage profile improvement, active and reactive power losses respectively. The heuristic method of genetic algorithm is given the best solutions in the fastest period of time. It is seen that with increment of DG size indices of VPI, PL and QL reduce and average of buses voltage increase. While, load model has no effect on the location of DG, but size of optimal DG and fitness function is strongly depends on the load model. It is shown in the results of simulation with the increment of DG size; its place in system may change. The results may be more realistic if other indices like cost, reliability of system and total harmonic distortion (THD) are considered.

References
Fig 4. Voltage profile comparison with one DG and two DG and without DG

[1]

The best solution which is obtained by genetic algorithm is bus 14 with DG size of 0.63 and difference between indices in the previous part has shown this issue. But, using two DG units with size of 0.3 the best bus for DG location by genetic algorithm is bus 14 and 31. Fig. 4

[2]

M. Sedighizadeh and M. Sadighi, A Particle Swarm Optimization for Sitting and Sizing of Distributed Generation in Distribution Network to Improve Voltage Profile and Reduce THD and Losses, Universities power engineering conference, 2008, pp. 1-5. C. Tautiva, Optimal Placement of Distributed Generation on Distribution Networks, Universities power engineering conference, 2009, pp. 1-5. International Journal Publishers Group (IJPG)

Meysam Kalantari et al.: Optimization of Distributed Generation Placement for Minimizing Power Losses and Voltage Profile Improvement Using Genetic Algorithm.

381

S. Chaitusaney and A. Yokoyama, Prevention of Reliability Degradation from RecloserFuse Miscoordination Due To Distributed Generation, IEEE Transactions on Power Delivery, vol. 23, 2008, pp. 25452554. [4] W. Prommee and W. Ongsakul, Optimal MultiDistributed Generation Placement by Adaptive Weight Particle Swarm Optimization, International conference on Control, Automation and Systems, 2008, pp. 16631668. [5] H. Hedayati, S.A. Nabaviniaki, and A. Akbarimajd, A Method for Placement of DG Units in Distribution Networks, IEEE Transactions on Power Delivery, vol. 23, 2008, pp. 1620-1628. [6] R.K. Singh, N.B.D. Choudhury, and S.K. Goswami, Optimal Allocation of Distributed Generation in Distribution Network with Voltage and Frequency Dependent Loads, ICIIS, 2008, pp. 1-5. [7] R.K. Singh and S.K. Goswami, Optimal Siting and Sizing of Distributed Generations in Radial and Networked Systems Considering Different Voltage Dependent Static Load Models, IEEE International Conference on Power and Energy, 2008, pp. 1535-1540. [8] P. Alemi and G.B. Gharehpetian, DG Allocation Using an Analytical Method to Minimize Losses and to Improve Voltage Security, IEEE International conference on Power and Energy, Malaysia, IEEE, 2008, pp. 15751580. [9] D. Singh, D. Singh, and K.S. Verma, Multiobjective Optimization for DG Planning With Load Models, IEEE Transaction on Power Systems, vol. 24, 2009, pp. 427436. [10] D. Singh, R.K. Misra, and D. Singh, Effect of Load Models in Distributed Generation Planning, IEEE Transaction on Power Systems, vol. 22, 2007, pp. 22042212. [11] A.F. Khoshbakht and M. Raoofat, Optimal Allocation of DGs and RCSs to Improve Distribution Network Reliability and Network Energy Loss, IEEE International Conference on Power and Energy, 2008, pp. 1586-1591. [3] Meysam Kalantari was born in 1984 Iran, Tehran. He recieved his B.Eng from University of Tehran (UT), Iran, in 2008. He is cuurrently a student of M. Sc. at department of electrical engeneering, Iran University of Science and Technology, Tehran. His areas of interested include: Distributed Generation, Reactive Power Control, heuristic Algorithm. Ahad Kazemi, was born in Tehran, Iran, in 1952. He received his MSc degree in electrical engineering from Oklahoma State University, U.S.A in 1979. He is currently an associate professor in electrical engineering department of Iran University of Science and Technology, Tehran, Iran. His research interests are reactive power control, power system dynamics, stability and

control and FACTS devices.

Mohammad Saleh Zakerinia was born in 1985 Iran, Tehran. He recieved his B.Eng from University of Tehran (UT), Iran, in 2008. He is cuurrently a student of M. Sc. at department of industrial engeneering, University of Tehran, Iran.

International Journal Publishers Group (IJPG)

You might also like