Conventional and Intelligent Methods for DG Placement Strategies

Yaser Soliman Qudaih, Student Member, IEEE, Syafaruddin Non member and T. Hiyama, Senior Member, IEEE
and voltage improvement [4]-[5] and others discuss about the DG contribution to primary frequency control [6]. In [7], authors compare the network performance between the installation of DG and compensating capacitor for loss minimization. However a plenty of researches discuss the replacement of capacitor such as in [8-10], that inspire related studies for DG allocation. In addition, many other studies discussed auxiliary systems such as Energy Capacitor System (ECS), to be involved in solving the problems of DG influences as in [11]. Intelligent system by means the artificial neural network (ANN) has been satisfactorily used to solve the estimation and optimization tasks of engineering problems. The eminences of ANN method over the other optimization methods are simple computational techniques and high pattern recognition abilities [12, 13]. Moreover, this method doesn’t need requirement of knowledge on internal system behavior; requires less computational effort and provides compact solutions for multi-objective problems [14]. In some cases, only training process is required and the optimum point is directly determined without solving any non-linear mathematical equations or statistical assumptions as in the conventional optimization methods [15]. For these reasons, the ANN methods is suitable to estimate the voltage profile and total power losses under connection of distributed generation units where their outputs are intermittent and fluctuated based on environmental conditions. In this paper the simulation based DG size and allocation have been investigated in order to approach the wealthy application of the DGs in reducing the power loss and enhancing the voltage profile. All data have been supplied to the Matlab/Smulink model, which has been built to meet with 33-bus distribution network [16]. The model could be considered as a handy tool to deal with various types of applications in the distribution network, such as power distribution system reconfiguration and some control and protection applications. As a case study in this research, the model can easily present both diesel generator and wind generator penetration to the system in an efficient and practical ways. Moreover, power flow calculations have been made to evaluate the system losses and system voltage profile by adopting the technique used in [2], namely the loss sensitivity factor method for optimal DG allocation and sizing. According to the same reference, a look up table can be created with only one power flow calculation and that table can be used to restrict the size of the DG at different busses. RBF neural network has been designed for the same purpose.

Abstract—Since the talks started about the benefits of utilizing distributed generations (DG) to the electrical power system, the number of researches about this phenomena became tremendous. In this paper, the wealth in DG diversity has been investigated. Power system losses and voltage profile have been put as a target under investigation to evaluate the efficient employment of the DGs in the electrical system. A 33-bus distribution system with diesel generator in one case and with wind generator in another case have been simulated and tested. Simulation results are included to show the wealthy effect in DG diversity in terms of power system loss reduction and voltage profile enhancement. Radial basis function (RBF) neural network has been constructed to represent the target system as an alternative way to solve the problem of DG placement. Result shows a unique implementation of neural network to replace the conventional model. Index Terms— RBF neural network, power distribution system, distributed generation, power loss reduction, voltage enhancement.



ART of the electrical system is the generation, where electric power is generated by means of power plants or relatively smaller size generators which may called as DG. The generated power will be delivered to the customers via transmission and distribution networks. Power distribution system is one of the most vital parts in the system as it is forming the last point before delivering power to the customers. As DG systems are not centrally planned, their influences are more expected on the distribution level of the system. The optimal allocation of the DGs has been discussed as an important factor related to the effect of the DG penetration. In [1], the authors explain the background to the technical constrains faced by DG projects, they use linear programming to determine the optimal allocation. Reference [2]-[3] discuss analytically, the size and allocation of the DG to reduce the power losses of the distribution networks. Other authors achieve to find the optimal allocation for reliability, losses,

Yaser Soliman Qudaih is with the Department of Computer Science and Electrical Engineering, Kumamoto University, Kumamoto 860-8555, Japan (e-mail: Syafaruddin is with the Department of Computer Science and Electrical Engineering, Kumamoto University, Kumamoto 860-8555, Japan (e-mail: Takashi Hiyama is with the Department of Computer Science and Electrical Engineering, Kumamoto University, Kumamoto 860-8555, Japan (e-mail:

II. LOSS MINIMIZATION MATHEMATICAL STRUCTURE According to the strategy in which this paper is prepared, the mathematical structure could be divided into three types: nto A. Formulation for Optimization A set of simplified feeder-line flow formulations is line employed [17]. Considering the single-line diagram depicted line in Fig. 1, the recursive equations (1-3) are used to compute the ) power flow. Total power loss in the base case and in the case of diesel generator will be calculated by summing the whole losses of all sections of the feeder as:

In case of the wind turbine application as a DG, the energy d t loss will be calculated as explained below or by following (9). Energy loss will be calculated for the period of time (t), ( equivalent to the time of running the simulation using the s Matlab model.

Fig. 1. Single-line diagram of the basic feeder in radial distribution system line

For the sake of simplicity, some assumptions have been taken into account that may considered upon application that application, are, the distribution lines are represented as series impedance and load demand as constant and balanced power sink of the value ( Zi, i + 1= Ri, i + 1+ jXi, i + 1) and S L = PL + jQL respectively. The real and reactive power flow at the receiving end of branch i+1 and the voltage magnitude at the receiving end is expressed respectively by the following equations:

The objective function has been defined as in (10)

Pi +1 = Pi − PLi +1 − Ri i ,i +1

Pi 2 + Qi2 Vi

(1) (2 )

Qi +1 = Qi − Q Li +1 − Xii ,i +1
2 2

Pi 2 + Qi2 Vi

Vi +1 = Vi − 2( Rii ,i +1 .Pi + Xi i ,i +1 .Qi ) + ( Rii2,i +1 + Xii2,i +1 ) Pi 2 + Qi2 Vi


Equations (1)-(3) are known as the Distflow equations. Hence, if P0, Q0 and V0 at node number 1 are estimated, then the same equations at the other nodes can be calculated by applying the above branch equations. That is known as forward update. Similarly a backward update is applied by the following set of equations:

Pi −1 = Pi + PLi +1 + Rii ,i +1

Pi '2 + Qi'2 Vi

(4 ) (5)

B. Matlab/Simulink Structure The model which has been made for this purpose is shown in Fig. 2. The illustrated results have been evaluated as follow: . 1. The base case: The constructed model was made according to the data taken from [12], and the simulation [1 was run to find the total losses in the system and the syste voltage at every node. All results were tabulated. ll result 2. With the diesel generator: A synchronous three-phase instantaneous model of the diesel generator has been used. The diesel generator of the same size was located at every node of the system, then the total power losses and voltage profile was determined and tabulated. After that, the optimum size of the diesel generator derived by the method in [2] was located in sequence at every node and the same calculations have been made and tabulated. 3. With the wind turbine: an induction generator wind turbine with a specific size and random wind speed changes has been used. The simulation ran for specific r time to calculate the energy extracted by this unit for the specified period. Total energy losses has been calculated by subtracting the energy consumed in the load from the energy generated as in (9), also the voltage at every node , was calculated and tabulated.

Qi −1 = Qi + QLi +1 + Xii ,i +1
2 2 '

Pi '2 + Qi'2 Vi

Vi −1 = Vi + 2( Rii ,i −1 .Pi + Xii ,i −1 .Qi ) + ( Rii2,i −1 + Xii2,i −1 )
' '

Pi '2 + Qi' 2 Vi

(6 )
Fig. 2. Proposed Matlab representation of the distribution network .

Pi = Pi + PLi and Qi = Qi + QLi .

The power loss of the line section connecting between buses i and i+1 is calculated as:

C. RBF Neural Network Structure In this study, we develop two structures of RBF neural network which is for voltage and total losses estimation. The otal

basic configuration of the proposed network is shown in Fig. 3. In this figure, there are 33 input signals for both estimation . purposes that represent the number of buses for distribution network. On the other hand, there are 33 output signals for voltage estimation and a single output signal for the total losses estimation. For each tasks, the development of RBF structure follows three important stages. They are establishment of training data set, training process and validation. Firstly, the training data set was taken from 33 . 33-bus IEEE test system. This data set is to cover the entire domain of voltage and total losses as a function of input power between 0 and 4 MW connected to each bus. For this assumption, there are 133 training data set. The second stage is the training ing process. During the training process, the input vector which will result in lowering the network error is used to create a new hidden neuron. If the current error after the neuron insertion is low enough, the training stops. In this study, the parameter of training process: the mean squared error goal (GOAL), spread of radial basis functions (SPREAD), maximum number of neurons (MN) and the number of neurons to add between displays (DF) are 0.003,1.0,133,1, respectively. The outcomes of the training process are the ectively. training error and number hidden neurons that represent the structure of RBF network. As results, there are 119 of hidden neurons and 0.000933 of training error for RBF (voltage RBF-1 estimation), while 129 of hidden neurons and 0.00062 of 9 training error for RBF-2 (total losses estimation). 2

decided from the figure, that such a wind turbine (1 MW rating power in this case study), will be restricted to be connected at some locations, bus no. 20, 21 and 22 in this case.

Fig. 4. Total power losses of the system with a diesel generator of the same sses size and of optimal size compared with each other and with the base case.

Fig. 5. Total energy losses of the system with a wind turbine of the same size compared with the base case.

Voltage profile was checked comparing three cases. One is the base where no DG is connected to the network, the other one, in the presence of optimal size diesel generator and the last in the presence of the wind turbine. That is illustrated in turbin Fig.6.

Fig. 3. ANN RBF network structure

III. RESULTS AND DISCUSSIO DISCUSSION Results were obtained in different stages to be matched with the steps where this paper has been built up. Thus, total power losses in kW unit were calculated, in the case of a diesel generator of the same size placed, turn by turn, at every node of the network and in the other case where an optimal size generator (Size restriction), have been placed. Comparing the result of these two cases illustrated in Fig.4, and certainly, compared with the base case where no DG was placed. Fig. 5, is showing the result comparing the loss in the system energy with and without a wind turbine unit, the losses were measured in kWh for 50 sec. of simulation. Clearly, it can be
Fig. 6. Voltage profile enhancement comparison

The third stage of obtaining the result was a validation process to confirm the RBF network accuracy. In this stage a accuracy random signal is defined as the mean of input power of about 2.5 MW with a suitable variance. The graphs in Fig. 7 compare the conventional and RBF results where the input result signal applied at bus 33. (Min. loss=141.22 kW for 1.0538 MW DG optimum size). Same comparison has been obtained to measure the system voltage in the base case as shown in Fig. 8.






[13] Fig. 7. Conventional/RBF comparison for a DG of random size variations at bus 33 [14]




H. N. Ng, M. M. A. Salama, and A.Y. Chikhani, “Classification of Capacitor Allocation Techniques,” IEEE Trans. on Power Delivery, vol. Trans 15, pp. 387-392, 2000. Y. T. Hsiao, C. Y. Chien, “Optimisation of Capacitor Allocation Using An Interactive Trade-off Method,” IEE Proceedings, Generation, Transmission and Distribution, vol. 148, pp. 371-374, 2001. , 371 M. Delfanti, G.P. Granelli, P. Marannino, and M. Montagna, “Optimal Capacitor Placement Using Deterministic and Genetic Algorithms,” Algorithms, IEEE Trans. on Power Systems, vol. 15, pp. 1041-1046, 2000. , 1041 Y. Qudaih and T. Hiyama, “Reconfiguration of Power Distribution System Using Multi Agent and Hierarchical Based Load Following Operation with Energy Capacitor System,” In Proc. 2007 The 8th International Power Engineering Conference (IPEC), Singapore, pp. onference Singapore 263-267. A. Mellit, and S. A. Kalogirou, "Artificial intelligence techniques for Artificial photovoltaic applications: A review," Progress in Energy and review Combustion Science, vol.34, no.5, pp.574-632, 2008. , pp.574 A. Mellit, S. A. Kalogirou, L. Hontoria, and S. Shaari, “Artificial intelligence techniques for sizing photovoltaic systems: A review,” Renewable and Sustainable Energy Reviews, vol.13, no.2, Reviews pp.406-419, 2009. F. Giraud, and Z.M. Salameh, “Analysis of the effects of a passing cloud on a grid-interactive photovoltaic system with battery storage using interactive neural networks,” IEEE Transactions on Energy Conversion, vol.14, Conversion no.4, pp. 1572-1577, 1999. A. Mellit, M. Benghanem, and S.A. Kalogirou, “An adaptive wavelet“ network model for forecasting daily total solar-radiation,” Applied etwork solar Energy, vol.83, no.7, pp.705-722, 2006. 722, M. E. Baran and F.Wu, “Network reconfiguration in distribution system for loss reduction and load balancing,” IEEE Trans. Power Del., vol. 4, no. 2, pp. 1401–1407, Apr. 1989. R.Srinivasa, S.V.L. Narasimham, and M. Ramalingaraju, “Optimization of Distribution Network Configuration for Loss Reduction Using Artificial Bee Colony Algorithm,” In Proc. 2008 World Academy of Science, Engineering and Technology, vol. 35, pp. 709-715. , 709

Fig. 8. Voltage profile comparison in the base case

IV. CONCLUSION DGs can participate in reducing the power loss of th the system and enhancing the voltage. The proposed work compares between conventional way as Matlab/Simulink structure and intelligent method as ANN RBF network structure. Results show the efficient representation of RBF structure to the target power system with high accuracy and high flexibility of application. V. REFERENCES
[1] Keane. A, and M. O'Malley, "Optimal Allocation of Embedded Optimal Generation on Distribution Networks," IEEE Trans. on Power Systems, EE vol. 20, pp. 1640-1646, Aug. 2005. C. Wang and M.H. Nehrir, "Analytical approaches for optimal placement of distributed generation sources in power systems IEEE systems," Trans. on Power Systems, vol. 19, pp. 2068-2076, Nov.2004. 2076, N.Acharia, P. Mahat, and N.Mithulananthan, “An analytical approach An for DG allocation in primary distribution network, Electrical Power network,” and Energy Systems, vol. 28, pp. 669-678, Dec. 2006. 678, C. Borges, DM. Falcao, “Optimal distributed generation allocation for Optimal reliability, losses, and voltage improvement,” Electrical Power and Energy Systems vol. 28, pp. 413-420, July. 2006. J.C. Hernándeza, A. Medinaa, and F. Jurado, “Optimal allocation and Optimal sizing for profitability and voltage enhancement of PV systems on feeders, " Renewable Energy, vol. 32, pp. 1768-1789 Aug. 2007. 1789, J. Morren, S. Haan, and J.A. Ferreira, “Contribution of DG units to Contribution primary frequency control,” European Transaction on Electrical Power Power, vol. 16, pp. 507-521, Sep. 2006. S. Rhim, T. Rahman, I. Musirini, S. Azmi, M. Mohamed, M. Hussain, M. Faridun, “Comparing the Network Performance Between the Installation of DG and Compensating Capacitor Using EP,” International Journal of Power, Energy and Artificial Intelligence vol. Intelligence, 1, pp 14-21, Aug. 2008.

Yaser Soliman Qudaih He graduated from University of Engineering and Technology (UET), T Lahore, Pakistan as an electrical engineer in 1996. He has been working in his country Palestine as an electrical engineer in Gaza International Airport. He continued his higher education in the Department of Computer Science and Electrical Engineering, Kumamoto University, Japan. He ring, received his M. S. from Japan in 2008. Currently he is a Ph. D student in Kumamoto University. Syafaruddin He received his B.Eng degree in Electrical Engineering from Universitas Hasanuddin, Indonesia, in 1996, M.Eng degree in Electrical Engineering from University of Queensland, Australia, in 2004 and D.Eng degree from Kumamoto University, Japan in 2009. He is currently working in Kumamoto University as project assistant professor for Graduate School of Science and Technology. His research interests chnology. include distributed generation planning, maximum power point tracking control of photovoltaic system, power system real-time simulation and neuroreal fuzzy logic control application in power system. Takashi Hiyama (M`86, SM`93) received his B. E., M. S. and Ph. D. degree all in electrical . engineering from Kyoto University in 1969, 1971 and 1980, respectively. Since 1989, he has been a professor at the Department of Computer Science and Electrical Engineering, Kumamoto University, Japan. His current research interests include the intelligent systems applications to power system operation, control and management. He is a senior member of IEEE, a member of IEE of Japan, and , Japan Solar Energy Society. Currently he is the Dean of the Graduate School of Science and Technology, Kumamoto University, Japan.







Sign up to vote on this title
UsefulNot useful