Optimal Location and Sizing of Distributed Generation in Distribution Network Using Adaptive Neuro-Fuzzy Logic Technique

The growing gap between electric power generated and that demanded is of utmost concern especially in developing economy, hence calling for measures to argument the existing power generated of which DG is a more viable aspect to explore in curtailing this challenges; although been confronted with issue of location and sizing. This research applied Adaptive neuro fuzzy logic technique to optimize DG location and size. A 24 bus radial network was used to demonstrate this process and having a suitable location and size at optimal position reduces power losses and also improves the voltage profile at the buses. The method was simulated using ANFIS toolbox MATLAB R2013b (8.2.0.701) 64-bit software and tested using Gwagwalada injection sub-station feeder 1 system. The results obtained were compared to that obtained using ANN. It was observed that adaptive neuro fuzzy logic technique performed better in terms of reducing power losses compared to ANN technique. The percentage reduction in the power loss at the buses cumulatively is 48.96% for ANN while adaptive neuro fuzzy logic technique is 49.21%. The voltage profile of the networks after optimizing the DG location and sizes using adaptive neuro fuzzy logic technique were also found to be much improved with the lowest bus voltage improved from 0.9284 to 1.05pu.



Abstract-The growing gap between electric power generated and that demanded is of utmost concern especially in developing economy, hence calling for measures to argument the existing power generated of which DG is a more viable aspect to explore in curtailing this challenges; although been confronted with issue of location and sizing.This research applied Adaptive neuro fuzzy logic technique to optimize DG location and size.A 24 bus radial network was used to demonstrate this process and having a suitable location and size at optimal position reduces power losses and also improves the voltage profile at the buses.The method was simulated using ANFIS toolbox MATLAB R2013b (8.2.0.701) 64-bit software and tested using Gwagwalada injection sub-station feeder 1 system.The results obtained were compared to that obtained using ANN.It was observed that adaptive neuro fuzzy logic technique performed better in terms of reducing power losses compared to ANN technique.The percentage reduction in the power loss at the buses cumulatively is 48.96% for ANN while adaptive neuro fuzzy logic technique is 49.21%.The voltage profile of the networks after optimizing the DG location and sizes using adaptive neuro fuzzy logic technique were also found to be much improved with the lowest bus voltage improved from 0.9284 to 1.05pu.Index Terms-Distributed Generation, Optimization, Power Loss, Voltage Stability Index.

I. INTRODUCTION
Electricity power has remained an essential prerequisite for the progress of any country's economy whether developed, emerging or developing.Increasing human activities due to technological advancement coupled with population growth, has made the demand for power more than doubling by the decades, thereby broadening the gap between power generated and the demand by the consumer.
Conservatively, the existing power system in Nigeria is one in which power is generated conventionally at remote stations and transmitted at high voltage through the transmission station to the distribution network, and subsequent delivery to the end consumers at a lower voltage level.Thus, the capability of humans to develop sources of energy necessary to realize useful task has played a critical role in the recurrent improvement of the standard of living generally.(e-mail: iamsteveben@gmail.com).
Hence, with the increase in power consumption, achieving the power demand of consumers at all location within the power network economically and dependably as possible has become the ultimate aim of the power system.The traditional practice of the electricity power generation system utilizes the conventional energy sources among which in Nigeria commonly used are the hydro and thermal with gradual advances in solar [1].

II. CONCEPT OF DISTRIBUTED GENERATION (DG)
Over the years, the increasing cases of installed peripheral power generating facilities had encouraged the publication of several articles especially with the authorities getting aware and recognizing the potential therein with DG.
DG by definition is an electric power unit or source coupled directly to the distribution network or the consumers' side of the meter [2].CIGRE describe DG as a concept of generation, which is characterized as follows; (CIGRE, 1999): it is not centrally strategic or planned; it is not centrally dispatched at present; it is usually connected to the distribution networks; it is smaller than 50-100MW [3].In [4], DG is described as: "electric power generation source linked directly to the distribution network or on the customer side of the meter".
The notion of DG is to integrate into the distribution network of the power system a sustainable energy systems ranging from few KW -50 MW to the distribution network to continuously supply electrical energy to the end users as the conception of DG contrasts with the customary centralized power generation concept, where the electricity is generated in large power stations and is transmitted to the end users through transmission and distributions lines [5].[6] DG put to use small-scale technologies to generate electricity adjacent to the end users of power.This technology do comprises of modular generators (and sometimes renewable-energy), and they offer a number of prospective benefits.In several circumstances, DG can offer lower-cost electricity and higher power dependability and security with fewer environmental consequences than can traditional power generators.
Using DG in a distribution network has some advantages as specified in [7], a lessening in line losses, emission pollutants, total costs due to improved efficiency & peak saving.Improvement of voltage profile, power quality, system reliability and security and the disadvantages are [8], reverse power flow, injected harmonics, increased fault currents depending on the location of DG units.DG also has several benefits like energy costs through combined heat and power generation, avoiding electricity transmission Optimal Location and Sizing of Distributed Generation in Distribution Network Using Adaptive Neuro-Fuzzy Logic Technique Evans C. Ashigwuike, and Stephen Adole Benson costs and less exposure to price volatility [9] As a result of importance accorded to the distribution network as the final link between the transmission network and the consumers, its performance, and quality of service provided is measured in terms of frequency of disruptions, and also the maintenance of suitable voltage levels at the consumer's locality [10], [11].Since the utilization of DG is gradually gaining calls for implementation in Nigeria, it is not enough to place the DG randomly within the distribution network.In other to achieve its desired objective, efforts need to strategize for optimal placement and sizing of DG within the distribution network, hence the reason for this research.

III. MATERIAL AND METHOD
The power flow problem is a very familiar problem in the field of power systems engineering, where voltage magnitudes and angles for one set of buses are sought after, given that voltage magnitudes and power levels for another set of buses are known and that a model of the network configuration is accessible.A power flow solution procedure is a numerical technique that is employed to resolve the power flow problem [12] A. Real and Reactive Power Injected in a Bus An electric power system is fundamentally made up of generators, transformers, transmission lines and loads.A modest power system is exemplified in Fig. 1 and Fig. 2. Hence, the network formed by these static components can be seen or taken as a linear network and represented by the matching admittance matrix or impedance matrix.In power flow calculation, the generators and loads are treated as nonlinear components and cannot be embodied in the network.In formulating the real and reactive power entering a bus, we need to describe the following variables.Let the voltage at the i th bus be designated by: and self-admittance at the bus-i as: Likewise, the mutual admittance between the buses i and j can be written as: Assuming the power system contains a total number of n buses.The current injected at bus-i is given as, It is to be noted we shall assume the current entering a bus to be positive and that leaving the bus to be negative.As a result the real and reactive power entering a bus will also be presumed to be positive.The complex power at bus-i is then given by sin cos sin cos sin cos sin cos sin cos sin cos (6) Hence, rearranging (6) and substituting, (5) gives the real and reactive power as

B. Voltage Stability Index (VSI)
VSI is presented in order to evaluate and estimate the stability limit.VSI is an imperative tool for assessing the proximity of a given operating point to voltage instability.The objective of the VSI is to quantify how close a particular point is to the steady state voltage stability margin.They can be used on-line or offline to aid operators in real time operation of power system or in designing and planning operations.
This index is used in conjunction with the power loss in the distribution network in the placement and sizing of distributed generation.Elements such as reactive power generating devices, tap changing transformers are optimally adjusted at each operating point to reach the objective of reducing voltage stability index at each individual bus as well as minimizing the global voltage stability indices.The system can be operated in the stable region by minimizing voltage stability index of buses and lines.
The VSI is given as: where, Z= line impedance Qj= reactive power at receiving end Vi= sending end voltage Xij = Line reactance

C. Artificial Neural Network (ANN)
ANNs refer to a class of models stirred by the biological nervous system.The models are composed of many computing elements, usually denoted neurons; each neuron has a number of inputs and one output [13], [14] & [15].It is based on nodes with connections between them as the neurons in brain and synapses as seen in fig. 3.In this case, certain nodes are assigned input of which in the normal sense is random.But deliberate attempt is made in the ANN to arrange these nodes in a more orderly form so that the input node can be clearly notify.Each node further has weight so that when two or more nodes are connected to same node, the weight can help decide which connection is more important.ANN training consists of three layers, namely input layer, hidden layer and output layer.To get an output, the hidden layer is attached a transfer function to judge the input after decision, the nodes choose it output node.

D. Adaptive Neuro Fuzzy Logic Technique
As indicated by the topic, Adaptive Neuro-Fuzzy logic technique is applied in carrying out the research.Been a sugeno model that is put in a framework of adaptive system, it facilitates learning and adaptation [16].It viability and robustness are features that help get result that has a high level of precision as compare to other technique been used by other researchers in the aspect of DG placement and sizing.
In recent time, adaptive neuro-fuzzy logic approach has become a popular method in areas including control.This is because, ANN are good at recognizing pattern but they are not good at explaining how they reach their decision and fuzzy logic is good at explaining the decision but cannot automatically acquire the rules for making the decision [21].Description of the adaptive neuro-fuzzy logic principles is given [17], [18], [19], and [20].
The adaptive neuro-fuzzy hybridization results in a hybrid intelligent system that synergizes these two techniques by combining the human-like reasoning style of fuzzy systems with the learning and connectionist structure of artificial neural networks [18].It incorporates the human-like reasoning style of fuzzy systems through the use of fuzzy sets and a linguistic model consisting of a set of IF-THEN fuzzy rules.The main strength of adaptive neuro-fuzzy systems is that they are universal approximators with the ability to implore interpretable IF-THEN rules [13], [19].
Hence in executing the adaptive neuro-fuzzy logic technique; the ANN system is used to obtain an optimal value of the Power loss using the variables obtained from the power flow studies of the bus network.The power loss and the voltage stability index which will also be calculated for each bus subsequently serve as an intermittent input for the adaptive neuro-fuzzy inference system.The research compares the results obtained with the propose method which is a hybrid of two artificial intelligences with that obtained with the ANN technique.
The data for this research was obtained from a local electricity distribution company.A 24 bus network of the Abuja Electricity Distribution Network representing a subnetwork or feeder of Gwagwalada injection sub-station taken for consideration to demonstrating the proposed technique.

IV. RESULT
The research considered a 24 bus radial distribution network of Gwagwalada injection sub-station 11KV feeder 1 Shown in fig. 4. Based on the research procedure, the DG location and size within a 24 bus network was determined considering the total power loss at the buses and the VSI of the buses which serve as a vital index in determining the stability and or the weakest point within the network.Before the optimization process, the voltage profile of the network is as seen in table I.The size of the Distributed Generation at each bus is given in fig. 5.

A. Implementing Artificial Neural Network Technique
Applying the ANN technique taking into account the power loss at each bus as well as VSI using backward propagation, bus 13 was identified to be the suitable location for siting of a DG with a size of 141.35KW.Table II shows the impact of the installation of the DG at the location, as the improvement in the voltage profile within the voltage limit can be seen.Given that the base voltage of the network in p.u is 1.0000, before the DG was installed, the voltages at each of the respective buses were less than the value of the base voltage with minimum value of 0.9284 at bus 13.The DG installed improved the voltage profile to a new maximum limit of 1.0500 with a minimum value of 0.9500.
As a result, in the improved voltage profile, there is also a reduction in the real power loss within the network as can be noted in the individual buses within the network shown in table III.Taking the location of installation that was identified (bus 13); a considerable loss reduction can be seen.Prior to the DG installation, it can be seen that the power losses at the buses was higher in buses farther away from the injection substation a challenge which is often an attribute of most radial distribution network.But with the DG been installed, the losses at these was reduced considerably with impact of reduction most felt at buses 4, 10, 13, 18, 21 and 24.

B. Implementing Adaptive Neuro Fuzzy Logic Technique
In applying adaptive neuro fuzzy logic technique, ANN was used to obtained an optimal value of the power loss at the buses using backward propagation as well as the size of the distributed generation; the output of the ANN in conjunction with the VSI then serve as an intermittent input variable for the fuzzy inference system to determine the location of the DG.
Matlab R2013b was used and the simulation carried out using Adaptive neuro-fuzzy inference system toolbox.A concise training structure of the ANFIS procedure is summarized in Fig. 6.The procedure is initiated by obtaining a training data set and checking data sets.These data sets are basically input and output variable for the training and in vector form.With this vector, the ANFIS is trained.The training data set is used to find the basis parameters for the membership functions.A threshold value for the error between the actual and desired output is determined [17].Adaptive Neuro fuzzy logic technique provided an output that is more optimal than that obtained using ANN in section A. Although not with a wide margin, but the essence of a robust output is essential as the aim is to optimize the end results which in this case, is to reduce the power loss and improve the voltage profile.This technique also did identify bus 13 as the optimal location for the DG placement with a more optimal DG size of 151KW that is; an increase of 6.83% to the value obtained using ANN.Table IV shows the outcome of the voltage profile using adaptive neuro fuzzy logic technique: In this application, the voltage profile was improved to the maximum value of 1.05 p.u at buses 1, 2, 7, 11, 13, 14, 16, and then 18 through to 23 within constraint.
The real power loss at the bus using adaptive neuro fuzzy also, show slight reduction cumulatively compare to that obtained using ANN especially at the bus of installation (bus 13).It can be observed in Table V that the inclusion of a DG in the distribution network reduces the power losses at buses 4, 10, 13, 18, 21 and 24 considerably.A reason affirmed due to the distance of these buses from the injection substation initially.The farther a location from the source of supply, the more the chances of voltage drop with a corresponding increase in real power losses.

C. Comparison of Both Techniques
The variations in output using ANN and adaptive neuro fuzzy logic technique can be seen in fig.7 and fig.8 representing voltage profile improvement and power loss reduction at the buses.It will be seen that the power losses is more prominent at buses 4, 10, 13, 18, 21 and 24.With the installation of a DG at bus 13 using both ANN technique and the adaptive neuro fuzzy logic technique, it is observed that the losses at the afore mentioned buses were considerably reduced.Although, the reduction in power losses realized using both techniques is not large apart, the adaptive neuro fuzzy logic provided a more robust and optimal result at bus 13, 16   ANN technique presents a 48.96% reduction in power loss at the buses cumulatively in the distribution network while the adaptive neuro fuzzy logic gives a 49.21% reduction in the power loss deduced from table VII.With reference to bus of DG installation, the voltage improvement at the bus has a 13.09% improvement with a power loss reduction of 82.91%.Table VI shows that prior to the optimal location of the DG, the minimum value of the voltage obtained in p.u was 0.9284 which was improved to 1.05p.u.With the optimal location of the DG, the overall minimum voltage of buses within the network becomes 0.9500p.u.
Adaptive neuro fuzzy logic technique offers a more robust technique in the optimization process as a result of combining the good qualities of ANN and fuzzy logic.The place of DG location and sizing cannot be over emphasized in power system.Optimal location of DG indicates an essential role for minimizing the power loss reduction in a distribution system.

V. CONCLUSION
This work presents the formulation and application of adaptive neuro fuzzy logic algorithm to support in reducing the distribution network power losses and improve voltage profile by optimizing the location and size of a DG.As realized from the results the adaptive neuro fuzzy logic technique gave a better power loss reduction compared to ANN technique.The percentage reduction in power loss is 49.21% and 48.96% respectively.
The technique chose bus 13 as the optimal DG location and reduced the power losses by 12.9231 KW.The technique also improved the lowest bus voltage from a value of 0.9284pu to 1.05pu.Therefore, the adaptive neuro fuzzy logic technique proved more suited for optimization.Accordingly, the objective of the research was attained effectively and the implemented adaptive neuro fuzzy logic technique proved to be a better technique for optimizing the location and size of a DG in power networks with the goal of reducing power losses and improving voltage profiles of the network.

Fig. 1 .
Fig.1.Single-line diagram of a simple network

TABLE I :
VOLTAGE PROFILE AT BUSES WITHOUT DG INSTALLATION

TABLE II :
VOLTAGE PROFILE IMPROVEMENT AT BUS DUE TO OPTIMAL LOCATION OF DG USING ANN

TABLE III :
POWER LOSS REDUCTION AT BUSES DUE TO OPTIMAL LOCATION OF DG USING ANN

TABLE IV :
IMPROVEMENT OF VOLTAGE PROFILE DUE TO OPTIMAL LOCATION OF DG USING ADAPTIVE NEURO FUZZY LOGIC

TABLE V :
POWER LOSS REDUCTION DUE TO OPTIMAL LOCATION OF DG and 23.The outcomes of the individual technique are summarized in the table VI and table VII below: