OPTIMAL LOW-VOLTAGE DISTRIBITION NETWORK DESIGN USING GENETIC ALGORTIHMS Belgin TRKAY Taylan ARTA˙ e-mail: turkay@elk.itu.edu.tr e- mail: taylanartac@yahoo.com Istanbul Technical University, Electrical & Electronics Faculty, Department of Electrical Engineering, 80626,Maslak, Istanbul, Turkey Keywords: Distribution Networks, Optimization, Genetic Algorithms ABSTRACT In this paper, a new genetic algorithm (GA) has been developed for optimal distribution network design. Distribution transformer station location and size selection is evaluated together with feeder route and size selection in the study. The model used in the design includes realistic rules that are derived from actual networks. I. INTRODUCTION As energy consumption increases, the need to expand energy production and distribution is inevitable. To meet the increase timely, many studies have been made in this area. In this paper the optimum distribution network design is examined. The goal of designing a new distribution network or the expansion of an existing distribution network is to provide the consumers with quality electrical energy while minimizing the cost. For this purpose, different algorithms and optimization techniques have been developed in this area. Some of the researches in the past have focused only on distribution transformer location selection [1-4] or feeder route selection [2-5]. Others that included both distribution transformer location selection and feeder route selection have optimised these models separately, resulting in a two-step optimization and a lower chance in reaching the global optimum. Most of the algorithms developed optimize these two models simultaneously [6-13]. The inclusion of these two models in the same algorithm results in a greater chance in reaching the global optimum result. Few researches have included the reliability model and the cost of energy interruption into the algorithm [14-17]. This attains a better result. It is important that the distribution network should not experience long breaks of energy interruption. In other words, the distribution network should be reliable. Unfortunately, the data needed for reliability analysis has not been obtained by most countries, making it very hard to insert the reliability model in the algorithm. In distribution network design, the selected optimization technique is nearly as important as the algorithm selected. The optimization techniques used for distribution networks can be categorized in two main headings: i) Mathematical programming techniques ii) Heuristic techniques (Includes evolutionary techniques) Previous studies mainly used mathematical programming while recent studies mainly use heuristic techniques. Both techniques have their advantages and disadvantages. Mathematical programming techniques have obtained global optimum with an acceptable fault. The main problems in this technique are to model real problems and the problems encountered in calculation. If too many variables are used, the mathematical programming takes too long computation time, forcing the programmer to neglect some of the variables. For this, the non-linear characteristics are linearized. Evolutionary techniques are stochastic methods shaped according to the laws of nature or evolution. GAs is the most used evolutionary technique. The main problem in GA is defining the genetic operators in order to meet the global optimum. The power of GAscome from being able to model some features that classical mathematical programming methods would neglect. As a result, a more realistic result, which is nearer to the global optimum, is obtained. The ability to reach a realistic optimal result compared to other optimiziaton methods has made the selection of GA preferable. Without violating the technical constraints, a GA optimization program developed in MATLAB which includes the model proposed for distribution network