ON-LINE CAPACITOR SWITCHING FOR DISTRIBUTION SYSTEM

Abstract

In this thesis, an artificial neural network based algorithm for determining optimal capacitor switching pattern for a given loading condition in a radial distribution system has been developed. Traditionally, combinatorial methods have been used to decide the optimal switching patterns. However, these combinatorial methods take significant amount of time for practical size power distribution system involving thousands of feeders. Hence, these algorithms may not be very suitable for on-line application in a modem distribution automation system. A multi-layer perceptron neural network with error-back-propagation training algorithm has been used to determine the optimal capacitor switching pattern. It has been found that the time taken by ANN based method is quite less compared to the time taken by the traditional combinatorial methods