Prediction of Losses Allocation using Artificial Neural Network and Electricity Tracing Technique

Abstract

This paper presents a method for prediction of transmission losses allocation in deregulated power market using Artificial Neural Network (ANN). The developed ANN requires Bialek's tracing technique for training and testing data, where the contributed losses from generators and loads are firstly obtained through upstream and downstream looking algorithm. The Bialek's tracing technique was implemented in IEEE 14-bus power system under various reactive loadings in order to obtain the tracing results. Subsequently, the contributed losses were used in the neural network training using Levenberg-Marquardt backpropagation algorithm, which was then followed by testing and post-regression analysis to measure its performance and accuracy. To achieve the desired ANN design, three cases were considered in the experiment: different number of neurons, different number of hidden layers and varying the momentum constant. Simulation has justified the proposed technique's capability in solving losses allocation problem under various loading conditions of the system. The accuracy in terms of contributed losses prediction is satisfying in addition to better correlation coefficient after training and testing have been conducted.