Classification of Power Quality Events Using Radial Basis Function Neural Network

Abstract

Power quality is a measure issue in power system. This is the measure of system reliability, equipment security, and power availability in power system to the industry or end user. The power quality events/problems are caused at generation, transmission and distribution level due to generation to load mismatch, short circuit faults, equipment failure, etc. This dissertation is introducing the generation of power quality events and classification of these events using radial basis function neural network. The power quality events are generated by developing an electric power distribution model using SimPowersystem in MATLAB/Simulink. This creates power quality events such as sag, swell, interruption, harmonics, transient, noises etc. Classification of power quality signals is difficult task and neural network is a non-linear, data driven self adaptive method that is a promising tool for classification. Amongst the neural networks radial basis function neural network (RBFNN) suppose to be a good selection with respect to other neural networks because of its faster learning capability and more compact structure. RBFNN is a non linear parametric approximation model based on combination of Gaussian function is applied to classify PQ events in power system. This RBFNN is made more effective by using Kmean clustering algorithm, K nearest neighbour algorithm and pseudo inverse method. A gradient descent learning method is used for the model to increase the accuracy of classification of proposed RBFNN model.