Preprocessing and training effects in voltage stability assessment using neural networks

We present the effects of preprocessing and training parameters in stability index computation using neural network. Two method of index computation was done. In first method active and reactive power are given as net inputs and bus voltage is set as target. From the predicted bus voltage, stability index is computed. In the second method P, Q, V and power factor is given as input and L-index is given as the net output. We show that preprocessing, the raw data with more number of input parameters makes more effective index computation. We also propose the optimum training parameters of the network, based on experimental observation.