Elman Neural Network Backpropagation Based Evaluation of Critical Busbars in Power Systems with Renewable Sources

Abstract

We present an efficient analysis for evaluation of critical busbars in electric power systems by judging the performance of various backpropagation schemes of Historical Elman Neural Network. The objective of this study is to find the most efficient scheme that yields fastest convergence under supervised learning. The study is conducted on the standard IEEE 30-bus test system supplemented by renewable source of generation. Out of six backpropagation schemes tried in this work, it is observed that gradient descent backpropagation with momentum and adaptive learning rate performs exceedingly well in terms of fast convergence irrespective of number of hidden layers and neuron assignment. It is claimed that this study would be very helpful to the power system utilities and researchers in reducing the burden on the utility of conducting routine power flow simulations.