Short Term Load Forecasting (STLF) using Generalized Neural Network (GNN) trained with adaptive GA

Abstract

The paper is mainly focus to develop an integration of GNN and wavelet based models for STLF. The model is trained by using error back-propagation algorithm, but there are certain inherent drawbacks of back-propagation algorithm. To overcome the drawbacks of back propagation algorithm such as slow learning, stuck in local minima, needs error gradient etc. genetic algorithm (GA) is proposed. The performance of GA is further improved by making an adaptive GA with the help of fuzzy system. The adaptive GA changes the GA parameters such as cross over probability and mutation rate during execution by using fuzzy system. The GNN-W-AGA is used to forecast electrical load and compared with GNN-W trained with backprop and actual data. © 2013 Springer International Publishing.