Short Term Hourly Load Forecasting Using Combined Artificial Neural Networks

This paper presents a new approach for short term hourly load forecasting (STLF) using combined artificial neural network (CANN) module. The CANN module is developed for STLF using two different algorithms - evolutionary programming (EP) and particle swarm optimization (PSO). In this paper, a set of neural networks has been trained with different architecture and training parameters. The artificial neural networks (ANNs) are trained and tested for the actual load data of Chennai city (India). EP and PSO based optimal linear combinations are applied to combine selected networks and to obtain CANN module, to produce better results, rather than using a single best trained ANN. The obtained test results indicate that the proposed approach improves the accuracy of the load forecasting.