A practical approach to electric load forecasting using artificial neural networks with corrective filtering

This paper presents the practical application of an artificial neural network to the power system load forecasting problem. This work examines the training, testing, and operation of a simple neural network. Furthermore, a method for improving the prediction accuracy of a forecasting neural network is proposed. This approach views the forecasting problem as a knowledge-based discrete time filtering problem. Encouraging results have been obtained using this method for forecasting the peak monthly load of a power utility, over a number of years.