Short-term load forecasting techniques using ANN

This paper presents some practical techniques of using artificial neural networks for the short-term load forecasting problem. The model described in this paper is a backpropagation based multilayer perceptron including the temperature factor. In order to expedite the training process, the quasi-Newton method is employed. An intelligent treatment to holiday factors in order to improve the forecasting accuracy is discussed. The average forecasting error of this system is 3.2%.