Selection of Input data of Artificial Neural Network to Improve Performance for Long-term Load Forecasting

Abstract

The paper describes a method to select input data of artificial neural network (ANN) to improve performances of training and prediction for long-term load pattern. The paper focuses on forecasting weekly peak load during a year. Generally, the ANN has been used to forecast short-term load because the training data can be less and the data with high correlation can be easily obtained. However, it is difficult to decide the ANN structure and to select proper input for long-term load forecasting via the ANN. In the paper, the effect of several input data is investigated and the smoothing method is applied to improve the training performance. The results of the paper are obtained by applying the proposed ANN model to forecast load data in South Korea.

 Keywords: Artificial neural network, long