Analysis of Short-Term Load Forecasting Using Artificial Neural Network Algorithm According to Normalization and Selection of Input Data on Weekdays

Abstract

Short-term load forecasting (STLF) is essential for stable and efficient power system operation. Recently, many papers have been published on application of STLF using artificial neural networks (ANNs). Input data selection, normalization method and the number of hidden neurons are very important factors in modeling ANNs. In order to improve the accuracy of STLF using ANNs, several input data selections and various input data normalization methods are analyzed. The past load, the past temperature and the temperature of the forecasting day are used as input data for STLF. In the case studies, the accuracy of forecasting in ANNs with various normalization methods and several input data selections are compared. Result of case studies show that the accuracy of STLF is good when using the maximum-minimum normalization and using the input data selection of recent two days.