Comparison of Machine Learning Based Methods for Residential Load Forecasting

Abstract

Load forecasting is a vital factor in electrical power systems due to technical and economic implications. Its forecasting is, however, a very challenging task due to many variable factors like occupant behavior, household size, loads present in the household and many other factors. In this article, three different machine learning algorithms have been applied for load forecasting on a load dataset of an Estonian household which was measured for a whole month. The simulation results of the Support Vector Machine (SVM) based forecasting model gives best results when compared with the real data. The validation results indicate that it has the lowest Root Mean Square Error (RMSE) value of 187.73 which is 28% and twice less than tree based and linear regression, respectively.