Performance of Genetic Algorithm-Support Vector Machine (GA-SVM) and Autoregressive Integrated Moving Average (ARIMA) in Electric Load Forecasting

Abstract

The main business focus of an electric power service provider is to meet the consumers’ demand in time and quality as required. The increase of electric load demand is influenced by various factors, for example the development of technology, business, region, standard of life, climatic and weather changes, or even consumers behavior. They must be considered by the power service provider in order to anticipate the load increase beyond the company’s capability and the existing power generator capacity. This study focuses on comparing the performances of two methods in electric load demand forecasting. The Genetic Algorithm-Support Vector Machine (GA-SVM) and the Autoregressive Integrated Moving Average (ARIMA) methods are applied for the prediction of daily load in Malang city, Indonesia, which is under the service coverage of the Indonesian national electricity provider, PT PLN Sub Unit P3B Jawa Timur-Bali. Two specific influencing factors, temperature and precipitation, are considered. The performance comparison is based on the error parameters of Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). The results of study indicate that the use of GA-SVM method provides better performance than that of the ARIMA method.