SHORT-TERM PEAK LOAD FORECASTING USING PSO-ANN METHODS: THE CASE OF INDONESIA

Abstract

The purpose of this study was to investigate the model for predicting the electricity load and usage in Indonesia. This study combined two artificial intelligence methods, one is particle swarm optimization (PSO) and the other one is artificial neural network (ANN). The combination of PSO and ANN (known as hybrid particle swam optimization algorithm (HPSO-ANN)) is attempted to obtain better short-term load forecasting accuracy, especially in the case of daily peak electrical load forecasting. Daily peak loads were analyzed using data from Indonesian state electricity company for West Java area in Indonesia from 2005 to 2012. Data were analyzed for every 30 minutes and classified into the types of days (i.e., weekdays (Monday to Friday), weekend (Saturday to Sunday), and national holidays). To measure the level of accuracy of the prediction, the simulation results were compared with feed forward back propagation methods and actual data from Indonesian power company. The HPSO-ANN method provided the best results with an accuracy level above 98 percent. This analysis provided information on how much waste of electrical energy could be reduced by selecting the appropriate strategies in forecasting according to the load and day characteristics.

Keywords

• Artificial neural network
• Back propagation
• MAPE
• Particle swam optimization
• Short-term peak load forecasting