Forecasting medium-term electricity demand in Thailand: comparison of ANN, SVM, DBN, and their ensembles

Abstract

Electricity demand forecasting is an important research area, most of the research focuses on forecasting the electricity consumption that is the critical process for planning the electric utilities to avoid a blackout in peak time. This paper focuses on forecasting the medium term (1-month ahead and 1-year ahead) of electricity peak demand in Thailand by using three machine learnings and ensemble method. The machine learnings include artificial neural network (ANN), support vector machine (SVM), and deep belief network (DBN). For the comparative performance between each model, mean absolute percentage error (MAPE) is used as the measurement. The result implies that the ensemble model of ANN and DBN is the best method for 1-month ahead with MAPE 1.44%, and ANN is the best method for 1-year ahead forecasting with MAPE 1.47%.