A hybrid FCW-EMD and KF-BA-SVM based model for short-term load forecasting

Abstract

This paper proposes a hybrid short-term load forecasting method, which is based on the fuzzy combination weights as well as the empirical mode decomposition process (FCW-EMD), and support vector machine optimized via the Bat algorithm as well as the Kalman filtering process (KF-BA-SVM). The subjective weight is presented as a new theory and is applied to capture the inherent correlation effectively among hourly loads. Based on the proposed objective weights and subjective weights, the fuzzy combination weights theory (FCW)-a new similar day selection theory is presented, which improves the accuracy of the similar day selection, and correspondingly, makes the original data for EMD processing decrease dramatically. BA is introduced to optimize parameters of the SVM model for further improving the forecasting accuracy. Using the decomposed load series via empirical model decomposition (EMD) as inputs to SVM and further correcting the output of SVM via KF, a hybrid FCW-EMD and KF-BA-SVM short-term load forecasting method is established. Numerical case studies on the load forecasting of a transformer substation in south China show that the proposed hybrid forecasting model outperforms other forecasting methods and effectively improves the prediction accuracy.