Short-term Electric Load Forecasting Based on Wavelet Neural Network, Particle Swarm Optimization and Ensemble Empirical Mode Decomposition

Abstract

Short-term load forecasting (STLF) is essential to the operation and planning of a utility company. In this paper a novel hybrid method for STLF based on ensemble empirical mode decomposition (EEMD), wavelet neural network (WNN) and particle swarm optimization (PSO) is proposed. One of the drawbacks of EEMD is in the sifting process, which carries through the decomposition of irrelevant intrinsic mode functions (IMFs). In this paper we present a new threshold approach based on the Brownian distance covariance to select the real IMFs. The model proceeds in four steps: First, the signal is decomposed in its IMFs and one residual. Second, the new threshold is applied to identify the principal IMFs. Third, each IMF is fed into the WNN, and PSO is used to optimize WNN's parameters. Finally, sum up all the processed signals to obtain the forecasting results. The experimental results have been compared with two forecasting models; in which its results verify the higher accuracy of the proposed model.