Ensemble empirical mode decomposition based electrical power demand forecasting for industrial user

Abstract

Aiming to control the industrial power demand characterized by strong fluctuation and impact, this paper studies the multi-step forecasting problem of ultra-short term demand load. Based on the integrated empirical mode decomposition method, the signals with different frequencies are effectively separated by decomposition. Then, the long short memory neural network is used to independently predict different signal subsequences, and finally the subsequence prediction results are combined. The experimental results show that the proposed method can well predict the industrial demand load, and the indices of prediction accuracy, such as MAPE, MAE, and NRMSE, are all controlled within 2%, and are significantly better than several classical time series prediction model, as well as the latest literature algorithms. The transfer error is also eliminated in the method, which represents good prediction accuracy and stability to meet the demand of demand control.