Research on Short-Term Power Load Forecasting Method Based on CEEMDAN-ChOA-GRU

Short-term load forecasting (STLF) have a significant meaning for the planning, construction, and stable operation of new power systems. However, traditional load forecast methods generally have the problems of insufficient accuracy and low model applicability. In view of this, an STLF method based on the combination of complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and chimp optimization algorithm (ChOA) is proposed. The CEEMDAN-ChOA-GRU model proposed in this paper integrates the advantages of CEEMDAN and ChOA, which can effectively realize the accurate forecasting of power loads. First, CEEMDAN-GRU is shown to have the best performance in the comparison of different mode decomposition methods to improve GRU in STLF. Furthermore, ChOA is used to optimize GRU hyper-parameters, gaining higher forecast accuracy than existing models. In the best case studied, the mean error of this method is less than 2%, and R-Squared (R
²
) is greater than 0.98, which shows good performance and excellent application potential.