Short-Term Power Load Forecasting Model Based on EEMD-SE-ERCNN
Mingui Yang, Yong Liu, Yu Mu, Zhibo Li, Honqin Zhang, Wenhao Chen, Fei Rong · 2024
Given the strong nonlinearity and multiple influencing factors in power load data, this paper proposes a forecasting method based on ensemble empirical mode decomposition (EEMD), sample entropy (SE), and an enhanced residual convolutional neural network (ERCNN). First, EEMD is used to decompose historical loads into several simplified subsequences. Next, Sample Entropy (SE) is introduced to calculate the entropy values of the subsequences, and subsequences with similar entropy values are reconstructed into random, detailed, low-frequency, and trend components. Then, ERCNN with different structures are used to forecast different types of components. Finally, the different components are superimposed to obtain the final forecasting results. Experimental results show that the proposed model outperforms existing models in terms of forecasting performance, with the forecasting error reduced by up to 1.62%, 178.4 MW, and 272.8 MW compared to existing deep learning models.