Seasonal short-term load forecasting for power systems based on modal decomposition and feature fusion multi-algorithm hybrid neural network model

Abstract To enhance the level of refined load decomposition of the power system and make full use of seasonal change information to improve the prediction performance further, this paper proposes a seasonal short-term load combination prediction model of the power system based on modal decomposition and feature fusion multi-algorithm hybrid neural network model, i.e., the characteristics of load components are analysed for different seasons, and the corresponding models are established respectively. Firstly, The ICEEMDAN (Improved complete ensemble EMD) method decomposes the four-season load, and a new sequence is obtained based on the RCMFE (refined composite multiscale fuzzy entropy) reconstruction of each decomposition component. Secondly, the correlation between different decomposition components and different features is measured using the mRMR (Max-Relevance and Min-Redundancy) method to filter out the subset of features with high correlation and low redundancy. Finally, the BiLSTM (Bi-directional Long Short-Term Memory) model based on the Bayesian optimisation algorithm is used to predict different components of different seasonal loads separately, and the predicted values are accumulated to obtain the final results. According to the experimental findings, the proposed method can successfully balance the issues of prediction accuracy and prediction time while having a higher level of prediction accuracy than the current prediction methods. It is confirmed that the proposed method can effectively address the load power variation brought on by seasonal differences in different regions.