1. School of Electric Power, North China University of Water Resources and Electric Power, Zhengzhou 450045; 2. Zhengzhou Skong Electric Technology Co., Ltd, Zhengzhou 450001; 3. State Grid Hebei Electric Power Co., Ltd, Handan Power Supply Company, Handan, Hebei 056000
Abstract：In order to improve the accuracy of short-term load forecasting, a combined forecasting model based on wavelet decomposition (WT), improved particle swarm optimization (IPSO) and BP neural network is proposed. Firstly, we use wavelet decomposition to preprocess the load data, and decompose the historical data into cd1, cd2, cd3 and ca3; then we use neural network to model and predict the decomposed wavelet sequence; finally, we use wavelet to reconstruct the final forecast of the load sequence. In view of the accuracy of BP neural network samples and to increase the convergence speed and stability of the neural network, the improved particle swarm optimization method is used to optimize the network, forming a “decomposition prediction reconstruction” model. Compared with the wavelet decomposition BP neural network method, it has stronger training and learning ability, faster convergence speed, high prediction accuracy and strong adaptability.
2021, Vol. 22 
