Electric power data element trading price prediction model based on improved grid VMD Resnet-BiLSTM algorithm

Abstract In response to the problems of large fluctuations in electricity prices and difficulty in long-term forecasting caused by the large-scale integration of new energy in the electricity market, this paper constructs an improved power grid variational mode decomposition (VMD) residual shrinkage network bidirectional long short-term memory neural network (Resnet-BiLSTM) algorithm model (VMD Resnet-BiLSTM). Firstly, the maximum mutual information coefficient is introduced to optimize the Variational Mode Decomposition parameter, decomposing the original electricity price sequence into stationary intrinsic mode function components to reduce nonlinear interference; Secondly, the residual shrinkage module is used to pre train the convolutional neural network, extract deep features from the data, and generate a new dataset to reduce the learning node requirements of the network; Finally, combining the bidirectional time modeling capability, a Resnet BiLSTM hybrid network is constructed to improve the flexibility of feature extraction by scaling the translation factor through convolutional kernels. The experimental results show that in the comparative experiment of predicting Australian electricity price data from 2020 to 2024, the proposed model is significantly better than the selected baseline comparison model. Research has shown that the proposed model effectively improves the accuracy, stability, and robustness of electricity price forecasting through data decomposition and model parameter optimization.