Short-Term Load Forecasting for Virtual Power Plants Based on Improved Laguerre Orthogonal Neural Network Algorithm

In this paper, a method for offline analysis of the short-term load method of virtual power plant based on the combination of AR-GARCH model and improved Laguerre neural network combination algorithm (LOBNN) is proposed. Firstly, the combination modelling of optimal AR-GARCH model and improved Laguerre neural network is established. The trained correlation matrix is screened to get the strong correlation factors affecting the load changes, and data preprocessing and correlation analysis are performed. Secondly, the LOBNN model is improved by improving the weights solving algorithm, which is used to enhance the flexibility of the model. Then, a new iterative algorithm is designed to further improve the prediction accuracy of the model, and the new LOBNN model is obtained. Finally, the data are adjusted by substituting the AR-GARCH model and the improved LOBNN model, obtaining two sets of predicted values for the input data, comparing the output multivariate load prediction results with the real values and calculating the loss function, and adjusting the network parameters according to the loss function generation by generation.