Short-Term Power Consumption Dynamic Forecasting Using Optimal Weighting Combination

Short-term load forecasting is critical for grids to plan and execute demand response methods. The method proposed in this paper is a dynamic ensemble method based on the optimal weighting combination method. The model integrates different statistical models, machine models, and deep learning models to avoid the instability of single model prediction. By evaluating the impact of each candidate model on the overall error, the weights are dynamically adjusted. Finally, the load data from southeast China are used for verification, and the results show that the proposed method has better robustness and superior generalization.