Multi-Energy Load Forecasting for Integrated Energy Systems based on Causal-Wavelet Neural Networks

The efficient and stable operation of integrated energy systems, which encompass electricity, cooling, and heating devices, necessitates precise and effective multi-energy load forecasting methods. However, the significant fluctuations in multi-energy loads, coupled with the strong interdependence and complementarity among various energy sources, present considerable challenges for accurate forecasting within integrated energy systems. To address this, an in-depth analysis of the inherent causal relationships between load and influencing factors in conducted within each component following wavelet decomposition. Subsequently, an efficient forecasting model based on causality analysis and wavelet neural networks is developed for multi-energy load forecasting. This model integrates Spearman correlation with reversed information entropy causality inference for feature selection and employs a wavelet-neural network for individual component forecasting, utilizing a rapid training algorithm. Numerical testing using public datasets demonstrates that our proposed model significantly enhances the accuracy of multi-energy load forecasting.