Energy Management Decision-Support System for Energy Internet Enterprises Based on Machine Learning

This research addresses the issue of energy management decision - making support for the integrated energy system of energy internet enterprises and proposes an energy demand forecasting method based on the Transformer-GNN model. By combining the global dependency capturing ability of the Transformer and the spatial information mining ability of the graph neural network (GNN), the model can effectively handle the complex coupling relationships among electricity, cooling, and heating loads and take into account the impact of user demand response behavior on energy consumption. The research utilizes the actual data of an energy internet enterprise, including historical data on electricity, cooling, and heating loads, electricity price information, and meteorological data, to construct a demand response model and a load forecasting model. The experimental results indicate that the Transformer-GNN model considering demand response demonstrates high accuracy and stability in forecasting electricity, cooling, and heating loads.