Enhanced multi-energy load forecasting via multi-task learning and GRU-attention networks in integrated energy systems

With the diversification of users’ energy demands, accurate load forecasting provides strong support for the planning, scheduling, and operation of energy systems. However, due to the coupled nature of energy production and consumption, a single load forecasting method cannot effectively and accurately predict multi-energy load systems. Therefore, this paper proposes a multi-energy load forecasting method based on multi-task learning and GRU-Attention Networks (MTL-GAN). First, a GRU-based network is employed to extract complex high-dimensional features and capture temporal dependencies from the historical sequences of multi-energy load data. Considering the strong correlations between different loads, a parameter-sharing layer is introduced to extract the inherent multi-energy coupling relationships using hard-weight sharing. Then, a residual network combined with a convolutional block attention module is proposed to capture additional spatial coupling features, with each task separately extracting local and long-term spatiotemporal dependencies to forecast cooling, heating, and electrical loads. Finally, the proposed MTL-GAN method is applied to the IES load data from Arizona State University’s Tempe campus and compared with several state-of-the-art methods. The results show that MAPE, RMSE, and WMAPE decreased by an average of 43%, 51%, and 50%, respectively, demonstrating its efficiency and superiority.