Research on Predictive Methods Considering the Inherent Coupling Mechanisms of Multienergy Sources and Loads in Campus Systems

As integrated energy systems (IESs) evolve to meet growing demands for sustainable energy, accurate forecasting of multisource energy loads becomes critical. This study proposes a forecasting framework integrating Bi‐directional long short‐term memory (BiLSTM) networks with multitask learning (MTL) to jointly predict electricity, cooling, heating, and photovoltaic (PV) loads. Unlike single‐task approaches, the MTL framework enhances performance by leveraging shared information. To improve robustness, empirical mode decomposition (EMD) denoises nonstationary time‐series data, and the snow ablation optimizer (SAO) fine‐tunes hyperparameters. Experimental results on real‐world campus data show that the EMD–MTL–BiLSTM–SAO model outperforms existing methods, offering superior accuracy and reliability. These results demonstrate the model’s potential for optimizing future low‐carbon energy systems.