基于MTL-GRU-Attention的综合能源系统多元负荷预测

Multivariate Load Forecasting of Integrated Energy System Based on MTL-GRU-Attention Mechanism

针对综合能源系统多元负荷预测问题,提出一种基于多任务学习、门控循环单元和注意力机制的多元负荷预测方法。首先,运用门控循环单元建立多任务学习的共享层,充分挖掘冷、热、电负荷之间的耦合特征;然后,利用贝叶斯优化算法实现门控循环单元最优超参数的自适应选择;最后,使用注意力机制实现子任务对共享层中重要特征的差异化提取,以增强关键信息的影响。以亚利桑那州立大学坦佩校区的实测负荷数据作为算例,结果表明所提模型具有更高的预测精度。

Aimed at the problem of multivariate load forecasting of an integrated energy system(IES),a multivariate load forecasting method based on multi-task learning(MTL),gated recurrent unit(GRU),and Attention mechanism is proposed.First,a shared layer of MTL is established by using GRUs,and the coupling characteristics between cold,heat,and electric loads are fully exploited.Then,the Bayesian optimization algorithm is used to realize the adaptive se⁃lection of optimal hyperparameters of GRU.Finally,the Attention mechanism is used to realize the differential extrac⁃tion of important features in the shared layer by subtasks to enhance the influence of key information.The measured load data of the Tempe Campus of Arizona State University is taken as an example,and results show that the proposed model has a higher prediction accuracy.