A Multivariate Load Prediction Method of Integrated Energy Systems Based on MMoE-TCN

In order to ensure the stable and economic operation of the integrated energy system, accurate multivariate load forecasting is undoubtedly a necessary prerequisite. However, all kinds of energy in IES have different physical characteristics and influence on each other. Considering the lack of multivariate load data, this paper proposes a GRU-GAN data generation and MMOE-TCN multi-task learning load prediction method. Firstly, the GRU-GAN data generation model is established to generate massive data in accordance with the temporal law according to the limited historical load data, so as to achieve the purpose of data enhancement. Then, the MMoE multi-task learning model based on parameter soft sharing mechanism was constructed, and the shared feature information was reasonably allocated through expert subnets and gating units, and the sub-task model was constructed using temporal convolution network. Finally, the IES open data set of Arizona State University Tempe campus is used for example analysis, and the errors are lower than those of the comparison model, which verifies the superiority of the proposed method.