Abstract: To effectively deal with the influence of multiple uncertainties such as renewable energy output and load demand on the safe and stable operation of multi-energy microgrid system, this paper proposes a two-stage distributionally robust optimization model for multi-energy microgrid system considering covariate factors. First, based on the distributionally robust optimization method, a two-stage optimal scheduling model of a multi-energy microgrid system including photovoltaic generation units, combined cooling, heat and power units, cold, heat and electric load and thermal energy storage is constructed. Then, considering covariate factors, a Wasserstein ambiguity set based on multivariate decision tree regression is established to describe the internal relationship between the uncertainty of source and load, and between the uncertainty and covariate factors. Next, using linear decision rules and duality theorem, a mixed integer linear programming form of the model is given. Finally, the model is applied to a modified 33-node multi-energy microgrid system for example analysis. The results show that the introduction of covariate factors can effectively improve the economy of the model compared with the classical robust optimization model and the distributionally robust optimization model. In the Monte Carlo out-of-sample test, the proposed two-stage distributionally robust optimization model shows good reliability in the face of uncertainty fluctuations.