Multi-energy Micro-grid is an important way to implement the stable and optimal operation of a“double-high” power system. However

there are differences in energy endowments between MEMG systems

and it is not clear how to effectively synergize the differences in energy endowments between microgrids in each region to fully utilize their energy advantages. Therefore

this paper firstly refines the knowledge of the characteristics of multi-energy microgrids

such as price elasticity and uncertainty

and inputs this knowledge into the subsequent neural network model. Secondly

considering that the opaque decision-making process of the neural network model will pose a threat to the system stability

this paper embeds the MEMG regulation knowledge in the form of a weight-decreasing loss function to guide the training of the neural network model. Subsequently

due to the MEMG presenting the characteristics of multi-energy coupling and data heterogeneity

an amalgamation distillation cooperative interconnection regulation model consisting of multiple teacher networks is constructed so as to optimize the overall performance of the model. Finally

this paper employs multiple interconnected and improved MEMG22 nodes to validate that the data-driven model with knowledge guidance is less risky compared to the existing model. The MEMG regulation model using merged knowledge distillation not only optimizes the model runtime results but also reduces the computation time.