In order to balance the interests of various participants in a microgrid system containing two forms of energy, electricity and heat, a microgrid energy management model based on the improved grey wolf algorithm was proposed. Firstly, the microgrid structure and the functions of various entities within the microgrid were analyzed. In order to comprehensively consider the decision-making ability of source-grid-load, the master-slave game method was applied to the interaction among energy producer, microgrid operators, and load aggregators. Secondly, to address the characteristics of high dimensionality and nonlinearity in the upper-layer model, tent mapping was used to initialize the population, a nonlinear convergence factor is employed to balance the population search capability and the Levy flight strategy was utilized to reduce the risk of falling into local optimum. In the process of model solving, the improved gray wolf algorithm was used for the upper-level, and quadratic programming methods were used for the lower-level. The combination was explored to discover the best strategy that maximize the interests of each entity. Finally, the efficiency of the algorithm and the superiority of the proposed model in improving the participants' revenue and smoothing the load distribution are verified by an arithmetic example.