Microgrid is an effective solution to solve the problem of grid connection of distributed power generation by effectively using renewable resources such as local scenery. The configuration of microgrid directly determines its economy and power supply reliability. Considering the influence of wind speed and light intensity with strong randomness and load fluctuation on the microgrid,based on the theory of multi-state system, the multi-state modeling of wind power output and load at each time of typical day in each month is carried out. Aiming at economic and environmental protection, an improved non inferior sorting genetic algorithm is introduced to solve the multi-objective optimal allocation model. The rationality and effectiveness of the optimal allocation model based on the multi-state system theory are verified by an example of integrated energy system engineering in a province of East China.