In order to obtain the optimal capacity ratio of the micro power sources in the microgrid, and satisfy the output demand of the load, this paper establishes a capacity optimization configuration model for the grid-connected microgrid with wind, solar, diesel and battery, and takes the lowest comprehensive operating cost as the objective function, and the distributed power output and pollutant emissions as constraints. Using refracted opposition-based learning strategy, differential mutation, cross selection strategies and the dynamic step factor to improve the standard sparrow search algorithm to solve the model, and comparing with whale optimization algorithm, differential evolution, gray wolf optimizer, and sparrow search algorithm. Two typical days in Ningxia are selected for analysising of calculation examples, the required cost is 3.05%, 4.12%, 8.46% and 1.13% lower than the other four algorithms respectively. The simulation results show that the proposed model is reasonable, and the improved sparrow search algorithm has better optimization ability.