Abstract: This article analyzed the historical accident data from 2013 to 2015, discovered the laws of the data and the influence of factors to the accident situation. Aimed to minimize the rescue cost, based on the P-median model, combined with the data analysis results and the station passenger flow, the line network structure, connection with other traffic modes, an emergency maintenance personnel allocation model based on the minimum cost of rescue cost was established. The model was solved by using Genetic Algorithm(GA) with MATLAB. The article analyzed the case of emergency maintenance personnel of the signal system in the network of the 3rd Branch Company of Beijing Subway Operation. 24 emergency stations were selected to meet the needs of hard constraints. The allocation results could meet the requirements of the actual maintenance and balance.