Intelligent substation inspection robot is of great significance for ensuring the safe, stable and reliable operation of power system, as well as improving the inspection quality and energy efficiency. In practice, the path planning algorithms of inspection robot are usually based on static map, which cannot effectively deal with the unknown substation information and the random factors of system. This paper proposes a simulation-based optimization method to improve the path strategy of inspection robot. Firstly, through the real-time observation in patrol, random sampling method is used to generate running samples for simulation optimization. Then, based on the given planning strategy, the simulation samples are used to improve the path strategy to ensure the path strategy performance while reducing the computational burden of the robot. The simulation results show that the proposed method can effectively improve the rule-based strategy and greedy strategy in grid environment.