Given the problem of difficult localization of the error area on a large-scale power grid simulation, an error area identification method based on disturbance depth and minimum spanning tree is proposed. This method calculates the disturbance depth index based on the measured data, sets the corresponding threshold value, decouples and removes the small disturbance area, abstracts the remaining network as an undirected graph, and uses the simulation credibility index as the weight of the line to obtain the minimum power grid. The spanning tree and its branches correspond to the optimal cut set. The measured data is injected at the optimal cut set for hybrid simulation. Through multiple iterations of compression, identifying the error area is finally achieved. The effectiveness of the proposed method is verified by the IEEE10 machine 39-node system. The method can effectively reduce the cost of dynamic simulation verification, and provide important guidance for the next step of block decoupling and model parameter correction. This work is supported by the National Key Research and Development Program of China (No. 2018YFB0904500) and the Science and Technology Project of the Headquarters of State Grid Corporation of China (No. 18-GW-05).