In order to solve the problem that the existing anti-misoperation strategy ignores the multi-dimensional indicators of power grid operation and can not prevent potential misoperation, a multi-level and multi-dimensional anti-misoperation intelligent discrimination method for power grid based on rule engine and deep learning is proposed. Through the rule engine-driven misoperation identification method, taking full account of the dimensions of command execution, equipment status and personnel operation behavior, the minimum perfect hash table is screened out of useless events to improve the identification efficiency, and the multi-level potential misoperation identification method based on hybrid in-depth learning is used to analyze and process historical operation data in depth and identify potential misoperation. It can also effectively supplement and update the rule base of the rule engine, and improve the automation level of the system. The simulation results show that the proposed method can increase the success rate of misoperation recognition by 2.10% and 3.06% compared with the two comparison algorithms.