The invention relates to an electric system weak circuit identification method. Based on the Q learning algorithm, the identification method comprises searching different fault tangent combinations byadopting epsilon-greedy strategy according to the current running state information of a power grid, and computing the system transient safety stability under the fault tangent combination given by the verification through the power grid transient stable simulation; and then the Q values of different tangent combinations are continuously updated by combining a transient stable simulation computation result through the Q learning algorithm, thereby screening out the tangent combinations easily causing the system instability; and based on the Q values of different tangent combinations, a circuit weakness index is proposed, and the weak circuit considering the power grid transient safety stability is obtained through computation. The weak circuit considering the system transient safety stability can be effectively distinguished by utilizing the Q learning algorithm, the required simulation frequency is much less than the fault traversal identification method, and the workload of the power grid running experts and the technicist for performing stability analysis and tidal current adjusting under different running ways can be greatly reduced.