Abstract: This paper proposes a novel identification method of weak lines in power grid based on double Q-learning（DQL） algorithm. Firstly, the Markov decision process was formulated for the weak lines identification problem considering the power grid transient stability. Then, the DQL was adopted to decide the line fault that most likely lead to the transient instability through the interaction with power grid transient simulation. Finally, a line weakness index based on the Q values was proposed to figure out the weak lines which have great influence on the transient security and stability of power grid. In order to improve the stability and training speed of the algorithm, the Q target table was introduced to DQL to separate the selection and assessment of line fault, and the prioritized experience replay was also adopted. The simulation results of 10-machine-39-bus system and 16-machine-68-bus system showed that the proposed weak line identification method based on DQL algorithm can effectively identify the weak lines of power system considering the transient stability constraint with fewer simulation times.