To solve the problem of power system outages caused by extreme weather,a reasonable strategy for restoring the power supply is needed to reduce losses. Therefore,a dynamic optimization method of black start recovery path considering the probability of device failure is proposed. Firstly,to accurately measure the probability of line fault under disaster conditions,a line fault model is founded by taking into account the spatiotemporal coupling of typhoons,rainfall,and landslides. Secondly,the black-start power evaluation model is established by combining the optimal weights of motor characteristics. Furthermore,the online recovery strategy is developed to determine the black-start recovery scheme,which can be solved by the improved Monte Carlo tree search algorithm and the Dijkstra algorithm. Finally,the IEEE 30-node distribution network system is used to verify the applicability of the dynamic decision system under different scenarios. The results show that the dynamic decision system can reduce the riskiness of the recovery scheme by 22.5% while keeping the power supply recovery time stable,effectively improving the rapidity and reliability of the power supply scheme.