By analyzing the mechanism of cascading failure， the topological relations of network structure， the self-critical features of lines in the transfer of system flow， combining with the probabilistic characteristics of various stages of cascading failure， a probabilistic assessment model of power system cascading failure based on line clusters is proposed. Through means of simulating cascading failure process， the indices such as overload and loss of system-load are generated to evaluate risk of cascading failure， in order to find the vulnerable spot of power system and then to generate prevention strategies to reduce the failure risk. Taking IEEE 39-bus system and Gansu power network as examples， the feasibility and effectiveness of the proposed probabilistic assessment model are validated.