Abstract: To improve the speed of risk assessment model of power system cascading failure and solve the difficult problem of on-line application, a rapid risk assessment model based on dangerous branches set is proposed in this paper. The first K shortest paths between two endpoints are searched by using bidirectional search algorithm, which can determine the flow transfer range after branch outages. The criterion of dangerous branches is defined by considering the influence of branch flow increment and flow over-limit degree, so the branch that causes cascading failure can be recognized rapidly by the first K shortest paths and the next safety analysis object can be determined. The range of safety analysis can be reduced by above two steps, which improves the risk assessment speed. The simulation on IEEE 39 system proves that the model has faster risk assessment speed and higher assessment accuracy, which is easy to realize on-line application.