In recent years, a large area of power system blackouts occur frequently at home and abroad, and these accidents are caused by a chain of faults. Therefore, the analysis of its development mechanism and the prediction of its development path have become the focus of scholars concerned. At present, the prediction method is relatively single and the factors affecting the chain failure cannot be considered comprehensively. In this paper, a time-varying hidden Markov prediction model is proposed, which improves the state transition matrix in the Markov model. Branch protection/circuit breaker rejection, misoperation probability, hardware failure probability of system branch, branch service life and other factors are fully considered. Firstly, the probability matrix of fault state transition between branches is calculated based on the system topology. Secondly, the matrix is improved in different ways. Then, by comparing the predicted results of different models, the chain fault propagation path is obtained through optimization and improvement. Finally, IEEE 36 model is utilized to verify the feasibility and accuracy of the proposed method.