Abstract： Aimed at the problem of fault diagnosis in the hybrid system, a particle filter algorithm for hybrid system state estimation and discrete modal identification was proposed. The proposed algorithm was spread to joint estimation of state and parameters, modified Bayes algorithm was utilized to give fault judgment, and fault diagnosis of hybrid system was implemented. The results of the simulation of the twotank state estimation and fault diagnosis show that the proposed algorithm can not only accurately diagnose the hybrid system fault, maintain better state and parameter estimation accuracy when fault occurs, but also be used for adaptive filtering, reliability prediction, fault tolerant control and many other fields.