Abstract: Hualong Pressurized Reactor（HPR1000） is the advanced Generation-Ⅲ nuclear power plant, which is independently developed by China with complete intellectual property right. The containment spray system（CSS） is one of the important safety systems for HPR1000, so it is essential to carry out the reliability assessment for the CSS of HPR1000. However, the traditional reliability assessment methods have some shortcomings in dealing with the event polymorphism and the non-deterministic logical relationship between events. In order to solve this problem, firstly, the failure modes, failure causes, and failure effects for the system components were analyzed by failure mode and effect analysis（FMEA） method. Then, the fault tree（FT） which could describe the deterministic logical relationship between failure modes, failure causes, and failure effects was established by fault tree analysis（FTA） method, and finally, the Bayesian network（BN） method was used to express the event polymorphism and non-deterministic logical relationship between events by correcting the conditional probability tables based on the FT, and the reliability assessment model for the CSS of HPR1000 was established. The established reliability assessment model was applied to calculate the failure probability of CSS of HPR1000 in East China and the results were 1.787e-05. In addition, the orientation for improving the CSS reliability of HPR1000 was clarified by identifying all its critical events.