Abstract: The online reliability evaluation of nuclear power unit in real time is the key to ensure the safe operation of the unit. In view of the fact that the static reliability analysis method cannot meet the online reliability requirements of conventional island water supply systems for nuclear power,an improved GO-FLOW reliability analysis for time correlation method is presented. At the same time,because the failure of the system components will cause the change of the reliability parameters,thus affecting the accuracy of reliability analysis. A fault diagnosis model of water supply system is established by using artificial fish swarms algorithm-support vector. When the system fails,the entropy weight Topsis and grey correlation degree are used to evaluate the state of fault components,and the component reliability parameters in the GO-FLOW reliability analysis method are modified,so that the online reliability evaluation of the water supply system has been realized. Simulation results show that the reliability of the entire water supply system can accurately evaluated online,and it has wide applicability.