Passive containment cooling system is innovatively used in AP1000 reactor design to enhance the safety. Since the system operation is based on natural circulation, physical process failure induced by uncertainties of physical parameters becomes one of the important failure modes (e.g. natural circulation cannot establish or keep and system design function cannot be accomplished because some parameters such as air temperature deviate from their design values), which should be considered in system reliability evaluation. As the heat sink, air temperature with high uncertainty has important effect on system reliability. In this article, we analyze the pressure variation in the containment along with the air temperature based on system thermal–hydraulic model, and the effect of air temperature on system operation is closely related to the thermal–hydraulic performance of the system. Moreover, the system thermal–hydraulic capacity is influenced by the system component configuration, so we evaluate the system physical process failure probability by Monte Carlo simulation and analyze the effect of air temperature distribution under different system component configurations. Finally, we evaluate the whole system reliability considering the logical relationship between physical process failure and equipment fault by fault tree method. The results illustrate that air temperature distribution has important influence on the system reliability, the system failure probability may be difference by several orders and the main contributors may be different at different plant locations, so climate should be considered in system design and reliability analysis.

In a nuclear power plant, the consequences of a multi-unit event occurring concurrently are more serious than those of a single-unit event. The first step in the probabilistic safety analysis of multi-units is to analyze the initiating events and calculate the frequency of initiating events for simultaneous events of multiple units. The difficulty in using the fault tree model is that the known data are all frequency data from a single unit and cannot be logically multiplied. In this paper, taking a dual unit as an example, we used the formula to convert the probability of failure of the second unit within 72 h and then build a fault tree model. After analyzing the results of the dual unit, the most frequent cut set was the common cause of failure of the main transformer and of the switching failure of the main and auxiliary external power. The final calculation of the frequency of simultaneous loss of off-site power events for the dual units within 72 h was 3.22 × 10⁻⁴/year. After comparing with the single-unit results, it was found that the common cause failure of each unit’s independent equipment was the main reason for the occurrence of a loss of off-site power. Shared equipment in a single unit was ranked low in all the cut sets (such as the stability of the external grid for the main and auxiliary power systems) but was ranked high in multiple units. The calculation results of the frequency of initiating events of double units were two orders of magnitude lower than those of a single unit. However, the consequences of simultaneous events of multiple units were higher than those of single reactors. Therefore, attention should be paid to the risk of a simultaneous loss of off-site power event of multiple units.