Abstract: Traditional data-driven fault diagnosis methods are difficult to accurately diagnose the faults of reactor coolant system(RCS) for nuclear power plant(NPP) in the noisy environment. To address this issue, a gated recurrent unit(GRU) model for fault diagnosis of the RCS for NPP was established via the technical routes: firstly, the initial gated recurrent unit(GRU) model for fault diagnosis of the reactor coolant system(RCS) for nuclear power plant(NPP) was established using GRU method. Then, the initialization parameters for the GRU model were modified by time-based back propagation and adaptive moment estimation optimization algorithm, and the GRU model for fault diagnosis of the RCS for NPP was developed. Furthermore, the GRU model for fault diagnosis of the RCS was applied in the fault diagnosis of the RCS. Finally, the effectiveness of GRU model was validated through the comparative analysis of the diagnostic accuracy and robustness of the GRU, back propagation neural network(BPNN), support vector machine(SVM), and extreme gradient boosting(XGBoost) model for the fault diagnosis of the RCS. The research showed that the developed GRU model for fault diagnosis of the RCS for NPP can accurately diagnose the faults of the RCS in the noisy environment.