Abstract: In order to improve the reliability of the secondary circuit control system, a steam generator is taken in this paper as the research object, and a fault tolerant control system of the water level is accordingly designed based on BP neural network.Fault tolerant control simulation is also conducted on the Matlab/Simulink platform, where the lumped parameter dynamic models of the steam generator are joined, and the calculated water level of the steam generator, with the fault of constant deviation, constant gain, or locked, is simulated respectively on the water level sensor.The experimental results show that the sensor faults could be diagnosed accurately and rapidly with the fault diagnosis model, and the fault tolerant control should be implemented when sensor faults occur.Meanwhile, the water level of the steam generator could be regulated close to the target value, and the system works steadily.In brief, the proposed fault tolerant control system based on BP neural network is effective to correct the water level sensor fault of the steam generator, and the system can be applied into the design of the steam generator water-feeding control system, which significantly improves the system reliability.