The switching power supply of nuclear power plant is responsible for the conversion, processing and regulation of electric energy. In the nuclear power application scenario, its working environment is very special and complex. When the performance of the switching power supply deteriorates, it can not only cause failures in other components within the system, but also potentially lead to serious production accidents and safety accidents. Therefore, the reliability of the switching power supply plays a decisive role in the safe and stable operation of the nuclear power unit system. With the wide application of switching power supply in nuclear instrumentation control systems, the demand for state evaluation of switching power supply is increasing rapidly. This paper proposes a state evaluation method for switching power supplies based on the extraction of fault characteristic parameters. Characteristic values are obtained from voltage, ripple, oscillation and other signals. Based on the large number of eigenvalues obtained above, a failure model of switching power supply is established, and fault characteristic parameters are processed based on BP (Back Propagation) neural network to evaluate the state of switching power supply. The research presented in this paper can accurately monitor the degradation status of switching power supply, ensure the safe and stable operation of nuclear power system, and have significant engineering significance for the fault monitoring and health management of switching power supply in nuclear power plant.