The application provides a method and device for improving the rated current breaking times of a direct-current
solid-state
circuit breaker and application thereof, and aims at the problem that the performance of an existing direct-current
solid-state
circuit breaker is degraded after multiple current impacts on a
voltage-dependent
resistor, the
junction temperature of a power device is difficult to measure, and the breaking times are limited due to over-temperature damage, the application comprises a
junction temperature prediction module based on a
physical information neural network, an auxiliary
energy absorption module and a power device
voltage clamping module. The
junction temperature prediction module accurately predicts the junction temperature of the power device in real time by fusing a physical
thermal model and a neural network. The
voltage clamping module suppresses the turn-off
overvoltage. The auxiliary
energy absorption module is started when the junction temperature is close to the limit, and absorbs the remaining energy. The design reasonably allocates the turn-off energy, reduces the
impact times and intensity of the voltage-dependent
resistor, significantly improves the service life (such as about 10 times under the
impact of 200 A), and further greatly improves the rated breaking times of the
solid-state
circuit breaker, and is low in cost, high in reliability, and suitable for scenes such as a direct-current
power grid.