Low-sample fault diagnosis method of high-voltage circuit breaker based on Deep Learning

Abstract

High-voltage vacuum circuit-breakers are essential for power systems. They ensure a stable power grid. Due to their complex nature, vacuum circuit breakers are susceptible to failure over time. Failure can occur due to operational, faulty, contact or actuator problems. This paper presents a fault diagnosis method for high voltage vacuum circuit breakers based on CGAN and CNN. It addresses the lack of fault data and limited diagnostic accuracy. The operating principles of high-voltage vacuum circuit breakers are first examined, followed by the construction of a platform to replicate fault scenarios. The data will be collected to create datasets on different types of faults. CGAN generates diverse fault data to improve model generalisation. A one-dimensional convolutional neural network is used for fault diagnosis. This model identifies and classifies a wide variety of fault types in high-voltage vacuum circuit breakers with an accuracy rate of over 98%. This ensures safe operation of power systems.

-