An intelligent fault diagnosis method for breakers based on instance segmentation and temperature probability density analysis

Abnormal thermal conditions leading to substation equipment failures can compromise the reliable and stable operation of power systems. To improve the efficiency and accuracy of thermal fault diagnosis for circuit breakers within smart grids, an automatic thermal fault diagnosis algorithm is proposed, utilizing advanced applications of both the Internet of Things (IoT) and artificial intelligence. Firstly, precise segmentation of breakers is achieved through improved instance segmentation, optimized for the characteristics of circuit breaker images. Compared with the baseline model, the mask [email protected] of the improved model increases from 0.667 to 0.793, with an absolute improvement of 0.126 and a relative improvement of 18.9%, while the floating-point operations (FLOPs) are reduced by 0.9 G. Comparison with other instance segmentation models shows that this model has the highest accuracy. Secondly, a thermal fault diagnosis algorithm based on kernel density estimation (KDE) and mean filtering is proposed. This algorithm utilizes KDE to establish the probability density function of circuit breaker temperatures. It identifies temperature parameters by locating extreme points on the function curve, and hot spot temperatures are extracted using mean filtering. The algorithm achieves fault diagnosis and fault point localization based on industrial application rules. The experimental results demonstrate that the proposed algorithm can effectively identify thermal faults in breakers. It offers a viable approach for automatically diagnosing thermal faults of high-voltage equipment within the framework of smart grid applications.