Laser & Infrared, Volume. 54, Issue 3, 423(2024)
Multi-feature aggregated characterization of circuit breaker thermal fault diagnosis rating methods
In response to the demand for accurate assessment of infrared thermal fault characteristics of power equipment, a multi-feature aggregated characterization of circuit breaker thermal fault diagnosis rating method is proposed, and the data test is carried out using infrared images of high-voltage circuit breakers as examples. Firstly, on the basis of the background separation of high-voltage circuit breaker infrared images, the equipment is accurately divided into regions to extract the temperature information of each region. Secondly, the Mean-shift and the improved region growth method are applied to fuse and accurately extract the area of the fault heat-emitting region. Then, a multi-dimensional aggregated characterization matrix is designed to combine the heat-emitting area, hot spot temperature, hot spot temperature difference, heat-emitting location, temperature rise of two identical positions of the same equipment and other eigenvalues into a multi-feature vector matrix, and the on-site case data is adopted to construct a correlation library of this vector matrix and HV circuit breaker fault types, levels and treatment opinions. Finally, 1002 sets of multi-feature vectors from 350 infrared images of high-voltage circuit breakers are trained and tested. The results show that the F-measure and Kappa coefficients of the multi-feature vector data extracted by this method using GWO-SVM classifier test are 96% and 95.43%, respectively, which can achieve the all-types of diagnostic rating and accurate localization of thermal faults in high-voltage circuit breaker equipment.
Get Citation
Copy Citation Text
SANG Jin-hai, XU Zhi-hao, LI Hong-bin, KANG Bing, DING Gui-li, WANG Zong-yao, ZHANG Xing-wang. Multi-feature aggregated characterization of circuit breaker thermal fault diagnosis rating methods[J]. Laser & Infrared, 2024, 54(3): 423
Category:
Received: May. 18, 2023
Accepted: Jun. 4, 2025
Published Online: Jun. 4, 2025
The Author Email: XU Zhi-hao (zhxuhi@whu.edu.cn)