Diagnosis of Spark Discharge Faults by Characteristic Gas Composition Detection

Feng Jing,Jiakun Wang,Lei Sun,Jiandong Jin,Xingang Wang,Qiongyaun Wang,Jifeng Chu,Aijun Yang

DOI: https://doi.org/10.1109/cieec58067.2023.10166146

2023-01-01

Abstract:Due to internal electric field distortion, SF6-insulated power equipment will generate partial discharge in the operation process. Thus, the operating state of power equipment could be assessed rapidly by the detection of SF6 discharge decomposition products. Semiconductor sensors have enormous advantage of rapid response, small size and high sensitivity, which can be easily integrated in portable equipment. In this paper, a sensor array consisted of 8 commercial semiconductor sensors was employed to detected the SF6 discharge decomposition products. The response-recovery experiments were performed over 12 days to demonstrate the long-term stability. Furthermore, the correlation between sensor response and the discharge types is proved by the two-dimensional t-sne scatter diagram. Ultimately, the classification accuracy of KNN after 10-dimensional features extracting can reach 92.36%. To summarize, this work explored the relationship between sensor response and discharge faults, which also proved the enormous potential for the recognition of SF6 discharge faults.