Quantitative Detection of Sulfur Hexafluoride Decomposition Product With a Gas Sensor Array

• Published in: IEEE sensors journal 2024-10, Vol.24 (20), p.33160-33170
• Main Authors: Li, Tengfei, Tan, Zejie, Tian, Bing, Zhao, Liang, Liu, Zhong, Nie, Shaoxiong, Wang, Hairong
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Carbon monoxide; Cost analysis; Decomposition; Evaluation system; Gas chromatography; Gas detectors; Gas sensors; Gases; Metal oxide semiconductors; microelectromechanical system (MEMS) metal oxide semiconductor (MOS) sensors; Microelectromechanical systems; Micromechanical devices; Multilayers; Qualitative analysis; quantitative detection; sensor array; Sensor arrays; Sensor phenomena and characterization; Sensor systems; Sensors; Sulfur; Sulfur dioxide; Sulfur hexafluoride; sulfur hexafluoride (SF₆) decomposition products; Switchgear
• ISSN: 1530-437X; 1558-1748
• DOI: 10.1109/JSEN.2024.3450884

Detection and analysis of the sulfur hexafluoride (SF6) decomposition product is an effective way to monitor the operating status of gas-insulated switchgear (GIS) equipment. To overcome the large size, high cost, and difficult maintenance of the traditional detection methods like gas chromatograph, and photoacoustic spectroscopy, a sensor array composed of microelectromechanicalical system (MEMS) metal oxide semiconductor (MOS) gas sensors was proposed to detect its response to different concentrations of carbon monoxide (CO) and sulfur dioxide (SO2) in SF6 background in this work. The Kendall coefficient was employed to evaluate the selectivity of MEMS sensors to characteristic gases. A number of algorithms were used and compared in qualitative and quantitative detection, and results indicated all the algorithms had classification accuracy of over 90%. Among them, the multilayer perception (MLP) model was the best one for concentration prediction, with which {R} ^{{2}} of 0.90 and 0.74 can be realized for CO and SO2, and RMSE are 0.91 and 0.79 for them, respectively. After comparing prediction results of the MEMS sensor array tested both in air and SF6 backgrounds, the MEMS sensor array was verified to be more suitable for quantitative detection of CO and SO2 in the SF6 background, and its qualitative recognition and quantitative measurement accuracy could be improved by 29.86% and 38.64%, respectively. Due to the small size and high prediction accuracy, the MEMS sensor array has the potential for large-scale deployment in online monitoring of GIS operations.