Abstract
Accurate and real-time detection of characteristics gases is crucial for Dissolved Gas Analysis (DGA) in electrical transformers. Conventional assessments of Metal-Oxide Semiconductor (MOS) sensor selectivity generally analyze the response to singular interfering gases at identical concentrations, failing to represent the realistic intricacies of multi-component gases environments in DGA. Cross-sensitivity complicates the assessment of selectivity. This study proposes a method for re-assessing the selectivity of MOS sensor by considering cross-sensitivity in multi-component gases. Utilizing the Kendall correlation coefficient to analyze the relationship between characteristic gas concentrations and dynamic sensor responses, we determined sensor selectivity, optimal operating temperatures, and sensor interchangeability. Furthermore, we use a new term of quasi-2D sensor characteristics, which facilitated the creation of a complementary sensor array that significantly enhances gas selectivity and effectively mitigates MOS sensor saturation challenges. The complementary array, assessed by Linear Discriminant Analysis (LDA) projections and kernel density estimation, demonstrated enhanced selectivity relative to previous sensor arrays in the detection of multi-component gases. This method not only improves sensor selectivity but also mitigates cross-sensitivity in multi-component gas environments. Future research could focus on assessing a substantial array of MOS sensors, considering particular cross-sensitivity backgrounds, to improve artificial intelligence algorithms for accurate and real-time DGA detection.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call