A Machine-Learning-Algorithm Enhanced Multi-Functional Gas Sensor for Self-Humidity Compensation and Partial...

Area of Science:

• Materials Science
• Sensor Technology
• Electrical Engineering

Background:

• Gas-insulated switchgear (GIS) is susceptible to partial discharges (PDs) in high electric fields.
• Nitrogen dioxide (NO2) concentration is a key indicator of PD type and fault severity.
• Environmental humidity significantly impacts GIS insulation and NO2 sensor signals, necessitating simultaneous detection and signal decoupling.

Purpose of the Study:

• To develop a self-calibrated sensor system for simultaneous humidity and NO2 detection in GIS.
• To decouple humidity interference in NO2 sensing for accurate PD analysis.
• To achieve real-time, humidity-adaptive calibration for precise fault diagnosis.

Main Methods:

• Development of a multifunctional WS2/ZnO sensitive material.
• Implementation of a novel self-humidity compensation algorithm, DF-MT1DCL, combining 1D-CNN and LSTM.
• Testing the sensor's response to NO2 and humidity across wide detection ranges at room temperature.
• Application of the sensor-algorithm system to PD monitoring in a simulated GIS environment.

Main Results:

• The WS2/ZnO sensor demonstrated simultaneous response to NO2 (100 ppb-10 ppm) and humidity (10.8-94.3% RH).
• The DF-MT1DCL algorithm achieved high prediction accuracy for humidity (R²=99.1%) and NO2 (R²=93.5%).
• The integrated system achieved 100% classification accuracy for PD types in a simulated GIS.

Conclusions:

• The WS2/ZnO sensor combined with the DF-MT1DCL algorithm offers robust resistance to humidity interference for NO2 detection.
• This synergistic approach enables accurate PD type recognition, advancing intelligent sensing for power equipment health monitoring.