SF6 gas leakage identification system for GIS equipment based on infrared imaging image enhancement

Abstract This study proposes a GIS equipment SF6 gas leakage identification system based on infrared imaging image enhancement. It combines Transformer and generative adversarial network technology to significantly improve the accuracy and real-time performance of leakage detection. Experimental data show that within the typical leakage interval of SF6 gas concentration of 20–60 ppm, the signal-to-noise ratio of the system linearly increases from 20 dB to 75 dB and sharply increases from 30 dB to 170 dB, with R 2 exceeding 0.95, confirming the relationship between concentration and enhancement effect. strong correlation. Compared with the traditional method, the recognition rate of infrared imaging enhancement technology in the low concentration range is 33%–100% higher than that of the traditional method, and the recognition rate advantage of 40% is stably maintained in the whole concentration range. The comparison of image enhancement algorithms shows that the method based on deep learning has a sharpness of 42.7, but the computational complexity is high. The system response time test shows that the image acquisition module only needs 50 ms, while the processing module takes 200 ms. In dynamic environment, the VCE enhancement algorithm performs best in the speed range of 5–25 Hz, and the fluctuation range of recognition accuracy is 35 percentage points smaller than that of LSE method. In addition, the GIS module realizes the leakage location error < 1.5 K thermal gradient through spatial mapping technology, and the confidence level reaches 0.6. Combined with the diffusion path prediction model, it effectively supports emergency decision-making.