Volume 57

Research on CO2 Pipeline Leakage Detection Method Based on Lightweight Network Zhang Lijun, Li Yuchun, Nian Xingwang, Tan Qingxun, Wang Zongke, Zhang Jingwei, Zhou Xinyan, Liu Jigang, Pan Cong

Abstract

In CCUS (Carbon Capture, Utilization and Storage) technology, CO2 pipeline transportation serves as a crucial link, and its safety directly impacts the reliability of the entire industrial chain. When a CO2 pipeline leaks, the sudden release of high-pressure CO2 from the pipeline into a low-pressure environment triggers the Joule-Thomson effect. This physical characteristic creates a significant temperature difference between the leakage area and the surrounding environment. The core advantage of infrared video technology lies in its high sensitivity to temperature field distribution, enabling the clear identification of low-temperature characteristic signals in the leakage area. Therefore, this paper employs infrared video technology for data collection, utilizes image denoising methods to remove impurities from infrared images, and finally adopts the lightweight network MobileNets to detect CO2 pipeline leaks under three different operating conditions: varying temperatures and pressures, in order to enhance the accuracy of CO2 pipeline leak detection and reduce response time. Experimental verification shows that, compared with Vgg-16 and ShuffleNet_V2 models, the MobileNetV3_Large model achieves a 100% detection accuracy rate and a faster response time relative to traditional classical models. This demonstrates that lightweight network can effectively distinguish between leakage temperature and pressure differences during CO2 pipeline leakage, achieving efficient and accurate detection results that can better meet application needs. It plays a positive role in promoting and enhancing CCUS pipeline safety.

Keywords CCUS, CO2 pipeline leakage, Lightweight network, Infrared video, Image denoising