Abstract: In the operation of natural gas stations, emergency response to sudden gas leakage accidents highly depends on the accurate acquisition of critical information such as leakage location and volume. Currently, most natural gas stations are equipped with TDLAS methane detectors capable of sensing minor leaks. However, due to the absence of integrated gas plume tracking and search algorithms, autonomous leakage source identification remains unachievable. To address this issue, this study established a global source model based on real-time concentration (GSRC) and a spherical source model based on real-time concentration (SSRC), with their effectiveness and advantages evaluated through numerical simulations and field tests. The results demonstrate that the GSRC and SSRC models, based on real-time gas plume concentration and search algorithms, enable cloud-based laser methane detectors to autonomously search and locate leaks. Furthermore, key model parameters were determined through in-depth analysis of influencing factors. Meanwhile, a full-scale CFD leakage dispersion model was developed for a natural gas station in western China, and the simulation found that the SSRC model has the smallest search path length and average search time, and the highest search efficiency. Field tests conducted after embedding the SSRC model into the detector's industrial control platform revealed that 86% of leakage incidents (within 10~60 m³/h leakage rates) could be located within 4 min, with positioning errors ≤2 m in 44% of cases and ≤ 4 m in 72% of cases. The developed leakage source search models significantly enhance localization accuracy, providing critical technical support for emergency response and recovery operations.

Key words: natural gas station, leakage source localization, computational fluid dynamics, global leak source release concentration model, spherical source release concentration model