The concentration data obtained by the unmanned aerial vehicle

With the rapid pace of global urbanization and rising energy demands, efficient gas leak detection is vital for public safety. This study proposes an efficient and sensitive gas leak detection method based on reinforcement learning to enhance localization speed and robustness. The approach includes critical area identification, reinforcement learning model training, and leak point localization. Simultaneously introducing noise and missing data to test the robustness of the model. This method has also been applied to the pipeline network center of Chongqing Jiangjin Natural Gas Co., Ltd., further confirming its effectiveness and adaptability in complex environments. In the case of multiple leakage points, the average relative deviation of each leakage point in the X and Y directions calculated by this method is 2.1 percentage points, and the average deviation distance is 0.75 meters. These results demonstrate the method's adaptability to complex environments, contributing to a comprehensive and efficient environmental management system.