Efficient and Accurate Leakage Points Detection in Gas Pipeline Using Reinforcement Learning-Based Optimization

Obtaining gas concentration leak data plays a crucial role in managing the safe maintenance and timely, effective, and accurate identification of chemical facilities' leak locations. Traditional random detection paths cannot meet the efficiency and accuracy requirements for obtaining key information on leak areas in modern society. In this study, we propose an efficient method that combines reinforcement learning and path evaluation, aiming to determine the concentration of gas leaks. This method consists of two parts: optimizing real-time paths based on concentration gradient changes and evaluating path performance using a strategy iteration algorithm. Furthermore, we validate the feasibility and effectiveness of this method by comparing it with actual measurement data and verify the performance of coordinate calculations in our study. Research results indicate that the points selected by this method can fully reveal concentration changes in the area and calculate more accurate leak coordinates with fewer sampling points, resulting in smaller relative errors. The method proposed in this study helps to improve the accuracy of instrument monitoring, promote intelligent monitoring of urban lifelines, and ensure the safety of the urban environment and its residents.