Abstract: Hazardous gas detection systems play an important role in preventing catastrophic gas-related accidents in process industries. Even though effective detection technology currently exists for hazardous gas releases and a majority of process installations have a large number of sensitive detectors in place, the actual operating performance of gas detection systems still does not meet the expected requirements. In this paper, a riskbased methodology is proposed to optimize the placement of hazardous gas detectors. The methodology includes three main steps, namely, the establishment of representative leak scenarios, computational fluid dynamics (CFD)-based gas dispersion modeling, and the establishment of an optimized solution. Based on the combination of gas leak probability and joint distribution probability of wind velocity and wind direction, a quantitative filtering approach is presented to select representative leak scenarios from all potential scenarios. The commercial code ANSYS-FLUENT is used to estimate the consequence of hazardous gas dispersions under various leak and environmental conditions. A stochastic mixed-integer linear programming formulation with the objective of minimizing the total leak risk across all representative leak scenarios is proposed, and the greedy dropping heuristic algorithm (GDHA) is used to solve the optimization model. Finally, a practical application of the methodology is performed to validate its effectiveness for the optimal design of a gas detector system in a high-sulfur natural gas purification plant in Chongqing, China. The results show that an appropriate number of gas detectors with optimal cost-effectiveness can be obtained, and the total leak risk across all potential scenarios can be substantially reduced. This methodology provides an effective approach to guide the optimal placement of pointtype gas detection systems involved with either single or mixed gas releases.

Key words: Leak scenario, Leak risk, Gas detection, Detector placement, Mixed-integer linear programming

摘要： Hazardous gas detection systems play an important role in preventing catastrophic gas-related accidents in process industries. Even though effective detection technology currently exists for hazardous gas releases and a majority of process installations have a large number of sensitive detectors in place, the actual operating performance of gas detection systems still does not meet the expected requirements. In this paper, a riskbased methodology is proposed to optimize the placement of hazardous gas detectors. The methodology includes three main steps, namely, the establishment of representative leak scenarios, computational fluid dynamics (CFD)-based gas dispersion modeling, and the establishment of an optimized solution. Based on the combination of gas leak probability and joint distribution probability of wind velocity and wind direction, a quantitative filtering approach is presented to select representative leak scenarios from all potential scenarios. The commercial code ANSYS-FLUENT is used to estimate the consequence of hazardous gas dispersions under various leak and environmental conditions. A stochastic mixed-integer linear programming formulation with the objective of minimizing the total leak risk across all representative leak scenarios is proposed, and the greedy dropping heuristic algorithm (GDHA) is used to solve the optimization model. Finally, a practical application of the methodology is performed to validate its effectiveness for the optimal design of a gas detector system in a high-sulfur natural gas purification plant in Chongqing, China. The results show that an appropriate number of gas detectors with optimal cost-effectiveness can be obtained, and the total leak risk across all potential scenarios can be substantially reduced. This methodology provides an effective approach to guide the optimal placement of pointtype gas detection systems involved with either single or mixed gas releases.

关键词: Leak scenario, Leak risk, Gas detection, Detector placement, Mixed-integer linear programming