Above-ground natural gas riser leak dispersion and leakage rate prediction based on CFD and BPNN approaches

Abstract

Natural gas risers, typically installed along exterior building walls, are susceptible to corrosion, construction activities, and mechanical impacts, thereby posing significant safety risks. Therefore, it is crucial to examine gas leakage and dispersion behaviors following pipeline failure and to develop predictive models for leakage rates. This study investigates above-ground natural gas risers in urban environments by employing a transient species transport model to simulate gas leakage and dispersion. The effects of pipeline operating pressure and leakage diameter on gas distribution, as well as the overall leakage and diffusion characteristics, are systematically analyzed. Moreover, a predictive model for leakage rate is developed based on the distance between monitoring points and the leakage source, incorporating measured gas concentration data. The results indicate that in the initial stage of a natural gas leak, a substantial amount of gas accumulates near the leakage hole, forming a momentum-driven jet that spreads rapidly horizontally with limited lateral dispersion capacity. As the leakage persists, the dominance of jet-driven dispersion gradually transitions to turbulent diffusion with buoyancy. At around 20 s, the dispersion process attains a quasi-steady state. Gas dispersion is influenced by turbulent diffusion, entrainment, and buoyancy, which together determine the overall spreading behavior. The sampling plane, aligned with the air intake of the Methane Detection Vehicle and situated within 20 m of the leakage source, shows a peak methane concentration approximately 3 m from the leakage hole. Although variations in pipeline operating pressure and leakage diameter influence the overall concentration distribution, they do not alter the position of the concentration peak. The proposed backpropagation neural network (BPNN) model demonstrates robust predictive performance for leakage rates in long-distance monitoring scenarios, thereby supporting the practical application of mobile methane detection vehicle technologies.