Emission Source Detection and Leak Rate Estimation Using Point Measurements of Concentration

Abstract

With the advent of global climate change, it has become incumbent on governments and industries to monitor and limit greenhouse gas emissions to prevent a catastrophic rise in the average global temperature. The Paris agreement [Paris 2015] aims to lower global greenhouse gas emissions by 40% (in comparison to greenhouse gas levels observed in 1990) by 2030. Methane is a greenhouse gas whose 100- year global warming potential is 25 times that of carbon dioxide [GWP] and whose atmospheric concentration has been increasing since 2007 [Nisbet 2016, Theo Stein, et al. 2021]. Thus, there is an increased requirement on industries from government regulators to detect, localize, quantify and mitigate both fugitive and vented emissions of methane. There are several different technologies that are available for automated methane emissions management. These include arial and ground-based mobile sensing units that are based on optical-gas imaging, satellite-based imagery [Jacob et al. 2016] and stationary metal-oxide based sensors. A key criterion that often needs to be satisfied is continuous monitoring for early detection and mitigation of fugitive leaks. Fixed metal-oxide based sensors [Yuliarto et al. (2015), Zeng et al. (2019), Yunusa et al. (2014), Potyrailo et al. (2020), Wang et al. (2010) and Feng et al. (2019)] are low-cost sensors that can be used for continuous monitoring of a site and are typically used for detection of leaks and alerting. The main challenge is to extend utility of these sensors to not only detect presence of fugitive and vented emissions, but also be able to estimate the number of leak sources and their probable locations and the total volume of hydrocarbon leaked over a period. This paper describes an approach used for detecting anomalies in emission data, identifying possible emission sources, and estimating emission leak rates using point measurements of concentration collected over a period along with measurements of wind speed and direction. This involves multiple analytics that combine concentration and wind-condition time-series data with physics models to predict the different outcomes.