We propose a generic, modular framework for emission event detection, localization, and quantification on oil and gas facilities that uses concentration data collected by point-in-space continuous emissions monitoring systems (CEMS). The framework uses a gradient-based spike detection algorithm to estimate emission start and end times (event detection) and pattern matches simulated and observed concentrations to estimate emission source location (localization) and rate (quantification). We test the framework on a month of single-source controlled releases ranging from 0.50 to 8.25 hours in duration and 0.18 to 6.39 kg/hr in size conducted at the Methane Emissions Technology Evaluation Center in Fort Collins, Colorado. All controlled releases are identified and 82% are localized correctly. For emissions < 1 kg/hr, the framework underestimates by 37.2% on average, with 90% of rate estimates within a factor of [-4.6, 2.8] or a percent difference of [-78.1%, 178.6%]; for emissions > 1 kg/hr, the framework overestimates by 1.5% on average, with 90% of rate estimates within a factor of [-2.0, 1.8] or a percent difference of [-49.6%, 77.4%] from the true rates. Potential uses for the proposed framework include near real-time alerting for rapid emissions mitigation and emission quantification for data-driven inventory estimation on production-like facilities.
We propose a generic, modular framework for emission event detection, localization, and quantification on oil and gas production sites that uses concentration data collected by point-in-space continuous monitoring systems (CMS). The framework uses a gradient-based spike detection algorithm to estimate emission start and end times (event detection) and pattern matches simulated and observed concentrations to estimate emission source location (localization) and rate (quantification). We test the framework on a month of non-blinded, single-source controlled releases ranging from 0.50 to 8.25 hours in duration and 0.18 to 6.39 kg/hr in size. All controlled releases are identified and 82% are localized correctly. 5.5% of predicted events are false positives. For emissions <= 1 kg/hr, the framework underestimates by 37.2% on average, with 90% of rate estimates within a factor of [-4.6, 2.8] or a percent difference of [-78.1%, 178.6%] from the true rate. For emissions > 1 kg/hr, the framework overestimates by 1.5% on average, with 90% of rate estimates within a factor of [-2.0, 1.8] or a percent difference of [-49.6%, 77.4%]. Potential uses for the proposed framework include near real-time alerting for rapid emissions mitigation and emission quantification for data-driven inventory estimation on production sites.
Characterizing methane emissions on oil and gas facilities often relies on a forward model to describe the atmospheric transport of methane. Here we compare two forward models: the Gaussian plume, a commonly used steady-state dispersion model, and the Gaussian puff, a time varying dispersion model that approximates a continuous release as a sum over many small "puffs". We compare model predictions to observations from a network of point-in-space continuous emissions monitoring systems (CEMS) collected during a series of controlled releases. Specifically, we use the Pearson correlation coefficient and mean absolute error (MAE) as metrics to assess the fit of the model predictions to the observed concentrations, with the former assessing the fit of pattern and the latter the fit of amplitude. The Gaussian puff outperforms the Gaussian plume using both metrics with average correlation coefficients of 0.38 and 0.31 and average MAEs of 0.70 and 0.74, respectively. We also investigate how the frequency at which puffs are generated affects the accuracy and computational cost of the Gaussian puff and propose guidelines for choosing an appropriate value. Finally, we provide open-source implementations of the Gaussian puff model in Python and R that are tailored for use on oil and gas facilities.
Characterizing methane emissions on oil and gas sites often relies on a forward model to describe the atmospheric transport of methane. Here we compare two forward models: the Gaussian plume, a commonly used steady-state dispersion model, and the Gaussian puff, a time varying dispersion model that approximates a continuous release as a sum over many small “puffs”. We compare model predictions to observations from a network of point-in-space continuous monitoring systems (CMS) collected during a series of controlled releases. Specifically, we use the Pearson correlation coefficient and mean absolute error (MAE) as metrics to assess the fit of the model predictions to the observed concentrations in terms of pattern and amplitude, respectively. The Gaussian puff outperforms the Gaussian plume using both metrics with average correlation coefficients of 0.38 and 0.31 and average MAEs of 0.70 and 0.74, respectively. We provide computationally efficient and scalable implementations of the Gaussian puff model. Compared to regulatory-grade, Gaussian puff-based models like CALPUFF, our implementations have higher spatial and temporal resolution and require only essential and practically available meteorological information. These features enable near real-time methane mitigation applications on oil and gas sites and might be useful for near-field atmospheric transport modeling applications more broadly.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.