Computationally efficient quantification of unknown fugitive emissions sources

• Main Authors: Carol A. Brereton, Lucy J. Campbell, Matthew R. Johnson
• Format: Article
• Language: English
• Published: Elsevier 2019-07-01
• Series: Atmospheric Environment: X
• Online Access: http://www.sciencedirect.com/science/article/pii/S2590162119300383

Fugitive emissions or unintentional losses of gas (e.g. leaks) are a significant source of greenhouse gases within the oil and gas sector. Previous work has demonstrated the potential of a scalar transport adjoint method for using sparse sensor data to locate and quantify multiple simultaneous unknown fugitive emission sources within a bluff-body dominated facility environment. This paper builds directly on that work and demonstrates the significant computational time reductions that can be achieved by modifying this approach to use a database of pre-computed retro-tracers (PRT). The computational cost, as well as estimated source emission rates and locations, were compared for both an open field release and multiple releases in a bluff-body dominated domain when using the PRT method versus the concurrent gas transport computations from previous work. For the open-field release, given the same wind input there were no significant differences in results of the two approaches. For the bluff-body dominated multiple source case (a domain representative of an actual gas plant), using simplified wind fields for the PRT database generation allowed major sources to be successfully located. The emission rates were computed within −75% to −32% of their actual value. When the wind direction coverage was increased to 110° from ∼60°, the emission rate computations improved to within approximately −30% to −25%. The total computational cost for both methods was of a similar order of magnitude when including the initial database generation for the PRT method, but non-reusable computational time was reduced by a factor of 200–600 times making the PRT method feasible on a standard desktop computer once the database is generated. This is a noteworthy achievement as it raises the possibility of continuous or near-continuous characterization of unknown fugitive emissions sources within a complex facility which could allow sources to be identified as they arise. Keywords: Fugitive emissions, Source quantification, Source characterization, Leak location, Leak quantification, Adjoint, Optimization, Computational time