An original deconvolution approach for oil production allocation based on geochemical fingerprinting. (1st November 2022)

• Record Type: Journal Article
• Authors: Sandoval, Leonardo; Riva, Monica; Franco, Placido; Colombo, Ivo; Galimberti, Roberto; Guadagnini, Alberto
• Is Part Of: Fuel. Volume 327(2022)
• Publication Date: 2022-11-01
• Subjects: Reservoir geochemistry -- Production allocation -- Geochemical fingerprinting -- Maximum Likelihood -- GC–MS; Fuel -- Periodicals; Coal -- Periodicals; Coal; Fuel; Periodicals; 662.6
• DOI: 10.1016/j.fuel.2022.124715
• Languages: English
• ISSNs: 0016-2361

Abstract

Graphical abstract: Highlights: A novel deconvolution algorithm for production allocation is introduced. An improved experimental procedure for GC targeting production allocation is proposed. Extensions to traditional production allocation algorithms are formulated. Abstract: We tackle oil commingling scenarios and develop an original deconvolution approach for geochemical production allocation. This yields robust assessment of the proportions of oils forming a mixture originating from commingling oils associated with diverse reservoirs or, wells. Our study starts from considering that production allocation performed by means of geochemical fingerprinting is relevant in the context of modern and sustainable use of georesources, with the added benefit of favoring shared facilities and production equipment. A geochemical production allocation workflow is typically structured according to two steps: (i) determination of the chromatograms associated with the mixture (and eventually with each of the End Members, EMs, constituting the fluids in the mixture), and (ii) the use of a deconvolution algorithm to estimate the mass fraction of each EM. Concerning the latter step, we introduce an original approach and the ensuing deconvolution algorithm (hereafter termed PGM) that does not require additional laboratory efforts in comparison with traditional approaches. We also present extensions of widely used deconvolution algorithms, which we frame in a (stochastic) Monte Carlo context to improve their robustness and reliability. The new PGM approach is assessed jointly with a suite of typically used approaches and algorithms against new laboratory-based commingling scenarios. The latter are based on the design and introduction of a novel and low-cost experimental method. The results of the study (i) constitute a unique and rigorous comparison of the traditionally employed production allocation deconvolution algorithms, (ii) document the critical importance of the number of features of the chromatograms used during a quantitative deconvolution, and (iii) suggest that our new PGM approach is very robust and accurate compared to existing approaches.