To main content

A Differentiable Simulator for Methane and VOC Emissions in Coupled Liquid–Gas Tank Systems

Abstract

Methane and other volatile organic compounds (VOCs) emitted during storage, loading, and transport of crude oil and condensate represent a growing environmental and operational concern. Industry faces tightening reporting requirements, and collective initiatives such as the Oil and Gas Methane Partnership 2.0 (OGMP 2.0) seek greater transparency and accuracy in methane accounting. Beyond compliance, reducing methane and VOC releases is a high‑leverage lever for near‑term climate and air‑quality/work‑environment improvements, directly supporting company‑ and sector‑level decarbonization targets. The simulation tool VOCSim is the leading numerical model for estimating such emissions. Built on nearly 40 years of field sampling from tanker loading, it has supported almost 100 industry projects at the research institute SINTEF, advancing understanding and control of VOC emissions. VOCSim models multi‑component, two‑phase transport with diffusion and phase equilibrium to estimate emitted‑gas flow rates and composition. Despite its success, the legacy code and user interface hinder extension to new models and limit the tool to forward‑only runs. Modern engineering workflows, however, require sensitivity analysis and optimization to guide decision‑making and process planning. VOCSim.jl is a new open‑source replacement written in Julia. It is built on the Jutul framework, a fully differentiable multiphysics simulator using implicit finite‑volume methods with automatic differentiation (AD). End‑to‑end AD enables discrete adjoints for efficient computation of sensitivities of outputs with respect to inputs. In this presentation, we introduce the numerical model in VOCSim.jl, demonstrate validation against VOCSim, and present initial results for industry‑relevant use cases. We show how the modular, AD‑based framework will enable flexible model extensions and supports engineering workflows for sensitivity analysis, uncertainty quantification, history matching, and parameter optimization, thereby improving operational decisions and reducing emissions through use cases co‑developed with industry.

Category

Conference lecture

Language

English

Author(s)

Affiliation

  • SINTEF Digital / Mathematics and Cybernetics
  • SINTEF Ocean / Energi og transport
  • Norwegian University of Science and Technology

Presented at

17th World Congress on Computational Mechanics (WCCM) and 10th European Congress on Computational Methods in Applied Sciences and Engineering (ECCOMAS)

Place

Munich

Date

19.07.2026 - 24.07.2026

Organizer

International Association for Computational Mechanics (IACM) and European Community on Computational Methods in Applied Sciences (ECCOMAS)

Date

22.07.2026

Year

2026

View this publication at Norwegian Research Information Repository