Uncertainty Quantification of an Explicitly Coupled Multiphysics Simulation of In-Situ Pyrolysis by Radio Frequency Heating in Oil Shale

Abstract

Summary In-situ pyrolysis provides an enhanced oil recovery (EOR) technique for exploiting oil and gas from oil shale by converting in-place solid kerogen into liquid oil and gas. Radio-frequency (RF) heating of the in-place oil shale has previously been proposed as a method by which the electromagnetic energy gets converted to thermal energy, thereby heating in-situ kerogen so that it converts to oil and gas. In order to numerically model the RF heating of the in-situ oil shale, a novel explicitly coupled thermal, phase field, mechanical, and electromagnetic (TPME) framework is devised using the finite element method in a 2D domain. Contemporaneous efforts in the commercial development of oil shale by in-situ pyrolysis have largely focused on pilot methodologies intended to validate specific corporate or esoteric EOR strategies. This work focuses on addressing efficient epistemic uncertainty quantification (UQ) of select thermal, oil shale distribution, electromagnetic, and mechanical characteristics of oil shale in the RF heating process, comparing a spectral methodology to a Monte Carlo (MC) simulation for validation. Attempts were made to parameterize the stochastic simulation models using the characteristic properties of Green River oil shale. The geologic environment being investigated is devised as a kerogen-poor under- and overburden separated by a layer of heterogeneous yet kerogen-rich oil shale in a target formation. The objective of this work is the quantification of plausible oil shale conversion using TPME simulation under parametric uncertainty; this, while considering a referenced conversion timeline of 1.0 × 107 seconds. Nonintrusive polynomial chaos (NIPC) and MC simulation were used to evaluate complex stochastically driven TPME simulations of RF heating. The least angle regression (LAR) method was specifically used to determine a sparse set of polynomial chaos coefficients leading to the determination of summary statistics that describe the TPME results. Given the existing broad use of MC simulation methods for UQ in the oil and gas industry, the combined LAR and NIPC is suggested to provide a distinguishable performance improvement to UQ compared to MC methods.