Hyper Fast THMC Well Integrity Simulator for Powering Optimization, Sensitivity and Risk Analysis

A semi-analytical approach to three-dimensional Thermo-Hydro-Mechanical-Chemical (THMC) well integrity simulations is discussed. The method extends previous plane-strain thermo-poroelastic models under a set of realistic assumptions and combines analytic and numerical solutions in a single framework. Non-analytical components of the model are resolved via one-dimensional Finite Difference schemes, and coupled with pen-and-paper solutions via simple numerical integration schemes. Time-dependent material properties of hydrating cement are taken into account in the incremental problem formulation and all model components are fully coupled. The proposed formulation is demonstrated to be in very good agreement with sophisticated FEM software, while the simulation run times are reduced by several orders of magnitude. Natural applications of the methodology proposed include high-volume workflows relying on thousands or more simulation runs (sensitivity studies, optimisation, Monte Carlo methods etc.). 1 INTRODUCTION Well integrity is essential in many engineering applications, including CO2 storage, well plugging & abandonment, geothermal energy, oil & gas and gas fuel storage (CH4 and H2). There is a clear need from the industry for a computational well integrity assessment tool which is efficient, robust, can work with limited data sets and/or high uncertainties, and fits well within a risk-based statistical approach. From the point of view of computational mechanics, well integrity is primarily controlled by the internal stresses within different materials comprising the well structure; these typically include metal casings surrounded by cement sheaths and the formation. Historically, two primary approaches have been undertaken to model the relevant stress responses. On one hand, the Finite Element Method (FEM) is widely used due to its flexibility and generality, in particular the natural capability of working with realistic geometries. Alternatively, idealised axisymmetric analytic models (often limited to two dimensions) have been studied by multiple authors Onaisi et al. (2017); Gholami et al. (2016); Joulin et al. (2022); Li et al. (2016); Bai et al. (2015). FEM and two-dimensional analytic models mark two ends of a spectrum; FEM is general and flexible but computationally expensive, while analytic models are ‘near-free’ computation-wise but oversimplified.