General-purpose Inverse Modeling Framework for Energy Transition Applications Based on Adjoint Method and Operator-Based Linearization

Abstract

This study investigates the application of inverse modeling in numerical geo-energy scenarios such as petroleum, geothermal, and CCS projects. The study aims to enhance model accuracy and predictive capabilities for real-world applications. The focus lies on the implementation of the inverse modeling framework within the open-source simulator called Delft Advanced Research Terra Simulator (DARTS), developed using the adjoint method and Operator-Based Linearization (OBL) to assemble derivatives efficiently. The adjoint method's efficiency and analytical gradient solution make it a preferred choice for gradient evaluation in inverse modeling. The work transitions from forward simulation to inverse modeling, elucidating the objective function definition, optimization theory, and the adjoint method's process. Prototype development in MATLAB and its translation to C++ are presented, showcasing the method's superiority. Application examples, like the data-driven proxy model and energy transition projects, demonstrate the framework's versatility and effectiveness in handling diverse observations and solving complex energy transition challenges.