A review on optimization algorithms and surrogate models for reservoir automatic history matching

Reservoir history matching represents a crucial stage in the reservoir development process and purposes to match model predictions with various observed field data, including production, seismic, and electromagnetic data. In contrast to the traditional manual approach, automatic history matching (AHM) significantly reduces the workload of reservoir engineers by automatically tuning the reservoir model parameters. AHM can be viewed as an automated solution to an inverse problem, and the selection of optimization algorithms is crucial for achieving effective model matching. However, the optimization process requires running numerous simulations. Surrogate models, achieved through simplification or approximation of the realistic model, offer a significant reduction in computational costs during the simulation process. In this paper, we provide an overview of commonly prevalent optimization algorithms and surrogate models in the AHM process, presenting the latest advancements in these methods. We analyze the strengths and limitations of these approaches and discuss the future challenges and directions of AHM, aiming to provide valuable references for further research and applications in this field.