Gradient-based Optimization Techniques for Automatic History Matching in Reservoirs

Summary In this abstract, we reviewed the previously existing automatic history-matching methods for reservoirs. These methods include gradient based methods, global optimization algorithms, data assimilation methods, and reinforcement learning methods. All of these methods are briefly introduced, and gradient-based methods are explained. This abstract shows that the best method for optimization is the adjoint method with a combination of the limited memory Broyden-Fletcher-Goldfarb-Shanno (L-BFGS) method for large-scale data history matching.