An Improved Parameter Dimensionality Reduction Approach Based on a Fast Marching Method for Automatic History Matching

History matching is a critical step in reservoir numerical simulation algorithms. It typically hindered by difficulties associated with the high-dimensionality of problem and gradient calculation approach. Here, multi-step solving method proposed which, first, Fast marching (FMM) used to calculate pressure propagation time determine single-well sensitive area. Second, mathematical model for history implemented using Bayesian framework. Third, an effective decomposition strategy adopted parameter dimensionality reduction. Finally, localization matrix constructed based on area data modify objective function. This has been verified through water drive conceptual example real field case. The results have shown that can generate more accurate information predictions compared traditional analytical methods other gradient-free