The value of history matching in field development

Decision-making in field development is usually based on models that predict future reservoir behavior. From history matching one obtains reservoir models with different levels of agreement with the production data and, consequently, different subsurface representations. This variability is indicative of the uncertainty associated with the reservoir's description. There is the question how field management decisions may depend on the variability of the possible model prediction across the uncertainty range. We present herein an approach to evaluate how the quality of history matching may impact the reservoir decision. We assessed the effect of performing stochastic well placement optimization in models that match the observed production data with different convergence levels in two industry benchmark cases. We combined Particle Swarm Optimization and Neighborhood Algorithm Bayes to assess the uncertainty estimation of optimal solutions. In the first case study, a new vertical well location is optimized, showing that the well-matched models provide a range of similar field oil and water productions at the optimized truth case. In the second case study, we optimized the direction of a new horizontal well. The results show the solutions from the poorly-matched and well-matched models clustered in a narrow range of field oil and water productions.