This study describes an automated history matching and optimization workflow using an integrated hydraulic fracturing reservoir simulator and applies the workflow in four cases. The automated workflow solves a formal mathematical optimization problem to minimize misfit with observations from any point in the lifecycle of a hydraulically fractured well, or to maximize a quantity of interest, such as net present value. The workflow uses a proxy model to improve computational speed and employs experimental design and Bayesian sampling techniques to generate points with which to train the proxy model. We first apply the automated workflow to two simple history matching cases. In the first case, we fit only only to production data, which results in a nonunique history match. In the second case, we incorporate both production data and fracture length, resulting in a unique history match. Next, we apply the automated workflow to a Bakken dataset with five history matching parameters and find that some parameters are well-constrained by matching the dataset, while other parameters are not as well constrained. In the final case, we use the best-fit point resulting from the Bakken history match in an automated forward optimization workflow to maximize a combination of low-price and high-price NPV objectives, which we developed using stylized generic economic assumptions.
