Reservoir production optimization based on surrograte model and differential evolution algorithm

Abstract Production Optimization is a significant method for oilfields to control water cut and stabilize oil production. When the oilfield enters the high or ultra-high water cut stage, it becomes particularly important to use production optimization methods for improving the water-flooding efficiency. Currently, the commonly used production optimization methods are based on reservoir simulators. Such methods require lots of forward simulations during optimizing, which results in low computational efficiency. It's not applicable to the reservoir without numerical simulation models. Thus, a new production optimization method based on the reservoir proxy model is proposed in this work. Firstly, the dynamic production data of the oilfield are collected and preprocessed for training the Extreme Gradient Boosting model, and constructing the proxy model for water cut prediction of producers and the reservoir. Then, an optimal control model for minimizing the water cut of the reservoir can be constructed based on the proxy model. Finally, an optimal injection-production scheme can be obtained by using the differential evolution algorithm. For the evaluation and verification purposes, the proposed method is applied to a well block from SL oilfield, China. Empirical results demonstrated that the proposed method can effectively improve the water-flooding efficiency.