An Improved Approach for Ensemble-Based Production Optimization

Abstract

Abstract For oil reservoirs under water and/or gas drive, it is challenging and rewarding to effectively manage the water/gas fronts to maximize the sweep efficiency. In the past, a wide variety of approaches were developed and implemented for optimizing the reservoir performance by smartly adjusting operational controls. With recent developments in monitoring and control, the opportunity to optimize reservoir production through frequent adjustment of well controls has substantially increased. In this paper, we propose an improved ensemble-based approach for production optimization using Conjugate Gradient method (CGEnOpt). The proposed method results in faster convergence rate as a result of using the conjugate gradient directions instead of steepest ascent search directions. The net present value (NPV) of a single reservoir is optimized by obtaining the gradient of the NPV from an ensemble of control variables. The effectiveness of the proposed method is demonstrated on a 3-D synthetic reservoir problem with 8 producers and 3 injectors. The production and injection rates are used as control variables to maximize NPV of the reservoir. The results indicate that the control settings from the proposed method result in a significantly higher NPV of the reservoir as compared to the reference case. The number of iterations needed for the proposed method is significantly reduced compared to the standard ensemble-based optimization. The proposed method is able to allocate the production and injection rates effectively which results in a substantial decrease in the amount of water produced. The proposed method is completely adjoint-free and is independent of the reservoir simulator used for the prediction simulation run which makes it fairly robust. Thus, it can be used with any existing simulator with minimal code development. The CGEnOpt technique is innovative for not requiring explicit computation of the gradient which makes it easy to implement.