An Efficient Methodology for Dynamic Multi-Objective Optimization of Water-Flooding Strategy

The efficient development of oilfield mostly depends on a comprehensive optimization of subsurface flow. The development effect of water-flooding is affected by technology, economy and other aspects, so its development objective is not invariable. To account for several discrete or even contradicting objectives, dynamic multi-objective optimization evolutionary algorithm (DMOEA) presents multiple optimum solutions for decision-making processes. The primary goal of this work is to optimize well placement and control parameters based on multiple design objectives using reservoir production potential formula and surrogate-assisted dynamic multi-objective optimization evolutionary algorithm. A new workflow is introduced to optimize water-flooding strategy in presence of multiple conflicting criteria and time-depending constraints. The workflow consists of two optimization stages. First, we construct an improved reservoir production potential formula which considers factors such as oil saturation, pressure, fluid flow capacity, etc. The influence of dynamic seepage capacity and static reserve distribution of oil on reservoir production capacity is comprehensively evaluated by this formula. Optimal well placement can be guided based on production potential. Then, a robust computational framework that couples Deep Neural Network (DNN) and dynamic multi-objective optimizers to optimize the aforementioned objectives in water-flooding processes simultaneously. DNN is trained and employed as surrogate model of the high-fidelity simulator in the optimization workflow and DNSGA-II-A is employed to optimize control parameters by maximizing the overall oil production and NPV, and minimizing the water cut. The Pareto front arising from the above process provides many water-flooding scenarios yielding to practical decision-making capabilities. The performance of the proposed workflow is validated in Shengli Oilfield. The results demonstrate that the method can ensure the more reasonable optimization of the whole process of water-flooding. This work can provide not only the economic and technical solutions but the correct optimization responses according to the multiple design objectives. Besides, the robustness and convergence speed of this method is better than other algorithms. Compared with the traditional single-objective optimization algorithm, the proposed method can comprehensively consider the relationship between various development objectives, to give reasonable optimal solutions. Compared with the traditional static optimization algorithm, it can track the changing Pareto optimal front in time, to provide a diversified optimal solution set according to the needs of reservoir engineers. The major contribution of this work is the introduction of a new approach that can effectively balance the needs of various objectives such as benefit, cost, and risk in the life-cycle of water-flooding and make a rapid response. The presented reliable method could provide certain significance for the efficient optimization of well placement and control parameters in the oilfield.