Data-driven evolutionary algorithm for oil reservoir well-placement and control optimization

TL;DR

This paper introduces a novel data-driven evolutionary algorithm, GDDE, that reduces simulation runs and accelerates well-placement and control optimization in oil reservoirs using neural networks and surrogate models.

Contribution

The paper proposes the GDDE algorithm combining PNN classifiers and RBF surrogate models to improve efficiency in reservoir optimization tasks.

Findings

Significant reduction in simulation runs compared to traditional methods.

Effective optimization on 2D reservoir and Egg model cases.

Enhanced convergence speed and solution quality.

Abstract

Optimal well placement and well injection-production are crucial for the reservoir development to maximize the financial profits during the project lifetime. Meta-heuristic algorithms have showed good performance in solving complex, nonlinear and non-continuous optimization problems. However, a large number of numerical simulation runs are involved during the optimization process. In this work, a novel and efficient data-driven evolutionary algorithm, called generalized data-driven differential evolutionary algorithm (GDDE), is proposed to reduce the number of simulation runs on well-placement and control optimization problems. Probabilistic neural network (PNN) is adopted as the classifier to select informative and promising candidates, and the most uncertain candidate based on Euclidean distance is prescreened and evaluated with a numerical simulator. Subsequently, local surrogate model is built by radial basis function (RBF) and the optimum of the surrogate, found by optimizer, is evaluated by the numerical simulator to accelerate the convergence. It is worth noting that the shape factors of RBF model and PNN are optimized via solving hyper-parameter sub-expensive optimization problem. The results show the optimization algorithm proposed in this study is very promising for a well-placement optimization problem of two-dimensional reservoir and joint optimization of Egg model.