Field development optimization using a cooperative micro-particle swarm optimization with parameter integration schemes

Abstract

Non-gradient-based methods can be efficient to find optimal solutions for discrete decision variables of oil field development such as drilling schedule, number of wells, well locations, well types and its conversion times, and operational settings. Particle swarm optimization (PSO), which is a population-based stochastic global search method, has gained considerable popularity in optimization because it has simple and fast search using the speed of a few particles. However, its application to optimal oil field development is still computationally challenging due to many wells, reservoir heterogeneity, discrete variables, and nonlinear constraints. In this study, we introduce cooperative micro-particle swarm optimization (COMPSO), which will enhance the optimization efficiency of PSO for high-dimensional and complex problems by breaking a high-dimensional problem into low-dimensional subcomponents and using a small number of populations. We also propose an advanced parameterization of the decision variables for the efficient optimization of field development. A large 2-dimensional (2D) synthetic field and the 3D PUNQ-S3 benchmark model are used to demonstrate the efficiency of COMPSO and our parameterization schemes. The proposed method shows faster convergence and higher net present value (NPV) solutions than PSO. In the two examples, COMPSO only requires 39% and 16% of reservoir simulation runs to reach the same NPV compared with PSO, respectively. Our parameterization schemes additionally save 56% and 18% of the number of reservoir flow simulations of the COMPSO in the two examples, respectively.