Optimization Of Oil Well Placement Using Particle Swarm Algorithm

A very important task in oil field development is to determine the oil-well placement and to evaluate the economic benefit of different designs in production, in order to obtain an optimal well pattern. To maximize the economic benefits, a reasonable well placement requires to minimize the number of oil-wells and to maximize the use of oil reserves. Therefore, the main tasks in the well network deployment are to determine the placement ways of oil, water, gas wells, their numbers and spacing size. Aiming at the optimization of oil well placement, the main contents of this paper are as follows:First, the well network deployment depends on the types of reservoir displacement and fluid properties and various economic parameters in reservoir, many different variables or high dimensional attributes are involved. How to select and combine these variables effectively is very important for the oil yield maximization. The traditional optimization methods are often not suitable for such multivariable complex problem for its strict objective function requirement. To solve this problem, this paper introduces the PSO (Particle Swarm Optimization) algorithm to optimize the oil well placement. The advantage of PSO algorithm includes simple calculation, few parameters setting, fast convergence, easy implementation. The new method has robustness, no specific requirement on the property of the optimization function for high dimensional variables, which makes it very suitable for solving complex optimization problems.Second, the performance of both PSO and GA is compared and it is shown that PSO algorithm is more suitable for oil-well pattern optimization with more accurate and faster convergence speed. In this paper, three benchmark functions (Sphere, Rastrigin, Rosenbrock) were used to test the optimization precision and implementation capacity,(a) Analyze the influence of parameters on algorithm performance. For the same function, the convergence precision and running speed of PSO algorithm will be different with different parameter combinations,(b) The performance of the PSO and GA algorithm on three functions is compared and analyzed with different dimension respectively. The statistical result of the experiment data shown that PSO final precision, convergence speed and the ability to escape from local optima are much better than GA algorithm, along with the increasing function dimension. Thus, the outperformance of PSO algorithm is verified for both unimodal functionand multimodal complex.Third, the PSO algorithm is applied to optimize the well placement in oilfield, accordingto the actual problem and the corresponding mathematical modeling. A software system foroil-well pattern optimization is built. Based on a set of substratum data, a reservoir model with3D permeability field is constructed, with the cumulative yield as the objective function tooptimize the well placement. The experimental results show that the oil wells optimizationwith PSO can achieve better effect and feasibility for oil fields.