Optimization of field development using particle swarm optimization and new well pattern descriptions
- The optimization of the type and location of new wells is an important issue in oil field development. Computational algorithms are often employed for this task. The problem is challenging, however, because of the many different well configurations (vertical, horizontal, deviated, multilateral, injector or producer) that must be evaluated during the optimization. The computational requirements are further increased when geological uncertainty is incorporated into the optimization procedure. In large-scale applications, involving hundreds of wells, the number of optimization variables and the size of the search space can be very large. In this work, we developed new procedures for well placement optimization using particle swarm optimization (PSO) as the underlying optimization algorithm. We first applied PSO to a variety of well placement optimization problems involving relatively few wells. Next, a new procedure for large-scale field development involving many wells was implemented. Finally, a metaoptimization procedure for determining optimal PSO parameters during the optimization was formulated and tested. The particle swarm optimization is a population-based, global, stochastic optimization algorithm. The solutions in PSO, called particles, move in the search space based on a "velocity." The position and velocity of each particle are updated iteratively according to the objective function value for the particle and the position of the particle relative to other particles in its (algorithmic) neighborhood. The PSO algorithm was used to optimize well location and type in several problems of varying complexity including optimizations of a single producer over ten realizations of the reservoir model and optimizations involving nonconventional wells. For each problem, multiple optimization runs using both PSO and the widely used (binary) genetic algorithm (GA) were performed. The optimizations showed that, on average, PSO provides results that are superior to those using GA for the problems considered. In order to treat large-scale optimizations involving significant numbers of wells, we next developed a new procedure, called well pattern optimization (WPO). WPO avoids some of the difficulties of standard approaches by considering repeated well patterns and then optimizing the parameters associated with the well pattern type and geometry. WPO consists of three components: well pattern description (WPD), well-by-well perturbation (WWP), and the core PSO algorithm. In WPD, solutions encode well pattern type (e.g., five-spot, seven-spot) and their associated pattern operators. These pattern operators define geometric transformations (e.g., stretching, rotation) applied to a base pattern element. The PSO algorithm was then used to optimize the parameters embedded within WPD. An important feature of WPD is that the number of optimization variables is independent of the well count and the number of wells is determined during the optimization. The WWP procedure optimizes local perturbations of the well locations determined from the WPD solution. This enables the optimized solution to account for local variations in reservoir properties. The overall WPO procedure was applied to several optimization problems and the results demonstrate the effectiveness of WPO in large-scale problems. In a limited comparison, WPO was shown to give better results than optimizations using a standard representation (concatenated well parameters). In the final phase of this work, we applied a metaoptimization procedure which optimizes the parameters of the PSO algorithm during the optimization runs. Metaoptimization involves the use of two optimization algorithms, where the first algorithm optimizes the PSO parameters and the second algorithm uses the parameters in well placement optimizations. We applied the metaoptimization procedure to determine optimum PSO parameters for a set of four benchmark well placement optimization problems. These benchmark problems are relatively simple and involve only one or two vertical wells. The results obtained using metaoptimization for these cases are better than those obtained using PSO with default parameters. Next, we applied the optimized parameter values to two realistic optimization problems. In these problems, the PSO with optimized parameters provided comparable results to those of the default PSO. Finally, we applied the full metaoptimization procedure to realistic cases, and the results were shown to be an improvement over those achieved using either default parameters or parameters determined from benchmark problems.
|Type of resource
|electronic; electronic resource; remote
|1 online resource.
|Stanford University, Department of Energy Resources Engineering
|Horne, Roland N
|Mukerji, Tapan, 1965-
|Horne, Roland N
|Mukerji, Tapan, 1965-
|Statement of responsibility
|Jérôme Emeka Onwunalu.
|Submitted to the Department of Energy Resources Engineering.
|Thesis (Ph. D.)--Stanford University, 2010.
- © 2010 by Jerome Onwunalu
- This work is licensed under a Creative Commons Attribution Non Commercial 3.0 Unported license (CC BY-NC).
Also listed in
Loading usage metrics...