A Robust Optimization Tool Based on Stochastic Optimization Methods for Waterflooding Project

Abstract

The increasing demand of crude oil has led to the increasing needs to improve recovery. Waterflooding is one of the most common secondary recovery techniques applied to many reservoirs. The method of waterflooding is to inject water into the reservoir to maintain reservoir pressure. In general, the determination of well locations is structured to form certain pattern (e.g. five spot). However, it is not the only factor affecting the optimum configuration. Besides reservoir rocks, fluid characteristics and well configurations, injection and production strategies (e.g. rates or bottom hole pressures) can also significantly affecting the overall recovery. As reported by other researchers, due to the complex nature of reservoir characterizations and fluid properties, the optimization solution of those parameters tends to have many best possible solutions. Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) were chosen because their robustness to avoid local optima and finding the global optima of an objective function. The objective function used to drive the solutions were Recovery Factor (RF) and Net Present Value (NPV) of a waterflooding project. An Optimization tool, (OpTool) was developed to prove that stochastic optimization methods could fit best in optimizing the waterflooding project. This tool is capable to communicate with reservoir simulators with known input/output formats to populate cases that would be evaluated using the well-defined objective functions. A synthetic case was built to show the OpTool capabilities to tackle waterflooding optimization problems. This tool is capable to determine an optimum pattern design (e.g. well spacing) and production and injection strategy to maximizing the RF and NPV of the project.

Untitled

“…The multiscale simulation was used to evaluate the response of the model, while stochastic simplex approximate gradient was used to compute the gradient of the objective function by implementing forward simulation reaction. Stochastic optimization based on evolutionary technology was also implemented by Ambia [35] to optimize the waterflooded performance index such as the NPV and recovery factor. A synthetic model was built to determine the optimum well pattern, spacing, production and injection scheme that will improve the NPV and recovery factor.…”

“…Review Summary. Optimization Algorithm Reference Data driven proxies Self-Optimizing Control [33], [146], [151], [153] - [155] Correlation based models [82], [112] - [126] Reduced order models [100], [127] - [130] Model Predictive control [132] - [139] Machine learning [157] Mean Variance optimization [18], [27] Robust optimization, Sequential Quadratic programming (SQP) [14] - [32], [35] - [39] Optimal Control theory [10], [85]- [93], [94] - [111] Ensemble Kalman Filter [14], [38], [71] - [82]…”

Revisiting EOR Projects in Indonesia through Integrated Study: EOR Screening, Predictive Model, and Optimisation

“…Stochastic methods are procedures to find the near-optimum solution of a problem based on random number generation (Ambia, 2012). Certain rules are applied to the random numbers so that they would act as "searching agents" in the defined search space with the objective of finding a near-optimum solution.…”