One of the major goals that field planning engineers and decision makers have to achieve in terms of reservoir management and hydrocarbon recovery optimization is the maximization of return on financial investments. This task yet very challenging due to high number of decision variables and some uncertainties, pushes the engineers and technical advisors to seek for robust optimization methods in order to optimally place wells in the most profitable locations with a focus on increasing the net-present value over a project life-cycle. The quest to deliver a good quality advice is also dependent on how some uncertainties – geologic, economic and flow patterns – have been handled and formulated all along the optimization process. With the enhancement of computer power and the advent of remarkable optimization techniques, the oil and gas industry has at hand a wide range of tools to get an overview on value maximization from petroleum assets. Amongst these tools, genetic algorithms which belong to stochastic optimization methods have become well known in the industry as one the best alternatives to apply when trying to solve well placement and production allocation problems, though computationally demanding. The aim of this work is to present a novel approach in the area of hydrocarbon production optimization where control settings and well placement are to be determined based on a single objective function, in addition to the optimization of wells’ trajectories. Starting from a reservoir dynamic model of a synthetic offshore oil field assisted by water injection, the work consisted in building a data-driven model that was generated using artificial neural networks. Then, we used Matlab’s genetic algorithm toolbox to perform all the needed optimizations; from which, we were able to establish a drilling schedule for the set of wells to be realized, and we made it possible to simultaneously get the well location and configuration (vertical or horizontal), well type (producer or injector), well length, well orientation – in the horizontal plane –, as well as well controls (flow rates) and near wellbore pressure with respect to a set of linear and nonlinear-constraints. These constraints were formulated so as to reproduce real field development considerations, and with the aid of a genetic algorithm procedure written upon Matlab, we were able to satisfy those constraints such as, maximum production and injection rates, optimal wellbore pressures, maximum allowable liquid processing capacity, optimal well locations, wells’ drilling and completion maximum duration, in addition to other considerations. We have investigated some scenarios with the intention of proving the benefits of development strategy that we have chosen to study. It was found the chosen scenario could improve NPV by 3 folds in comparison to a base case scenario. The positioning of the wells was successful as all producers were placed in zones having initial water saturation less than 0.4., and all injectors were placed high water saturation zones. Moreover, we established a procedure regarding well trajectory design and optimization by taking into account, minimum dogleg severity and maximum duration for a well to be drilled and completed with respect to a time threshold. The findings as well as the workflow that will be presented hereafter could be considered as a guideline for subsequent tasks pertaining to the process of decision making, especially when it has to do with the development of green oil and gas fields and will certainly help in the placement of wells in less risky and cost-effective locations.

中文翻译：

---

联合优化流量分配、井位和井眼轨迹的油田开发策略

油田规划工程师和决策者在油藏管理和油气采收优化方面必须实现的主要目标之一是实现财务投资回报的最大化。由于大量的决策变量和一些不确定性，这项任务仍然非常具有挑战性，促使工程师和技术顾问寻求稳健的优化方法，以便将油井最佳地放置在最有利可图的位置，重点是增加净现值超过一个项目生命周期。寻求提供优质建议还取决于在整个优化过程中如何处理和制定某些不确定性——地质、经济和流动模式。随着计算机能力的增强和卓越的优化技术的出现，石油和天然气行业拥有广泛的工具，可以概览石油资产的价值最大化。在这些工具中，属于随机优化方法的遗传算法已成为业内众所周知的尝试解决井位和生产分配问题时应用的最佳替代方法之一，尽管计算要求很高。这项工作的目的是在碳氢化合物生产优化领域提出一种新方法，除了优化井的轨迹外，还可以根据单个目标函数来确定控制设置和井位。以注水辅助海上合成油田储层动力学模型为出发点，这项工作包括构建一个使用人工神经网络生成的数据驱动模型。然后，我们使用 Matlab 的遗传算法工具箱来执行所有需要的优化；从中，我们能够为要实现的一组井建立钻井计划，并且我们可以同时获得井位和配置（垂直或水平）、井类型（生产井或注入井）、井长、水平面内的井定向，以及与一组线性和非线性约束相关的井控制（流速）和近井筒压力。这些约束的公式化是为了重现真实的现场开发考虑，并且借助在 Matlab 上编写的遗传算法程序，我们能够满足这些约束，例如，最大生产和注入速率、最佳井筒压力、最大允许液体处理能力、最佳井位、钻井和完井的最长持续时间，以及其他考虑因素。我们调查了一些情景，目的是证明我们选择研究的发展战略的好处。结果发现，与基本案例情景相比，所选情景可以将 NPV 提高 3 倍。井的定位是成功的，因为所有生产井都被放置在初始含水饱和度小于 0.4 的区域，并且所有注入器都被放置在高含水饱和度区域。此外，我们建立了一个关于井轨迹设计和优化的程序，考虑到，相对于时间阈值，钻井和完井的最小狗腿严重程度和最大持续时间。下文将介绍的研究结果和工作流程可被视为与决策过程相关的后续任务的指南，特别是当它与绿色油气田的开发有关时，肯定会有助于将油井放置在风险较低且成本效益较低的位置。