油藏经营管理优化决策方法研究

【作者】 梁保升；

【导师】 杜志敏；

【作者基本信息】 西南石油学院 ， 油气田开发工程， 2003， 硕士

【摘要】 油藏经营管理贯穿于整个油田开发生命周期，以提高最终采收率、获取最佳的开采效益为目的，属于复杂的多目标优化决策问题。油气藏的复杂多变性、不确定性以及高投资高风险性决定了要实现油藏经营管理，必须进行正确的优化决策。因此可以说，优化决策是油藏经营管理成功与否的关键。然而，对于如何把已有的其它学科的算法，结合油藏经营管理的实际问题进行优化决策，寻找出贴切油藏经营管理优化决策的方法，以便从整体上进行宏观把握这些方面，研究的并不多。本文在研究油藏经营管理优化决策内容、工作过程的基础上，通过大量的文献调研，借鉴数学、管理运筹学、计算机等多学科的最新成果，尝试研究出适合油藏经营管理优化决策的通用流程模式，建立了运作模型，并探索出合理的、可操作性强的模型求解方法，为油藏经营管理实现真正的智能管理集成增添了关键的一环。 通过本文系统研究，得到如下结论和认识： (1) 从管理学角度分析了影响油藏经营管理优化决策的因素，给出了优化决策的基本过程，并分析了整个油藏经营管理周期各个阶段优化决策的主要内容。 (2) 将油藏经营管理优化决策影响因素进行总体分类，提出两大类指标体系——开发评价指标体系和经济评价指标体系，形成了比较完整的评价体系。 (3) 通过研究对比各种优化决策方法，针对油藏经营管理优化决策的特点，选取了适合油气用实际操作的三种方法——层次分析法、复合形调优算法和遗传算法。给出了基于层次分析法的油藏经营管理优化决策通用模式。 (4) 大部分复杂的决策问题都同时含有许多定性与定量因素，层次分析法可以很好的解决这类问题。同时，在工作量增加不大的情况下，便于进行敏感性的分析和评价。以白庙气田的开发设计方案为例，表明层次分析法是一种集成定量和定性因素的实用性很强的方法。 (5) 对于确定性措施规划问题，建立多目标规划模型，求解采用复合摘要 形调优算法。以华北南马庄马2断块为例，考虑了增油和控水两种情况，通过对比分析得出最优方案。同时证实了该算法是一种比较有效地处理有约束条件的多目标优化的直接搜索方法。 (6)实际油藏构造和流动机理的模糊性和不确定性，使得传统的模型和算法无法或者很难描述和求解。为了更好的处理实际生产中参数的不确定性，根据数学规划论中处理随机现象的机理，建立多目标随机规划模型，模型求解采用基于随机规划的遗传算法。遗传算法是一种启发式蒙特卡洛方法，有强大的全局最优解搜索功能。研究了该算法原理及实现的关键技术，设计出基于随机模拟的遗传算法的程序步骤。更多还原

【Abstract】 Reservoir management runs through the whole life of petroleum recovery. It aims to enhance oil recovery and to acquire optimum exploitation benefits, In operation, reservoir management faces complicated multi-objective optimum decision-making. It is essential to make optimum decision to realize reservoir management for the complexity, uncertainty and high investment risk in petroleum industry. Therefore, optimization and decision-making is crucial to successful reservoir management. However, in order to control as a whole, it is not wide or systematic on the aspects of how to introduce algorisms existed in other fields into petroleum discipline, of combining actual problems and theoretic methods, of seeking out proper ways to treat with scenarios in reservoir management. Via large and wide literature investigation, this paper, based on the studies of contents, performance process of optimization and decision-making, referred to multi-disciplinary up-to-date achievements such as mathematics, management sciences, and computer science, tries to search out the general process for reservoir management optimization and decision-making, sets up operation models and probes proper model solution methods with strong maneuverability, and adds a key to intelligent reservoir management integration.Through systematic study, this paper achieves following results and recognitions:(1) Influence factors in reservoir management optimization and decision-making are analyzed from the view of management. The general process for optimization and decision-making is put forwarded. And the chief work during each period of the whole reservoir management life is made clear.(2) After classification, influence factors are categorized into two systems: one is development evaluation indices system, the other economic evaluation indices system. Thus, a relatively complete evaluation structure is shaped.(3) Based on the studies of all kinds of methods in optimization and decision-making, combined with the characters in reservoir management, threemethods which match the actual reservoir operation well, Analytic Hierarchy Process, Polytope Algorithm, Genetic Algorithm, are recommended. General pattern for reservoir management optimization and decision-making on the basis of Analytic Hierarchy Process is developed.(4) Almost all the complex decision-making problems contain qualitative and quantitative factors simultaneously. Analytic Hierarchy Process can deal with such problems well. Besides, with little more work, factor sensitivity analysis is carried out. The application case, optimum reservoir planning selection of Baimiao gas field, verifies that Analytic Hierarchy Process is easily available indeed in operation by integrate qualitative and quantitative aspects.(5) Multi-objective model is established for determinative measure programming. Polytope Algorithm is adopted to solve such model. In the case of Nanmazhuang Ma 2 Block, two conditions, oil production increase and water production control are considered. And an optimum plan is determined after comprehensive comparison. At the same time, this case also shows proof that Polytope Algorithm is an effective direct search method to treat multi-objective optimization with constraining conditions.(6) The ambiguity and uncertainty of reservoir actual structure and fluid flow mechanics make it difficult for conventional modeling and algorithms to describe and solve. Thus, according to mechanics of dealing with stochastic phenomena in programming theory, multi-objective stochastic programming model is developed to dispose parameter uncertainty. As a heuristic Monte Carlo approach with powerful global searching, Genetic Algorithm based on stochastic programming is utilized. This paper studies fundamental theory, key technologies. Program procedure of Genetic Algorithm based on stochastic simulation is advanced.更多还原