典型试验区石油开发指标的变化规律预测及效益评价
|
摘要
随着我国国民经济快速的发展,石油消费总量日益增长,1991年我国石油消费量增长率首次大于原油产量增长率,自1993年开始,我国由石油净出口国转变为净进口国,石油消费缺口日益增大。根据国务院2012发布的《中国的能源政策》白皮书称,中国生产原油20748万吨,同比增长1.9%；进口原油27109万吨,同比增长7.3%；原油对外依存度56.4%,分析认为这一数据将继续攀升。石油资源作为一种重要的非再生自然资源,影响着每个国家和地区的经济发展、政治稳定、军事安全和外交政策,因此我国将石油资源列为影响社会经济可持续发展的战略资源之一。历经五十余年的发展,我国逐步形成“两种资源,两种市场”的战略格局,一方面,加强与国际石油生产国的沟通与合作,积极利用国际石油资源来弥补自身消费缺口；另一方面,立足国内,加强对国内生产油田的勘探、开发和利用,充分实现油田稳产增产工作,提高油田的利用率和自给率。
然而,国内油田生产形势却较为严峻。据资料统计,我国大部分陆上油田都进入了开采的中后期,油田产量递减日益加快,综合含水率大都已经接近90%,可采储量采出程度也相对较高,“老字辈”大型油田都已进入特高含水期和高采出程度的双高阶段,开采经济效益相对较低。对于这些进入开采中后期的主力油田而言,如何优化选取量化指标客观真实地反映油田开发现状?如何实现对油田开发指标变化规律的预测?如何反映油田识别和判定油田开发的潜力和前景?如何综合全面地评价油田的开发技术效益和经济效益?这些都是油田值得思考和探索的问题。本文选择大庆油田某典型试验区作为研究对象,以石油开发指标变化规律预测及效益评价为研究内容,不仅具有较好的理论研究价值,同时也具备较强的实践应用参考意义。
针对上述提及的问题,本文的创新点主要体现在以下几个方面：
第一,如何优化选取量化指标客观真实反映油田开发效益?油田在生产过程中积累了大量的数据,但这些生产数据往往存在冗余大、不完整和属性关系复杂等特点,难以直接有效地加以利用。那么,如何通过有效的方法对这些繁杂的数据进行识别和处理,以提取对油田开发评价有用的信息,并优化选取得到科学合理的评价指标,将是本文研究的重点所在。在评价指标优化选取的方法上,本文将硬计算和软计算两种方法相融合,基于相关分析创建SVM-FCM特征选择模型。首先将评价模式结果分为5类(很好,好,中等,差,很差),接着提出影响油田开采效益的相关技术指标：生产天数、泵深、动液面、含水率、产油量和产量递减率,最后通过计算SVM灵敏度和相关系数的绝对值,优选出影响油田开采效益的关键指标为产量递减率。
第二,如何实现对油田开发指标变化规律的预测?基于遗传算法优化BP神经网络,构建石油产量预测模型。BP神经网络具有较强的并行存储和并行计算的能力,以及较强的自学习能力和非线性映射能力,适用于复杂多变的油田地质数据。但是,BP神经网络算法本质上是局部搜索算法,预测结果的准确在很大程度上取决于样本数据的准确性,因此如果利用遗传算法的全局搜索能力,优化神经网络的网络权值和阈值,能较好地提高神经网络的收敛速度和预测精度,以实现两者互补,发挥各自长足,提高模型的仿真精度与自适性。首先本文将油田综合含水率、产量递减率、流动系数和采收率作为输入因子,并确定平均月产油量为关键预测输出指标；然后对原始数据进行归一化处理,消除各因素之间量纲的差异对预测带来的影响；最终结合BP神经网络和遗传算法确定各个因子的个体编码,实现油田产量的预测。结果表明,平均月产油量的仿真预测结果与原始产量训练数据具有较好的一致性,这就说明基于遗传算法优化BP神经网络来预测石油产量,预测准确性较高。
第三,如何反映油田识别和判定油田开发的潜力和前景?建立了反映剩余油饱和度和剩余油储量丰度的量化指标体系,从而构建出剩余油潜力评价模型和水驱开发效果评价模型,并运用变异系数法确定各个指标的权重,以有效地对典型试验区开发潜力进行科学分类。
第四,如何综合全面地评价油田的开发技术效益和经济效益?我国大多数油田都面对着多井、低产、高含水和高采出的开发局面,如果从技术效益或经济效益单方面地对油田做出评价,难免会以偏概全,很难做到定量评价。那么,如何通过精细准确的评价方法综合全面地对油田效益进行定性定量表述呢?油田开发效益受到市场经济、地质条件、政策计划和生产工艺设备等多种因素的制约和影响,本文基于模糊数学的综合评价,根据模糊数学隶属度理论,运用定性和定量评价相结合的方法对油田整体开发效益做出一个综合的评价。
本文的主要研究内容和章节安排如下：
第一章,绪论部分。本章简要介绍了选题的研究背景和意义,概述了研究内容与结构,以及本文主要的创新点。
第二章,评价指标的优化选取。本章对属性约简的历史发展进程和实践应用价值做了简单的介绍,概述了属性约简常用的方法,其中包括基于差别矩阵的属性约简、启发式的属性约简、不完备性系统中的属性约简、分布式和并行式的属性约简算法,以及特征转化和子集选取。基于上述理论研究基础,并结合油田实际生产状况(原始属性集合包括：生产天数、泵深、动液面、含水率和产油量),本章在构建油田效益评价指标过程中引入了基于硬计算(Hard Computing, HC)模式和软计算(Soft Computing, SC)模式融合的属性约简思路,采用基于相关分析的SVM-FCM特征选择,首先构建5个不同的SVM灵敏度分析器,计算原始属性集合每个指标的灵敏度值来进行初步约简,进而计算变量间相关系数的绝对值,以识别最优特征。我国油田效益评价指标在指标规范性、方法科学性和结果客观性等方面都存在着一定的不足,而这些不足恰是本章研究的重点所在。
第三章,开发技术效益评价。本章从综合含水率、产油量状况、累计产油量和增油量、产量递减率、井网密度五个方面客观全面地反映了大庆油田BED试验区的开采现状,从开发技术效益的角度对该试验区做出了评价。其中,在产量递减率指标描述过程中,分别计算出不同影响因素的变化系数,得出了影响产量递减的关键因素为：井网密度变化、生产压差变化、流动系数变化和含水系数变化。另外,在井网密度指标描述过程中,分别介绍了经济合理井网密度和经济极限井网密度的概念,在不同收益水平、不同投资开发年限、不同石油价格、不同税收水平等各方面条件下,计算出大庆油田BED试验区不同的井网密度界限值,给油田制定生产决策和方案选择时提供可行性建议。
第四章,基于遗传算法优化BP神经网络的原油产量预测。在实际操作应用过程中,高含水阶段油田的地质数据具有复杂多变的特点,因此本章在原油产量预测模型构建过程中,将BP神经网络和GA遗传算法相结合,利用遗传算法优化神经网络初始权值和阈值,这样就能有效地减少模型预测的不确定性,使得神经网络可以更快达到稳定状态,进而提高产量预测结果的准确度,最终得到最优预测结果。在案例分析中,遵循数据可得性、全面性、动态性、对比性和独立性等原则,本章选取了综合含水率、产量递减率、流动系数和注采比为输入因子,平均月产油量为输出因子,通过训练数据和测试数据的一致性表明,遗传算法优化下的BP神经网络具有较高的准确性,该方法比较适合特高含水期油田的产量预测。
第五章,经济效益动态评价。本章从经济效益的角度,基于大庆油田BED试验区2009年项目建设总投资估算,分别从投资回收期、净现值、内部收益率、投资利润率和利税率等角度对该试验区进行经济效益评价。另外,考虑到油田生产决策的特殊性,本章将经济效益转化为经济产量和经济储量,提出了经济动态下限产量和经济极限储量两个评价指标,使得经济效益动态评价更加直接化和可视化。其中,在经济动态下限产量和经济极限储量模型构建过程中,充分考虑了资金时间价值、市场价格波动和成本的变化,在盈亏平衡净现值为零的思路下,推导计算出试验区油田的经济界限值,这对增加经济预警意识,制定科学合理的投资决策具有重大的意义。
第六章,剩余油潜力评价。如何确定剩余油分布规律以及如何对其进行潜力评价是油田开发中后期井网密度综合调整挖潜的关键所在,因此本章首先介绍了剩余油潜力评价的评价思路,首先,通过油田油水相对渗透状况和含油饱和度描述评价区域的水淹状况,进而利用经济合理井网密度和单井经济储量描述评价区域的剩余油经济储量丰度,最终将两者进行联立评价得出剩余油潜力综合评价。研究结果表明,大庆油田BED试验区进入特高含水期阶段,其水淹状况属于“强水淹”级别,但其可采储量丰度远大于经济储量丰度,这就说明该区域还有较大的可采空间。
第七章,油田效益综合评价。前面分别从技术效益和经济效益的角度分别评价了油田开发效益,那么如何综合反映评价区域的整体开发效果呢?这就是本章的研究重点所在。本章首先将效益综合评价指标分为开发技术指标、生产管理指标和经济效益指标三大类,优化选取了井网完善状况等11个评价指标,并采用变异系数客观赋权的方法来确定各指标的权重,进而采用岭型隶属度函数模型构建各个评价等级,并采用评判矩阵进行单指标标准化评价,最终进行多指标加权综合评价,即分别对每一类指标内相关权重进行评价,从而得到油田整体开发效果。本章构建了适用于油田效益综合评价的指标体系,综合效益评价模型分析结果表明大庆油田BED试验区整体开发效果较好,采用逆向分析方法对评价结果进行解析,发现影响开发效果的核心因素为开发技术类指标中的“综合递减率”和“自然递减率”。因此,加强对综合递减率和自然递减率的控制,才能更好地实现老油田稳产、增产挖潜工作。
With the fast development of national economy, total oil consumption is increasing. In1991, the national growth rate of oil consumption was greater than the crude production. Since1993, China has changed from the net exporter to the net importer, and the oil consumption gap is still increasing. In the white paper of 《Chinese Energey Policy》 issued by the State Council in2012, Chinese crude oil production is207.48million tons that the year-on-year growth is1.9%, but the import quantity is271.09million tons that year-on-year growth is7.3%. Moreover, the foreign-trade dependence has reaches to56.4%, which will continue to rise in the future. As an important kind of non-renewable natural resource, oil resource has great influence on the economic development, political stability, military security and diplomatic policy. Therefore, China has regarded the petroleum resource as strategic resource, which affects the sustainable development of society and economy. After the developments in50years, our country has formed the strategy pattern of "Two kinds of resources, two kinds of markets", on the one hand, it's necessary to strengthen the communication and cooperation with international petroleum producer country, and make use of international resources to make up for our own consumption gap; on the other hand, our country should pay attention to the exploration, development and utilization of domestic production oilfields, the oilfields also need to keep stable or incremental production so as to improve the rate of utilization and self-sufficiency.
However, the domestic production situation is severe. Referred to the statistics, most overland oilfield has entered into the declining stage. In this stage, the production is decreasing, the comprehensive moisture content is close or over than90%and the degree of recoverable reserves is relatively in high level. For this reason, the traditional large-scale oilfields have accessed to the ultra-high water cut stage and high recovery degree stage, and its economic efficiency isn't very well. For these great oilfields, how to reflect the present exploration situation accurately and objectively through the quantitative indicators? How to realize the future trend prediction of development indexes? How to reflet and identify the exploration potential and prospect of targeted oilfield? How to make comprehensive evaluation on technology benefits and economic benefits? These questions are worth thinking and researching. The article chooses a typical trial zone in DaQing Oilfield as the objective of study, and makes the variation prediction and benefit evaluation of development indexes as research contents, which has good theoretical research value and strong practical application significance.
As to the above questions mentioned, the innovations are reflected in the following aspects:
Firstly, how to reflect the present exploration situation accurately and objectively through the quantitative indicators? A large number of statistics has been accumulated and recorded in the production process, but these statistics often are redundant, incomplete and complex, which are hard to be used directly. Therefore, how to identify these multifarious statistics through the effective methods for extracting useful information could achieve scientific and reasonable evaluation indexes system in the oilfield exploration evaluation process, and this will be the research key in the article. In the process of selecting evaluation indexes, the article integrates the soft computing and hard computing to create the SVM-FCM future selection model based on the correlation analysis. The evaluation results are divided into5kinds (very good, good, medium, poor and very poor), and then the article puts forward to related technical indicators that influences oilfield exploration benefit, such as production days, pump depth, dynamic fluid level, moisture content, oil production and production decline rate. Finally through the calculation of SVM sensitivity and the absolute value of correlation coefficient, the key indicator is selected and applied as production decline rate.
Secondly, how to realize the future trend prediction of development indexes? BP neural network has the strong ablity of parallel storage, parallel computing, elf learning and nonlinear mapping, which is for complicated geological statistics. However, the BP neural network algorithm is essentially a local search algorithm, and the accuracy of prediction results largely depend on the accuracy of the sample statistics. If using the global search ability of genetic algorithm to optimize network weights and threshold value, it will well improve the convergence rate and prediction accuracy, which the relationships between BP neural network and genetic algorithm are complementary and helpful to gain the accuracy and applicability in the predicition model. Using genetic algorithm optimizes BP neural network in the procution prediction model. The imput factors are determined as comprehensive moisture content, production decline rate, flow coefficient and recovery, on the opposite, the average monthly production is regarded as the key output indicator. The following step is the normalization of original statistics, and the normalization can eliminate the disadvantage influences because of dimensional differences among various indexes. At last, through combining with BP neural network and genetic algorithm to gain the individual coding of each factor, the article has realized the prediction of oilfield production. The results show that the simulation prediction of average monthly production has good consistency with the original training data, which means that using genetic algorithm to optimize the BP neural network has higher prediction accuracy.
Thirdly, how to reflet and identify the exploration potential and prospect of targeted oilfield? From the establishment of quantitative indicators system reflecting remaining oil saturation and remaining oil reserves abundance, the article builds the evaluation models of remaining oil potential and water-flooding development effectiveness. And then, the model applies variation coefficient method to get each index weight, and make scientific classification effectively on development potential in typical trial zone.
Fourthly, how to make comprehensive evaluation on technology benefits and economic benefits? Most oilfields face the situation of multi-well, lower production, high water cut and exploration. If the evaluations are made from technology benefits or economic benefit unilaterally, it will take a part for the whole that is difficult to do quantitative evaluation. Therefore, how to make a qualitative and quantitative description on oilfiled benefit through elaborate and accurate methods? Oilfield development benefit is influenced by many kinds of factors, including market economy, geological condition, political planning and processing equipment. Therefore, the comprehensive evaluation based on fuzzy mathematics degree theory makes an integral analysis on oilfield development benefit via qualitative and quantitative methods.
The major contents are introduced as following:
The first chapter is preface. The chapter briefly introduces the background and significance of the topic research, and also summaries the research content, structure and main innovation points.
The second chapter is the optimization selection of evaluation indexes. The article briefly introduces historical development and application value of attribute reduction, as shown, the commonly methods are included dissimilarity matrix, heuristic attribution, non-completeness system, distributed and parallel type algorithm, feature transformation and subset selection. Based on these theoretical foundations and actual production situations (original attribute set includes:production day, pump depth, dynamic fluid level, moisture content and oil production), this chapter combines the hard computing model with soft computing model in the process of constructing evaluation indexes system. The model builds5different SVM sensitivity analyzers and makes correlation analysis on SVM-FCM feature selection. After calculating the sensitivity value of every index in original attribute set to finish the initial attribution reduction, the absolute value of correlation coefficients for the variables will help to identify the optimal characteristics. The benefit evaluation indexes in our country have some limits in indicators normalization, methods science and results objectivity, and these limits are just the key of research.
The third chapter is benefit evaluation of development technique. In the view of development technology benefit, this part comprehensive reflects exploration situation of typical trial zone in DaQing Oilfield from comprehensive moisture content, production status, cumulative production and incremental production, production declining rate and well spacing density five aspects. In addition, in the process of describing production declining rate, the variation coefficient among different influenced factors shows that the key factors are the changes of well spacing density, production pressure difference, flow coefficient and water coefficient. In the chapter, the well spacing density indexes separately introduce the concept of economic reasonable well spacing density and economic limit well pattern density. In different conditions such as different income levels, different investment periods, different oil prices and different tax levels, the density threshold values will provide feasible suggestions for making production decisions and planning selection.
The forth chapter is production prediction based on the genetic algorithm to optimize BP neural network. In the actual operation process, the geological statistics of oilfield in the high water cut stage have the characteristics of complication, so in the model of production prediction, this chapter combines BP neural network with the genetic algorithm GA, and uses genetic algorithm to optimize weights and threshold value in BP, which will effectively reduce the prediction uncertainty and improve the statistics quality. In this way, the BP neural network can achieve stable state and better results accuracy than before, and finally gain the optimal prediction results. In the case study, considering the principles of availability, comprehensiveness, oscillatory, comparativeness and independence, the chapter selects comprehensive moisture content, production decline rate, flow coefficient and injection production ratio as input factors, and then the average monthly production as output factor. The consistency of training data and testing data shows that using the genetic algorithm to optimize BP neural network is equipped with high accuracy, therefore this method is more suitable for production prediction in high water cut stage.
The fifth chapter is dynamic evaluations of economic benefits. From the view of economic benefits, the chapter respectively uses investment payback period, net present value, internal rate of return, return on investment ratio, profit and tax investment ratio to make economic benefit evaluation for typical test zone in DaQing Oilfield, which is the foundation of projects total investment estimation in2009. Furthermore, taking the particularity of production decision into consideration, this part transforms economic benefits into economic production and economic reserve, and then put forward the dynamic economic limit production and economic limit reserve, so these two indexes make economic evaluation more directly and visually. In the calculation model based on the time value of money, market price fluctuation and cost changeability, these indexes utilize the profit and loss balance of net present value to achieve economic limit value, which has great significance in increasing economic warning consciousness and making scientific and reasonable investment decisions.
The sixth chapter is the potential evaluation of remaining petroleum. How to understand the distribution law of remaining petroleum and how to carry on the potential evaluation are the keys of comprehensive adjustment on well spacing density in the late exploration stage. The chapter introduces the evaluation thoughts:first of all, the water flooded conditions are described through oilfield relative permeability and saturation; at the same time, the economic reserves abundance of remaining petroleum is described through the economic spacing density and economic reserves of every single well; finally, the simultaneous methods are combined together to make a comprehensive evaluation of potential evaluation. The research results shows that BED Test Zone in DaQing Oilfield has entered into the high water cut stage, and its water flooding status reaches the "strong flooding" level. However, the recoverable reserves abundance is far higher than the the economic reserves abundance, which means that it has enough recoverable space.
The seventh chapter is comprehensive evaluation of overall exploration efficiency. The former part respectively makes an evaluation from the technology effectiveness and economic effectiveness, but how to reflect the comprehensive effectiveness of the overall regional? This is the key research contents in this part. The chapter divides the evaluation indexes into technology indexes, management indexes and economic indexes, and then selects11specific evaluation indexes including adjustment condition of well network etc. Besides, the weights of indexes are determined through the coefficient of variation, and the related evaluation grades are constructed by the model of ridge membership function. Finally, this part uses evaluation matrixes to make standardization evaluation of single index, and ultimately reaches to comprehensive evaluation of multi-indexes weighted, which is namely respectively the calculation of every index to get the comprehensive development effectiveness. The article constructs the indexes system that is suitable for comprehensive evaluation, and the evaluation model indicates that the development benefits are good. From the inverted analysis to the evaluation results, the key factors are "comprehensive decline rate" and "natural decline rate" in technology indexes affecting development effectiveness. Therefore, strengthening the control to "comprehensive decline rate" and "natural decline rate" can better achieve the stable or increasing oilfield production.
然而,国内油田生产形势却较为严峻。据资料统计,我国大部分陆上油田都进入了开采的中后期,油田产量递减日益加快,综合含水率大都已经接近90%,可采储量采出程度也相对较高,“老字辈”大型油田都已进入特高含水期和高采出程度的双高阶段,开采经济效益相对较低。对于这些进入开采中后期的主力油田而言,如何优化选取量化指标客观真实地反映油田开发现状?如何实现对油田开发指标变化规律的预测?如何反映油田识别和判定油田开发的潜力和前景?如何综合全面地评价油田的开发技术效益和经济效益?这些都是油田值得思考和探索的问题。本文选择大庆油田某典型试验区作为研究对象,以石油开发指标变化规律预测及效益评价为研究内容,不仅具有较好的理论研究价值,同时也具备较强的实践应用参考意义。
针对上述提及的问题,本文的创新点主要体现在以下几个方面：
第一,如何优化选取量化指标客观真实反映油田开发效益?油田在生产过程中积累了大量的数据,但这些生产数据往往存在冗余大、不完整和属性关系复杂等特点,难以直接有效地加以利用。那么,如何通过有效的方法对这些繁杂的数据进行识别和处理,以提取对油田开发评价有用的信息,并优化选取得到科学合理的评价指标,将是本文研究的重点所在。在评价指标优化选取的方法上,本文将硬计算和软计算两种方法相融合,基于相关分析创建SVM-FCM特征选择模型。首先将评价模式结果分为5类(很好,好,中等,差,很差),接着提出影响油田开采效益的相关技术指标：生产天数、泵深、动液面、含水率、产油量和产量递减率,最后通过计算SVM灵敏度和相关系数的绝对值,优选出影响油田开采效益的关键指标为产量递减率。
第二,如何实现对油田开发指标变化规律的预测?基于遗传算法优化BP神经网络,构建石油产量预测模型。BP神经网络具有较强的并行存储和并行计算的能力,以及较强的自学习能力和非线性映射能力,适用于复杂多变的油田地质数据。但是,BP神经网络算法本质上是局部搜索算法,预测结果的准确在很大程度上取决于样本数据的准确性,因此如果利用遗传算法的全局搜索能力,优化神经网络的网络权值和阈值,能较好地提高神经网络的收敛速度和预测精度,以实现两者互补,发挥各自长足,提高模型的仿真精度与自适性。首先本文将油田综合含水率、产量递减率、流动系数和采收率作为输入因子,并确定平均月产油量为关键预测输出指标；然后对原始数据进行归一化处理,消除各因素之间量纲的差异对预测带来的影响；最终结合BP神经网络和遗传算法确定各个因子的个体编码,实现油田产量的预测。结果表明,平均月产油量的仿真预测结果与原始产量训练数据具有较好的一致性,这就说明基于遗传算法优化BP神经网络来预测石油产量,预测准确性较高。
第三,如何反映油田识别和判定油田开发的潜力和前景?建立了反映剩余油饱和度和剩余油储量丰度的量化指标体系,从而构建出剩余油潜力评价模型和水驱开发效果评价模型,并运用变异系数法确定各个指标的权重,以有效地对典型试验区开发潜力进行科学分类。
第四,如何综合全面地评价油田的开发技术效益和经济效益?我国大多数油田都面对着多井、低产、高含水和高采出的开发局面,如果从技术效益或经济效益单方面地对油田做出评价,难免会以偏概全,很难做到定量评价。那么,如何通过精细准确的评价方法综合全面地对油田效益进行定性定量表述呢?油田开发效益受到市场经济、地质条件、政策计划和生产工艺设备等多种因素的制约和影响,本文基于模糊数学的综合评价,根据模糊数学隶属度理论,运用定性和定量评价相结合的方法对油田整体开发效益做出一个综合的评价。
本文的主要研究内容和章节安排如下：
第一章,绪论部分。本章简要介绍了选题的研究背景和意义,概述了研究内容与结构,以及本文主要的创新点。
第二章,评价指标的优化选取。本章对属性约简的历史发展进程和实践应用价值做了简单的介绍,概述了属性约简常用的方法,其中包括基于差别矩阵的属性约简、启发式的属性约简、不完备性系统中的属性约简、分布式和并行式的属性约简算法,以及特征转化和子集选取。基于上述理论研究基础,并结合油田实际生产状况(原始属性集合包括：生产天数、泵深、动液面、含水率和产油量),本章在构建油田效益评价指标过程中引入了基于硬计算(Hard Computing, HC)模式和软计算(Soft Computing, SC)模式融合的属性约简思路,采用基于相关分析的SVM-FCM特征选择,首先构建5个不同的SVM灵敏度分析器,计算原始属性集合每个指标的灵敏度值来进行初步约简,进而计算变量间相关系数的绝对值,以识别最优特征。我国油田效益评价指标在指标规范性、方法科学性和结果客观性等方面都存在着一定的不足,而这些不足恰是本章研究的重点所在。
第三章,开发技术效益评价。本章从综合含水率、产油量状况、累计产油量和增油量、产量递减率、井网密度五个方面客观全面地反映了大庆油田BED试验区的开采现状,从开发技术效益的角度对该试验区做出了评价。其中,在产量递减率指标描述过程中,分别计算出不同影响因素的变化系数,得出了影响产量递减的关键因素为：井网密度变化、生产压差变化、流动系数变化和含水系数变化。另外,在井网密度指标描述过程中,分别介绍了经济合理井网密度和经济极限井网密度的概念,在不同收益水平、不同投资开发年限、不同石油价格、不同税收水平等各方面条件下,计算出大庆油田BED试验区不同的井网密度界限值,给油田制定生产决策和方案选择时提供可行性建议。
第四章,基于遗传算法优化BP神经网络的原油产量预测。在实际操作应用过程中,高含水阶段油田的地质数据具有复杂多变的特点,因此本章在原油产量预测模型构建过程中,将BP神经网络和GA遗传算法相结合,利用遗传算法优化神经网络初始权值和阈值,这样就能有效地减少模型预测的不确定性,使得神经网络可以更快达到稳定状态,进而提高产量预测结果的准确度,最终得到最优预测结果。在案例分析中,遵循数据可得性、全面性、动态性、对比性和独立性等原则,本章选取了综合含水率、产量递减率、流动系数和注采比为输入因子,平均月产油量为输出因子,通过训练数据和测试数据的一致性表明,遗传算法优化下的BP神经网络具有较高的准确性,该方法比较适合特高含水期油田的产量预测。
第五章,经济效益动态评价。本章从经济效益的角度,基于大庆油田BED试验区2009年项目建设总投资估算,分别从投资回收期、净现值、内部收益率、投资利润率和利税率等角度对该试验区进行经济效益评价。另外,考虑到油田生产决策的特殊性,本章将经济效益转化为经济产量和经济储量,提出了经济动态下限产量和经济极限储量两个评价指标,使得经济效益动态评价更加直接化和可视化。其中,在经济动态下限产量和经济极限储量模型构建过程中,充分考虑了资金时间价值、市场价格波动和成本的变化,在盈亏平衡净现值为零的思路下,推导计算出试验区油田的经济界限值,这对增加经济预警意识,制定科学合理的投资决策具有重大的意义。
第六章,剩余油潜力评价。如何确定剩余油分布规律以及如何对其进行潜力评价是油田开发中后期井网密度综合调整挖潜的关键所在,因此本章首先介绍了剩余油潜力评价的评价思路,首先,通过油田油水相对渗透状况和含油饱和度描述评价区域的水淹状况,进而利用经济合理井网密度和单井经济储量描述评价区域的剩余油经济储量丰度,最终将两者进行联立评价得出剩余油潜力综合评价。研究结果表明,大庆油田BED试验区进入特高含水期阶段,其水淹状况属于“强水淹”级别,但其可采储量丰度远大于经济储量丰度,这就说明该区域还有较大的可采空间。
第七章,油田效益综合评价。前面分别从技术效益和经济效益的角度分别评价了油田开发效益,那么如何综合反映评价区域的整体开发效果呢?这就是本章的研究重点所在。本章首先将效益综合评价指标分为开发技术指标、生产管理指标和经济效益指标三大类,优化选取了井网完善状况等11个评价指标,并采用变异系数客观赋权的方法来确定各指标的权重,进而采用岭型隶属度函数模型构建各个评价等级,并采用评判矩阵进行单指标标准化评价,最终进行多指标加权综合评价,即分别对每一类指标内相关权重进行评价,从而得到油田整体开发效果。本章构建了适用于油田效益综合评价的指标体系,综合效益评价模型分析结果表明大庆油田BED试验区整体开发效果较好,采用逆向分析方法对评价结果进行解析,发现影响开发效果的核心因素为开发技术类指标中的“综合递减率”和“自然递减率”。因此,加强对综合递减率和自然递减率的控制,才能更好地实现老油田稳产、增产挖潜工作。
With the fast development of national economy, total oil consumption is increasing. In1991, the national growth rate of oil consumption was greater than the crude production. Since1993, China has changed from the net exporter to the net importer, and the oil consumption gap is still increasing. In the white paper of 《Chinese Energey Policy》 issued by the State Council in2012, Chinese crude oil production is207.48million tons that the year-on-year growth is1.9%, but the import quantity is271.09million tons that year-on-year growth is7.3%. Moreover, the foreign-trade dependence has reaches to56.4%, which will continue to rise in the future. As an important kind of non-renewable natural resource, oil resource has great influence on the economic development, political stability, military security and diplomatic policy. Therefore, China has regarded the petroleum resource as strategic resource, which affects the sustainable development of society and economy. After the developments in50years, our country has formed the strategy pattern of "Two kinds of resources, two kinds of markets", on the one hand, it's necessary to strengthen the communication and cooperation with international petroleum producer country, and make use of international resources to make up for our own consumption gap; on the other hand, our country should pay attention to the exploration, development and utilization of domestic production oilfields, the oilfields also need to keep stable or incremental production so as to improve the rate of utilization and self-sufficiency.
However, the domestic production situation is severe. Referred to the statistics, most overland oilfield has entered into the declining stage. In this stage, the production is decreasing, the comprehensive moisture content is close or over than90%and the degree of recoverable reserves is relatively in high level. For this reason, the traditional large-scale oilfields have accessed to the ultra-high water cut stage and high recovery degree stage, and its economic efficiency isn't very well. For these great oilfields, how to reflect the present exploration situation accurately and objectively through the quantitative indicators? How to realize the future trend prediction of development indexes? How to reflet and identify the exploration potential and prospect of targeted oilfield? How to make comprehensive evaluation on technology benefits and economic benefits? These questions are worth thinking and researching. The article chooses a typical trial zone in DaQing Oilfield as the objective of study, and makes the variation prediction and benefit evaluation of development indexes as research contents, which has good theoretical research value and strong practical application significance.
As to the above questions mentioned, the innovations are reflected in the following aspects:
Firstly, how to reflect the present exploration situation accurately and objectively through the quantitative indicators? A large number of statistics has been accumulated and recorded in the production process, but these statistics often are redundant, incomplete and complex, which are hard to be used directly. Therefore, how to identify these multifarious statistics through the effective methods for extracting useful information could achieve scientific and reasonable evaluation indexes system in the oilfield exploration evaluation process, and this will be the research key in the article. In the process of selecting evaluation indexes, the article integrates the soft computing and hard computing to create the SVM-FCM future selection model based on the correlation analysis. The evaluation results are divided into5kinds (very good, good, medium, poor and very poor), and then the article puts forward to related technical indicators that influences oilfield exploration benefit, such as production days, pump depth, dynamic fluid level, moisture content, oil production and production decline rate. Finally through the calculation of SVM sensitivity and the absolute value of correlation coefficient, the key indicator is selected and applied as production decline rate.
Secondly, how to realize the future trend prediction of development indexes? BP neural network has the strong ablity of parallel storage, parallel computing, elf learning and nonlinear mapping, which is for complicated geological statistics. However, the BP neural network algorithm is essentially a local search algorithm, and the accuracy of prediction results largely depend on the accuracy of the sample statistics. If using the global search ability of genetic algorithm to optimize network weights and threshold value, it will well improve the convergence rate and prediction accuracy, which the relationships between BP neural network and genetic algorithm are complementary and helpful to gain the accuracy and applicability in the predicition model. Using genetic algorithm optimizes BP neural network in the procution prediction model. The imput factors are determined as comprehensive moisture content, production decline rate, flow coefficient and recovery, on the opposite, the average monthly production is regarded as the key output indicator. The following step is the normalization of original statistics, and the normalization can eliminate the disadvantage influences because of dimensional differences among various indexes. At last, through combining with BP neural network and genetic algorithm to gain the individual coding of each factor, the article has realized the prediction of oilfield production. The results show that the simulation prediction of average monthly production has good consistency with the original training data, which means that using genetic algorithm to optimize the BP neural network has higher prediction accuracy.
Thirdly, how to reflet and identify the exploration potential and prospect of targeted oilfield? From the establishment of quantitative indicators system reflecting remaining oil saturation and remaining oil reserves abundance, the article builds the evaluation models of remaining oil potential and water-flooding development effectiveness. And then, the model applies variation coefficient method to get each index weight, and make scientific classification effectively on development potential in typical trial zone.
Fourthly, how to make comprehensive evaluation on technology benefits and economic benefits? Most oilfields face the situation of multi-well, lower production, high water cut and exploration. If the evaluations are made from technology benefits or economic benefit unilaterally, it will take a part for the whole that is difficult to do quantitative evaluation. Therefore, how to make a qualitative and quantitative description on oilfiled benefit through elaborate and accurate methods? Oilfield development benefit is influenced by many kinds of factors, including market economy, geological condition, political planning and processing equipment. Therefore, the comprehensive evaluation based on fuzzy mathematics degree theory makes an integral analysis on oilfield development benefit via qualitative and quantitative methods.
The major contents are introduced as following:
The first chapter is preface. The chapter briefly introduces the background and significance of the topic research, and also summaries the research content, structure and main innovation points.
The second chapter is the optimization selection of evaluation indexes. The article briefly introduces historical development and application value of attribute reduction, as shown, the commonly methods are included dissimilarity matrix, heuristic attribution, non-completeness system, distributed and parallel type algorithm, feature transformation and subset selection. Based on these theoretical foundations and actual production situations (original attribute set includes:production day, pump depth, dynamic fluid level, moisture content and oil production), this chapter combines the hard computing model with soft computing model in the process of constructing evaluation indexes system. The model builds5different SVM sensitivity analyzers and makes correlation analysis on SVM-FCM feature selection. After calculating the sensitivity value of every index in original attribute set to finish the initial attribution reduction, the absolute value of correlation coefficients for the variables will help to identify the optimal characteristics. The benefit evaluation indexes in our country have some limits in indicators normalization, methods science and results objectivity, and these limits are just the key of research.
The third chapter is benefit evaluation of development technique. In the view of development technology benefit, this part comprehensive reflects exploration situation of typical trial zone in DaQing Oilfield from comprehensive moisture content, production status, cumulative production and incremental production, production declining rate and well spacing density five aspects. In addition, in the process of describing production declining rate, the variation coefficient among different influenced factors shows that the key factors are the changes of well spacing density, production pressure difference, flow coefficient and water coefficient. In the chapter, the well spacing density indexes separately introduce the concept of economic reasonable well spacing density and economic limit well pattern density. In different conditions such as different income levels, different investment periods, different oil prices and different tax levels, the density threshold values will provide feasible suggestions for making production decisions and planning selection.
The forth chapter is production prediction based on the genetic algorithm to optimize BP neural network. In the actual operation process, the geological statistics of oilfield in the high water cut stage have the characteristics of complication, so in the model of production prediction, this chapter combines BP neural network with the genetic algorithm GA, and uses genetic algorithm to optimize weights and threshold value in BP, which will effectively reduce the prediction uncertainty and improve the statistics quality. In this way, the BP neural network can achieve stable state and better results accuracy than before, and finally gain the optimal prediction results. In the case study, considering the principles of availability, comprehensiveness, oscillatory, comparativeness and independence, the chapter selects comprehensive moisture content, production decline rate, flow coefficient and injection production ratio as input factors, and then the average monthly production as output factor. The consistency of training data and testing data shows that using the genetic algorithm to optimize BP neural network is equipped with high accuracy, therefore this method is more suitable for production prediction in high water cut stage.
The fifth chapter is dynamic evaluations of economic benefits. From the view of economic benefits, the chapter respectively uses investment payback period, net present value, internal rate of return, return on investment ratio, profit and tax investment ratio to make economic benefit evaluation for typical test zone in DaQing Oilfield, which is the foundation of projects total investment estimation in2009. Furthermore, taking the particularity of production decision into consideration, this part transforms economic benefits into economic production and economic reserve, and then put forward the dynamic economic limit production and economic limit reserve, so these two indexes make economic evaluation more directly and visually. In the calculation model based on the time value of money, market price fluctuation and cost changeability, these indexes utilize the profit and loss balance of net present value to achieve economic limit value, which has great significance in increasing economic warning consciousness and making scientific and reasonable investment decisions.
The sixth chapter is the potential evaluation of remaining petroleum. How to understand the distribution law of remaining petroleum and how to carry on the potential evaluation are the keys of comprehensive adjustment on well spacing density in the late exploration stage. The chapter introduces the evaluation thoughts:first of all, the water flooded conditions are described through oilfield relative permeability and saturation; at the same time, the economic reserves abundance of remaining petroleum is described through the economic spacing density and economic reserves of every single well; finally, the simultaneous methods are combined together to make a comprehensive evaluation of potential evaluation. The research results shows that BED Test Zone in DaQing Oilfield has entered into the high water cut stage, and its water flooding status reaches the "strong flooding" level. However, the recoverable reserves abundance is far higher than the the economic reserves abundance, which means that it has enough recoverable space.
The seventh chapter is comprehensive evaluation of overall exploration efficiency. The former part respectively makes an evaluation from the technology effectiveness and economic effectiveness, but how to reflect the comprehensive effectiveness of the overall regional? This is the key research contents in this part. The chapter divides the evaluation indexes into technology indexes, management indexes and economic indexes, and then selects11specific evaluation indexes including adjustment condition of well network etc. Besides, the weights of indexes are determined through the coefficient of variation, and the related evaluation grades are constructed by the model of ridge membership function. Finally, this part uses evaluation matrixes to make standardization evaluation of single index, and ultimately reaches to comprehensive evaluation of multi-indexes weighted, which is namely respectively the calculation of every index to get the comprehensive development effectiveness. The article constructs the indexes system that is suitable for comprehensive evaluation, and the evaluation model indicates that the development benefits are good. From the inverted analysis to the evaluation results, the key factors are "comprehensive decline rate" and "natural decline rate" in technology indexes affecting development effectiveness. Therefore, strengthening the control to "comprehensive decline rate" and "natural decline rate" can better achieve the stable or increasing oilfield production.
引文
[1]李鼎鑫.石油管线之争考验中国能源[N].经济参考报,2003.08.
[2]葛家理.石油企业经营发展战略研究[M].北京：石油工业出版社,1998.
[3]赵培华.中国路上水驱油田开发测井技术[J].测井技术,1998.05.
[4]曾雪梅,范景明.北二东开采层系重组试验区水驱开发效果试验方案[R].大庆油田有限责任公司第三采油厂,2009.06.
[5]丁卫平.基于量子蛙跳协同进化的粗糙属性快速约简[J].电子学报,2011：2597-2598.
[6]Pawlak. Z. Rough sets. International Journal of Computer and Information Sciences[J].1982, 341-356.
[7]刘清Rough逻辑及其在数据约简中的应用[J].软件学报,2001：415-419.
[8]Hand D, Mannila H, Smyth P.数据挖掘：概念与技术(范明.译)[M].北京：机械工业出版社,2003：193-205.
[9]Pawlak Z. Rough Sets.Theoretical Aspects of Reasoning about Data. Kluwer Academic Publishers, Dordrecht,1991.
[10]杨明.一种机遇改进差别矩阵的属性约简增量式更新算法[J].计算机学报,2007.
[11]杨波.机遇差别矩阵的完备属性约简算法[J].计算机工程,2011.
[12]张娟.基于差别矩阵的高效属性约简算法[J].武汉理工大学学报,2010.12.
[13]韩智东.用差别矩阵思想设计的基于正区域的高效属性约简算法[J].小型微型计算机系统,2011.
[14]Hu Xiaohua, Cercone N. Learning in Relational Database:a rough set approach[J]. Computational Inelligence,1995,11(2):323-337.
[15]胡可云.基于概念格和粗糙集的数据挖掘方法研究[D].博士学位论文.清华大学,北京.2001.
[16]王国胤.基于条件信息熵的决策表约简[J].计算机学报,2002,25(7)：759-766.
[17]刘洋.一种改进的基于差别矩阵的属性约简算法[J].微电子学与计算机,2007.
[18]杨常清.基于粗糙集的属性约简算法研究[D].硕士学位论文.西安电子科技大学,西安.2010.
[19]Zhang W X. Information Systems and Knowledge Discovery[M]. Beijing:Science Press, 2003:22-47.
[20]Leung Y. Knowledge Acquisition in Incomplete Information Systems-A Rough Set Approach [J].European J of Operational Research,2004:164-180.
[21]杨成福.粗糙集属性约简算法研究[D].硕士学位论文.电子科技大学,成都.2006.
[22]王炜.不完备信息系统属性约简的矩阵算法[J].计算机工程,2011.
[23]孙涛.用于粗糙集约简的并行算法[J].吉林大学学报(理学版),2006,44：211-216.
[24]覃政仁.一种基于Rough Set的海量数据分割算法[J].模式识别与人工智能,2006：249-256.
[25]李订芳.基于U/{a}划分的最小约简构造[J].武汉大学学报(理学版),2005：269-272.
[26]Larry M. Deuser. A Hybrid Multispectral Feature Selection Criterion[J]. IEEE Transactions On Computers,1971:1116-1117.
[27]Huan Liu. Feature Transformation and Subset Selection[J]. IEEE Intelligent Systems, 1998(3):26-27.
[28]Marco Bressan and Jordi Vrtria. On the Selcetion and Classification of Independent Features [J]. IEEE Transactions On Pattern Analysis and Machine Intelligence,2003(10): 1312-1317.
[29]K.Manumala and K.Selvi. Hybrid Soft Computing Techniques for Feature Selection and Parameter Optimization in Power Quality Data Mining[J]. Applied Soft Computing, 2011(11):5485-5497.
[30]L.A.Zadeh.Adefinition of Soft Computing in Greater Detail-WWW page,1991/1994.Available from
[31]郭嗣琮,陈刚.信息科学中的软计算方法[M].沈阳：东北大学出版社,2001.
[32]张颖,刘艳秋.软计算方法[M].北京：科学出版社,2002.
[33]刘晋寅,李洪兴.软计算及其哲学内涵[J].自然辩证法研究,16(5),2000,26-29.
[34]杨善林k-means算法中的k值优化问题研究[J].系统工程理论与实践,2006,2：97-101.
[35]邓祖新.sas系统和数据分析[M].北京：电子工业出版社.2002.
[36]余建英,何旭宏.数据统计分析与SPSS应用[M].北京：人民邮电出版社,2003.
[37]L.A. Zadeh.Fuzzy Sets[M].Information and Contro18,1965:338-353.
[38]L.A.Zadeh. Outline of aNew Approach to the Analysis of Complex Systems and Decision Processes[J].IEEE Transactions on Systems, Man,and Cybernetics3,1973:28-44.
[39]S.J.Ovaska and H.F.VanLandingham,and A.Kamiya.Fusion of Soft Computing and Hard Computing in Industrial Application:An Overview[J].IEEE Transactions on Systems,Man,and Cybernetics-Part C:Applications and Reviews32,2002,72-79.
[40]S.J.Ovaska and H.F.VanLandingham.Guest Editorial:Special Issue on Fusion of Soft Computing and Hard Computing in Industrial Application[J].IEEE Transactions on Systems,Man,and Cybernetics-Part C:Applications and Reviews32,2002,69-71.
[41]R.Sterritt and D.W.Bustard.Fusing Hard and Soft Computing for Fault Management in Telecommunications Systems [J]. IEEE Transactions on Systems, Man, and Cybernetics-Part C:Applications and Reviews32,2002,92-98.
[42]Terrence P.Fries and James H.Graham. Fusion of Soft and Hard Computing for Fault Diagnosis in Manufacturing Systems [J]. IEEE Transactions on Systems, Man, and Cybernetics-Part C:Applications and Reviews 32,2003,80-91.
[43]T.Furuhashi.Fusion of Fuzzy/Neuron/Evolutionary Computing for Knowledge Acquisition [J]. Proceedings of the IEEE89,2001,1266-1274.
[44]B.Sick.Fusion of Hard and Soft Computing Techniques in Indirect, Online Tool Wear Monitoring [J].IEEE Transactions on Systems, Man, and Cybernetics-Part C:Applications and Reviews 32,2002,80-91.
[45]诸克军.模糊C-均值中的最优聚类与最佳聚类数[J].系统工程理论与实践,2005,3：52-61.
[46]李传亮.油藏工程原理[M].北京：石油工业出版,2005.11.
[47]刁凤琴.石油储量动态经济评价研究[D].武汉：中国地质大学(武汉),2007.12.
[48]李斌,袁俊香.影响产量递减率的因素与减缓递减的途径[J].石油学报.第18卷第3期,1997.07.
[49]王任飞.大庆升平油田主要开发指标变化规律及递减影响因素研究[D].北京：中国地质大学(北京),2007.06.
[50]管纪昂,武若霞,韩东.用生产资料确定谢卡尔乔夫公式参数的方法[J].西安石油学院学报(自然科学版),2003,18(2),25-29.
[51]俞启泰.计算水驱砂岩油藏合理井网密度与极限井网密度的一种方法[J].石油勘探与开发,1986,13(4)：49-54.
[52]刘世良,郑应钊.确定老油田合理井网密度和极限井网密度的新方法[J].新疆石油地质,2004,25(3)：310-311.
[53]朱圣举,刘宝良.确定低渗透油藏合理井网密度和极限井网密度的新方法[J].特种油气藏,1999,6(1)：14-17.
[54]张树林,黄耀琴.净现值法——一种计算经济极限井网密度的新方法[J].地质科技情报,2004,23(1)：78-80.
[55]刘显德,高泓.基于SVR的组合预测模型及其应用.计算机工程与设计2009,30(19),4506-4509.
[56]胡高贤,龚福华.多元回归分析在低渗透油藏产能预测中的应用.油气田地面工程,2010,29(12),23-25.
[57]国亮.改进多元线性回归模型在某油田产量预测中的应用.西安电子科技大学学报(社会科学版),2009,19(3),71-75.
[58]董玉才,李红燕,朱连军,鞠桂玲,等.基于偏最小二乘回归的油田产量预测.信息系统工程,2009,(10)：27-30.
[59]钟雨师,叶志瑾,陈武,陈光海.油田原油产量改进的二次指数最优预测方法研究.钻采工艺,2008,31(4)：71-73.
[60]江厚顺,张伟.高5块低渗油气田产量预测方法分析.天然气地球科学,2010,21(3),385-388.
[61]朱小梅,杨先凤,张群燕LS-SVM-GA算法在油田产量预测中的应用研究.煤炭技术,2010,29(11),197-199.
[62]钟仪华,张志银,朱海双.特高含水期油田产量预测新方法.断块油气田,2011,18(5)：661-664.
[63]樊纪香,张宏,李辉,王兵团.BP网络和多元线性回归在产量预测中的应用,计算机工程与应用,2007,43(23),203-205.
[64]赵国忠,孟曙光.聚合物驱含水率的神经网络预测方法[J].石油学报,2004,25.
[65]许敏,赵粉霞.神经网络法在气井产能预测中的应用口].天然气工业,2000,20(3)：73-75.
[66]丁士圻,郭丽华.人工神经网络基础[M].哈尔滨工程大学出版社,2008年3月.
[67]韩荣,祁大足,吴赞美,闰更平.BP神经网络在预测石南31油田产量变化中的应用,内蒙古石油化工,2010,10,170-173.
[68]倪红梅,王维刚,罗仕明.人工神经网络在采油产量预测中的应用.陕西理工学院学报(自然科学版),2008,24(3),36-39.
[69]周晶晶,贺勇,诸克军.基于主成分BP神经网络的油田产量预测研究.统计与决策,2008,(13)：35-37.
[70]舒丹,曹谢东,汤力,范勇,等.人工神经网络在油田产量预测中的应用.智能系统学报,2008,3,149-152.
[71]翟亮亮,王连堂,王俊杰.基于Levenberg-Marquardt算法的油田产量预测模型.石油地质与工程,2010,24(3),53-56.
[72]李春生,谭民浠,张可佳.基于改进型BP神经网络的油井产量预测研究.科学技术与工程,2011,11(31),7766-7769.
[73]董玉安,赵蕊,叶欢,廖晶.遗传算法与神经网络结合预测低渗透油藏产油量——以宝浪油田为例.石油地质与工程,2011,25(5)：63-66.
[74]曲占庆.采油工程基础知识手册.北京：石油工业出版社,2002.08
[75]罗洪群,陈武.低渗透气田经济极限储量界限研究[J].西南石油大学学报(自然科学版),2011(3)：169-170.
[76]陆冬梅.论资金时间价值在现代企业理财中的运用[J].现代经济信息,2010：154.
[77]吴亚丽.海外油气项目经济极限值测算[J].中外能源,2011(6)：35-36.
[78]张登庆.敏感性分析在油田集输系统经济评价中的应用[J].油气田地面工程,2004(12)：6-7.
[79]张明泉.基于经济极限产量的油田生产优化决策研究[J].西南石油大学学报(社会科学版),2011(13)：80-82.
[80]国家计委.建设项目经济评价方法与参数(第二版)[M].北京：中国计划出版社,1993：13-18.
[81]袁庆峰.认识油田开发规律科学合理开发油田[J].大庆石油地质与开发,2004,23(5)：65-66.
[82]秦同洛.实用油藏工程方法[M].北京：石油工业出版社,1989.
[83]黄炳光,刘蜀知.实用油藏工程与动态分析方法[M].北京：石油工业出版社,1998.
[84]Huang Hongdu.Evaluation of surfactants from oxidized hydrocarbons in enhanced oil recovery[J].J Petrol Sci.Eng,1991,6 (3):193-200.
[85]Jayanti S, Britto L N, Dwarakanath V.Laboratory evaluation of custom-designed surfactants to remediate NAPL source zones[J]. Environ Sci.Tcchnol,2002,36:549-5497.
[86]Liu Q, Dong M, Ma S. Alkaline/surfactant flood potential in Western Candian heavy oil reservoirs[z]. SPE 99791,2006.
[87]张德芳,孙继伟,唐磊等.注水开发油田特高含水期剩余油潜力评价[J].河南石油：2001-11,15(6)：22-24.
[88]洪世铎.油藏物理基础[M].北京：石油工业出版社,1985.
[89]张凤莲,崔玉婷,郭振,等.确定低渗透油藏合理井网密度和极限进网密度的方法[J]大庆石油地质与开发,2008,27(2)：88-90.
[90]钱深华,李永伏,袁成章,油田加密调整经济合理井网密度的确定[J].大庆石油地质与开发,2000,19(4)：23-25.
[91]管纪昂,武若霞,韩东.用生产资料确定谢卡尔乔夫公式参数的方法[J].西安石油学院学报,2003-03,18(2)：25-27.
[92]任忠和,陈冰,陈舒薇,等.油田开发下限及经济极限的确定[J].大庆石油地质与开发,2005,24(1)：61-63.
[93]陈维华.高含水油田开发效果评价体系研究：[硕士学位论文].华东：中国石油大学,2006-05-01.
[94]刘义坤,毕永斌,隋新光.高含水后期油田开发指标预测[J].大庆石油地质与开发,2008,17(1)：61-62.
[95]谢丛姣,陈生泽.低渗透油田开发效果评价指标—以张天渠油田为例[J].新疆石油学院学报,2004-12,16(4)：29-33.
[96]姜瑞忠,刘小波,王海江等.指标综合筛选方法在高含水油田开发效果评价中的应用-以曾埕东油田威力为例[J].油气地质与采收率,2008-03,15(2)：99-101.
[97]范保国.水驱油田特高含水期开发效果评价体系[J].大庆石油地质与开发,2009-04,28(2)：53-58.
[98]中国石油天然气总公司开发生产局.SY/T 6219-1996.中华人民共和国石油天然气行业标准[M].北京：石油工业出版社,1996-12-25.
[99]时光新,王其昌,刘建强.变异系数法在小流域治理效益评价中的应用[J].水土保持通报,2000-12,20(6)：38-39.
[100]Jia,Z.Design and Analysis of Oilfield Development[M].Press of Harbin Industrial University.1994.
[101]Yang Erlong, Jing nuan, Liu Meijia. An Optimization Model for Stimulation of Oilfield at the Stage of High Water Content.Press of Harbin Industrial University.2004.
[102]张新征,张烈辉,熊钰,等.高含水后期油田开发效果评价方法及应用研究[J].大庆石油地质与开发,2005,24(3)：48-50.
[103]刘秀婷,程仲平,杨纯东,等.油田开发效果综合评价方法新探[J].中外能源,2006,11(5)：37-41.
[104]苏永华,何满潮,孙晓明.岩体模糊分类中隶属函数的等效性[J].北京科技大学学报,2007-07,29(7)：670-675.
[2]葛家理.石油企业经营发展战略研究[M].北京：石油工业出版社,1998.
[3]赵培华.中国路上水驱油田开发测井技术[J].测井技术,1998.05.
[4]曾雪梅,范景明.北二东开采层系重组试验区水驱开发效果试验方案[R].大庆油田有限责任公司第三采油厂,2009.06.
[5]丁卫平.基于量子蛙跳协同进化的粗糙属性快速约简[J].电子学报,2011：2597-2598.
[6]Pawlak. Z. Rough sets. International Journal of Computer and Information Sciences[J].1982, 341-356.
[7]刘清Rough逻辑及其在数据约简中的应用[J].软件学报,2001：415-419.
[8]Hand D, Mannila H, Smyth P.数据挖掘：概念与技术(范明.译)[M].北京：机械工业出版社,2003：193-205.
[9]Pawlak Z. Rough Sets.Theoretical Aspects of Reasoning about Data. Kluwer Academic Publishers, Dordrecht,1991.
[10]杨明.一种机遇改进差别矩阵的属性约简增量式更新算法[J].计算机学报,2007.
[11]杨波.机遇差别矩阵的完备属性约简算法[J].计算机工程,2011.
[12]张娟.基于差别矩阵的高效属性约简算法[J].武汉理工大学学报,2010.12.
[13]韩智东.用差别矩阵思想设计的基于正区域的高效属性约简算法[J].小型微型计算机系统,2011.
[14]Hu Xiaohua, Cercone N. Learning in Relational Database:a rough set approach[J]. Computational Inelligence,1995,11(2):323-337.
[15]胡可云.基于概念格和粗糙集的数据挖掘方法研究[D].博士学位论文.清华大学,北京.2001.
[16]王国胤.基于条件信息熵的决策表约简[J].计算机学报,2002,25(7)：759-766.
[17]刘洋.一种改进的基于差别矩阵的属性约简算法[J].微电子学与计算机,2007.
[18]杨常清.基于粗糙集的属性约简算法研究[D].硕士学位论文.西安电子科技大学,西安.2010.
[19]Zhang W X. Information Systems and Knowledge Discovery[M]. Beijing:Science Press, 2003:22-47.
[20]Leung Y. Knowledge Acquisition in Incomplete Information Systems-A Rough Set Approach [J].European J of Operational Research,2004:164-180.
[21]杨成福.粗糙集属性约简算法研究[D].硕士学位论文.电子科技大学,成都.2006.
[22]王炜.不完备信息系统属性约简的矩阵算法[J].计算机工程,2011.
[23]孙涛.用于粗糙集约简的并行算法[J].吉林大学学报(理学版),2006,44：211-216.
[24]覃政仁.一种基于Rough Set的海量数据分割算法[J].模式识别与人工智能,2006：249-256.
[25]李订芳.基于U/{a}划分的最小约简构造[J].武汉大学学报(理学版),2005：269-272.
[26]Larry M. Deuser. A Hybrid Multispectral Feature Selection Criterion[J]. IEEE Transactions On Computers,1971:1116-1117.
[27]Huan Liu. Feature Transformation and Subset Selection[J]. IEEE Intelligent Systems, 1998(3):26-27.
[28]Marco Bressan and Jordi Vrtria. On the Selcetion and Classification of Independent Features [J]. IEEE Transactions On Pattern Analysis and Machine Intelligence,2003(10): 1312-1317.
[29]K.Manumala and K.Selvi. Hybrid Soft Computing Techniques for Feature Selection and Parameter Optimization in Power Quality Data Mining[J]. Applied Soft Computing, 2011(11):5485-5497.
[30]L.A.Zadeh.Adefinition of Soft Computing in Greater Detail-WWW page,1991/1994.Available from
[31]郭嗣琮,陈刚.信息科学中的软计算方法[M].沈阳：东北大学出版社,2001.
[32]张颖,刘艳秋.软计算方法[M].北京：科学出版社,2002.
[33]刘晋寅,李洪兴.软计算及其哲学内涵[J].自然辩证法研究,16(5),2000,26-29.
[34]杨善林k-means算法中的k值优化问题研究[J].系统工程理论与实践,2006,2：97-101.
[35]邓祖新.sas系统和数据分析[M].北京：电子工业出版社.2002.
[36]余建英,何旭宏.数据统计分析与SPSS应用[M].北京：人民邮电出版社,2003.
[37]L.A. Zadeh.Fuzzy Sets[M].Information and Contro18,1965:338-353.
[38]L.A.Zadeh. Outline of aNew Approach to the Analysis of Complex Systems and Decision Processes[J].IEEE Transactions on Systems, Man,and Cybernetics3,1973:28-44.
[39]S.J.Ovaska and H.F.VanLandingham,and A.Kamiya.Fusion of Soft Computing and Hard Computing in Industrial Application:An Overview[J].IEEE Transactions on Systems,Man,and Cybernetics-Part C:Applications and Reviews32,2002,72-79.
[40]S.J.Ovaska and H.F.VanLandingham.Guest Editorial:Special Issue on Fusion of Soft Computing and Hard Computing in Industrial Application[J].IEEE Transactions on Systems,Man,and Cybernetics-Part C:Applications and Reviews32,2002,69-71.
[41]R.Sterritt and D.W.Bustard.Fusing Hard and Soft Computing for Fault Management in Telecommunications Systems [J]. IEEE Transactions on Systems, Man, and Cybernetics-Part C:Applications and Reviews32,2002,92-98.
[42]Terrence P.Fries and James H.Graham. Fusion of Soft and Hard Computing for Fault Diagnosis in Manufacturing Systems [J]. IEEE Transactions on Systems, Man, and Cybernetics-Part C:Applications and Reviews 32,2003,80-91.
[43]T.Furuhashi.Fusion of Fuzzy/Neuron/Evolutionary Computing for Knowledge Acquisition [J]. Proceedings of the IEEE89,2001,1266-1274.
[44]B.Sick.Fusion of Hard and Soft Computing Techniques in Indirect, Online Tool Wear Monitoring [J].IEEE Transactions on Systems, Man, and Cybernetics-Part C:Applications and Reviews 32,2002,80-91.
[45]诸克军.模糊C-均值中的最优聚类与最佳聚类数[J].系统工程理论与实践,2005,3：52-61.
[46]李传亮.油藏工程原理[M].北京：石油工业出版,2005.11.
[47]刁凤琴.石油储量动态经济评价研究[D].武汉：中国地质大学(武汉),2007.12.
[48]李斌,袁俊香.影响产量递减率的因素与减缓递减的途径[J].石油学报.第18卷第3期,1997.07.
[49]王任飞.大庆升平油田主要开发指标变化规律及递减影响因素研究[D].北京：中国地质大学(北京),2007.06.
[50]管纪昂,武若霞,韩东.用生产资料确定谢卡尔乔夫公式参数的方法[J].西安石油学院学报(自然科学版),2003,18(2),25-29.
[51]俞启泰.计算水驱砂岩油藏合理井网密度与极限井网密度的一种方法[J].石油勘探与开发,1986,13(4)：49-54.
[52]刘世良,郑应钊.确定老油田合理井网密度和极限井网密度的新方法[J].新疆石油地质,2004,25(3)：310-311.
[53]朱圣举,刘宝良.确定低渗透油藏合理井网密度和极限井网密度的新方法[J].特种油气藏,1999,6(1)：14-17.
[54]张树林,黄耀琴.净现值法——一种计算经济极限井网密度的新方法[J].地质科技情报,2004,23(1)：78-80.
[55]刘显德,高泓.基于SVR的组合预测模型及其应用.计算机工程与设计2009,30(19),4506-4509.
[56]胡高贤,龚福华.多元回归分析在低渗透油藏产能预测中的应用.油气田地面工程,2010,29(12),23-25.
[57]国亮.改进多元线性回归模型在某油田产量预测中的应用.西安电子科技大学学报(社会科学版),2009,19(3),71-75.
[58]董玉才,李红燕,朱连军,鞠桂玲,等.基于偏最小二乘回归的油田产量预测.信息系统工程,2009,(10)：27-30.
[59]钟雨师,叶志瑾,陈武,陈光海.油田原油产量改进的二次指数最优预测方法研究.钻采工艺,2008,31(4)：71-73.
[60]江厚顺,张伟.高5块低渗油气田产量预测方法分析.天然气地球科学,2010,21(3),385-388.
[61]朱小梅,杨先凤,张群燕LS-SVM-GA算法在油田产量预测中的应用研究.煤炭技术,2010,29(11),197-199.
[62]钟仪华,张志银,朱海双.特高含水期油田产量预测新方法.断块油气田,2011,18(5)：661-664.
[63]樊纪香,张宏,李辉,王兵团.BP网络和多元线性回归在产量预测中的应用,计算机工程与应用,2007,43(23),203-205.
[64]赵国忠,孟曙光.聚合物驱含水率的神经网络预测方法[J].石油学报,2004,25.
[65]许敏,赵粉霞.神经网络法在气井产能预测中的应用口].天然气工业,2000,20(3)：73-75.
[66]丁士圻,郭丽华.人工神经网络基础[M].哈尔滨工程大学出版社,2008年3月.
[67]韩荣,祁大足,吴赞美,闰更平.BP神经网络在预测石南31油田产量变化中的应用,内蒙古石油化工,2010,10,170-173.
[68]倪红梅,王维刚,罗仕明.人工神经网络在采油产量预测中的应用.陕西理工学院学报(自然科学版),2008,24(3),36-39.
[69]周晶晶,贺勇,诸克军.基于主成分BP神经网络的油田产量预测研究.统计与决策,2008,(13)：35-37.
[70]舒丹,曹谢东,汤力,范勇,等.人工神经网络在油田产量预测中的应用.智能系统学报,2008,3,149-152.
[71]翟亮亮,王连堂,王俊杰.基于Levenberg-Marquardt算法的油田产量预测模型.石油地质与工程,2010,24(3),53-56.
[72]李春生,谭民浠,张可佳.基于改进型BP神经网络的油井产量预测研究.科学技术与工程,2011,11(31),7766-7769.
[73]董玉安,赵蕊,叶欢,廖晶.遗传算法与神经网络结合预测低渗透油藏产油量——以宝浪油田为例.石油地质与工程,2011,25(5)：63-66.
[74]曲占庆.采油工程基础知识手册.北京：石油工业出版社,2002.08
[75]罗洪群,陈武.低渗透气田经济极限储量界限研究[J].西南石油大学学报(自然科学版),2011(3)：169-170.
[76]陆冬梅.论资金时间价值在现代企业理财中的运用[J].现代经济信息,2010：154.
[77]吴亚丽.海外油气项目经济极限值测算[J].中外能源,2011(6)：35-36.
[78]张登庆.敏感性分析在油田集输系统经济评价中的应用[J].油气田地面工程,2004(12)：6-7.
[79]张明泉.基于经济极限产量的油田生产优化决策研究[J].西南石油大学学报(社会科学版),2011(13)：80-82.
[80]国家计委.建设项目经济评价方法与参数(第二版)[M].北京：中国计划出版社,1993：13-18.
[81]袁庆峰.认识油田开发规律科学合理开发油田[J].大庆石油地质与开发,2004,23(5)：65-66.
[82]秦同洛.实用油藏工程方法[M].北京：石油工业出版社,1989.
[83]黄炳光,刘蜀知.实用油藏工程与动态分析方法[M].北京：石油工业出版社,1998.
[84]Huang Hongdu.Evaluation of surfactants from oxidized hydrocarbons in enhanced oil recovery[J].J Petrol Sci.Eng,1991,6 (3):193-200.
[85]Jayanti S, Britto L N, Dwarakanath V.Laboratory evaluation of custom-designed surfactants to remediate NAPL source zones[J]. Environ Sci.Tcchnol,2002,36:549-5497.
[86]Liu Q, Dong M, Ma S. Alkaline/surfactant flood potential in Western Candian heavy oil reservoirs[z]. SPE 99791,2006.
[87]张德芳,孙继伟,唐磊等.注水开发油田特高含水期剩余油潜力评价[J].河南石油：2001-11,15(6)：22-24.
[88]洪世铎.油藏物理基础[M].北京：石油工业出版社,1985.
[89]张凤莲,崔玉婷,郭振,等.确定低渗透油藏合理井网密度和极限进网密度的方法[J]大庆石油地质与开发,2008,27(2)：88-90.
[90]钱深华,李永伏,袁成章,油田加密调整经济合理井网密度的确定[J].大庆石油地质与开发,2000,19(4)：23-25.
[91]管纪昂,武若霞,韩东.用生产资料确定谢卡尔乔夫公式参数的方法[J].西安石油学院学报,2003-03,18(2)：25-27.
[92]任忠和,陈冰,陈舒薇,等.油田开发下限及经济极限的确定[J].大庆石油地质与开发,2005,24(1)：61-63.
[93]陈维华.高含水油田开发效果评价体系研究：[硕士学位论文].华东：中国石油大学,2006-05-01.
[94]刘义坤,毕永斌,隋新光.高含水后期油田开发指标预测[J].大庆石油地质与开发,2008,17(1)：61-62.
[95]谢丛姣,陈生泽.低渗透油田开发效果评价指标—以张天渠油田为例[J].新疆石油学院学报,2004-12,16(4)：29-33.
[96]姜瑞忠,刘小波,王海江等.指标综合筛选方法在高含水油田开发效果评价中的应用-以曾埕东油田威力为例[J].油气地质与采收率,2008-03,15(2)：99-101.
[97]范保国.水驱油田特高含水期开发效果评价体系[J].大庆石油地质与开发,2009-04,28(2)：53-58.
[98]中国石油天然气总公司开发生产局.SY/T 6219-1996.中华人民共和国石油天然气行业标准[M].北京：石油工业出版社,1996-12-25.
[99]时光新,王其昌,刘建强.变异系数法在小流域治理效益评价中的应用[J].水土保持通报,2000-12,20(6)：38-39.
[100]Jia,Z.Design and Analysis of Oilfield Development[M].Press of Harbin Industrial University.1994.
[101]Yang Erlong, Jing nuan, Liu Meijia. An Optimization Model for Stimulation of Oilfield at the Stage of High Water Content.Press of Harbin Industrial University.2004.
[102]张新征,张烈辉,熊钰,等.高含水后期油田开发效果评价方法及应用研究[J].大庆石油地质与开发,2005,24(3)：48-50.
[103]刘秀婷,程仲平,杨纯东,等.油田开发效果综合评价方法新探[J].中外能源,2006,11(5)：37-41.
[104]苏永华,何满潮,孙晓明.岩体模糊分类中隶属函数的等效性[J].北京科技大学学报,2007-07,29(7)：670-675.