【作者】 晁会霞；

【导师】 杨兴科；

【作者基本信息】 长安大学 ， 矿产普查与勘探， 2009， 博士

【摘要】 本文以鄂尔多斯盆地白豹地区为例,在数据挖掘的广义观点的指导下,基于油田多源数据分析与挖掘,将储层特征的数据分析与挖掘过程划分为确定储层数据分析与挖掘目标、建立储层项目数据库、数据预处理、储层分析建模、模型评估5个阶段。论文在大量参阅相关文献和研究成果的同时,综合收集该区原始岩矿数据和物性资料,针对研究区已有的海量数据资料分散,数据量大,闲置严重而信息贫乏的现状,确定储层数据分析与挖掘任务,提取研究区相关储层数据,构建研究区储层项目数据库,实现海量和多源(元)数据的融合与集成,建立了数据的共享,这对于提高储层定量化研究和促进油田的信息化发展有重要的理论与现实意义。以可视化数据挖掘技术为研究手段,对储层数据进行预处理、数据质量审核、数据统计分析与建模,对研究区进行系统全面的储层研究,并实现储层数据分析与挖掘过程和结果的可视化,最终通过这些参数的统计特征,空间分布规律来描述和预测储层,深化研究储层地质相关问题,指导油气田进一步的勘探开发。鄂尔多斯盆地上三叠统延长组第6段在白豹地区是最重要的勘探目的层之一,本文以长6储层为研究对象,从建立的研究区储层项目数据库出发,运用沉积学、石油地质学、地球物理学等知识,根据储层数据分析与挖掘的过程,进行物源分析与聚类分区、岩石类型识别与分布、孔渗建模、储层空间(二维、三维)展布研究及储层综合评价等方面的研究。主要研究内容和进展归纳如下:1.在对研究区物源定性分析的基础上,提取长6层位的655块薄片样品,将碎屑组分及填隙物成分等考虑在内,进行空间聚类分析,定量划分为3个物源区。从研究区长6储层空间聚类结果平面展布规律来看,以东北物源为主,西南物源为次,在元城—白马—五蛟一带受到两个物源方向的叠加影响,为混源区。物源区的划分为定量研究各个物源区的岩石学特征及储集性能等奠定了基础;2.岩石类型研究中,建立砂岩分类自动识别模型,使得对大数据量砂岩骨架组分的定量研究变得简单易行。岩石类型分析结果为:长6油层组以岩屑长石砂岩为主,次为长石砂岩,少量的长石岩屑砂岩。分别对不同物源分区的岩石类型进行进一步的分析,东北物源区以岩屑长石砂岩为主,其次为长石砂岩。混源区以岩屑长石砂岩为主,次为长石岩屑砂岩。西南物源区以长石岩屑砂岩为主,次为岩屑长石砂岩;3.基于岩芯分析资料和多种测井信息,应用神经网络技术对研究区长6储层参数进行预测,以期实现储层参数的精确解释。用神经网络的方法预测孔隙度和渗透率,得到的孔隙度、渗透率变化趋势与取芯井中孔、渗变化规律相符,预测精度比传统的回归方法有所提高。利用该方法建立测井响应与岩芯分析数据之间的对应关系,弥补了取芯井数量的不足,有助于我们更准确翔实地进行孔隙度和渗透率相关性与非均质性分析以及定量参数在空间上的分布规律等研究;4.在对储层储集性能的分析中,对孔隙度和渗透率相关性进行了分析,并分小层对长6油层组的非均质性进行研究。长6储层各个小层的孔隙度非均质性远小于渗透率非均质性,孔隙度非均质性不太明显,渗透率非均质性比较强,其中长6~3小层的非均质性(突进系数、级差和变异系数)明显高于长6~2和长6~1,长6~1的非均质性最弱。利用孔隙度和渗透率两个重要的物性参数,通过聚类分析对储层进行初级分类。为了研究岩石学特征与储层物性间的关系,将薄片样品与物性样品配对,通过相关图分析影响储集性能的多种因素,粒间孔、面孔率及绿泥石含量与孔隙度、渗透率成正比,而岩石密度、碳酸盐胶结物、粘土类胶结物含量与孔隙度、渗透率成反比;5.利用研究区井位信息和分层数据,通过数据处理流程,采用Rockworks建模软件进行三维地质建模。三维地层模型展示了地下真实面貌,使我们直观了解地下储层的几何形态与空间展布特征,地层模型所反映的信息与区域上的沉积—构造演化特征相一致。在建立地层模型的基础上,对研究区长6油层组分小层长6~3、长6~2、长6~1,利用Suffer软件建立储层三维构造模型及储层参数模型。三维构造模型表明,在区域西倾单斜背景上,发育有三排东西向展布的低幅度的鼻状隆起构造,隆起部位的部分井区属于工业油流井,说明鼻状隆起构造在一定程度上控制了油气的富集以及油藏的油水分异;6.本文最后进行储层综合定量评价,首先利用特征选择算法对评价参数进行筛选,然后根据灰关联分析来确定各影响因素的权重,进而运用最大值标准化法确定各项参数的评价分数,最后计算各项参数综合得分,在此基础上,运用聚类分析进行储层分类评价。对储层评价结果进行统计分析,所划分的各类储层特征明显,与研究区储层实际特征具有很好的一致性。研究区长6储层大部分为Ⅱ类和Ⅲ类,少数为Ⅳ类和Ⅰ类,长6~3小层的储层质量明显好于长6~1和长6~2小层。Ⅰ类储层和Ⅱ类储层作为研究区的最好和较好储层,是寻找油气的有利区带。更多还原

【Abstract】 Under the guidance of the extensive viewpoint of data mining and based on multivariatedata analysis and minning of oilfield,the paper takes Baibao oil field district of Ordos basinfor example,dividing the data mining course of reservoir characteristics into five stages:determination of reservoir data mining target,establishment of project database,datapreprocessing,modeling and evaluation.By referring to abundant references and correlateiveresearch results,comprehensively collecting rock and mineral data and physical property,according to present situation of data dispersion,massive data,left unused seriously butdeficient in knowledge,the task of reservoir data mining target is determined.We extractrelated reservoir data of Baibao oil field district in order to construct reservoir projectdatabase of study area,and then massive,multivariate (multiple) data fusion and integrationcan be achieved and the mechanism of data sharing is established.All these have importanttheoretical and practical significance for promoting quantitative study of reservoir andinformatization development of oil field.By adopting research method of data mining visualization techniques,reservoir datapreprocessing,quality audit,data statistic analysis and modeling are taken to comprehensivelyresearch the reservoirs in the study area,and visual process and result of data mining arerealized,finally through statistical characteristics and spatial distributions laws of geologicalparameters,the reservoirs are described and predicted,related geology problems of reservoirare further studied to guide the further exploration and development of oil and gas field.Member 6 of Yanchang Formation of upper-triassic in Ordos basin is one of the mostimportant exploration target stratum in Baibao district.By taking Chang 6 as the researchobject,according to reservoir database of Baibao district,by using the theory ofsedimentology,petroleum geology and geophysics,based on the course of reservoir datamining,the paper takes provenance analysis and clustering partition,rock type discriminationand distribution law,modeling of porosity and permeability,reservoir spatial distributioncharacteristics research and comprehensive evaluation of reservoir.The main researchcontents and research progress are proposed as follows:1.On the basis of qualitative provenance analysis of the study area,taking detritalcomposition and pore filling mineral composition into consideration,the paper takes spatialcluster analysis by extracting 655 samples of rock thin section database of Chang 6.The studyarea is divided into three provenance partitions.From planar distribution law of spatial clusterresult of Chang 6,the reservoir is mainly sourced from northeast,secondly from southwest,and in the belt of Yuan cheng-Baima-Wujiao,it is overlapping influenced by the two sourcedirection and formed mixing source region.The division of provenance partition lay afoundation for quantitative study of petrological characteristics and reservoir property of eachprovenance partition. 2.In the research of rock type,by establishing automatic recognition model,the papermakes it simple to quantitative study the large data of composition.The analysis results ofrock type show that in Chang 6 Formation in the study area,the most of reservoir rock isdebris feldspar sandstone and secondly is feldsparthic sandstone and small part of feldspathiclithic sandstone.By further analysis of rock type of different provenance partition,it showsthat northeast provenance is mainly debris feldspar sandstone,and minor is feldsparthicsandstone,the mixed source area is dominated by debris feldspar sandstone,secondary isfeldspathic lithic sandstone,and southwest provenance is mainly feldspathic lithic sandstone,and minor debris feldspar sandstone.3.Based on the core analysis data and multi-logging information,by using neuralnetwork technology,the paper predicts the reservoir parameter of Chang 6 in order to provideprecise interpretation of reservoir parameter.The changing trend of porosity and permeabilitypredicted by neural network equates with that of coring well data.In contrast with traditionalregression analysis,the prediction accuracy of neural network is somewhat increased,whichmeet the need of reservoir heterogeneity research.By using the method,the paper establishesthe relationship between well logging response and core analysis data,and then makes up forthe deficiency of number of coring well.Accordingly,we can take analysis of the correlationand heterogeneity of porosity and permeability,and spatial distribution law of quantitativeparametersmore accurate and more detailed.4.In the study of reservoir property,the paper analyzes the correlation between porosityand permeability and studies the reservoir heterogeneity of each sublayer of Chang 6.It isconsidered that porosity heterogeneity of Chang 6 is far less than permeability,and the formeris relatively homogeneous,and the heterogeneity of the latter is relatively strong.It shows thatthe permeability heterogeneity (level difference,mutation coefficient and coefficient ofvariation) of Chang 6~3 is larger than that of Chang 6~2 and Chang 6~1,while the heterogeneity ofChang6~1 is weakest.By using two important physical parameters-porosity and permeability,the paper then takes primary classification of Chang 6 reservoir by cluster analysis.In order tostudy the relation between petrologic characteristics and reservoir property,the paper matchesthe samples of rock thin section and reservoir property and then analyzes the influencingfactors through relation maps.Intergranular pore,areal porosity and the content of chlorite isrespectively proportional to porosity and permeability,and rock density,contents of carbonatecements and clay cement is respectively inversely proportional to porosity and permeability.5.By extracting well position information and layered data,through data processing flow,and by using modeling software of Rockworks,the paper takes the research ofthree-dimensional geological modeling.The three dimension strata model shows the realvisage underground,which can visually show the geometric shape of underground reservoirsand space distributing characteristics.The information reflected by strata model is consistentwith the characteristics of regional sediment-structural evolution.Based on the establishmentof strata model,the paper also sets up three-dimensional structure model and reservoirparameter model.The 3D structural model of Chang 6 shows that there exist three rows of low relief nose-like structure,extended from west to east on the background of monoclinalstructure.Some well in the nose fold have commercial oil flows.That indicates that to someextent,the nose-like structure controls hydrocarbon enrichment and oil and waterdifferentiation of oil reservoir.6.At last,the paper completely evaluates the reservoir of Chang 6.Firstly,selectevaluation parameters by using feature selection algorithm.Secondly,determine the weight ofeach influencing factors by grey relational analysis.Then determine the evaluation score ofeach parameter by the method of maximal value normalization.Finally,calculate the generalscore.Based on the procedure,reservoir classification evaluation was taken by cluster model.The statistics analysis based on the reservoir evaluation results shows that the reservoircharacteristics of each type is obvious and has a high coherence with the actual characteristicsof the reservoir.Statistics analysis indicates that Chang 6 reservoir mostly is classⅡandclassⅢ,and a few classⅣand classⅠ.The reservoir quality of Chang 6~3 sublayer isobviously better than Chang 6~1 and Chang 6~2.As the best and the better reservoir in the studyarea,classⅠand classⅡare the favorable zones for oil.更多还原