油藏描述中的地震相分析新技术

《油藏描述中的地震相分析新技术》由会员分享，可在线阅读，更多相关《油藏描述中的地震相分析新技术（123页珍藏版）》请在金锄头文库上搜索。

1、油藏描述中的地震相分析新技术 油藏描述中的地震相分析新技术绪论地震相分析技术在油藏描述中的作用波形分类地震相分析技术的特点和应用实例 -以Stratimagic软件为例地震相分析技术存在的问题和发展趋势绪论地震相是个“古老”，宽泛的名词，概念。60年代中期，有人开始使用，70年代初随着地震地层学的兴起，被广泛使用。地震相的定义，多种多样不统一，本人愿意定义为： 地震信号特征的一种表征形式，并且这种表征形式所表征的信号地震信号特征的一种表征形式，并且这种表征形式所表征的信号特征可以在横向或纵向上划分成单元或分类。特征可以在横向或纵向上划分成单元或分类。 地震信号外形，内部结构，振幅，相位，频率，

2、速度等均可以作地震信号外形，内部结构，振幅，相位，频率，速度等均可以作为地震相。现在一般将振幅，相位等表征地震动力学特征的信号为地震相。现在一般将振幅，相位等表征地震动力学特征的信号称为属性，而把反射外形等静力学特征的信号定义为地震相的较称为属性，而把反射外形等静力学特征的信号定义为地震相的较多。多。地震相可应用于地震地层学，岩性地震学以及油藏描述中的储层预测等。本次讲座主要集中在地震相的概念，以及工业界已应用的地震相分析方法的原理，实例的介绍上。油藏描述中的地震相分析新技术绪论地震相分析技术在油藏描述中的作用波形分类地震相分析技术的特点和应用实例 -以Stratimagic软件为例地震相分析

3、技术存在的问题和发展趋势地震相分析技术在油藏描述中的作用油藏描述的任务： - -储层储层/ /油藏分布预测油藏分布预测 - -储层储层/ /油藏物性的确定油藏物性的确定解决油藏描述的地震方法： - - 属性属性/ /地震相分析方法地震相分析方法 - -井约束井约束AI/EIAI/EI反演方法反演方法推荐的工作流程构造解释构造解释沿层选定目的段沿层选定目的段层段内地震相分析层段内地震相分析靶区的井约束反演靶区的井约束反演地震相和反演结果地震相和反演结果的综合解释的综合解释定性到定量的地震相分析定性到定量的地震相分析Does this channel-like feature repeat its

4、elf in other places?Can I map this reflector terminationand display it on a horizon map?Build a 3D volume that describesthe shape of this channel. Display it on any seismic data or map.What is Results from the Seismic Facies What is Results from the Seismic Facies Analysis Technology ?Analysis Techn

5、ology ?Comparison with Traditional MethodsMap of Amplitude Standard Deviation (the best Amplitude Map)Seismic Facies superimposed on the Amplitude mapEACH real trace is assigned a color according to whichmodel trace it is most closely correlated.The Seismic Facies Map地震地震方法用于油藏描述方法用于油藏描述的现状的现状Attrib

6、uties Analysis/Attributies Analysis/地震相地震相 地震属性分析方法地震属性分析方法所提取属性种类不断增加（所提取属性种类不断增加（2020，5050种，更多？）种，更多？）用户选择属性用户选择属性缺少合适的方法对多种属性解释缺少合适的方法对多种属性解释地质意义不明确。地质意义不明确。Well Calibration and InversionWell Calibration and Inversion 地震的井标定和反演地震的井标定和反演外推估算地震信号的横向变化通常是困难的外推估算地震信号的横向变化通常是困难的需要先验的初始模型需要先验的初始模型花费和计算

7、吞吐量仍是系统化工业化应用的障碍花费和计算吞吐量仍是系统化工业化应用的障碍先验约束往往出现误差先验约束往往出现误差地震相分析技术在油藏描述中的作用快速进行地震信号特征的分类，研究地震信号的变化规律从地震信号某种/多种特征的变化规律中确定反映地质体沉积，物性等变化的规律，从而直接进行沉积相研究，储层预测和物性预测等。地震相分析可以快速的为井约束AI/EI反演等确定靶区，指导反演结果的解释。新的地震相分析方法可以进一步确定地震微相，地震相的定量化等，从而进行油藏的精细描述。 Well 1Well 1Well 2Well 2Example - Channel Definition (1)Conven

8、tional Instantaneous / Average Amplitude MapsExample - Channel Definition (2)Seismic Facies Map of isolated channel region, using a Neural Network derived classificationModel - 12 classesBAExample - Channel Definition (3)Detail of Central ChannelDifferences in production from wells A and B are expla

9、inedWell B - central (clean) channelWell A - point bar地震相分析技术在油藏描述中的作用快速进行地震信号特征的分类，研究地震信号的变化规律从地震信号某种/多种特征的变化规律中确定反映地质体沉积，物性等变化的规律，从而直接进行沉积相研究，储层预测和物性预测等。地震相分析可以快速的为井约束AI/EI反演等确定靶区，指导反演结果的解释。新的地震相分析方法可以进一步确定地震微相，地震相的定量化等，从而进行油藏的精细描述。油藏描述中的地震相分析新技术绪论地震相分析技术在油藏描述中的作用波形分类地震相分析技术的特点和应用实例 -以Stratimagic软

10、件为例地震相分析技术存在的问题和发展趋势波形分类地震相分析技术的特点和实例 Attributies Analysis/Attributies Analysis/地震相地震相 地震属性分析方法地震属性分析方法所提取属性种类不断增加（所提取属性种类不断增加（所提取属性种类不断增加（所提取属性种类不断增加（20202020，50505050种，更多？）种，更多？）种，更多？）种，更多？）用户选择属性用户选择属性用户选择属性用户选择属性缺少合适的方法对多种属性解释缺少合适的方法对多种属性解释缺少合适的方法对多种属性解释缺少合适的方法对多种属性解释地质意义不明确。地质意义不明确。地质意义不明确。地质意义

11、不明确。 Well Calibration and InversionWell Calibration and Inversion 地震的井标定和反演地震的井标定和反演外推估算地震信号的横向变化通常是困难的外推估算地震信号的横向变化通常是困难的外推估算地震信号的横向变化通常是困难的外推估算地震信号的横向变化通常是困难的需要先验的初始模型需要先验的初始模型需要先验的初始模型需要先验的初始模型花费和计算吞吐量仍是系统化工业化应用的障碍花费和计算吞吐量仍是系统化工业化应用的障碍花费和计算吞吐量仍是系统化工业化应用的障碍花费和计算吞吐量仍是系统化工业化应用的障碍先验约束往往出现误差先验约束往往出现误差

12、先验约束往往出现误差先验约束往往出现误差Stratimagic-Stratimagic-地震地层解释地震地层解释/ /地震相分析软件地震相分析软件专门用于解释岩性，地层，油藏，地质相对比的新的地震解释技术源于ELF公司获得专利的波形分类技术，由CGG-FLAGSHIP开发为软件产品。2002年Paradigm购并Flagship后，进一步与其它的地震相分析技术结合，如Seisfacies, NexModel, VoxelGeo等，使其更加完整，功能强大。Stratimagic: a unique solutionStratimagic: a unique solution Stratimagi

13、c: Stratimagic: 独特的解决方案独特的解决方案波形分类地震相分析波形分类地震相分析 A process : characterization based on trace shape A process : characterization based on trace shape 一种处理：一种处理： 基于道形状的基于道形状的特征描述特征描述 Trace shape classification represents the true heterogenity of the Trace shape classification represents the true heter

14、ogenity of the seismic signal seismic signal 道形状分类代表了地震信号的真实的横向异常道形状分类代表了地震信号的真实的横向异常 A technology: self-organizing neural networks A technology: self-organizing neural networks 一项技术：自组织的神经网络一项技术：自组织的神经网络 An industrial shape-recognition process, robust and unaffected by An industrial shape-recogniti

15、on process, robust and unaffected by noise or spurious eventsnoise or spurious events 一个工业化的形状识别处理，它稳定，不受噪音和假同相轴的影响一个工业化的形状识别处理，它稳定，不受噪音和假同相轴的影响 A method: many years of operational success applied. A method: many years of operational success applied. 一种方法：成功地应一种方法：成功地应多年多年 to exploration, appraisal

16、and reservoir studiesin clastics and to exploration, appraisal and reservoir studiesin clastics and carbonatescarbonates for oil and gas, onshore or offshorefor oil and gas, onshore or offshore on 5 continents, from sea-bottom to 20.000 ft.on 5 continents, from sea-bottom to 20.000 ft. 可用于勘探评价和油藏研究，

17、碎屑岩和碳酸岩，油或气，陆上和海上。可用于勘探评价和油藏研究，碎屑岩和碳酸岩，油或气，陆上和海上。The Basic Assumption is Changes in any of the physical parameters of the subsurface are always reflected in a change in shape of the seismic trace.For example change in porosity will result in a differently shaped trace.“shape” is quantified in the ch

18、ange of sample value from sample to sample.What is the Seismic Facies Classification What is the Seismic Facies Classification Technology Mentioned Here?Technology Mentioned Here?What do you see ?Your brain is a neural network - SHAPE is used to decidehow many different types of vegetable are here.N

19、OT color(how many peppers?) or size (how many tomatoes?).Are these the same shape?Now What Do You See?Does this channel-like feature repeat itself in other places?Can I map this reflector terminationand display it on a horizon map?Build a 3D volume that describesthe shape of this channel. Display it

20、 on any seismic data or map.What is the Seismic Facies Classification What is the Seismic Facies Classification Technology Mentioned Here?Technology Mentioned Here?How to understand the meaning of seismic data through Facies Identification and Classification using Trace shape?-XX% amplitude+/-2msSam

21、pling to nearest 4ms sample generates+/-2ms unbiased noise on timeup to 25% biased noise on amplitudeFIXED VERTICAL SAMPLINGReduces sampling noise Takes full advantage of propagation beyond seismic sampleTRACE RECONSTRUCTIONTrace Reconstruction: a critical step.Trace Reconstruction: a critical step.

22、WHAT ARE BASIC NEURAL NETWORKS?Signal Flow: Input OutputSynapseINPUT SEISMICINTERVALOUTPUT TRACESDendritesCell BodySynapsesAxonLooking for seismic shape changesLooking for seismic shape changesNeural NetworkClustering analysisA Neural Network looks for a suite of traces that describe the progressive

23、 changes in the seismic shape.Looking for seismic shape changesLooking for seismic shape changesNeural Networkordered color changesClustering analysisabrupt color changesWhat Do We Classify?Whole cube?Significantly exceeds actual Significantly exceeds actual volume of interest (reservoir), good volu

24、me of interest (reservoir), good for early exploratory work onlyfor early exploratory work onlyAttribute maps?Demands prior knowledge, can be Demands prior knowledge, can be used to refine insight, but not to used to refine insight, but not to define itdefine itProblem: Which maps to use as Problem:

25、 Which maps to use as input?input?Problem: Some information could Problem: Some information could be bypassedbe bypassedTrace shape in interval?Focused on geological volume of Focused on geological volume of interestinterestSeismic signal shape includes all Seismic signal shape includes all attribut

26、esattributesComparison of Benefits and DrawbacksDoes this channel-like feature repeat itself in other places?Can I map this reflector terminationand display it on a horizon map?Build a 3D volume that describesthe shape of this channel. Display it on any seismic data or map.What is the Seismic Facies

27、 Classification What is the Seismic Facies Classification Technology Mentioned Here?Technology Mentioned Here?How to understand the meaning of seismic data through Facies Identification and Classification using Trace shape?工作流程工作流程( (work flow)work flow)I. Learning from the data, and only the dataI.

28、 Learning from the data, and only the data从地震数据中学习，且仅仅从地震数据从地震数据中学习，且仅仅从地震数据 The model traces The model traces 模型道模型道 These synthetic traces are constructed by the neural network These synthetic traces are constructed by the neural network process, using a learning set extracted from the seismic pro

29、cess, using a learning set extracted from the seismic interval. No well data is used at this stage. The user has no interval. No well data is used at this stage. The user has no influence on the selection of data, and there are no weighting influence on the selection of data, and there are no weight

30、ing criteria. The result is 100% repeatable.criteria. The result is 100% repeatable. 这些合成道是用从地震层段中提取出来的由神经网络处理建造的，这些合成道是用从地震层段中提取出来的由神经网络处理建造的， 这一阶段不这一阶段不需要井数据。用户在数据选择方面没有影响，没有加权标准，结果需要井数据。用户在数据选择方面没有影响，没有加权标准，结果. 100% . 100% 可重可重复。复。INPUT SEISMICINTERVALOUTPUT TRACESSynapsesDendritesCell BodyAxonWh

31、at are Basic Neural Networks?Signal Flow: Input Output SynapseThe ProcessThe ProcessThe Neural Network trains itself on the actual trace shapes within a 3D seismic interval, and constructs synthetic seismic traces that represent the signal diversity over the entire defined volumeTraces are refined b

32、y an iterative processuntil the best correlation to the real datais obtainedThe Seismic Facies MapThe Seismic Facies MapEACH real trace is assigned a color according to whichmodel trace it most closely correlates toNEURAL NETWORK PARAMETERSNumber of model traces (number of colours in the output faci

33、es map)Number of iterationsRate of learning (epsilon), Continuity (sigma)Reference surfaces, interval thickness, sub-sampling parameterOUTPUTINPUTPROCESSINGClassification Maps: Class Range 2 to 1003 Classes7 Classes15 ClassesIncreasing the number of classes results in greater detailSmall number of c

34、lasses identifies first order trace variabilityUnlike clustering, Neural Networks do not require preconceived ideas about the number of classesNumber of Iterations: Range 1 to 1001 Iteration20 Iterations50 Iterations100 IterationsCLASSIFICATION MAPS1 iteration unstable20 iterations stable results0It

35、erations100Rate of convergence50Neural Network Stabilises2010The seismic facies map The seismic facies map 地震相图地震相图 The map The map 地震相地震相图图 Each trace has been Each trace has been assigned the number assigned the number (and color) of the model (and color) of the model trace to which it has the tra

36、ce to which it has the best correlation.best correlation. 每一道赋给它与模型道最相关的每一道赋给它与模型道最相关的号码和颜色。号码和颜色。 By observing By observing the distribution of color the distribution of color on this map, we can on this map, we can assess the distribution of assess the distribution of seismic shapes seismic shapes

37、 throughout the throughout the interpreted area.interpreted area. 通过观察图上颜色的分布，我们通过观察图上颜色的分布，我们可以评定解释区域的地震形状的可以评定解释区域的地震形状的分布。反映了岩性，地层，地质分布。反映了岩性，地层，地质相的变化。相的变化。Projecting facies information on seismicProjecting facies information on seismic将相的信息投影到地震剖面上将相的信息投影到地震剖面上The classification result can be p

38、rojected directly above The classification result can be projected directly above the interval on which the process was applied, allowing the interval on which the process was applied, allowing a one-to-one visualization of the actual data traces and a one-to-one visualization of the actual data tra

39、ces and their corresponding assignement to one of the classes.their corresponding assignement to one of the classes.分类结果可以直接投影到处理过的层段的上，允许一对一的实际数据道及分类结果可以直接投影到处理过的层段的上，允许一对一的实际数据道及其中一个相应的赋值分类的可视化，其中一个相应的赋值分类的可视化， 为地震相的变化确定其具体反射为地震相的变化确定其具体反射特征。特征。利用专门的解释工具（利用专门的解释工具（Reflector Termination&Envelops)Re

40、flector Termination&Envelops)等，逐线等，逐线解释出地震相变化的位置和形状，如上超，下超，不整合等。解释出地震相变化的位置和形状，如上超，下超，不整合等。利用利用TerminationTermination和和EnvelopeEnvelope解释解释利用利用利用利用Termination Termination 和和和和Envelope Envelope 进进行解行解行解行解释释Where do we go from here? Fitting the facies map to well informationFitting the facies map to w

41、ell information The relevance of the facies map(s) relative to a The relevance of the facies map(s) relative to a geological setting can be assessed by fitting in well geological setting can be assessed by fitting in well information.information. Interval scopeInterval scope The process is obviously

42、 sensitive to data that is The process is obviously sensitive to data that is included (or excluded) from the volume of interest. included (or excluded) from the volume of interest. While intervals should be larger than the strict time-While intervals should be larger than the strict time-thickness

43、of interest (to catch e.g. tuning effects), it thickness of interest (to catch e.g. tuning effects), it is interesting to try different thicknesses.is interesting to try different thicknesses. Area of interestArea of interest Once a general-purpose map has indicated some Once a general-purpose map h

44、as indicated some major features, the process can be focused on the major features, the process can be focused on the zones of interest, to obtain a sharper, more detailed zones of interest, to obtain a sharper, more detailed picture. picture. II. II. 将地震相结果与井信息匹配进一步细分地震相将地震相结果与井信息匹配进一步细分地震相 Fitting

45、 well information: Fitting well information: 与井信息匹配与井信息匹配 Seismic signal at well position: Seismic signal at well position: 井位置的地震信号井位置的地震信号 For each well, select either a synthetic seismogram from a list, or the For each well, select either a synthetic seismogram from a list, or the actual seismic

46、trace at the penetration of the interval. Correlation to all actual seismic trace at the penetration of the interval. Correlation to all models is computed, and a color is assigned according to best models is computed, and a color is assigned according to best correlation.correlation. 对于每一口井，选择或者合成记

47、录，或者层段位置的地震道，与所有的模型道计算相关，对于每一口井，选择或者合成记录，或者层段位置的地震道，与所有的模型道计算相关，按照最好的相关赋给颜色。按照最好的相关赋给颜色。Fitting Well Information: Comparing Seismic ResponseFitting Well Information: Comparing Seismic Response Real trace at well location is compared with the modelsWhich is the best model?Where else can we seethis mo

48、del type?Substituting traces in the model table: Substituting traces in the model table: 替换模型道替换模型道 Stratimagic allows the user to substitute one of the model Stratimagic allows the user to substitute one of the model traces with the trace currently in the selector window, be traces with the trace c

49、urrently in the selector window, be it a seismic trace from the dataset, or a synthetic it a seismic trace from the dataset, or a synthetic seismogram computed from the well data.seismogram computed from the well data. Stratimagic Stratimagic 允许用户用现有选择窗口的道替换一个模型道，可以是来允许用户用现有选择窗口的道替换一个模型道，可以是来自数据体的地震

50、道，或者是由井数据计算的合成记录。自数据体的地震道，或者是由井数据计算的合成记录。Recalculating on an area of interest Recalculating on an area of interest 重新计算关心的区域重新计算关心的区域 To reveal more detail in the channel, we could increase the number To reveal more detail in the channel, we could increase the number of classes. However, it is more

51、efficient to process a new interval of classes. However, it is more efficient to process a new interval restricted to the prospective area.restricted to the prospective area.为了揭示河道内更细的细节，可以增加分类为了揭示河道内更细的细节，可以增加分类的数目，而更有效的方法是在限定的区域内处理新的层段。的数目，而更有效的方法是在限定的区域内处理新的层段。III. Quantifing Seismic Facies with

52、Petrophysical parameters地震相的定量化-岩石物性参数模拟 NexModelPiloted seismic facies analysis using NexModelTM and StratimagicTM NexModelTM Basic WorkflowLoad log dataCreate layered impedance modelfor wellLoad seismic at well locationCreate wellsyntheticTie well syntheticto seismic & optimiseBasic/Advancedmodell

53、ingCalibrate/quantifyseismic faciesExport newfacies groupsSUPERVISEDCLASSIFICATIONUNSUPERVISEDCLASSIFICATIONBRENTBRENTBRENTBRENTBRENTBRENT$reducing risk& adding value to subsurfaceevaluationA powerful starting pointThe NexModel synthetic is correlated to.The nearby seismic tracesThe nearby seismic t

54、racesThe Stratimagic NNT trace modelsThe Stratimagic NNT trace modelsDTRHOBGRSynthetic SeismogramSeismic tracesat borehole, with horizonsSynthetic traceStratimagicbest fitmodel TraceAcoustic impedencecolumnNexmodel Nexmodel & StratimagicModeledtrace AModeledtrace Bq qGuided classification in Stratim

55、agic, through model traces Guided classification in Stratimagic, through model traces substitution with Nexmodel synthetic traces. substitution with Nexmodel synthetic traces. Insertion of :Insertion of : the well A synthetic tracethe well A synthetic trace the pseudo-well synthetic traces the pseud

56、o-well synthetic tracesNexmodel Nexmodel & StratimagicNexmodel & StratimagicWell AWell BModeledtrace AModeledtrace BOther main functions Conventional Attributes analysis常规的属性分析工具常规的属性分析工具3D propagation3D 3D 自动追踪自动追踪 VoxelGeo3D3D可视化可视化StratiQC成图成图 Conventional Interpretation tools常规的解释工具常规的解释工具Add-on

57、 Value and Project database integration Add-on Value and Project database integration 高附加值与其它数据库的高度集成高附加值与其它数据库的高度集成 Preserving your investmentPreserving your investment 保护你的投资保护你的投资 Your current OpenWorks or GeoFrame project databases are Your current OpenWorks or GeoFrame project databases are acc

58、essedaccessed 你现有的数据库是你现有的数据库是OpenWorksOpenWorks还是还是GeoFrame? Stratimagic GeoFrame? Stratimagic 都可以都可以访问访问, , 包括地震数据体，包括地震数据体， 层位和井数据。层位和井数据。 Your users continue to use their familiar structural Your users continue to use their familiar structural interpretation toolsinterpretation tools 你的用户继续使用他们熟

59、悉的构造解释工具你的用户继续使用他们熟悉的构造解释工具 Put to work immediatelyPut to work immediately 立即投入工作立即投入工作 Without duplication, access large-volume seismic datasets Without duplication, access large-volume seismic datasets over the network.over the network. 不需要备份，可以通过网络访问大的地震数据体。不需要备份，可以通过网络访问大的地震数据体。沿层或层间提取多种属性，作为岩性，

60、地层，沿层或层间提取多种属性，作为岩性，地层，油藏解释的辅助手段油藏解释的辅助手段沿层或层间提取14大类30多种属性，得到属性平面图，作为描述岩性变化等辅助特性。利用叠合图（Mix Map)手段， 将所有属性结果，地震相图等任意叠合在一起，利用方便的色彩管理工具，突出它们的共同特征，定性的综合对比，为解释岩性，地层，油藏等提供更多的依据。 Dip map, final channel baseZoomed areaChannel thickness attributesChannel thickness attributesThis attribute highlights This attr

61、ibute highlights the features of interest the features of interest that have been seen on that have been seen on other maps.other maps.Average amplitude of peakevents in intervalFacies map (channel area alone)Facies map (channel area alone)Running the facies classification process on the Running the

62、 facies classification process on the restricted area of the channels yields a more restricted area of the channels yields a more representative characterizationrepresentative characterizationSeismic facies map 12 classesMixMap- Facies map- Avg. Ampl. PeaksGeological model , prospectivityGeological

63、model , prospectivity Prospective features were mapped as Prospective features were mapped as envelopesenvelopes (3D delineation of volume) (3D delineation of volume) Channel leveeChannel levee Paleo-channelPaleo-channel Point barPoint barEnlargement on attribute map of average amplitudeof interval

64、peaks Random profile Time Slice 1660 ms New FeaturesMulti-Attribute Seismic Facies Classification using NNTMulti-volume trace shape classification on constant and non-constant intervalsMulti-attribute map classificationIntegration with VoxelGeoViewing and manipulation of Stratimagic data in VoxelGeo

65、 with no workflow disruptionsSeismic volumesSeismic volumesSeismic facies and attribute mapsSeismic facies and attribute mapsWells (boreholes, tops, logs) Wells (boreholes, tops, logs) SeisFaciesMulti-attribute seismic analysisIntroducing a new software solution jointly developed by ENI Agip Divisio

66、n and Flagship GeoEni多属性全数据体分类多属性全数据体分类SeisFacies BenefitsFacies distribution (and its associated geological meaning) can be applied for:Preliminary screening of seismic Preliminary screening of seismic data in exploration activitydata in exploration activityRanking of prospectsRanking of prospectsC

67、onditioning of geostatistical Conditioning of geostatistical reservoir modellingreservoir modellingReservoir characterizationReservoir characterizationGeohazard risk assessmentGeohazard risk assessmentCLASSIFIED SEISMIC TRACES OR SAMPLES CAN BE DIRECTLY CALIBRATED FOR QUANTITATIVE DEFINITION OF RESE

68、RVOIR PROPERTIESSeisFaciesSeisFacies incorporates technologies and methods developed by ENI AGIP DIVISIONClassificationCalibrationFusion 3D SEISMIC TRACE CLASSIFICATION MULTIATTRIBUTE MAPS CLASSIFICATION MULTIATTRIBUTE BLOCK CLASSIFICATIONSeisFacies Classification ProcessesClassificationNNT & Hierar

69、chicalBLOCKSMAPSTRACESInput Data:Multiple 3D seismic volumesVariable or constant time intervalInput Data:Interval attribute mapsHorizon mapsClassification mapsInput Data:Multiple 3D seismic volumesVariable or constant time intervalPCA(Recommended)Output Data:3D Seismic Facies VolumeClassificationHie

70、rarchicalOutput Data:Attribute Facies MapClassificationHierarchicalPCA(Optional)Output Data:Seismic Facies MapZonation(Optional)PCA(Optional)AmplitudeCoherencyImpedancePCA 2PCA 1PCA 2PCA ComponentsAmplitudeCoherencyImpedanceInput Seismic VolumesPCA 1SeisFacies PCAAPPLICATION OF REGRESSION FUNCTION(S

71、)CONTROL VARIABLEDEFINITION OFREGIONSOUTPUT : CALIBRATED VOLUMESeisFacies Multi-Attribute SeisFacies Multi-Attribute CalibrationCalibrationClassificationSEMBLANCEVOLUMESEMBLANCE MAPIMPEDANCE MAP IMPEDANCE VOLUMEMIIXATTR RESULTS“ MIXED VOLUME” “ MIXED MAP “RESERVOIRCHARACTERIZATIONRECODINGThe SeisFac

72、ies “Fusion” ApproachTurbidite SystemBase of TurbiditeHorizon Slice: Facies BlockHorizon Slice: Fusion Semblance/ ImpedanceSemblanceImpedanceSeisFacies FusionExampleSeisFacies Conclusions Integrated component and extension to Stratimagic; Integrated component and extension to Stratimagic; shares Str

73、atimagics user interface and shares Stratimagics user interface and infrastructureinfrastructure Robust solution for multi-attribute classification and Robust solution for multi-attribute classification and calibration of seismic data, incorporating calibration of seismic data, incorporating technol

74、ogies and methods developed by ENI AGIPtechnologies and methods developed by ENI AGIP Enables effortless work on multiple versions of a Enables effortless work on multiple versions of a seismic survey, or a set of attributes computed over seismic survey, or a set of attributes computed over timetime

75、 Enables a detailed description of the reservoir, Enables a detailed description of the reservoir, resulting in better informed business decisions resulting in better informed business decisions based on more accurate prediction of reservesbased on more accurate prediction of reserves Improves reser

76、voir characterization within field Improves reservoir characterization within field development projects development projects A New Methodology Based on Seismic Facies Analysis and Litho-Seismic ModelingThe Elkhorn Slough Field Pilot Project The Elkhorn Slough Field Pilot Project Solano County Calif

77、orniaSolano County CaliforniaBy Manuel Poupon (Flagship Geosciences, today By Manuel Poupon (Flagship Geosciences, today Paradigm) and Kostia Azbel (CGG-Geoscience) Paradigm) and Kostia Azbel (CGG-Geoscience) Offshore, March 1999Offshore, March 1999 The Elkhorn Slough Field Pilot Project Solano Coun

78、ty California Scope of the Project Scope of the Project The Data: Winters Pinchout 3DThe Data: Winters Pinchout 3D The Play: Deep Water Fan/ChannelThe Play: Deep Water Fan/Channel Structural InterpretationStructural Interpretation Conventional Horizon AttributesConventional Horizon Attributes Geolog

79、ical Horizon AttributesGeological Horizon Attributes Stratigraphic InterpretationStratigraphic Interpretation Conventional Interval Attribute AnalysisConventional Interval Attribute Analysis Seismic Facies AnalysisSeismic Facies Analysis Modeling Seismic Facies TraceModeling Seismic Facies Trace Rev

80、ised Geological Model and Business ImpactsRevised Geological Model and Business Impacts ConclusionsConclusionsInterpretation of horizons & faultsBeforeAfterSeismic Classification: The Missing Link?Calibration to wellsInversion or Interval map analysisGeostatistical analysis, modelingInterpretation o

81、f horizons & faultsAnalyze surface attributesSeismic Facies ClassificationInterval attribute analysisInterpretation of geological shapesGeostatistical analysis modelingThe Data: Winters Pinchout 3D3D Survey - Solano Co., CaliforniaShot and processed by CGG-Shot and processed by CGG-Americas in 1995.

82、Americas in 1995.830 in-lines, 700 cross-lines - 830 in-lines, 700 cross-lines - 110x110 bin spacing (52 sq. 110x110 bin spacing (52 sq. miles).miles).Sample interval 2 msec - Record Sample interval 2 msec - Record length 6 sec.length 6 sec.Well Data4 wells drilled on a turbiditic play 4 wells drill

83、ed on a turbiditic play (based on 2D interpretation).(based on 2D interpretation).Well C: “70 Feet of Well C: “70 Feet of netpay in the A sand”.netpay in the A sand”.Recoverable reservesRecoverable reservesincreased to 14-18 increased to 14-18 BCFBCFDeep Water FanIncised channelPrograding sequencesB

84、asinal shalesABC DWECDThe Play: Deep Water Fan/ChannelWinters Sands“The Winters pinchout play is turbiditic in nature, sands “The Winters pinchout play is turbiditic in nature, sands being transported through channels incised into the shelf being transported through channels incised into the shelf a

85、nd deposited into deep water fans surrounded by shales” and deposited into deep water fans surrounded by shales” (K. Lanning 1998).(K. Lanning 1998).Refining the Channel/Fan AreaWe can do two things:1. Limit the area of analysis to the channel/fan only2. Use the wellbore trace to pilot the seismic f

86、acies map Well A: 15ft gasWell B: 45ft waterWell C: 70ft gasWell D: no sand ABCDTime map (Top of Winters sand)Mixed map (Time + Dip)Azimuth mapStructural InterpretationConventional Horizon Attributes Sediment layers dip toward the South West.Sediment layers dip toward the South West. Dip and Azimuth

87、 maps respectively highlight the Dip and Azimuth maps respectively highlight the channel and fan system.channel and fan system.Stratigraphic InterpretationConventional Horizon AttributesHorizon Amplitude Map: Two different geological Horizon Amplitude Map: Two different geological environments are e

88、xpressed with similar high amplitude environments are expressed with similar high amplitude valuesvaluesBright spotsChannel/Fan systemPrograding sequencesThis Neural Network Technology is licensed from TotalFinaElfSeismic Facies Analysis using NNT: What Is It?Seismic Facies: The description and geol

89、ogic interpretation of seismic reflection patterns including configurations, (continuous, sigmoidal, etc.), frequency, amplitude, and continuity.Neural Network Technology (NNT): The ability to analyze and classify trace shapes using a discriminating process.Seismic Facies Map: This is a similarity m

90、ap of actual traces to a set of model traces that represents the diversity of various trace shapes present in an interval. Model TracesInterval of interestChannel/Fan PlayProgradingsequencesSeismic Facies MapStratigraphic InterpretationUnpiloted Regional Seismic Facies AnalysisClassifying the 60-ms

91、interval above the reference horizon Classifying the 60-ms interval above the reference horizon using Neural Network shape recognition. Seismic facies using Neural Network shape recognition. Seismic facies map shows turbidites deposited along a NNW-SSE paleo-map shows turbidites deposited along a NN

92、W-SSE paleo-coastline. coastline. Several channels incising the shelf can also be identified.Several channels incising the shelf can also be identified.Stratigraphic InterpretationUnpiloted Channel Seismic Facies Analysis Classifying the 60-msec interval below the reference Classifying the 60-msec i

93、nterval below the reference horizon using Neural Network shape recognition. Seismic horizon using Neural Network shape recognition. Seismic facies map highlights the outline of an asymmetric fan.facies map highlights the outline of an asymmetric fan.Seismic response at Well CMain stream NW to SE or

94、tilted sea bottom towards SE?Asymmetric fanModel Traces15-25700Seismic Facies MapABCDiStratigraphic InterpretationPiloted Seismic Facies Analysis Using the seismic response at Well C as an indicator of Using the seismic response at Well C as an indicator of gas-charged sands and focusing over the ch

95、annel/fan gas-charged sands and focusing over the channel/fan area only, piloted seismic facies map highlights the area only, piloted seismic facies map highlights the distribution of the thicker reservoir sands.distribution of the thicker reservoir sands.Seismic traceat Well C used asmodel #9Model

96、tracesPiloted Seismic Facies MapOnly a limited area in the channel/fan system have seismic responses similar to Well C (thick sand) Well A & B (thin sand) are out of the main fan area, Well D (shaled out) has a distinct seismic facies.Random LinePetro-Acoustic ModelingModeling Seismic Facies Trace U

97、sing log traces from Well C and model trace #9, seismic Using log traces from Well C and model trace #9, seismic response is calibrated as 70 of gas-charged sands.response is calibrated as 70 of gas-charged sands.?Well Logs (DT, RHOB, GR)Synthetic SeismogramSand ReservoirSeismic traces at Well CMode

98、l Trace #9Petro-Acoustic ModelingSeismic Facies Calibration Using log traces at Well C and D to respectively calibrate Using log traces at Well C and D to respectively calibrate seismic response of gas-charged A sands and seismic seismic response of gas-charged A sands and seismic response of a no-s

99、and zone.response of a no-sand zone.Seismic Well CSeismic Well DNote: Model traces are not “True Amplitude” DataPetro-Acoustic ModelingPetro-acoustic Modeling of Reservoir Characteristics Perturbing Seismic Response from Well C using sand Perturbing Seismic Response from Well C using sand thickness,

100、 reservoir porosity, and fluid content as variable thickness, reservoir porosity, and fluid content as variable parameters. parameters. Decreased Sand ThicknessDecreased PorosityDecreased Gas SaturationWell CPetro-Acoustic ModelingPerturbation of Reservoir Characteristics Thickness variations are mo

101、deled from C to D wells (70 to Thickness variations are modeled from C to D wells (70 to 0).0).Synthetic traces and seismic traces are similarSynthetic traces and seismic traces are similarDecreased Thickness (70 to 0)Flattened seismic sectionInterval of interestWEPetro-Acoustic ModelingPerturbation

102、 of Reservoir Characteristics Synthetic model traces are generated between C and D Synthetic model traces are generated between C and D wells using a combination of reservoir thickness, water wells using a combination of reservoir thickness, water saturation and porosity. Synthetics are then used to

103、 pilot saturation and porosity. Synthetics are then used to pilot the seismic facies analysis.the seismic facies analysis.Synthetic model tracesSeismic Facies MapProposed E wellPost-Mortem Analysis of Well DRevised Geological Model Synthetic Classifying the 82-msec Synthetic Classifying the 82-msec

104、interval below the “despiked” interval below the “despiked” reference horizon using 20 model reference horizon using 20 model traces. traces. This new unpiloted seismic facies This new unpiloted seismic facies map highlights the outline of a late map highlights the outline of a late shale plug fan t

105、hat could explain shale plug fan that could explain the absence of A sand in the D the absence of A sand in the D well.well. Seismic Facies MapModel tracesBusiness ImpactsCost to Expose Pay Drilling program is directly Drilling program is directly related to the geological modelrelated to the geolog

106、ical modelMaximize Production RateIdentify the sweet spots High-Identify the sweet spots High-grade prospectsgrade prospects? ?Conclusions Exploration within the Elkhorn Slough field had been mostly Exploration within the Elkhorn Slough field had been mostly driven by amplitude anomalies, inversion

107、techniques and driven by amplitude anomalies, inversion techniques and coherency technology.coherency technology.Seismic Facies Analysis combined with litho-seismic modeling Seismic Facies Analysis combined with litho-seismic modeling of well data was applied to the Elkhorn Slough Field.of well data

108、 was applied to the Elkhorn Slough Field. This methodology is accurate, cost-effective, quick and often This methodology is accurate, cost-effective, quick and often reveals subtle geological features only expressed in the shape reveals subtle geological features only expressed in the shape of the s

109、eismic trace. of the seismic trace. The geological model was tested with the E well which found The geological model was tested with the E well which found 100 of gas sand. 100 of gas sand. Seismic facies map (present work)Coherency slice (K. Lanning 1998)100 of gas-charged sandTurbidite Characteriz

110、ation Using Multi-Attribute Volume Classification Data Offshore Angola3D Survey Offshore Angola, Africa650 in-lines, 650 cross-lines 6.25X6.25 bin spacing Sample 650 in-lines, 650 cross-lines 6.25X6.25 bin spacing Sample interval 2 msec - Volume used from 2500-2900ms Generated interval 2 msec - Volu

111、me used from 2500-2900ms Generated Attributes Amplitude, Dip, Azimuth, AI, Porosity and Attributes Amplitude, Dip, Azimuth, AI, Porosity and SemblanceSemblanceWell Data2 wells drilled on a turbidite play2 wells drilled on a turbidite play Well 2 good producer from Well 2 good producer from massive s

112、andsmassive sands Well 4 bad producer from poorly Well 4 bad producer from poorly sorted sandssorted sandsWell 4Well 2Part Two: Volume Classification The Data: Offshore Angola, Post Stack 3D dataThe Data: Offshore Angola, Post Stack 3D data Structural Setting: Extensional Faulting Structural Setting

113、: Extensional Faulting Depositional Setting: Turbidite Slope ChannelsDepositional Setting: Turbidite Slope Channels Stratigraphic InterpretationStratigraphic Interpretation Classic Seismic Trace Shape AnalysisClassic Seismic Trace Shape Analysis Attribute Map ClassificationAttribute Map Classificati

114、on Multi-Attribute Volume VisualizationMulti-Attribute Volume Visualization Multi-Attribute Volume ClassificationMulti-Attribute Volume Classification Subvolume Detection in 3D environmentSubvolume Detection in 3D environment ConclusionsConclusionsStructural SettingWell 4Well 2Well 4Well 2Top System

115、 (blue)Top Sequence A (blue)Top Channel A (violet)Intra Channel A (yellow)Base Channel A (red)Erosion Base Sequence A (green)Wells with GR Log Well 2 is a good producer in both Well 2 is a good producer in both upper and lower A Unit.upper and lower A Unit. Well 4 is a poor producer from Well 4 is a

116、 poor producer from upper A unit and shows a low GR upper A unit and shows a low GR response for the second unit of response for the second unit of corresponding massive sands in corresponding massive sands in Well 2.Well 2.Inline 2258 Well 4Inline 2120 Well 2Well 4Well 2Stratigraphic Interpretation

117、 WorkflowStratigraphic Interpretation WorkflowTRACESInput Data:Multiple 3D seismic volumesVariable or constant time intervalClassificationHierarchicalPCA(Optional)Output Data:Seismic Facies MapMAPSInput Data:Interval attribute mapsHorizon mapsClassification mapsClassificationHierarchicalOutput Data:

118、Attribute Facies MapPCA(Optional)ClassificationNNT & HierarchicalBLOCKSInput Data:Multiple 3D seismic volumesVariable or constant time intervalPCA(Recommended)Output Data:3D Seismic Facies VolumeZonation(Optional)Interval Attribute MapsInterval Attributes between Top Channel A and Intra Channel AInt

119、erval IsopachInterval IsopachThird DerivativeThird DerivativeFourth DerivativeFourth DerivativeAmplitude PeaksAmplitude PeaksPeak-Trough Peak-Trough RatioRatioFrequency PeaksFrequency PeaksAmplitude Amplitude Positive PolarityPositive PolarityWell 2Well 4Map Classification with NNT 14C Major Channel

120、 System Major Channel System (yellow and red classes) (yellow and red classes) with inner Channel (blue with inner Channel (blue classes).classes). Well 4 (poorly sorted Well 4 (poorly sorted sands) is located on blue sands) is located on blue classes and Well 2 classes and Well 2 (massive sands) on

121、 yellow (massive sands) on yellow classes.classes. Axis of eroding channel is Axis of eroding channel is clearly defined.clearly defined.Well 2Well 4Trace Shape Analysis on Upper UnitTrace shape analysis on a non-constant interval between Top Channel A and Intra Channel AWell 2Well 4Three major fami

122、lies of traces , faciesPrincipal Component Analysis - Why?Principal Component Analysis - Why?Why do we want to perform PCA ?To analyse data redundancy and To analyse data redundancy and bring several individual attribute bring several individual attribute volumes down to fewer PCA volumes down to fe

123、wer PCA volumes (Eigenvalues greater than volumes (Eigenvalues greater than 1) for further analysis and 1) for further analysis and classificationclassificationTo understand which attributes To understand which attributes contribute the most in describing contribute the most in describing the trend

124、in a data set the trend in a data set To get a better understanding of To get a better understanding of data/attribute dependencies and data/attribute dependencies and correlationscorrelationsTo eliminate noiseTo eliminate noiseWhat Attributes to Classify ?A possibility would be to classify PCA volu

125、mes, and here we have a cut off A possibility would be to classify PCA volumes, and here we have a cut off at Component 3.at Component 3.We have reduced the amount of data from 8 to 3 volumes to be handled.We have reduced the amount of data from 8 to 3 volumes to be handled.AI and Azimuth have low c

126、orrelation value and it will thus be interesting to classify them togetherAttribute Volumes, AICrossline 3858, Well 2Well 4 Random Line Well 2Well 4Well 2Acoustic Impedance Scale BarAttribute Volumes, AzimuthHorizon Slice +4 From Top Channel A Crossline 3858, Well 2Azimuth Scale BarWell 2Well 4-180+

127、180Manual Attribute ClassificationAzimuthAICLASSIFICATIONDATA COMPRESSION BY ZONATION (OPTIONAL)PRINCIPAL COMPONENT ANALYSIS (OPTIONAL)AI and Azimuth CrossplotSwCut-off 2 4600 Cut-off 1 4200 Analysis in Vanguard shows two major cut-offs for acoustic impedance values, corresponding to clean massive s

128、ands and intermediate sands and shalesAcoustic Impedance4200 46007000AzimuthCero+90-90Massive sands in red colorsIntermediate sands in greenShales in blue rangeAI and Azimuth Classification VolumeWell 4Well 2Crossline 3858, Well 2Well 4 Random Line Well 2Horizon Slice on Classified VolumeWell 2Good

129、producing sands showing four distinct depositional directionsWell 4A thin layer of intermediate sands (green) with the well 2 corresponding lower A unit with shales(Note that channel in green eroding massive sands in Well 4 !)Well 2Well 4Horizon slice through upper A unitHorizon Slice on Classified

130、VolumeWell 2Well 4Horizon Slice (+20) through lower unit. Well 2 shows two clear depositional directions.Note there are no sands in Well 4. Possible side track to massive sand units in close areas.Lets have a look at the random linesHorizon Slice on Classified VolumeRandom Line 1, Well 412Southern S

131、and unit shows mainly one depositional direction (red) and might be one single deposition, channel.AIAzimuthHorizon Slice on Classified VolumeUpper A unit sands extend (above Horizon Slice) towards the south and show various depositional directions. This may indicate stacked channels and several sys

132、tems.Horizon Slice is indicated on Random line 2 with a green line.12Random Line 2Well 2Well 43D Visualization in VoxelGeoClassified volume in 3D visualizationSubvolume Detection Class 5All detected bodies with6-way connectivityClass 52 Main Connected BodiesInteractive display of main detected bodie

133、s with 6-way connectivitySummary Map-based Trace Shape Analysis Interval Attribute Generation and Classification Multi-Attribute Volume Analysis / PCA Multi-Attribute Volume Classification Subvolume Detection on sand prone classStratimagic and SeisFacies Tools for advanced Stratigraphic interpretati

134、on in a complex channelled environment through:ConclusionCost to Expose PayDrilling program is directly related to the geological Drilling program is directly related to the geological modelmodelMaximize Production RateIdentify the sweet spotsIdentify the sweet spotsHigh-grade prospectsHigh-grade pr

135、ospectsVolumetricReserves calculation is directly affected by the Reserves calculation is directly affected by the geological model geological model Know the value of your prospect!Stratimagic and SeisFacies Tools for advanced Classification Methods油藏描述中的地震相分析新技术绪论地震相分析技术在油藏描述中的作用波形分类地震相分析技术的特点和应用实例

136、 -以Stratimagic软件为例地震相分析技术存在的问题和发展趋势地震相分析技术的存在的问题和发展趋势地震相分析技术的存在的问题和发展趋势 发展现状：发展现状： - - 定性描述多，定量化方法少定性描述多，定量化方法少 - - 单属性分类划相多，多属性综合划相少单属性分类划相多，多属性综合划相少 - - 从地震资料入手，精度受制于地震的分辨率从地震资料入手，精度受制于地震的分辨率 - - 与地质结合少，地震相的地质意义仍没有直接的对应关系与地质结合少，地震相的地质意义仍没有直接的对应关系 - - 分类方法众多，存在多解性分类方法众多，存在多解性 今后发展趋势：今后发展趋势： - - 多属性综合划相多属性综合划相 - - 地震相与地质相的自动解释和对应地震相与地质相的自动解释和对应 - - 多种方法综合划相结果的判定多种方法综合划相结果的判定 - - 碳酸岩地震相描述方法的研究碳酸岩地震相描述方法的研究 - - 地震相分析地震相分析/AI/EI/AI/EI反演的相互约束反演的相互约束 谢 谢