心尤有痛

Joseph

近日，复读《Geostatistical Reservoir Modeling》（Second Edition），至2.1.1 The Story这一节，原书28页。

2.1.1 The Story

Numerical reservoir models could be constructed without an in-depth understanding of the geologicstory responsible for the genesis of the reservoir. One could approach reservoir modeling as a purely statistical study. In this scenario, the data would be handed over from the geologists to the reservoirmodeler(s) working in isolation. Data would be statistically described, as well as pooled into distinctregions, and trends are mapped. Spatial continuity ischaracterized within the regions, and simulation andpost-processing could proceed as described in subsequent chapters of this book. The resulting modelscould then be validated only by their reproductionof input statistics and local data, described as minimum acceptance criteria (Boisvert, 2010; Leuangthong et al., 2004). These models would seem to beperfectly correct from a statistical perspective, andsome may prefer them for their apparent objectivity and lack of bias due to a preconceived conceptualmodel.

Yet, this approach would ignore important information not apparent from the local data alone andcould even violate concepts fundamental to the geology of the reservoir. Potential consequences includea lack of reproduction of geological features at a scaleless than the well spacing and a poor assessment ofmodel uncertainty. The geological information beyond the data may provide important informationto constrain extrapolation away from the data anduncertainty constraints. This geological informationis central to selecting analogs to aid in inference insparse data settings and to decide on appropriatehigh-resolution heterogeneity in mature, dense datasettings. A more intangible consequence relates tomodel communication and credibility. Models thatviolate or lack integration of fundamental geological knowledge may communicate misconceptions torecipients or the flow simulation results may not betrusted, resulting in issues in project team alignment.

Some important caveats on the value of the geological story should be mentioned. Firstly, the importance of the geological story will depend on theamount of data available. It is clear that in exploration and appraisal, with few wells and poor or noseismic information, the geological story forms thefoundation for all reservoir model decisions. On theother hand, for a mature project with dense well dataand high-resolution measurements from a seismicdata set, a more data-driven approach may be utilized with the geological story providing little morethan a consistency check. Secondly, the importanceof the geological story may shift with changes in project objectives and scale. For example, if enhancedrecovery is later applied in the mature field and theresults are sensitive to fine-scale intrawell heterogeneity, then the geological story may once agaibecome more important. Thirdly, it is difficult toknow what geological features are important priorto modeling; therefore, ignoring geological information even in a mature field may result in significantlost opportunity.

A reservoir modeler, within the integrated earthscience and engineering project team, should attempt to discover and piece together a plausibleand consistent story for the reservoir. This effort allows for the integration of not only the local hardand soft data, but also geological insight and concepts. Given the previously mentioned data paucity,incompleteness, and inaccuracies typically encountered, finding the story may be difficult. Generalconcepts with regard to addressing uncertainty (seeSection 5.3) can be employed. Uncertainty may becharacterized by (a) retaining multiple scenarios orparameter distributions and (b) multiple stories withvariations in the details. The simplest explanation(s)that matches the data and concepts should be retained. At times, less likely cases may be retained toaid in risk mitigation.

Issues in Telling the Story

Devising a reservoir story involves many decisionsand an assessment of the associated uncertainty.While doing this, we should be aware that the heuristics we naturally apply to solve these difficult problems may lead to bias (Tversky and Kahneman,1974). These include representativeness, availability, and adjustment and anchoring. For example,representativeness results in the assignment of probability of A being part of B by the similarity of A andB. Availability results in the assignment of greaterprobability to the familiar and does not represent lessfamiliar. Finally, adjustment and anchoring is a natural human tendency to seize onto a position andadjust to formulate an estimate. Firstly, often the initial anchor is unreliable, we may anchor to unrelatedor unreliable information; so secondly, we tend tounder adjust the estimate from the anchor.

Not only are there potential issues in the mannerthat we estimate, but once an estimate is made, confirmation bias may become an issue. Confirmationbias is the tendency to ignore or undervalue information that contradicts the currently supported view,and group polarization is the tendency of a group toevolve to the extreme views of the group. The resultof these cognitive issues is (1) we typically underestimate uncertainty or, more plainly stated, we thinkwe know more than we actually do, and (2) groups of experts may not work together in an optimum manner. When the project team’s efforts are moderated,then group wisdom may result in better estimates foreven the most difficult estimation problems.

Value of Telling the Story

A project team utilizes individual expertise to carefully formulate a comprehensive story that integratesall the available data, information, and expertise.This story continues to play a role in the project asa reality check, communication tool, and means tomake further inferences. It must be pieced togetherso that each contributing chapter is consistent anddefendable and serves as a tool for future data andconcept integration.A critical skill set is the ability to communicatebetween the disciplines and to integrate all salientinformation from the geologic perspective into thenumerical model. The following sections provide abrief treatment of the immense science of basin andreservoir geology. The focus is on providing a sampling of the concepts and nomenclature that formthe basis for the geological model of a petroleumreservoir with references for more intense study.Section 3.2 provides more details on ConceptualModels that represent this story. The following subsection discusses the models applied by geologists torepresent the story.

Joseph

这里所谈的很多事情实际已经远远超越了建模本身。

即如，讲故事这一段内，作者最后就说到了他们关注到的问题：我们经常会低估或简单面对不确定性，——他们还注释说，我们总是以为我们知道而其实我们实际上没有那么清楚；第二个问题，各专业的专家们其实没有协作经验。我们在编织故事的时候更倾向于把故事讲好而不是把困境预估清楚。

但是，这个故事仍然非常重要。它的重要性之一就是在数据、信息、专业之间的整合效用来说没有任何东西能够超越一个故事——至少在目前阶段而言，我们还没有办法真的把我们的认知完全嫁接到数据本身去实现，也没有必要去这样尝试。

理解这一段的人，可以算是同道中人了罢。

newHF

领导需要知道 1+1 = ?
不确定性扰乱领导决策的   ;)  :D

200610023135

文章很好！！！！！谢谢分享！！！！

200610023135

200610023135

谢谢分享，文章很好。

wuhouweiwu

顶一个