塔里木盆地下侏罗统阿合组下砂砾岩段致密砂岩成岩相划分及测井识别--以库车坳陷依奇克里克地区为例

Diagenetic facies division and logging identification of tight sandstone in the lower conglomerate member of Lower Jurassic Ahe Formation in Tarim Basin:Case study of Yiqikelike area in Kuqa Depression

塔里木盆地库车坳陷依奇克里克地区下侏罗统阿合组储层成岩作用复杂、非均质性强,具有低孔、低渗特征。利用岩心观察、普通薄片和铸体薄片鉴定、扫描电镜等多种地质资料,依据成岩作用及成岩矿物将目的层划分为5类成岩相:致密压实相、碳酸盐胶结相、不稳定组分溶蚀相、溶蚀微裂缝相和微裂缝相。通过交会图处理常规测井资料,发现不同的成岩相具有不同的测井响应,但由于不同的成岩相测井响应存在信息重叠,因此并不能通过交会图识别不同成岩相。利用BP神经网络对测井信息进行数据挖掘,将成岩相测井识别从低维线性不可分问题映射到高维非线性可分,训练出的学习模型准确率较高,并通过与薄片鉴定结果和孔渗数据的对比,验证了学习模型的准确性,进而为缺乏取心井段的储层成岩相测井识别提供依据。

Reservoir diagenesis of Lower Jurassic Ahe Formation in Yiqikelike area,Kuqa Depression,Tarim Basin is complex,highly heterogeneous and characterized by low porosity and low permeability.According to the diagenesis and diagenetic minerals,the target strata were divided into five types of diagenetic facies:tightly compacted facies,carbonate cemented facies,unstable-components dissolution facies,dissolution micro-frac⁃ture facies and micro-fracture facies by observation of core,thin section,cast thin section and scanning electron microscope.Different diagenetic facies have different logging responses through cross-plot processing of conven⁃tional logging data.However,it is not possible to identify different diagenetic facies by cross-plot because of the overlap of logging information of different diagenetic facies.By using BP neural network to mine logging infor⁃mation,the training model has a high accuracy rate.By comparing with the thin section identification results and pore and permeability data,the accuracy of the learning model is verified,thus providing a basis for the logging identification of reservoir diagenetic facies in the interval lacking of coring.