中亚阿姆河右岸东部地区侏罗系盐下碳酸盐岩储层特征及预测新方法

Characteristics and new prediction methods of Jurassic subsalt carbonate reservoirs in the eastern right bank of Amu Darya,Central Asia

以岩心、薄片及测井资料为基础,分析了中亚阿姆河右岸东部地区侏罗系牛津阶储层的岩性、储集空间及物性特征,利用三维地震数据,通过正演地震模拟、波形聚类技术、分频RGB融合技术以及集成学习等方法,对储层分布及厚度进行了半定量预测。研究结果表明:①阿姆河右岸东部地区侏罗系牛津阶储层岩性主要为生屑灰岩、颗粒灰岩及泥晶灰岩,为裂缝-孔洞型储层,储集空间以生物格架孔、粒内孔和裂缝为主,裂缝为后期溶蚀流体提供了通道;储层平均厚度为41.6 m,平均孔隙度为4.65%。②研究区储层多发育于XVhp层,根据其在XVhp层的分布位置可分为顶部型、底部型和两期型,顶部型储层主要分布于研究区中部及东部,底部型储层主要集中于西部,而两期型储层大量发育于西北部;研究区西部和东北部储层发育较好,厚度为45~75 m,多沿断裂发育带分布,东南部储层发育较差,厚度小于30 m,距离断层更远。③利用集成学习方法计算研究区储层厚度时,采用异质集成方法中的Stacking方法进行模型运算,以基于箱线图的离群值剔除方法剔除极端数据,以交叉验证法评估模型的预测性能,得到的储层厚度与11口实钻井储层厚度的变化趋势相符,相关系数为0.74。

Based on core,thin section,and well logging data,the lithologies,reservoir space,and physical properties of Jurassic Oxfordian reservoir in the eastern right bank of Amu Darya in Central Asia were analyzed.3D seismic data were used to carry out semi-quantitative prediction of reservoir distribution and thickness through methods such as forward seismic modeling,waveform clustering,frequency-division RGB fusion and ensemble learning.The results show that:(1)The lithologies of Jurassic Oxfordian reservoir in the eastern right bank of Amu Darya are mainly bioclastic limestone,grainstone and micritic limestone,and the reservoirs are characterized by fractured-vuggy type.The reservoir space is dominated by biological framework pores,intragranular pores and fractures,and fractures provide a channel for later dissolution fluids.The average thickness of the reservoir is 41.6 m,and the average porosity is 4.65%.(2)The reservoirs in the study area are mostly developed in the XVhp layer,and can be divided into top-type,bottom-type,and dual-type according to their distribution position in the XVhp layer.The top-type reservoirs are mainly distributed in the central and eastern parts of the study area,the bottom-type reservoirs are concentrated in the west,while the dual-type reservoirs are extensively developed in the northwest.The reservoirs in the western and northeastern parts of the study area are well developed,with a thickness of 45-75 m,mostly distributed along fault zones.The reservoirs in the southeastern part are poorly developed,with a thickness less than 30 m,and are further away from faults.(3)When using ensemble learning method to calculate the reservoir thickness in the study area,the stacking method in the heterogeneous ensemble approach was used for model computation.Outliers were eliminated using a boxplotbased method,and the predictive performance of the models was evaluated through cross-validation.The calculated reservoir thickness showed a high degree of agreement with the thickness observed in 11 wells drilled in the study area,and the correlation coefficient is 0.74.