石油勘探与开发2026，Vol.53Issue(2)：369-383,15.DOI:10.11698/PED.20250167
塔河油田奥陶系古岩溶暗河结构划分与充填程度智能定量预测
Structural classification of Ordovician paleokarst conduits and intelligent quantitative prediction of filling degree in Tahe Oilfield,NW China
摘要
Abstract
Based on 3D seismic,logging,and core data from the Tahe Oilfield,Tarim Basin,this study carried out logging-based identification of cave filling facies in drilled Ordovician paleokarst conduits and quantitative calculation of filling degree,and analyzed the internal structure of paleokarst conduits.On this basis,a quantitative prediction of the filling degree of conduit networks in plan view was achieved by constructing a nonlinear relationship model.The results show that,according to differences in petrophysical fabric,filling facies in drilled caves can be classified into host-rock facies within caves,sandy-muddy cemented conglomeratic clastic facies,transported sandstone facies,chemical sedimentary filling facies and unfilled cave facies.Using a convolutional neural network algorithm,the filling degree of 156 drilled caves in the study area was quantitatively calculated,among which caves with a filling degree greater than 80％account for 39.7％,whereas those with a filling degree less than 20％account for only 16.0％.The genetic types of paleokarst conduits were divided into 7 categories:main-stream conduits,tributary conduits,outflow conduits,along-stream conduits,turnaround conduits,sinking-river conduits and labyrinthine conduits;and six conduit morphologies were identified:sinkholes,hall-shaped chambers,underflow loops,horizontal underflow passages,corridor passages and medium-dip passages.On this basis,a backpropagation neural-network-based quantitative prediction method for conduit filling degree was established using geological controlling factors.The prediction results indicate that the filling within paleokarst conduits shows obvious spatial differentiation:the probability of filling is relatively high in underflow loop segments,zones of increased potential energy,and medium-dip passage segments,whereas the spaces above hall-shaped chambers,the upper parts of medium-dip connecting passages,and downstream outlets of conduits have relatively low filling probabilities.The latter should therefore be regarded as key potential targets for future fine-scale development of paleokarst conduit reservoirs.关键词
古岩溶暗河/神经网络/充填程度/暗河结构/中下奥陶统/塔河油田/塔里木盆地Key words
paleokarst conduits/neural network/filling prediction/conduit structure/Middle-Lower Ordovician/Tahe Oilfield/Tarim Basin分类
能源科技引用本文复制引用
高济元,王诺宇,李雨阳,蔡忠贤,张恒,蒋林,汪彦,王仕林..塔河油田奥陶系古岩溶暗河结构划分与充填程度智能定量预测[J].石油勘探与开发,2026,53(2):369-383,15.基金项目
油气勘探开发理论与技术湖北省重点实验室开放基金(TTPED-2021-12) （TTPED-2021-12）
中国石化西北油田分公司科研项目(KY2021-S-094) （KY2021-S-094）
中国科学院战略性先导科技专项(A类)(XDA14010302) （A类）
