摘要
传统的储层渗流屏障识别方法大多为定性或半定量,对于单砂体内部规模较小的储层渗流屏障的识别精度及划分效率相对较低。以沧东凹陷枣南孔一段油藏为例,基于岩心及测井资料,引入人工智能机器学习方法,提出一种基于SOM神经网络聚类和KNN算法的储层渗流屏障定量识别方法。该方法主要通过SOM算法逐点对取芯井多测井曲线进行聚类,获取能够表征储层质量差异的测井相神经单元,然后与岩性及构型进行对比,建立取芯井测井相神经单元定量划分储层渗流屏障标准,最后通过KNN算法将取芯井测井相神经单元模型传播到非取芯井,并对非取芯井进行储层渗流屏障识别与划分。结果表明,该方法对储层渗流屏障的识别结果与岩心的符合率超过90%,同时快速地对全区非取芯井渗流屏障进行划分,有效地提升储层渗流屏障的识别精度与效率,这也为类似的研究提供一种新的思路和方法。
Traditional methods for identifying reservoir flow barriers are mostly qualitative or semi-quantitative,and have limitations in terms of accuracy and efficiency,especially for identifying small-scale reservoir flow barriers within a single sand body.In this paper,a quantitative identification method for reservoir flow barriers is proposed based on core and logging data from the first member of Kongdian Formation in Zaonan Oilfield,Cangdong Sag.This method utilizes a machine learning approach incorporating the SOM neural network clustering and KNN algorithm.The method begins by applying the SOM algorithm to cluster the multi-logging curves of the core-taking well point by point,identifies the logging phase neural units that effectively characterize the quality difference of the reservoir.These neural units are then compared with lithology and configuration data to establish a quantitative standard for dividing the reservoir flow barriers within the core-taking well logging phase neural units.Finally,the KNN algorithm is employed to propagate the core-taking well logging phase neural unit model to non-core-taking wells.This allows for the identification and division of reservoir flow barriers in these wells.The results demonstrate that the proposed method achieves an identification accuracy of over 90%in agreement with the core data for reservoir flow barriers.Additionally,the flow barriers in non-core-taking wells are efficiently and accurately divided.Consequently,the proposed method significantly improves the accuracy and efficiency of reservoir flow barrier identification.This approach presents a new idea and method for similar research in the field,providing a valuable contribution to the field of reservoir characterization.
作者
斯扬
蔡明俊
张家良
芦凤明
王芮
黄金富
孟瑞刚
SI Yang;CAI Mingjun;ZHANG Jialiang;LU Fengming;WANG Rui;HUANG Jinfu;MENG Ruigang(PetroChina Dagang Oilfield Company,Tianjin 300280,China)
出处
《中国石油大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2023年第4期35-47,共13页
Journal of China University of Petroleum(Edition of Natural Science)
基金
中国石油重大科技专项(2018E-11-06)。
关键词
渗流屏障
自组织神经网络
K最近邻算法
枣南孔一段
沧东凹陷
reservoir flow barriers
self-organizing neural network
K-nearest neighbor algorithm
the first member of Kongdian Formation in Zaonan Oilfield
Cangdong Sag
