【作者】 张静；

【导师】 赵永军；

【作者基本信息】 中国石油大学(华东) ， 地质资源与地质工程， 2015， 硕士

【摘要】 大牛地气田位于鄂尔多斯盆地,属于典型的低孔、低渗致密砂岩气藏,其勘探及开发越来越受到重视,致密砂岩气藏储层参数的测井解释模型建立一直是测井解释的难点。由于致密砂岩气藏与常规砂岩相比有很大的不同,孔隙度和渗透率相对较低,孔隙结构复杂,储层的储集性能也相对较差,这些特点都会给致密含气砂岩储层的测井解释方法研究带来困难。本文首先进行岩心归位,充分利用已有的测井资料、岩心资料、物性等资料,在对储层的岩石学、物性、电性特征进行深入研究的基础上,针对研究区致密砂岩储层低孔、低渗的特征,系统的对储层四性关系进行了研究,明确了储层物性的主要控制因素,为测井解释模型的建立打下了基础。为了更加准确的求取储层参数,本文采用分层位、分岩性来建立储层参数测井解释模型,对于砂岩类型的划分,采用了神经网络判别来进行岩性识别,取得了较好的分类效果。对于储层参数的求取,通过使用岩心资料采用了多种统计方法建立了物性参数模型,包括线性一元、多元回归以及非线性的BP神经网络和支持向量机。通过对多种模型的效果对比,选出适合研究区的测井解释方法,最后对解释模型效果进行验证,证明能够挖掘致密砂岩储层电性和物性间非线性关系的支持向量机方法预测储层物性参数有较好的预测效果,提高了大牛地致密砂岩气藏储层参数测井解释模型的精度。更多还原

【Abstract】 Daniudi gas field located in the Ordos Basin,in the north-east slope of Shanxi,which is a typical low porosity and low permeability tight sandstone gas reservoir.The exploration and development get more and more attention,log interpretation model of tight sandstone gas reservoir’s parameters have been difficult.Because of compared with conventional sandstone,tight sandstone gas reservoirs are very different,relatively low porosity and permeability,complex pore structure,reservoir properties of reservoir is relatively poor,Research on logging interpretation is difficult for tight gas sandstone reservoirs.Firstly,on the basis of core correction,fully use of existing logging data,core data,properties and other information,to study the lithology,physical properties and electrical properties characteristics,based on the study area for tight sandstone reservoir low porosity and permeability characteristics,the system for four sexual relationships were studied,get the main controlling factor in reservoir properties and all of these established the foundation for log interpretation model.In order to predict reservoir parameters accurately,we use different layers and different lithology to establish reservoir parameters interpretation models.For sandstone type is divided by using a neural network,which achieved good classification results.reservoir parameters are obtained by using the core data through a variety of statistical methods to establish the physical parameters model,including linear one yuan,multiple regression and nonlinear BP neural network and support vector machines.By comparing a variety of models,the suitable log interpretation method is chosen for the study area,through the interpretation model validation results,indicating that the use of support vector machine to predict tight sandstone reservoir parameters have better predictive results,which can improve Daniudi tight sandstone gas reservoirs logging interpretation model accurately.更多还原