模糊优选神经网络储层识别技术在长庆中部气田马五_1段的应用  

Application of fuzzy optimization neural network in Ma5_1,reservoir identification in the central Changqing gas-field

作　　者：匡建超[1] 曾剑毅[1] 王众[1] 

机构地区：[1]成都理工大学商学院,四川成都610059

出　　处：《油气地质与采收率》2008年第5期5-7,12,共4页Petroleum Geology and Recovery Efficiency

基　　金：2007年四川省学术和技术带头人培养基金(川人办发[2008]24号);2008年成都理工大学科研基金项目(Sxyzc08-09)

摘　　要：鄂尔多斯盆地奥陶系马家沟组岩性复杂,属于致密储层,所以储层识别是该层系天然气开发中所面临的关键问题和难点之一。传统的储层识别方法准确率较低,因此提出了利用基于粒子群算法的模糊优选神经网络对储层中的气、水、干层进行识别。选用长庆中部气田19口井分层测试92个已知样本,通过对物性、测井和储渗特征等参数的分析,选取电阻率、自然伽马、产能系数、储渗因子和介质类型因子5个主成分控制特征参数作为样本输入,以样本储层的产能赋值作为输出,构建了基于粒子群算法的模糊优选神经网络的储层识别模型。通过试算,优选了2个模型,回判正确率分别达到96.2%和92.3%,储层识别正确率达到100%。The Ordovician Majiagou Formation in Ordos Basin belongs to dense reservoir with very complex lithology.Reservoir identification is the key issue and also one of the difficulties in natural gas development in this area.Considering the low accuracy and precision of traditional reservoir identification method,the fuzzy optimization neural network based on particle swarm algorithm was applied to identify the gas strata,water strata and dry zone in the reservoir.Ninty-two samples were acquired by zonal testing from the selected nineteen wells in central gas-field of Changqing.According to the analysis of parameters of physical property,logging,storage-permeability characteristics etc,five dominant control factors(resistivity,natural gamma-ray,productivity coefficient,storage-permeability factor and media type factor)were selected as the input of the model,and reservoir capacity assignment was selected as output.Then,a new reservoir identification model was constructed.Two optimization models were selected based on the trial calculation,and the recognition probability of each reached 96.2% and 92.3% respectively.In addition,the correct rate of identification reached 100%.

关 键 词：储层识别 粒子群算法 模糊优选神经网络 长庆中部气田 鄂尔多斯盆地 

分 类 号：TE112.23[石油与天然气工程—油气勘探]