基于横波测井资料的神经网络火山岩流体性质识别  

Volcanic Reservoirs Fluid Identification by Neural Network Based on Shear Wave Log Data

作　　者：边会媛[1] 潘保芝[2] 王飞[2] 

机构地区：[1]中国地质大学,北京100083 [2]吉林大学地球探测科学与技术学院,吉林长春130026

出　　处：《测井技术》2013年第3期264-268,共5页Well Logging Technology

基　　金：国家自然科学基金(41174096);国家重大专项(2011zx05009-001)联合资助

摘　　要：利用偶极横波测井数据求取火山岩储层的4个气层识别指标:压缩系数、泊松比、横纵波速度比以及等效弹性模量差比。这4个气层识别指标在一定程度上都能指示气层的存在,但是单个指标不能将气层、差气层、气水同层、水层、干层区分开,为此引入Kohonen神经网络方法综合识别流体性质。在试气层段提取各个气层指标的数据作为神经网络的输入,流体性质作为输出,构成Kohonen神经网络所需样本数据,建立神经网络气层自动识别方法,通过合层技术自动输出解释剖面。在松南火山岩气田应用,与试气结论相比,预测符合率为83.3%。Based on dipole shear wave logging data, the 4 indicators, such as Poisson＇s ratio, the coefficient of compressibility, the ratio of compressional wave velocity to shear wave velocity and the relative ratio of equivalent elastic modulus, are calculated to identify gas in volcanic reservoirs. To a certain extent,the 4 indicators can identify gas layers, but single indicator cannot discriminate gas layers,water layers, water and gas layers or a dry layers. So Kohonen neural network is introduced to determine the fluid property. In tested formation,various gas indicators are extracted as the input data of Kohonen neural network, and the fluid property as the output of Kohonen neural network. Then the sample datum of Kohonen neural network are made up, so gas automatic identification method using Kohonen neural network is established. Interpretation sections are automatically calculated by merging layers of the same fluid property. We get a satisfactory result in the volcanic reservoir in Songnan gas field, and the forecast coincidence rate is 83. 3%.

关 键 词：声波测井 偶极横波 气层指标 火山岩 KOHONEN神经网络 

分 类 号：P631.84[天文地球—地质矿产勘探]