一种识别非典型气层的新方法--决策树的分类模型  

A new method of identifying the atypical gas reservoir based on the classification model of decision tree

作　　者：戴诗华 李雄炎[2,3] 于红岩[2,3] 周金煜[4] 陈亦寒[2,3] 

机构地区：[1]中国石油西部钻探测井公司,克拉玛依834000 [2]油气资源与探测国家重点实验室 中国石油大学(北京),北京102249 [3]地球探测与信息技术北京市重点实验室 中国石油大学(北京),北京102249 [4]长庆油田公司勘探开发研究院,西安710021

出　　处：《地球物理学进展》2010年第3期885-890,共6页Progress in Geophysics

基　　金：国家高技术研究发展计划(863项目:2009AA062802)项目资助

摘　　要：非典型气层由于成因复杂,受储层厚度较薄、围岩的影响及测井仪器分辨率的限制,在测井曲线上表现出许多模糊性,致使三孔隙度和电阻率曲线在气层的敏感性降低,尤其是与气水同层、水层、干层的差异性不大,从而使一系列基于该资料的识别方法对非典型气层的识别无明显效果.而利用多参数测井信息对非典型气层的识别其实质是一个复杂的、非线性的、高维数的模式识别问题,因此可以采用模式识别方法对非典型气层进行识别.而决策树具有学习能力强、学习过程透明、结果可理解性强,能提供友好的人机交互机制,因此以多方面的测井信息为基础,利用决策树提取非典型气层的预测模型,并综合储层实际特征,对预测模型进行修正.实际应用结果表明,决策树提取的预测模型对非典型气层的识别具有较好的效果.Because of the effects of the reservoir thickness, the adjacent beds and well logging instrument＇s resolution, the atypical gas reservoir＇s log information is fuzzy. It results in the three porosities and resistivity curves in the lower sensitivity in the atypical gas reservoir, so some methods based on the three porosities and resistivity curves can not identify accurately the atypical gas reservoir. In fact, it is a complicated, nonlinear and high- dimensional pattern recognition question that we use log information to identify the atypical gas reservoir, so we can use the pattern recognition method to identify the atypical gas reservoir. The decision tree＇s learning ability is strong, the learning process is transparent, the learning result is understandable and the kind human-computer interaction mechanism is supplied. Therefore, we use the decision tree to get the predictive model of the atypical gas reservoir based on the lots of log information. At the last but least, we improve the predictive model based on the real reservoir characterization. The practical application results show that the predictive model can identify accurately the atypical gas reservoir which is established by the decision tree.

关 键 词：决策树 非典型气层 测井评价 分类模型 预测模型 油气 识别 

分 类 号：P631[天文地球—地质矿产勘探]