利用数据挖掘技术识别深层火山岩气层

石炭系火山岩的低渗透气藏具有的埋藏深、成因复杂、类型多、分布广、三孔隙度曲线响应特征不明显等特点，并且主要的四种火山岩气藏：玄武岩、安山岩、英安岩、流纹岩的密度和纵波时差测井值差别较大，因此使基于三孔隙度测井资料的一系列识别气层的有效方法在该盆地石炭系火山岩气层的识别中无明显效果.数据挖掘技术从数据的角度出发，在进一步丰富岩心、测井和试油资料的前提下，利用数据挖掘技术中的聚类分析和关联分析获取核心参数和数据之间的内在联系，用决策树提取预测火山岩气层的模型.该方法充分利用已有的数据资料，用数学分析方法遍历寻找对识别火山岩气层有用的信息，而不仅仅依靠三孔隙度和电阻率曲线，并且消除了岩性的影响，因此获得了较高的识别率.

Applying data mining techniques to identify deep volcanic gas reservoir

Volcanic gas reservoirs of low permeability in the carboniferous of a basin are characterized by deep burial depth, complex genesis, many types, wide distribution and non-evident expressions in three-porosity log curves, as well as the variedvalues of density and compression wave of main four volcanic gas reservoirs (such as the basalt, andesite, dacite and rhyolite). So a number of effective methods based on three porosity log data cannot work in identifying the volcanic gas reservoirs of the carboniferous of the basin. We propose a method to identify the volcanic gas reservoirs based on the data mining techniques. This method makes the core data, log data and well test data abundant, uses the cluster analysis and correlation analysis to obtain the sensitive parameters and the internal relations between the data and applies the decision tree to acquire the prediction model to identify the volcanic gas reservoirs. The method acquires the most useful information from the data not only depending on the three porosity and resistivity log curve and eliminates the impact of lithology, so it gets a better recognition rate. The practical application of the method shows that it can indicate the existence of natural gas in the deep volcanic rocks of the basin.