样本分解交汇图法与决策树方法识别苏丹某油田流体性质

Fluid Nature Identification of an Oil-gas Field in Sultan Using Stylebook Hierarchical Decomposition Cross-plot and Decision Tree Methods

识别储层流体性质是测井评价储层的内容之一,识别结果影响着对油气藏的认识和开发方案部署。本文以多种测井参数作为储层流体的分类参数,采用样品分解思想的交汇图和决策树方法对苏丹某油田流体性质进行分类建模。结果表明,(1)单个交汇图最高识别正确率为80%,而样品分解思想将交汇图分为两个层次,可提高识别准确率,但是仍存在不足之处;(2)决策树方法通过计算参数权重,依据权重大小自动挑选参数,识别准确率达到了86.7%;而交汇图和决策树法相结合,识别准确率可达到100%,弥补了样品分解交汇图的不足;(3)原来认为苏丹某油田是具有统一油气界面的油气藏,应用决策树结合交汇图方法后,认为该油田应为层状油气藏;油田据此调整开发方案,取得了良好效果。

Identifying the natural properties of the fluid is one of the significant tasks involving reservoir evaluation,and the consequential results would directly affect the understanding and development of oil-gas filed.By taking the advantage of a variety of logging curves as parameters,the stylebook hierarchical decomposition cross-plot and decision tree methods are utilized to identify the fluid natural properties in a Sudanese oil-gas field.Finally a classification modeling is obtained.The results indicate that（1） The accuracy of a single cross-plot could be reached as high as 80%;however,there is still an abundant room for the improvement if the stylebook hierarchical decomposition cross-plot is used in spite of shortcomings.（2） The parameters could be automatically selected in decision tree by calculating the weight of the parameters,thus the accuracy could be further reached at 86.7%.Combination implementation of the cross-plot and decision tree is able to improve the accuracy up to 100%,and the shortcomings are able to be made up.（3） According to the conventional understanding,an uniform oil-gas interface is expected to occur in the Sudanese filed,however with the applications of the decision tree,the filed now is regarded as a layered reservoir,the oil-field has achieved better results by adjusting the development strategy which is depended on the results of the decision tree.