软计算与硬计算融合的含油性属性与样本约简

随着石油勘探领域的不断扩大,含油性识别的研究对象也越来越复杂,传统的基于单一硬计算或软计算的方法在含油性识别中面临着严峻挑战.首先提出了软计算与硬计算融合的4种模式,然后运用GA-FCM对含油性的测井属性进行约简,将约简后的测井属性结合软计算与硬计算融合的分离模式对某油田Oilsk81,Oilsk83,Oilsk85三口井进行含油性模式识别,去掉出错率较高的样本,达到样本约简的目的;最后利用判别分析法对约简后的样本集进行检验分析.实验表明:第一,在这几个油区可以用声波时差和含油饱和度两个测井属性进行含油性识别;第二,将出错率高的样本进行约简可以提高样本集识别的正确率.

Attribute Value and Sample of Oil-Bearing Formation Reduction Based on Fusion of Soft Computing and Hard Computing

As the domain of oil and gas exploit expend quickly,the objects of oil-recognition become more and more complex,so traditional methods based on only soft computing or hard computing will face more challenge.In this paper we first propose four patterns that soft computing and hard computing integrate together.And then we use GA-FCM to simplify the well-log attributes in oil-recognition.The simplified well-log attributes with separate pattern of soft computing and hard computing used to identify the oil-bearing formation in Oilsk81,Oilsk83,Oilsk85 oil well.The aim is to drop out the samples of high wrong rate and simplify it.Lastly,the simplified sample sets is tested by discriminate method.It is proved that we can use AC and So to recognize oil-bearing formation.To simplify the high wrong rate samples will enhance the exactness of recognition.