摘要
油气水层数据统计是一种非线性分类统计问题,由此建立了多项logistic回归油气水层模式识别模型。表征油气水层各因素之间存在着复杂的耦合关系,采用粗糙集属性约简算法对原始样本数据进行属性约简,消除因素间的耦合关系对识别结果的影响。选取大庆油田某地区的20口油井的数据作为建模样本数据,另10口油井的数据为测试样本数据,实验表明基于粗糙集的多项logistic回归模型对建模样本的解释正确率为100%,对测试样本的解释正确率为90%,远高于非属性约简的多项logistic回归模型,为油气水层模式识别提供了一种新方法。
Pick to the oil and gas layer data statistics is a kind of nonlinear classification statistical problems, a multinomial logistic regression pattern recognition model of oil gas water is established. There are complicated coupling relationships between factors and characterization of oil gas water, using attrib-ute reduction algorithm of rough set on the original sample data to carry on the attribute reduction, elimi-nating the influence of coupling relationship between factors on the recognition result. Select an area of Daqing oilfield 20 wells data as the modeling sample data, the other 10 wells data as test sample data,ex-periments show that the correct rate to the explanation of the modeling sample based on multinomial logis-tic regression model based on rough set is 100%, the right rate to the interpretation of the test sample was 90%, much higher than that a multinomial logistic regression model in non attribute reduction , which provides a new method for oil and gas reservoir pattern recognition.
出处
《工业仪表与自动化装置》
2015年第3期28-32,共5页
Industrial Instrumentation & Automation
基金
甘肃省科技厅项目"石油化工企业应急演练系统"(1204GKCA004)
甘肃省财政厅专项资金立项资助(甘财教[2013]116号)
关键词
油气水层
模式识别
粗糙集属性约简
多项logistic回归模型
oil and gas reservoir
pattern recognition
rough set attribute reduction
multinomial logistic regression model
