机器学习方法在储层分类中的应用
|
摘要
油气田开发中有效储层和非有效储层的样本点存在混合带时,两类储层的划分是一个难点问题.从统计学上来看,其本质是一个含噪声的小样本二分类问题,可以采用机器学习方法,充分挖掘有试油成果的样本点的数据信息.分别利用线性判别分析、支持向量机、多层感知机神经网络建立储层分类模型,利用10次10折交叉验证法进行模型评估与优选,并利用全部样本点建立了有效的储层分类模型,最后将模型推广应用到样本分布的三种不同情形.结果表明,线性支持向量机模型具有最好的分类效果和很强的泛化能力,对于区分有效储层和非有效储层是有效的,可以在油气田开发中进行推广.
When there are mixed zones in the sample points of effective reservoirs and noneffective reservoirs in oil and gas field development,the division of the two types of reservoirs is a difficult problem.Statistically,it is essentially a small sample binary classification problem with noise,and the machine learning method can be used to fully mine the data information of the sample points with oil test results.The reservoir classification model is established by linear discriminant analysis,support vector machine and multilayer perceptron neural network,and 10 times 10-fold cross-validation method is used to evaluate and optimize models.And an effective reservoir classification model is established by using all sample points.Finally,the model is applied to three different situations of sample distribution.The results show that the linear support vector machine model has the best classification effect and strong generalization ability.It is effective to distinguish effective reservoirs from non-effective reservoirs and can be popularized in the development of oil and gas fields.
When there are mixed zones in the sample points of effective reservoirs and noneffective reservoirs in oil and gas field development,the division of the two types of reservoirs is a difficult problem.Statistically,it is essentially a small sample binary classification problem with noise,and the machine learning method can be used to fully mine the data information of the sample points with oil test results.The reservoir classification model is established by linear discriminant analysis,support vector machine and multilayer perceptron neural network,and 10 times 10-fold cross-validation method is used to evaluate and optimize models.And an effective reservoir classification model is established by using all sample points.Finally,the model is applied to three different situations of sample distribution.The results show that the linear support vector machine model has the best classification effect and strong generalization ability.It is effective to distinguish effective reservoirs from non-effective reservoirs and can be popularized in the development of oil and gas fields.
引文
[1]郭睿.储集层物性下限值确定方法及其补充[J].石油勘探与开发,2004(5):140-144.
[2]刘毛利,冯志鹏,蔡永良,等.有效储层物性下限方法的研究现状和发展方向[J].四川地质学报,2014,34(1):9-13.
[3]路智勇,韩学辉,张欣,等.储层物性下限确定方法的研究现状与展望[J].中国石油大学学报(自然科学版),2016,40(5):32-42.
[4]刘之的,张永波,孙家兴,等.有效储层物性下限确定方法综述及适用性分析[J].地球物理学进展,2018,33(3):1-9.
[5]丛琳,刘洋;马世忠,等.动静态资料在致密砂岩有效储层物性下限确定中的应用[J].当代化工,2014,43(8):1599-1601,1619.
[6]邵长新,王艳忠,操应长.确定有效储层物性下限的两种新方法及应用-以东营凹陷古近系深部碎屑岩储层为例[J].石油天然气学报,2008(2):414-416,653.
[7]柴明锐,程丹,张昌民,等.机器学习方法对砂砾岩岩屑成分的预测-以西北缘X723井百口泉组为例[J].西安石油大学学报(自然科学版),2017,32.(5):22-28,61.
[8]谭锋奇,李洪奇,孟照旭,等.数据挖掘方法在石油勘探开发中的应用研究[J].石油地球物理勘探,2010,45(1):85-91,164,172.
[9]李洪奇,郭海峰,郭海敏,等.复杂储层测井评价数据挖掘方法研究[J].石油学报,2009,30(4):542-549.
[10]石广仁.数据挖掘在石油勘探数据库中的应用前景[J].中国石油勘探,2009,14(1):1,60-64.
[11]Ning SHI,Hongqi LI,Weiping LUO.Data mining and well logging interpretation application to a conglomerate reservoir[J]:Applied Geophysics,2015,12(2):263-272.
[12]李道红.线性判别分析新方法研究及其应用[D].南京航空航天大学,2005.
[13]刘忠宝,王士同.改进的线性判别分析算法[J].计算机应用,2011,31(1):250-253.
[14]杨绪兵.线性判别分析及其推广性研究[D].南京航空航天大学,2003.
[15]汪海燕,黎建辉,杨风雷.支持向量机理论及算法研究综述[J].计算机应用研究,2014,31(5):1281-1286.
[16]C Cortes,V Vapnik.Support-vector networks[J].Machine Learning,1995,20(3):273-297.
[17]刘全稳.测井神经网络技术综述[J].石油地球物理勘探,1996(S1):64-69,147.
[18]Aminzadeh F.Applications of AI and soft computing for challenging problems in the oil industry[J].Journal of Petroleum Science and Engineering,2005,47(1):5-14.
[19]Platt JC.Fast training of support vector machines u-sing sequential minimal optimization[J].Advances in kernel methods:support vector learning table of contents,1998,185-208.
[20]操应长,王艳忠,徐涛玉,等.东营凹陷西部沙四上亚段滩坝砂体有效储层的物性下限及控制因素[J].沉积学报,2009,27(2):230-237.
[21]付金华,罗安湘,张妮妮,等.鄂尔多斯盆地长7油层组有效储层物性下限的确定[J].中国石油勘探,2014,19(6):82-88.
[2]刘毛利,冯志鹏,蔡永良,等.有效储层物性下限方法的研究现状和发展方向[J].四川地质学报,2014,34(1):9-13.
[3]路智勇,韩学辉,张欣,等.储层物性下限确定方法的研究现状与展望[J].中国石油大学学报(自然科学版),2016,40(5):32-42.
[4]刘之的,张永波,孙家兴,等.有效储层物性下限确定方法综述及适用性分析[J].地球物理学进展,2018,33(3):1-9.
[5]丛琳,刘洋;马世忠,等.动静态资料在致密砂岩有效储层物性下限确定中的应用[J].当代化工,2014,43(8):1599-1601,1619.
[6]邵长新,王艳忠,操应长.确定有效储层物性下限的两种新方法及应用-以东营凹陷古近系深部碎屑岩储层为例[J].石油天然气学报,2008(2):414-416,653.
[7]柴明锐,程丹,张昌民,等.机器学习方法对砂砾岩岩屑成分的预测-以西北缘X723井百口泉组为例[J].西安石油大学学报(自然科学版),2017,32.(5):22-28,61.
[8]谭锋奇,李洪奇,孟照旭,等.数据挖掘方法在石油勘探开发中的应用研究[J].石油地球物理勘探,2010,45(1):85-91,164,172.
[9]李洪奇,郭海峰,郭海敏,等.复杂储层测井评价数据挖掘方法研究[J].石油学报,2009,30(4):542-549.
[10]石广仁.数据挖掘在石油勘探数据库中的应用前景[J].中国石油勘探,2009,14(1):1,60-64.
[11]Ning SHI,Hongqi LI,Weiping LUO.Data mining and well logging interpretation application to a conglomerate reservoir[J]:Applied Geophysics,2015,12(2):263-272.
[12]李道红.线性判别分析新方法研究及其应用[D].南京航空航天大学,2005.
[13]刘忠宝,王士同.改进的线性判别分析算法[J].计算机应用,2011,31(1):250-253.
[14]杨绪兵.线性判别分析及其推广性研究[D].南京航空航天大学,2003.
[15]汪海燕,黎建辉,杨风雷.支持向量机理论及算法研究综述[J].计算机应用研究,2014,31(5):1281-1286.
[16]C Cortes,V Vapnik.Support-vector networks[J].Machine Learning,1995,20(3):273-297.
[17]刘全稳.测井神经网络技术综述[J].石油地球物理勘探,1996(S1):64-69,147.
[18]Aminzadeh F.Applications of AI and soft computing for challenging problems in the oil industry[J].Journal of Petroleum Science and Engineering,2005,47(1):5-14.
[19]Platt JC.Fast training of support vector machines u-sing sequential minimal optimization[J].Advances in kernel methods:support vector learning table of contents,1998,185-208.
[20]操应长,王艳忠,徐涛玉,等.东营凹陷西部沙四上亚段滩坝砂体有效储层的物性下限及控制因素[J].沉积学报,2009,27(2):230-237.
[21]付金华,罗安湘,张妮妮,等.鄂尔多斯盆地长7油层组有效储层物性下限的确定[J].中国石油勘探,2014,19(6):82-88.