一种致密碎屑岩储层产能预测的新方法——以新场气田沙溪庙组为例
|
摘要
针对致密碎屑岩储层产能预测精度较低这一难题,首次提出基于支持向量回归的产能预测方法。该方法在处理小样本问题上具有独到的优势,能够处理分类和回归预测问题。以新场气田沙溪庙组致密碎屑岩储层为例,选取33个已知样本(其中23个用于模型构建,10个用于精度检验),以储层厚度、裂缝张开度、裂缝孔隙度、裂缝渗透率、声波时差差值、测井孔隙度、电阻率差7个影响因素作为支持向量机输入,以样本储层的自然产能作为输出,构建基于支持向量回归的产能预测模型。模型预测的标准误差和平均绝对误差分别为0.0416和0.0229,表明该模型具有较高的预测精度,为致密碎屑岩储层产能的准确预测探索了又一新方法,对同类地区的研究和开发在一定程度上具有指导作用。
It is always difficult to predict production capacity in reservoir accurately.This paper proposes a method of prediction capacity based on support vector regression(SVR) to solve this problem.Support vector machine(SVM),which is a new machine study arithmetic based on statistical learning theory,can solve small-sample problems better.It succeeds in classification and prediction.33 samples tested(23 of them were used to form the model and the others were used to examine the precision of the model) from the compact clastic reservoir in Shaxumiao Formation of Xinchang gas field were selected,and the capacity prediction model based on SVR was established by using h,Δt-Δtmaε,φf,Kf,φand ΔR as input and using the capacity of reservoir as output.The normative error of capacity prediction model is 0.0416,and the average absolute error is only 0.0229.Therefore,the model formed is a new method for predicting capacity in clatic reservoir rock,and will provide useful reference for similar research in other regions.
It is always difficult to predict production capacity in reservoir accurately.This paper proposes a method of prediction capacity based on support vector regression(SVR) to solve this problem.Support vector machine(SVM),which is a new machine study arithmetic based on statistical learning theory,can solve small-sample problems better.It succeeds in classification and prediction.33 samples tested(23 of them were used to form the model and the others were used to examine the precision of the model) from the compact clastic reservoir in Shaxumiao Formation of Xinchang gas field were selected,and the capacity prediction model based on SVR was established by using h,Δt-Δtmaε,φf,Kf,φand ΔR as input and using the capacity of reservoir as output.The normative error of capacity prediction model is 0.0416,and the average absolute error is only 0.0229.Therefore,the model formed is a new method for predicting capacity in clatic reservoir rock,and will provide useful reference for similar research in other regions.
引文
[1]黄天虎,段永刚.气井产能预测综述[J].石油地质与工程,2007(2):43-47.
[2]王昀,张津林.基于地震资料的储层产能预测方法[J].石油天然气学报,2007(3):87-89.
[3]杨志明,刘广利.不确定性支持向量机原理及应用[M].北京:科学出版社,2007.
[4]向丹,施泽进,黄大志.鄂尔多斯盆地北部古生界气藏气井产能评价[J].成都理工大学学报(自然科学版),2005(1):54-57.
[5]匡建超,徐国盛,王允诚等.致密碎屑岩储层裂缝和产能预测的单井建模[J].矿物岩石,2001(2):62-67.
[6]匡建超,王允诚,胡远来等.致密碎屑岩储层裂缝和产能的横向识别[J].矿物岩石,2001(4):75-78.
[7]乐友喜,袁全社.支持向量机方法在油气储层参数预测中的应用[J].天然气工业,2005(12):45-47.
[8]Vapnik V N.The nature of statistical learning theory[M].NewYork:Springer2 Verlag,2000:193-245.
[9]张彦周,刘叶玲,谢宝英.支持向量机在储层厚度预测中的应用[J].勘探地球物理进展,2005(6):423-424,431.
[10]毛志强,李进福.油气层产能预测方法及模型[J].石油学报,2000(5):58-61.
[2]王昀,张津林.基于地震资料的储层产能预测方法[J].石油天然气学报,2007(3):87-89.
[3]杨志明,刘广利.不确定性支持向量机原理及应用[M].北京:科学出版社,2007.
[4]向丹,施泽进,黄大志.鄂尔多斯盆地北部古生界气藏气井产能评价[J].成都理工大学学报(自然科学版),2005(1):54-57.
[5]匡建超,徐国盛,王允诚等.致密碎屑岩储层裂缝和产能预测的单井建模[J].矿物岩石,2001(2):62-67.
[6]匡建超,王允诚,胡远来等.致密碎屑岩储层裂缝和产能的横向识别[J].矿物岩石,2001(4):75-78.
[7]乐友喜,袁全社.支持向量机方法在油气储层参数预测中的应用[J].天然气工业,2005(12):45-47.
[8]Vapnik V N.The nature of statistical learning theory[M].NewYork:Springer2 Verlag,2000:193-245.
[9]张彦周,刘叶玲,谢宝英.支持向量机在储层厚度预测中的应用[J].勘探地球物理进展,2005(6):423-424,431.
[10]毛志强,李进福.油气层产能预测方法及模型[J].石油学报,2000(5):58-61.
版权所有：© 2023 中国地质图书馆 中国地质调查局地学文献中心