首页 | 本学科首页   官方微博 | 高级检索  
     

多元回归和神经网络在多影响因素下优选压裂候选井层中的应用
引用本文:曾凡辉,郭建春,赵金洲.多元回归和神经网络在多影响因素下优选压裂候选井层中的应用[J].石油工业计算机应用,2007,15(4):6-8.
作者姓名:曾凡辉  郭建春  赵金洲
作者单位:西南石油大学油气藏地质及开发工程国家重点实验室
摘    要:针对乌里雅斯太凹陷储层非均质性强、隔层遮挡性差、压裂井投产后效果相差悬殊,优选增产效果好的候选井层难的特点,本文提出了利用多元回归和神经网络优选待选压裂井层的方法。根据前期压裂井的有效资料,选择了对压裂效果影响较大的9个因素作为基本参数,建立了压裂井层的数据库。计算结果表明:多元线性回归不能满足优选压裂井层的需要;二次回归和神经网络方法能够满足选井选层的非线性问题,两者拟合误差均为0,预测误差平均值为0.57%和0.47%,能够满足工程的需要。

关 键 词:压裂  选井选层  多元回归  BP神经网络

APPLICATION OF MULTIPLE REGRESSION AND NEURAL NETWORK IN SELECTING A CANDIDATE ZONE TO BE FRACTURED UNDER MULTI-FACTOR EFFECT
Abstract:A method of optimizing a candidate zone to be fractured with multiple regression and neural network is put forward in the paper because there are some problems in Wuliyasi Sag, such as strong reservoir heterogeneity, poor interlayer shielding, big difference between fractured wells and it is hard to select good candidate wells.According to the old available data of fractured wells, 9 factors which have important effect on fracturing are selected as the basic parameters to establish a databank of fractured wells or zones.The calculated results show that quadric regression and neural network methods, which fitting errors are all 0 and the mean value of prediction errors are 0.57%and 0.47%, can meet the requirement of nonlinear problem in optimizing the fractured wells or zones instead of multielement linear regression.
Keywords:fracturing  selection of wells and zones  multiple regression  BP neural network
本文献已被 维普 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号