渤海海域井壁取心裂解烃S_(2)烃类损失恢复回归分析  

Regression analysis of hydrocarbon loss restoration of cracking hydrocarbon S2 from sidewall coring in the Bohai Sea

作　　者：李阳[1] 郭明宇 倪鹏勃 李鸿儒 符强 黄子舰 LI Yang;GUO Mingyu;NI Pengbo;LI Hongru;FU Qiang;HUANG Zijian(Exploration and Development Research Institute,Liaohe Oilfield of CNPC,Panjin,Liaoning 124010,China;Tianjin Branch of CNOOC(China)Co.,Ltd.,Tianjin 300452,China;China French Bohai Geoservices Co.,Ltd.,Tianjin 300457,China;Panjin Zhonglu Oil&Gas Technology Service Co.,Ltd.,Panjin,Liaoning 124010,China)

机构地区：[1]中国石油辽河油田分公司勘探开发研究院 [2]中海石油(中国)有限公司天津分公司 [3]中法渤海地质服务有限公司 [4]盘锦中录油气技术服务有限公司

出　　处：《录井工程》2024年第2期49-56,共8页Mud Logging Engineering

基　　金：中海石油(中国)有限公司重大科技专项“渤海油田上产4000万吨新领域勘探关键技术”(编号:CNOOC-KJ 135 ZDXM36 TJ 08 TJ)。

摘　　要：地化录井受工程、地质条件及人为因素的影响,往往造成岩石样品从井底到地表的烃类损失,不能很好地反映地下储层的真实含油气信息,因此需要一种合理准确的方法进行烃类损失恢复。针对渤海海域不同地区不同层位的岩屑值(自变量)与壁心值(因变量)之间的关系,基于最小二乘法、梯度下降法及其衍生算法,以多元线性回归和非线性回归两种方式来拟合研究区井壁取心数据。多元线性回归模型可使用标准方程法、岭回归、LASSO(Least Absolute Shrinkage and Selection Operator)及弹性网进行回归拟合,非线性回归模型可使用梯度下降法和分段函数的拟合方法。对不同回归分析方法进行分析对比可知,岭回归在计算线性关系的烃类损失方面具有较好的效果,决定系数r^(2)均超过0.7;基于岭回归分段函数拟合和非线性回归模型y=x/(b+kx)适合非线性烃类损失恢复。与传统的烃类损失恢复方法相比,使用量化的方式对研究区烃类进行恢复,更加科学全面,具有广泛的应用前景。Geochemical logging is often affected by engineering,geological conditions,and human factors,resulting in hydrocarbon losses of rock samples from the bottom of the well to the surface.It fails to reflect the real oil and gas content information of underground reservoirs well,and a reasonable and accurate method is needed to restore hydrocarbon losses.In view of the relationship between cuttings values(independent variables)and wall core values(dependent variables)at different horizons in different areas of the Bohai Sea,based on least square method,gradient descent method and its derivative algorithm,the paper fits the data of sidewall coring in the study area using two methods:multiple linear regression and non⁃linear regression.The model of multiple linear regression can be fitted by regression using normal equation,ridge regression,LASSO(Least Absolute Shrinkage and Selection Operator)and elastic net.The non⁃linear regression model can use the gradient descent method and the piecewise function fitting method.The analysis and comparison of different regression analysis methods show that ridge regression has good results in calculating the hydrocarbon losses of linear relationship,with the determination coefficient r^(2) exceeding 0.7.Piecewise function fitting based on ridge regression and nonlinear regression model y=x/(b+kx)are suitable for non⁃linear hydrocarbon loss restoration.Compared with the traditional method of hydrocarbon loss restoration,using the quantitative method to restore hydrocarbons in the study area is more scientific and comprehensive,and has a broad prospect of application.

关 键 词：烃类损失恢复 裂解烃 回归分析 多元线性回归模型 非线性回归模型 井壁取心 

分 类 号：TE132.1[石油与天然气工程—油气勘探]