基于机器学习的多地震属性沉积相分析
|
摘要
为研究苏里格气田的沉积环境及沉积相展布规律,以苏里格气田召30区块为研究对象,结合对沉积相较为敏感的均方根振幅、平均瞬时频率和有效带宽3种地震属性,同时利用研究区丰富的水平井资料,运用机器学习中的半监督模糊C均值方法,得到召30区块盒8段沉积相展布特征。结果表明,相比传统的模糊C均值方法,该方法能够清晰地刻画盒8段南北向条带状分布的4条河道,并且忠实于测井信息,预测结果更符合先验地质认识,并且改善了地质人员在无井区域的沉积认识,可为同类区块储层预测方面提供一定的借鉴。
The Block Zhao30 is taken to analyze the depositional environment and the distribution pattern of sedimentary facies in the Sulige Gasfield.Three seismic attributes that are sensitive to sedimentary facies are taken into consideration in this analysis,including root mean square amplitude,average instantaneous frequency and effective bandwidth.The semi-supervised fuzzy C-means method in machine learning is used to gain the distribution pattern of sedimentary facies in He8 segment of Block Zhao30 by combining with the multiple horizontal well data in this region.Research indicates that comparing with the interpretation of traditional fuzzy C-means method,this new method could clearly characterize the four river channels with north-south stripe distribution in the He8 segment.The analysis shows a perfect agreement with logging interpretation and prior geological understanding,which significantly improves the geological understanding of sedimentary facies in free area.This research could provide certain reference for the reservoir prediction in similar blocks.
The Block Zhao30 is taken to analyze the depositional environment and the distribution pattern of sedimentary facies in the Sulige Gasfield.Three seismic attributes that are sensitive to sedimentary facies are taken into consideration in this analysis,including root mean square amplitude,average instantaneous frequency and effective bandwidth.The semi-supervised fuzzy C-means method in machine learning is used to gain the distribution pattern of sedimentary facies in He8 segment of Block Zhao30 by combining with the multiple horizontal well data in this region.Research indicates that comparing with the interpretation of traditional fuzzy C-means method,this new method could clearly characterize the four river channels with north-south stripe distribution in the He8 segment.The analysis shows a perfect agreement with logging interpretation and prior geological understanding,which significantly improves the geological understanding of sedimentary facies in free area.This research could provide certain reference for the reservoir prediction in similar blocks.
引文
[1]袁照威,陈龙,高世臣,等.基于马尔科夫-贝叶斯模拟算法的多地震属性沉积相建模方法——以苏里格气田苏10区块为例[J].油气地质与采收率,2017,24(3):37-43.
[2]赖生华,柳伟明,赵永刚,等.利用地震信息研究沉积体系平面分布特征:以鄂尔多斯盆地延长探区高家河三维区盒八段为例[J].石油实验地质,2017,39(2):272-277.
[3]周志华.机器学习[M].北京:清华大学出版社,2016:293-317.
[4]ANTOINE C,LORENZO P,ERWAN G,et al.Machine learning as a tool for geologists[J].The Leading Edge,2017,36(3):215-219.
[5]郭继刚,郭凯,宫鹏骐,等.鄂尔多斯盆地延长组储层致密化及其影响下的致密油充注特征[J].石油实验地质,2017,39(2):169-179.
[6]唐俊,王琪,马晓峰,等.Q型聚类分析和判别分析法在储层评价中的应用[J].特种油气藏,2012,19(6):28-31.
[7]唐衔,侯加根,邓强,等.基于模糊C均值聚类的流动单元划分方法——以克拉玛依油田五3中区克下组为例[J].油气地质与采收率,2009,16(4):34-37.
[8]MAJID B,MOHAMMADAR.Seismicfacies analysis from well logs based on supervised classification scheme with different machine learning techniques[J].Arabian Journal of Geoscience,2015,8(9):7153-7161
[9]BRENDON H.Facies classification using machine learning[J].The Leading Edge,2016,35(10):906-909.
[10]张阳,邱隆伟,李际,等.基于模糊C均值地震属性聚类的地震相分析[J].中国石油大学学报(自然科学版),2015,39(4):53-61.
[11]FORESTIER G,WEMMERT C.Semi-supervised learning using multiple clusterings with limited labeled data[J].Information Sciences,2016,361-362(C):48-65.
[12]高世臣,袁照威.地震属性在沉积相预测中的方法研究——序贯随机模式识别[J].地球物理学进展,2016,31(3):1066-1072.
[13]郭智,贾爱林,何东博,等.鄂尔多斯盆地苏里格气田辫状河体系带特征[J].石油与天然气地质,2016,37(2):197-204.
[14]袁照威,强小龙,高世臣,等.苏里格气田不同沉积相建模方法及空间结构特征评价[J].特种油气藏,2017,24(1):77-83.
[15]QI J,LIN TF,ZHAO T,et al.Semi-supervisedmultiattribute seismic faciesanalysis[J].Interpretation,2016,4(1):SB91-SB106.
[2]赖生华,柳伟明,赵永刚,等.利用地震信息研究沉积体系平面分布特征:以鄂尔多斯盆地延长探区高家河三维区盒八段为例[J].石油实验地质,2017,39(2):272-277.
[3]周志华.机器学习[M].北京:清华大学出版社,2016:293-317.
[4]ANTOINE C,LORENZO P,ERWAN G,et al.Machine learning as a tool for geologists[J].The Leading Edge,2017,36(3):215-219.
[5]郭继刚,郭凯,宫鹏骐,等.鄂尔多斯盆地延长组储层致密化及其影响下的致密油充注特征[J].石油实验地质,2017,39(2):169-179.
[6]唐俊,王琪,马晓峰,等.Q型聚类分析和判别分析法在储层评价中的应用[J].特种油气藏,2012,19(6):28-31.
[7]唐衔,侯加根,邓强,等.基于模糊C均值聚类的流动单元划分方法——以克拉玛依油田五3中区克下组为例[J].油气地质与采收率,2009,16(4):34-37.
[8]MAJID B,MOHAMMADAR.Seismicfacies analysis from well logs based on supervised classification scheme with different machine learning techniques[J].Arabian Journal of Geoscience,2015,8(9):7153-7161
[9]BRENDON H.Facies classification using machine learning[J].The Leading Edge,2016,35(10):906-909.
[10]张阳,邱隆伟,李际,等.基于模糊C均值地震属性聚类的地震相分析[J].中国石油大学学报(自然科学版),2015,39(4):53-61.
[11]FORESTIER G,WEMMERT C.Semi-supervised learning using multiple clusterings with limited labeled data[J].Information Sciences,2016,361-362(C):48-65.
[12]高世臣,袁照威.地震属性在沉积相预测中的方法研究——序贯随机模式识别[J].地球物理学进展,2016,31(3):1066-1072.
[13]郭智,贾爱林,何东博,等.鄂尔多斯盆地苏里格气田辫状河体系带特征[J].石油与天然气地质,2016,37(2):197-204.
[14]袁照威,强小龙,高世臣,等.苏里格气田不同沉积相建模方法及空间结构特征评价[J].特种油气藏,2017,24(1):77-83.
[15]QI J,LIN TF,ZHAO T,et al.Semi-supervisedmultiattribute seismic faciesanalysis[J].Interpretation,2016,4(1):SB91-SB106.