Strong tolerance random forest algorithm in seismic reservoir prediction
Song Jianguo1,2, Yang Lu1, Gao Qiangshan3, Liu Jong4
1. School of Geosciences, China University of Petroleum(East China), Qingdao, Shandong 266580, China; 2. Laboratory of Marine Mineral Resource Evaluation and Detection, Qingdao National Laboratory for Marine Science and Technology, Qingdao, Shandong 266071, China; 3. Lunar and Planetary Science Research Center, Institute of Geochemistry, Chinese Academy, Gui-yang, Guizhou 550081, China; 4. Research Institute of Petroleum Exploration and Production, SINOPEC, Beijing 100083, China
Abstract:Since it is difficult to effectively define noise and signal on seismic and logging data,seismic re-servoir prediction needs good noise tolerance algorithms.The random forest (RF) algorithm with strong noise tolerance is proved by adding noise to training samples.However this does not mean RF has good noise tolerance in seismic reservoir prediction as well.First we extract noise samples from original seismic data with strong noise in the Survey F3,and extract denoised samples from the data processed by the dip-steered median filter.Then we establish random forest regression models between seismic attributes and the porosity parameter.After processing the original seismic data and the filtered seismic data with the noise sample model and denosied sample model,we estimate four different porosity parameter cubic data.The results reveal that the two data sets obtained with the noise model are more disturbed by noise,and the other two data sets obtained with the denoised model are much less affected by noise.On these data sets,reservoir geological characteristics can be effectively characterized which proves the random forest model has strong robustness and perfect tolerance to abnormal data differing from the sample data.The key issue in the application of the random forest algorithm to seismic reservoir prediction is that the training data does not contain noise.In other words,the input variable of sample data being denoised is much more significant,whereas whether seismic data were denoised or not has less effects on the prediction result.
李伟,岳大力,胡光义等.分频段地震属性优选及砂体预测方法——秦皇岛32-6油田北区实例.石油地球物理勘探,2017,52(1):121-130.Li Wei,Yue Dali,Hu Guangyi et al.Frequence segmented seismic attribute optimization and sandbody distribution prediction:an examples in North Block,Qinhuangdao 32-6 Oilfield.OGP,2017,52(1):121-130.
[3]
赵学松,高强山,唐传章等.基于支持向量回归机与井导向的三角洲岩性油气藏储层参数预测.石油地球物理勘探,2016,51(5):976-982.Zhao Xuesong,Gao Qiangshan,Tang Chuanzhang et al.Delta stratigraphic reservoir parameter estimation based on support vector regression machine and well logging data.OGP,2016,51(5):976-982.
[4]
乐友喜,杨丽.储层地震预测基础理论方法研究.天然气地球科学,2009,20(4):563-570.Yue Youxi,Yang Li.Basic theories and methods of seismic reservoir prediction. Natural Gas Geosciences,2009,20(4):563-570.
[5]
印兴耀,周静毅.地震属性优化方法综述.石油地球物理勘探,2005,40(4):482-489.Yin Xingyao,Zhou Jingyi.Summary of optimum methods of seismic attributes.OGP,2005,40(4):482-489.
[6]
赵小辉,于宝利,曹小璐.属性融合技术在微小断裂识别中的应用.石油地球物理勘探,2017,52(增刊2):164-169.Zhao Xiaohui,Yu Baoli and Cao Xiaolu.Minor fault identification with seismic multi-attribute fusion.OGP,2017,52(S2):164-169.
[7]
杨平,孙赞东,许辉群等.叠后储层地震滤波质量控制方法.地球物理学进展,2014,29(1):234-239.Yang Ping,Sun Zandong,Xu Huiqun et al.Quality control of reservoir seismology filtering method.Progress in Geophysics,2014,29(1):234-239.
Breiman L.Consistency for a Simple Model of Ran-dom Forests.Technical Report 670,UC Berkeley,CA,2004,URL www.stat.berkeley.edu/breiman/.
[11]
方匡南.随机森林组合预测理论及其在金融中的应用.福建厦门:厦门大学出版社,2012.
[12]
Dietterich T.An experimental comparison of three methods for constructing ensembles of decision trees:bagging,boosting and randomization.Machine Learning,2000,40(2):139-157.
Biau G.Analysis of a random forests model.Journal of Machine Learning Research,2012,13(2):1063-1095.
[15]
Prasad A M,Iverson L R,Liaw A.Newer classification and regression tree techniques:Bagging and random forests for ecological prediction.Ecosystems,2006,9(2):181-199.
[16]
宋建国,高强山,李哲.随机森林回归在地震储层预测中的应用.石油地球物理勘探,2016,51(6):1202-1211.Song Jianguo,Gao Qiangshan and Li Zhe.Application of random forests for regression to seismic reservoir prediction.OGP,2016,51(6):1202-1211
[17]
王志宏,韩璐,戚磊.随机森林分类方法在储层岩性识别中的应用.辽宁工程大学学报:自然科学版,2015,34(9):1083-1088.Wang Zhihong,Han Lu,Qi Lei.Random forests classification method in the application of reservoir litho-logy recognition.Journal of Liaoning Technical University:Natural Science,2015,34(9):1083-1088.
[18]
赵保成,谭志详,邓喀中.利用随机森林回归模型计算主要影响角正切.金属矿山,2016,45(3):172-175.Zhao Baocheng,Tan Zhixiang,Deng Kazhong.Calculation of the tangent of major influence angle based on random forest regression model.Metal Mine,2016,45(3):172-175.
[19]
温廷新,张波,邵良杉.煤与瓦斯突出预测的随机森林模型.计算机工程与应用,2014,50(10):233-237.Wen Tingxin,Zhang Bo,Shao Liangshan.Prediction of coal and gas outburst based on random forest model.Computer Engineering and Application,2014,50(10):233-237.
[20]
Kuhn S,Cracknell M J,Reading A M.Lithological mapping via random forests:Information entropy as a proxyforinaccuracy.ASEGExtendedAbstracts2016:25th International Geophysical Conference and Exhibition,904-907.
[21]
程国建,周冠武,王潇潇.概率神经网络方法在岩性识别中的应用.微计算机信息,2007,23(6-1):288-289.Cheng Guojian,Zhou Guanwu,Wang Xiaoxiao.The probability neural networks for lithology identification.Microcomputer Information,2007,23(6-1):288-289.
[22]
宋延杰,张剑风,闫伟林等.基于支持向量机的复杂岩性测井识别方法.大庆石油学院报,2007,31(5):18-21.Song Yanjie,Zhang Jianfeng,Yan Weilin et al.A new identification method for complex lithology with support vector machine.Journal of Daqing Petroleum Institute,2007,31(5):18-21.
[23]
张长开,姜秀娣,朱振宇等.基于支持向量机的属性优选和储层预测.石油地球物理勘探,2012,47(2):282-285.Zhang Changkai,Jiang Xiudi,Zhu Zhenyu et al.Attributes selection and reservoir prediction based on support vector machine.OGP,2012,47(2):282-285.
Tingdahl K M,De Groot P F M.Post-stack dip-and azimuth processing.Journal of Seismic Exploration,2003,12(2):113-126.
[26]
赵丽娅,傅群,潘海滨.Opendtect软件倾角控制模块在地震解释中的应用.海洋地质动态,2008,24(4):33-37.Zhao Liya,Fu Qun,Pan Haibin.Application of Opendetct software dip-steering to seismic interpretation.Marine Geology Letters,2008,24(4):33-37.
[27]
Overeem I,Weltje G J,Bishop-Kay C et al.The Late Cenozoic Eridanos delta system in the Southern North Sea Basin:a climate signal in sediment supply?.Basin Research,2001,13(3):293-312.
[28]
Sørensen J C,Gregersen U,Breiner M et al.High frequency sequence stratigraphy of upper Cenozoic deposits.Marine Petroleum Geology,1997,14(2):99-123.