Title

Supervised learning and unsupervised learning for hydrocarbon prediction using multiwave seismic data

单位

(1.山东省沉积成矿作用与沉积矿产重点实验室,山东科技大学地球科学与工程学院,山东青岛266590;2.山东大学岩土与结构工程研究中心,山东济南250061;3.海底科学与探测技术教育部重点实验室,中国海洋大学地球科学学院,山东青岛266100;4.山东正元建设工程有限责任公司,山东济南250101;5.山东科瑞机械制造有限公司,山东东营257000;6.山东省煤田地质局,山东济南250104)

Organization

(1.Shandong Provincial Key Laboratory of Depositional Mineralization & Sedimentary Minerals,College of Geological Sciences & Engineering,Shandong University of Science and Technology,Qingdao 266590,China;2.Geotechnical and Structural Engineering Research Center,Shandong University,Jinan 250061,China;3.Key Lab of Submarine Geosciences and Prospecting Techniques,MOE and College of Marine Geosciences,Ocean University of China,Qingdao 266100,China;4.Shandong Zhengyuan Construction Engineering,Jinan 250101,China;5.Kerui Machinery Manufacturing Co.Ltd.,Dongying 257000,China;6.Shandong Bureau of Coal Geology,Jinan 250104,China)

摘要

利用地震数据直接预测含油气储层分布情况是油气地震勘探的终极目标之一。含油气储层对纵、横波的敏感度存在差异,这种差异突出了含油气储层的地震特征。鉴于此,发展了一种基于无监督与监督学习下的多波地震油气储层分布预测方法。首先,利用不同卷积核卷积升维形成各类纵、横波地震属性,然后,利用聚类分析法进行无监督学习,通过聚类分析分别对纵、横波地震属性降维,再利用聚合法求取能突出油气储层特征的多波地震聚合属性,最后以降维后的聚合属性作为支持向量机的学习集进行含油气储层地震预测。实际应用结果表明,所预测的含油气储层边界更加清晰,与实际情况基本吻合。

Abstract

For oil and gas seismic exploration,predicting the distribution of oil and gas reservoirs with seismic data is the primary objective.It is helpful to highlight the characteristics of oil and gas reservoirs using the sensitivity difference between the P-wave and PS-wave.A method to predict the distribution of oil and gas reservoirs based on unsupervised and supervised learning is proposed.First,convolution using different convolved nuclei is conducted on P-waves and PS-waves to generate a variety of seismic attributes.Then,unsupervised learning is carried out using the cluster-analysis method to reduce the dimensions of the seismic attributes of P- and PS- waves.That is,the aggregation attributes from multiwave seismic data,which can be used to extract the characteristics of oil and gas reservoirs,is calculated using the polymerization method.Finally,using the aggregation attributes with the reduced dimension as the learning set for a support-vector machine,supervised learning is performed for oil and gas reservoir prediction.The application results show that the predicted results were in agreement with the actual situation and had a clearer boundary of the seismic responses from the hydrocarbon reservoir.

关键词：

多波地震; 地震属性; 数据挖掘; 无监督学习; 监督学习; 支持向量机; 深度学习; 储层预测;

Keywords:

multiwave seismic data,; seismic attribute,; data mining,unsupervised learning,; supervised learning,; support vector machine,deep learning,seismic hydrocarbon prediction;

基金项目

国家高技术研究发展计划863项目(2013AA064201,2012AA061202)、国家自然科学基金项目(41174098,41374126)共同资助。

DOI

10.3969/j.issn.1000-1441.2018.04.015

无监督与监督学习下的含油气储层预测