下刚果盆地A区块Madingo组烃源岩TOC含量的地球物理定量预测  

Geophysical quantitative prediction of TOC content in source rocks of Madingo Formation in Block A,Lower Congo Basin

作　　者：季少聪 杨香华[1,2] 朱红涛[1,2] 邓运华[3] 康洪全[3] 王波[1,2] 

机构地区：[1]中国地质大学(武汉)资源学院,湖北武汉430074 [2]中国地质大学(武汉)构造与油气资源教育部重点实验室,湖北武汉430074 [3]中国海洋石油研究总院,北京100027

出　　处：《石油地球物理勘探》2018年第2期369-380,共12页Oil Geophysical Prospecting

基　　金：国家科技重大专项"西非和南美海域重点区油气地质评价及关键技术研究"(2017ZX05032-001)资助

摘　　要：TOC含量是烃源岩有机质丰度和生烃潜力评价的重要内容,由于受油基泥浆污染,下刚果盆地A区块可用于实测Madingo组烃源岩TOC含量的样品较少,且分布不均匀,难以进行定量评价。而地球物理资料蕴含着烃源岩的多种地球化学信息,可以有效地定量预测TOC含量。以研究区实测TOC含量和测井资料为基础,通过交会分析寻找与TOC相关性较好的测井参数,分别利用多元回归分析法、改进的ΔlgR法和BP神经网络法预测TOC值,并比较预测值与实测值的相关性,优选计算方法并进行单井TOC含量的测井定量预测。结合三维地震数据建立测井预测TOC含量与井旁道地震属性之间的神经网络模型,计算TOC数据体。结果表明:研究区与实测TOC相关性较好的测井参数包括密度、自然伽马和声波时差;BP神经网络法预测效果最好,单井预测TOC结果和实测TOC值相关系数高达0.9542;研究区TOC三维定量预测结果呈"垂向上层状分布,平面上北东低、南西高"的特征。TOC content is an important content of evaluation of organic matter abundance and hydrocarbon generation potential.There are few samples available for testing TOC content in source rocks of Madingo Formation in Block A,Lower Congo Basin due to oil-based mud pollution.And the distribution of these samples is nonuniform,so it is difficult to carry out quantitative evaluation.Geophysical data contain a variety of geochemical information in source rocks and can be used to quantitatively and effectively predict TOC content.Based on measured TOC content and logging data in study area,we find out some logging parameters with good correlations with TOC by cross-plot analysis.The predicted TOC content was calculated by multiple regression analysis,BP neural network and modifiedΔlgR.And the best algorithm for single well TOC content prediction based on logging data is selected.A neural network model near the well is established based on the relation between the predicted TOC content of single well and seismic attributes on 3 Dseismic data,and this model is used to calculate a TOC data volume.The results show that well logging parameters with good correlations with measured TOC include density,natural gamma and interval transit time.BP neural network method has the best prediction and the correlation coefficient between predicted TOC content and measured TOC content of single well is up to 0.9542.

关 键 词：TOC含量 BP神经网络 地震属性 定量预测 Madingo组 下刚果盆地 

分 类 号：P631[天文地球—地质矿产勘探]