多属性融合在陵水凹陷烃源岩研究中的应用  

Source-Rock Characterization with Multi-Attribute Fusion in Lingshui Sag

作　　者：刘仕友 徐冲[2] 孙万元 邢军辉[2] 徐晓宇 Liu Shiyou;Xu Chong;Sun Wanyuan;Xing Junhui;Xu Xiaoyu(CNOOC (China) Zhanjiang Branch, Zhanjiang, Guangdong 524057, China;Ocean University of China, Qingdao, Shandong 266100, China)

机构地区：[1]中海石油(中国)有限公司湛江分公司,广东湛江524057 [2]中国海洋大学,山东青岛266100

出　　处：《特种油气藏》2019年第2期23-27,共5页Special Oil & Gas Reservoirs

基　　金：国家科技重大专项"海洋深水区油气勘探关键技术"(2016ZX05026);国家自然科学基金"南海西北部盆地构造沉积特征对青藏高原隆升的响应"(41530963)

摘　　要：针对深水区钻井资料少,中深层地震资料品质差,常规手段无法有效进行烃源岩总有机碳含量(TOC)预测的问题,运用ΔlogR法求取测井TOC曲线,运用相关性分析进行地震属性优选,运用多元线性回归和概率类神经网络2种方法进行多属性融合预测TOC,以陵水凹陷为例建立了一套适用于深水区少井条件下的多属性融合预测烃源岩TOC的技术流程。研究结果体现了概率类神经网络融合预测TOC的优越性,该研究对深水区少井条件下的烃源岩预测具有一定的借鉴意义。Due to the lack of drilling data in deep-water area and the relatively low-quality of seismic data in medium-deep strata, traditional methods are limited in the prediction of source-rock total organic carbon content ( TOC ). The logging TOC curve was obtained by Δlog R method and the correlation analysis was adopted to optimize the seismic attributes. Both the multiple linear regression and the probabilistic neural network were applied to predict TOC with multi-attribute fusion. A technical flow of source-rock TOC prediction with multi-attribute fusion for deep-water and less well conditions was established by taking Lingshui Sag as an example. Result shows that the correlation coefficient of source-rock TOC prediction with probabilistic neural network fusion is up to 90%, which is more superior than that of multiple linear regression fusion method. This research could provide certain reference for the source-rock characterization under less well in deep-water area.

关 键 词：烃源岩 总有机碳含量 多属性融合 概率类神经网络 陵水凹陷 

分 类 号：TE122[石油与天然气工程—油气勘探]