基于属性优选的气烟囱型气云识别方法在渤海东部地区的应用  

Application of gas chimney-type gas cloud recognition method based on attribute optimization in the eastern Bohai Bay Basin

作　　者：姚永强 熊煜 高磊 蒋志恒 贺电波 YAO Yongqiang;XIONG Yu;GAO Lei;JIANG Zhiheng;HE Dianbo(Tianjin Branch of CNOOC(China)Ltd.,Tianjin 300452,China)

机构地区：[1]中海石油(中国)有限公司天津分公司,天津300452

出　　处：《海洋地质前沿》2022年第9期86-91,共6页Marine Geology Frontiers

基　　金：中海油“七年行动计划”重大科技专项课题“渤海油田上产4000万吨新领域勘探关键技术”(CNOOC-KJ 135 ZDXM 36 TJ 08 TJ)。

摘　　要：渤海地区气云表现形式多样,目前还没有系统的分类。根据不同气云带的分布范围,将渤海地区的气云从深至浅划分为气烟囱型、亮点型和麻坑型3类,并分析其成因机理。针对气烟囱型气云,利用单一地震属性进行气云识别和预测,具有一定局限性,且神经网络方法就是一个“黑匣子”,无法判断属性在计算过程中发挥的作用。文中提出利用具有“多属性融合神经网络”技术体系,对不同的属性组合进行分类,突出对气云敏感的属性,从而对气云准确识别,精细刻画气云的空间分布范围。该方法在渤东地区蓬莱A油田取到较好的应用效果,为下一步寻找大中型油气藏提供了依据。Gas cloud in the eastern Bohai Bay Basin shows various forms and there is no systematic classification at present.According to the distribution of different gas clouds in vertical direction,gas cloud of Bohai Sea could be classified into three types from deep to shallow:gas chimney type,bright spot type,and flax pit type;and their formation mechanism were analyzed.For the gas chimney type,conventional methods use a single seismic attribute to identify and predict the gas cloud,which have certain limitations,and previous neural network method is a"black box",unable to judge the effect of attributes in the calculation process.We developed a"multi-attribute fusion neural network"technology system to classify different attribute combinations,highlight the sensitive attributes of gas cloud,to accurately identify the gas cloud and finely determine the spatial distribution range of gas cloud.This method has been successfully applied in Penglai A Oilfield in the eastern Bohai Bay Basin,which provides a basis for further exploration of large and medium-sized reservoirs.

关 键 词：渤海地区 气云带分类 多属性融合 神经网络 

分 类 号：P744.4[天文地球—海洋科学]