模式识别技术预测火山岩相  

Volcanic facies prediction with pattern recognition

作　　者：郭淑文[1] 王振升[1] 牟智全[1] 徐晓星 邢兴[1] 彭雪梅[3] 

机构地区：[1]中国石油大港油田勘探开发研究院,天津300280 [2]渤海钻探工程有限公司泥浆技术服务分公司,天津300280 [3]中国石油大港油田勘探事业部,天津300280

出　　处：《石油地球物理勘探》2017年第A01期60-65,共6页Oil Geophysical Prospecting

摘　　要：东关潜山中生界火山岩岩性复杂,岩相变化大,地震反射特征与围岩相似,常规地震属性无法预测火山岩相带边界。通过地震地质模型正演,分析火山岩岩性及厚度变化对地震反射波形的影响。试验结果表明:随着火山岩岩性相带的变化,地震反射总体特征有变化,局部特征不明显;振幅变化受中生界顶界不整合面的影响较大;局部范围内振幅的变化与厚度变化相关;地震反射波负半周的反射特征与火山岩岩性及厚度变化相关性更强。根据模型分析结果,对东关潜山地震数据地震波负半周反射开时窗,利用GeoEast系统进行波形聚类及地震属性提取,综合应用钻井火山岩厚度数据及地震属性,采用GeoEast系统神经网络模式识别方法,对东关潜山中生界火山岩平面分布进行预测,有效提高了火山岩地震预测的精度。Dongguan buried hill Mesozoic volcanic rock lithology is very complex and seismic reflection characteristics are similar, so a single seismic attribute cannot be used to predict volcanic boundary. We analyze the influence of volcanic rock lithology and thickness changes on seismic reflection waveform by modeling, and obtain the following observations : a. The change of seismic reflection characteristics is not obvious as the change of volcanic facies; b. Mesozoic top boundary unconformity has great influence on amplitude; c. Amplitude variation is related to volcanic rock thickness in some locations; d. Negative half cycles of seismic waves are related to volcanic rock thickness. According to our analysis, the neural network pattern recognition provided by GeoEast is applied to predict the distribution of Dongguan volcanic rocks combining with the waveform clustering and seismic attributes within a negative half cycle window of seismic waves. The accuracy of volcanic lithology prediction is then effectively improved.

关 键 词：GeoEast系统 火山岩 模型正演 地震属性优选 神经网络 地震相 

分 类 号：P631[天文地球—地质矿产勘探]