利用残差网络和地震发射层析成像的微地震事件检测  

Detection of microseismic events based on residual network and seismic emission tomography

作　　者：王维波[1] 官强 高明[1] 盛立[1] WANG Weibo;GUAN Qiang;GAO Ming;SHENG Li(College of Control Science and Engineering,China University of Petroleum(East China),Qingdao,Shandong 266580,China)

机构地区：[1]中国石油大学(华东)控制科学与工程学院,山东青岛266580

出　　处：《石油地球物理勘探》2022年第2期251-260,I0001,共11页Oil Geophysical Prospecting

基　　金：国家自然科学基金项目“旋转导向钻井工具精确轨迹跟踪的智能自主容错控制系统研究”(62033008);山东省自然科学基金项目“网络化随机系统的控制、状态估计与故障诊断”(ZR2020YQ49)联合资助。

摘　　要：地震发射层析成像(Seismic emission tomography,SET)是一种适用于地面微地震监测的震源定位方法,该方法利用地面众多站点监测的信号对储层特定范围分层成像,通过图像判定微地震事件并确定震源坐标。传统处理方法通常采用人工看图判断一段信号的SET是否包含有效微地震事件。然而人工判别方法难以完成对海量监测数据的全部处理,无法充分发挥SET方法的优势。针对此问题,采用残差网络对微地震监测数据的SET数据进行处理,实现微地震事件自动识别。首先,利用合成数据和实际油井的水力压裂地面微地震监测数据进行SET,构建SET图像样本数据集;然后对残差网络进行训练和测试,得到事件识别准确率最高的残差网络模型;再使用训练好的残差网络对不同信噪比的合成信号,以及多口油气井的地面微地震监测信号的SET数据进行事件检测。测试结果表明,基于残差网络和SET的方法能够有效检测微地震事件,且具有较强的抗噪能力和泛化能力。Seismic emission tomography(SET)is a hypocentral location method suitable for surface microseismic monitoring.This method makes use of the monitoring signals from many stations on the ground to image a specific area of a reservoir layer by layer.The images are used to determine whether there are microseismic events and where the hypocenter coordinates are.In traditional processing methods,whether the SET images of a section of signal contain an effective microseismic event is usually judged by human experiences.It is difficult to process all the massive monitoring data manually,and thus it cannot make full use of the advantages of the SET method.To solve this problem,a residual network is proposed to process SET images of microseismic monitoring data,which can detect microseismic events automatically.Firstly,a large number of SET images are produced using synthetic data and actual surface microseismic monitoring data of an oil well with hydraulic fracturing,and these SET images are used to construct a sample data set for training and testing a residual network.In this way,a residual network model with the highest accuracy of event detection is obtained.Then,the trained residual network model is employed to detect and locate microseismic events from other SET images produced by synthetic signals with different signal-to-noise ratios and surface microseismic monitoring data of other oil&gas wells with hydraulic fracturing.The test results prove that the proposed method can detect microseismic events effectively and has good noise suppression and generalization abilities.

关 键 词：微地震 事件检测 震源定位 地震发射层析成像 残差网络 模型训练 

分 类 号：P631[天文地球—地质矿产勘探]