北京大学学报自然科学版 ›› 2020, Vol. 56 ›› Issue (3): 406-416.DOI: 10.13209/j.0479-8023.2020.018

上一篇    下一篇

基于波形聚类分析的微地震监测事件类型判别及应用

翟尚1, 喻志超1, 谭玉阳2, 黄芳飞3, 刘玲3, 胡天跃1,†, 何川1,†   

  1. 1. 北京大学地球与空间科学学院, 北京大学石油与天然气研究中心, 北京 100871 2. 中国科学技术大学地球和空间科学学院, 合肥 230026 3. 中国地质调查局广州海洋地质调查局, 广州 510760
  • 收稿日期:2019-05-08 修回日期:2019-07-08 出版日期:2020-05-20 发布日期:2020-05-20
  • 通讯作者: 胡天跃, E-mail: tianyue(at)pku.edu.cn; 何川, E-mail: chuanhe_pku(at)163.com
  • 基金资助:
    中国地质调查局天然气水合物专项(DD20190232-6)和国家重点研发计划(2017YFC0307605, 2017YFC0307702)资助

Microseismic Monitoring Events Classification Based on Waveform Clustering Analysis and Application

ZHAI Shang1, YU Zhichao1, TAN Yuyang2, HUANG Fangfei3, LIU Ling3, HU Tianyue1,†, HE Chuan1,†   

  1. 1. Institute of Oil & Gas, School of Earth and Space Sciences, Peking University, Beijing 100871 2. School of Earth and Space Sciences, University of Science and Technology of China, Hefei 230026 3. Guangzhou Marine Geological Survey, China Geological Survey, Guangzhou 510760
  • Received:2019-05-08 Revised:2019-07-08 Online:2020-05-20 Published:2020-05-20
  • Contact: HU Tianyue, E-mail: tianyue(at)pku.edu.cn; HE Chuan, E-mail: chuanhe_pku(at)163.com

摘要:

以不同类型微地震监测事件在波形相似性上的差异为基础, 结合发生位置、走时规律和偏振方向等方面的特征, 提出一种基于波形聚类分析的微地震监测事件类型判别方法。首先使用常规的微地震事件识别算法, 快速地得到待分类的疑似事件; 然后进行波形聚类分析, 结合事件的属性特征, 实现对不同类型微地震事件及噪声事件的分类和判别。分类结果可用于波形模板匹配, 识别同类的低信噪比微地震事件; 还可将所有同类事件作为一个整体, 采用全局优化手段提高初至拾取的精度。

关键词: 层次聚类, 属性提取, 波形互相关, 微地震事件, 层次聚类, 属性提取

Abstract:

Based on the difference of waveform similarity between different types of microseismic monitoring events and combined with their characteristics in occurrence location, traveling time and polarization direction etc., a method for classifying microseismic monitoring events based on waveform clustering analysis is proposed. Firstly unclassified events can be identified rapidly using conventional microseismic event detection methods, then similar events are grouped based on waveform clustering analysis, finally the types of microseismic events or noise events are determined combining the attribute characteristics. Classified microseismic events can be further used for template matching technique to finely detect similar events with low signal-to-noise ratio. Meanwhile the global optimization approach which aims to improve the accuracy of arrival time picking can be also performed by taking similar microseismic events as a whole. 

Key words: feature extraction, waveform cross correlation, microseismic event, hierarchal clustering, feature extraction