北京大学学报自然科学版 ›› 2021, Vol. 57 ›› Issue (3): 425-434.DOI: 10.13209/j.0479-8023.2021.034

上一篇    下一篇

基于大数据量的初至层析成像算法优化

吕雪梅1, 张献兵1,†, 康平2, 胡天跃1,†   

  1. 1. 北京大学地球与空间科学学院, 北京 1000871 2. 中国石油青海油田公司, 敦煌 736200
  • 收稿日期:2020-04-22 修回日期:2020-05-23 出版日期:2021-05-20 发布日期:2021-05-20
  • 通讯作者: 张献兵, E-mail: zxb(at)pku.edu.cn, 胡天跃, E-mail: tianyue(at)pku.edu.cn
  • 基金资助:
    国家重点研发计划(2018YFA0702503)、国家自然科学基金(41674122)和国家科技重大专项(2016ZX05004003)资助

Algorithm Optimization of First-Break Tomography Statics Based on Large Datasets

LÜ Xuemei1, ZHANG Xianbing1,†, KANG Ping2, HU Tianyue1,†   

  1. 1. School of Earth and Space Sciences, Peking University, Beijing 100871 2. Qinghai Oilfield, PetroChina, Dunhuang 736200
  • Received:2020-04-22 Revised:2020-05-23 Online:2021-05-20 Published:2021-05-20
  • Contact: ZHANG Xianbing, E-mail: zxb(at)pku.edu.cn, HU Tianyue, E-mail: tianyue(at)pku.edu.cn

摘要:

三维陆地地震资料采集技术向着高密度和宽方位的方向发展, 导致产生海量地震数据。初至旅行时层析成像的经典算法需要过多内存和计算时间, 不适合处理大的地震数据量。为解决这一问题, 提出一种基于优化数学公式的初至旅行时层析成像算法, 既减少Frechet矩阵和Hessian矩阵的存储对内存的需求, 也减少Hessian矩阵的求逆时长, 能快速和高效地解决大数据量的层析反演问题, 适合高密度和宽方位采集的陆地地震资料和精细参数化的网格模型, 并且不影响精度, 易于实现并行计算。模型试验和实际数据都验证了该方法的有效性, 当初至旅行时达到一定的数量时, 该方法能提供可靠的层析成像结果, 进行静校正。

关键词: 层析静校正, 近地表, 速度模型, 大数据量

Abstract:

The development of 3D land seismic data acquisition in the direction of wide azimuth and high density will lead to huge dataset. The classical first arrival time tomography algorithms are not suit for processing huge seismic dataset due to very high memory request and computing time cost. In order to solve this problem, the authors develop an optimal mathematical formula from the classical first break travel time tomography method to avoid the memory occupation that required by Frechet derivative matrix and Hessian matrix, and reduce the time cost of computing Hessian matrix inversion. This method can efficiently solve the tomography inversion problem for huge datasets. It is suitable for huge and high-density land seismic exploration, and not affect the dataset and model accuracy. It is easy for parallel processing. Both the model and real data examples confirm the effectiveness of this method. It can provide reliable tomographic results for static correction when the first breaks reaches a certain amount.

Key words: tomography statics, near surface, velocity model, huge datasets