中国石油勘探 ›› 2009, Vol. 14 ›› Issue (2): 66-69,86,8.

• • 上一篇    下一篇

多参数优化在模拟退火法波阻抗反演中的应用

赵一民,吴亚东,王振国,徐永梅   

  1. 中国石油勘探开发研究院;中国地质大学(北京);中国石油大港油田分公司
  • 出版日期:2009-04-15 发布日期:2009-04-15

Simulated AnnealingWave Impedance Inversion throughMulti-Parameter Optimization

Zhao Yimin,Wu Yadong,Wang Zhenguo and Xu Yongmei   

  1. Zhao Yimin1,2,Wu Yadong1,Wang Zhenguo1,Xu Yongmei3//1.Research Institute of Petroleum Exploration & Development,PetroChina,Beijing 100083;2.China University of Geosciences,Beijing 100083;3.PetroChina Dagang Oilfield Company,Tianjin 300000
  • Online:2009-04-15 Published:2009-04-15

摘要: 多参数优化的模拟退火法波阻抗反演是通过地震资料信噪比、相邻道的相关性、初始模型偏差、反射系数门槛值等约束条件,采用能避免目标函数值陷入局部极值区而获得全局最优解的模拟退火法来实现的。介绍了方法的基本原理,给出了它在实际地震资料处理中的实现步骤,即:高分辨率处理、合成记录及层位标定、子波提取、初始模型建立、模拟退火反演和多参数约束条件的调整。实际资料应用表明:与常规波阻抗反演相比,该方法可以提高地震资料反演的分辨率;原始地震数据相干体切片和波阻抗相干体切片的断层位置均与构造图的断层位置一致,但与原始地震数据相干体切片相比,波阻抗相干体切片可为解释断层提供较好的依据。

关键词: 模拟退火, 波阻抗反演, 多参数优化, 提取子波, 初始模型

Abstract: Simulated annealing of multi-parameter optimization is the simulated annealing wave impedance inversion with some constraints, including S/N ratio, adjacent-channel correlation, deviation of initial model, and threshold value of reflection coefficient. The method can avoid the objective function value into a local extremum and has global optimum solutions. The principles of this method are introduced in this paper. The procedures are high-resolution processing, synthetic seismogram and horizon calibration, wavelet extraction, initial model establishment, simulated annealing inversion, and adjustment of multi-parameter restraints. The actual data processing indicates that: 1) compared to conventional wave impedance inversion, the method can improve the resolution of seismic data inversion; 2) the fault position in coherent slices of original data and coherent slices of wave impedance are coincide with that in structure maps, but the latter can provide a better criteria for fault interpretation than the former.

Key words: simulated annealing, wave impedance inversion, multi-parameter optimization, wavelet extraction, initial model;

中图分类号: