基于概率统计方法的有限元正演模拟及CMP叠加处理技术研究

At present, China's western mountain is the most potential oil and gas exploration and development zone in our county. The complex surface condition and geological structure make the seismic record low signal-to-noise ratio, complex wavefields, and effective information difficult to distinguish, which cause the processing and interpretation with the great difficulty, thus we need to study the special seismic processing technology in the region , and a forward modeling to assist the processing and interpretation. However, at present, forward modeling record is basically pure theory record, without the complex random interference background just as that in the field seismic record and sometimes white noise is simply added that lacks correlation with the theoretical record, It can be eliminated after superposition,but the random interference background in real record can't be eliminated. Those forward modeling methods can not guide field actual data processing and interpretation to its utmost extent. Based on the seismic interference background with the nature of randomness, this project aims to study: (1) The forward modeling with random disturbance, the typical decomposition method as the foundation, combined with finite element forward modeling, in order to construct with complex random interference background of CMP gather. The random disturbance background has coherent characteristic with CMP gather. The forward modeling record using the high precision method is closer to the actual field record.(2)the CMP gather stack processing, based on the probability statistical properties of random disturbance background, according to the probability statistical criteria, so as to realize the optimal stack processing.

中国西部山地是我国目前油气田最具潜力的勘探发展区，该地区复杂的地表条件和地质结构使得地震记录信噪比低、面貌复杂、有效信息难以分辨，给处理和解释带来很大难度，因此需要研究该地区特殊的地震处理技术，借助正演模拟来辅助处理和解释。但是，目前正演模拟的记录基本上没有野外记录中那些复杂随机干扰背景。有时只是加上与理论记录没有相关性的白噪音，经过叠加后基本都能够消除，实际记录却不能消除，因此不能最大程度地指导野外实际资料的处理和解释。基于地震干扰背景具有随机性的这种情况，本项目拟研究：①以随机干扰典型分解方法为基础，结合有限元正演模拟及GPU/CPU协同并行计算技术，快速构建具有复杂随机干扰背景的共中心点道集（CMP道集）。随机干扰背景具有与CMP道集相干特性的面貌。利用该方法得到的高精度的正演模拟记录更接近野外实际记录。②以随机干扰背景的概率统计性质为基础，根据概率统计判断准则，实现最优叠加处理。

中国西部山地是我国目前油气田最具潜力的勘探发展区，该地区复杂的地表条件和地质结构使得地震记录信噪比低、面貌复杂、有效信息难以分辨，给处理和解释带来很大难度，需要以正演模拟来辅助处理和解释。但是，目前正演模拟的记录基本上都是纯理论记录，没有野外记录中那些复杂随机干扰背景。有时只是加上与理论记录没有相关性的白噪音，经过叠加后基本都能够消除，不能最大程度地指导野外实际资料的处理和解释。基于地震干扰背景具有随机性的这种情况，本项目主要研究了以下两个方面：①正演模拟方面，以随机干扰典型分解方法为基础，结合有限元正演模拟及GPU/CPU协同并行计算技术，快速构建具有复杂随机干扰背景的共中心点道集（CMP道集）。该项随机干扰背景具有与CMP道集相干特性有关的面貌。利用该方法得到的高精度的正演模拟记录更接近野外实际记录。②CMP道集叠加处理方面，以随机干扰背景的概率统计性质为基础，根据概率统计判断准则，实现最优叠加处理，通过四年的研究顺利完成了预定的任务。