基于稀疏表示的增维叠前地震反演方法

Multi-dimensional pre-stack seismic inversion based on sparse representation

常规的叠前和叠后反演技术难以满足刻画薄储层的需求。由于道集叠加而产生的干涉效应对地震信号所蕴含的真实信息造成不同程度的弱化、畸变甚至湮灭,因此为有效地保存真实信息,应尽量避免叠加。另外,叠前道集的地震属性在随炮检距变化之外还受频率因素的影响,而以高分辨率谱分解为基础展开的FAVO反演方法能够提升流体识别能力。为此,以不叠加为核心思想,提出基于稀疏表示的增维叠前地震反演方法。首先,从叠前道集出发,对包含目的层段的角度内道集进行筛选和提取,依托信号的稀疏表示理论对抽选的单角度地震数据开展高分辨率时频分解;其次,基于贝叶斯理论直接建立波阻抗与地震数据的映射关系,利用非线性最优化算法对初始模型进行扰动;最后,以前一个频率反演的结果作为后一个频率反演的约束,逐级进行反演,最终得到不同角度的高分辨率增维反演结果。以A研究区为例,在有效角度范围内对单角度地震数据分频约束反演,拓宽了数据维度,提升了反演精度,增强了对薄储层的描述能力。

It is difficult to characterize thin reservoirs through conventional pre-stack and post-stack inversion.Interference effects generated by stacking may fade,distort,and suppress effective signals;thus,we suggest avoiding stacking to preserve signals.Pre-stack attributes vary with offset and frequency;this means that FAVO inversion based on high-resolution spectral decomposition may be capable of addressing the issue of fluid detection.We present multi-dimensional pre-stack seismic inversion based on sparse representation,and the point is no stacking.The first step is to extract angle gathers at the zone of interest and then perform high-resolution time-frequency decomposition for each single angle gather extracted in terms of sparse representation.The second step is to establish the mapping relationship between impedance and seismic data in accordance with the Bayesian theory and add some perturbation,calculated using non-linear optimization,to the initial model.The last step is to take the result at the previous frequency as the constraint to perform inversion at the next frequency until the final output at each angle is obtained.The case study in Prospect A shows that constrained inversion at each angle and frequency generates multi-dimensional results with high precision;this facilitates thin reservoir characterization.