低频采集反射地震数据约束全波形反演关键问题研究

Full waveform inversion (FWI) is a highly non-linear geophysical inverse problem. Using multi-scale and layer-stripping inversion strategies, as well as imposing reasonable model constrains may effectively stabilize the FWI optimization process. Key horizons on prestack migrated images are used to build the block geological structure model. Multi-attributes extraction and classification are used to calibrate the local smoothness of velocity model under the block structure model constrains. Through dispersion analysis and optimization, and direct methods for large-scale sparse matrix LU decomposition, a fast, multi-shot, frequency domain finite difference forward modeling algorithm is developed. A local smoothness based objective function is designed for constrained full waveform inversion. The corresponding constrained inversion workflow incorporates several optimized inversion strategies, including multi-scale, layer-stripping, and gradient preconditioning. Through analysis of real low frequency seismic data, the characteristics of seismic signal and typical noise are extracted. Two kind of Synthetic models are built to mimic the situation of real low frequency seismic acquisition work. After synthetic data processing and prestack migration, attribute analysis is used to extract local model smoothness constrains. Several groups of constrained waveform inversion tests are carried out to reveal the role and effects of local model smoothness constrains on waveform inversion. This research will help to build the two-way positive feedback mechanism between the new constrained FWI algorithm and prestack migration, to increase th utilization ratio of multi-discipline geoscience data,and to insure the geolological plausibility of inversion results. This work will provide new theoretical and practical basis for high resolution velocity model building and reservoir characterization under deep, compelex geological regions and unconventional exploration environment.

本项研究通过叠前偏移数据体标志层位解释与块状构造建模、宏观构造约束下地震属性提取与分选优化，实现基于多属性加权的模型局部光滑度估算；通过有限差分法离散化格式频散分析与优化、大型稀疏矩阵LU分解直接解法实现频率域多炮快速模拟；设计基于模型局部光滑度的约束反演目标函数，建立多尺度约束波形反演流程；提取实际低频地震资料信号与噪声特征，设计两组正演模型建立逼近实际低频采集情况的合成地震记录，在合成记录偏移剖面提取模型局部光滑度约束关系，在此基础上逐一改变多组反演参数进行约束波形反演测试，进而分析不同情况下模型局部光滑度约束关系对多尺度波形反演影响规律。本项研究可以建立约束全波形反演与叠前偏移方法之间双向良性反馈机制，进一步提高多源地学信息利用率和波形反演结果地质合理性，为我国海洋深水油气藏、陆地中深层油气藏以及非常规油气勘探开发进程中的高精度速度建模与储层评价方案优化提供技术支持。

低频地震采集方法能保护10Hz以下的有效低频成分，这对于多尺度全波形反演特别重要。约束反演是提高波形反演稳定性、精度和效率、降低反演多解性的最为有效和可行方法之一。本项研究提出利用振幅类属性实现速度模型局部光滑度信息提取方法，建立了测井数据与地震多属性数据联合提取多参数低频背景模型方法流程，完善了基于分频对数谱属性的低频速度模型建模方法，利用模型实验验证基于地震属性的反演约束建模方法可行性。实现适用于变密度与变Q值模型的二维和三维频率域省时交错网格有限差分阻尼波场正演模拟方法，推导频率域阻尼波动方程交错网格离散化方程，利用模拟退火法优化离散化波动方程的频散关系，提高正演精度，利用理论模型正演验证方法正确性。完善低频信号随采、提取与补偿方法与流程技术，建立多频率并行多尺度约束反演方法，确定迭代化反演过程的低频信号与低频模型内在映射机制，形成面向目标的速度参数时移差异性多尺度全波形反演方法。.本项研究为建立约束全波形反演与叠前偏移方法之间双向良性反馈机制，进一步提高多源地学信息利用率和波形反演结果地质合理性，为我国海洋深水油气藏、陆地中深层油气藏以及非常规油气勘探开发进程中的高精度速度建模与储层评价提供优化的解决方法。