【作者】 彭星；

【导师】 戴永寿；

【作者基本信息】 中国石油大学 ， 信息与通信工程， 2011， 硕士

【摘要】 地震子波的估计质量直接影响到高分辨率、高信噪比、高保真度的地震勘探数据处理结果。针对实际地震数据处理过程中,提取的地震子波是否准确无法判断、子波提取方法的性能优劣无法衡量的问题,提出一种基于反褶积的信息反馈控制子波提取的方法。其中对反褶积方法进行深入分析,首先采用基于稀疏假设的混合相位反褶积方法获取反射系数序列,与参数化模型方法提取的子波褶积获得合成地震记录,然后通过地震剖面属性分析优化获得关键属性,利用合成地震记录和原始地震记录中的关键属性比较结果判断提取子波的准确性和子波提取技术的有效性,最后利用比较结果修正子波,得到高精度的地震子波。由反褶积方法得到的反射系数序列的精度直接影响到得到的合成地震记录的精度。通过对经典反褶积以及现有改进反褶积方法的研究,针对各种反褶积方法的假设条件及适用范围不同而导致提取的反射系数序列精度不高的问题,采用基于稀疏假设的混合相位子波反褶积方法,将反射系数序列的求解问题转化成最优化问题,通过建立目标函数和反复迭代求解最佳反射系数序列。与经典的最小平方反褶积方法相比,理论分析和数据仿真实验结果表明该方法在子波为混合相位时求取的反射系数序列精度较高,与子波褶积后得到的合成地震记录与原始地震记录十分相似。基于信息反馈的地震子波提取方法另一个关键问题是反馈信息和反馈机制的确定。针对地震剖面属性具有复杂性和交叉性的特点,采用地震剖面属性分析优化方法,得到反映地震剖面特性的关键属性(波形、时延、主频、频宽),确定反馈信息。通过分析合成地震记录过程中各个环节对地震剖面关键属性的影响,建立反馈机制,根据不同的关键属性的差异确定子波提取方法的改进方向,达到提高子波精度的目的。在保证反褶积提取的反射系数序列的精度较高和通过地震剖面属性分析优化得到的反馈信息准确的前提下,采用本文提出的方法对线性(累积量矩阵法)与非线性(累积量拟合法)相结合的参数化模型方法提取的子波进行验证和子波修正。理论分析和实验数据仿真表明,在原始子波已知时,利用修正后的子波合成新的地震记录并进行二次子波提取,将二次子波与原始子波进行比较,可验证子波提取方法的有效性,间接验证了反褶积提高地震记录分辨率的特性;在原始子波未知时,通过合成地震记录与原始地震记录的关键属性比较法可有效的指导改进子波提取方法,提高子波精度。更多还原

【Abstract】 The seismic wavelet estimation quality directly affects high resolution, high signal-to-noise ratio and high fidelity seismic exploration data processing. In the practical seismic data process, the precision of seismic wavelet extraction cannot be judged, and the effects of wavelet estimation technique performances cannot be evaluated. This paper proposed a high precision method of information feedback to control seismic wavelet estimation based on the deconvolution. Under the thorough analysis of deconvoluton method, the synthetic seismic records can be obtained by convolution of the reflection coefficient sequences estimated by the sparse hypothesis mixed phase methods and the seismic wavelet estimated by parametric model method. Then the linchpin attributes can be got by seismic profiles attribute analysis optimization. By comparing and analyzing the linchpin attributes of the synthetic seismic records with the ones of original seismic records to evaluate the extracted wavelet accuracy and the wavelet extraction technology performance. The third step is using the comparison results to fix wavelet and get a high precision seismic wavelet.The accuracy of reflection coefficient sequences by deconvolution methods directly affects the synthetic seismic records. Through the study of classic deconvolution and existing improved deconvolution method,the issue is aimed at various deconvolution methods because different assumed conditions and scope of application resulting the accuracy of extraction reflection coefficient sequences is not high. This paper adopts the mixed phase wavelet deconvolution method based on sparse hypothesis in order to transform the problem of solving reflection coefficient sequences into optimization problem, by establishing objective function and repeate iteration to obtain reflection coefficient sequences. Compare with classic minimum square deconvolution method, theory and data simulation experiment results show the method can get high accuracy reflection coefficient sequences when wavelet is mixed phase, convolution with wavelet can get synthetic seismic records very similar to raw seismic records.Another key problem of the wavelet extraction method based on information feedback is the feedback information and feedback mechanism determination. Feedback information and feedback mechanism are the two keys of seismic wavelet extraction based information feedback. According to the complex and crossed seismic profiles attribute, the feedback information is confirmed from the linchpin attributes (waveform, delay, main-frequency, bandwidth), which reflect the features of seismic profiles through the optimization of seismic profiles attributes. By analyzing all aspects on the key attributes of seismic profiles in the process of synthetic seismograms, the feedback mechanism system was established. According to each key attribute difference wavelet extraction method improvement direction was confirmed to achieve the purpose of wavelet accuracy.In the premise of higher reflection coefficient sequences precision estimated by deconvolution and feedback information accuracy optimized by attributes analysis of seismic profile, the method is used to validate and fix the wavelet estimated by the parameterized modeling method combining of linear (cumulant matrix method) and nonlinear (cumulant fitting method).Under the primitive wavelet known conditions, the theoretical and experimental data simulation results show that the comparison between the second wavelet estimated by synthetic seismic records and the original wavelet can verify the wavelet extraction method combining linear and nonlinear validity. It can also verify that deconvolution can improve the characteristics of seismic records resolution. Under the primitive wavelet unknown conditions, using the comparison result of the seismic profile linchpin attributes can improve wavelet extraction method effects and improve wavelet precision.更多还原