基于谱多流行聚类的Shearlet变换微地震数据噪声压制研究

发布时间：2018-07-01 19:59

本文选题：地面微地震监测 + Shearlet变换　；参考：《吉林大学》2017年硕士论文

【摘要】：微地震监测技术是一种先进的水力压裂裂缝监测技术,它已经成为改造低渗透油气藏储层结构的一项非常重要的手段。井中微地震监测和地面微地震监测是微地震监测技术的两种不同实现方式。其中,地面微地震监测技术凭借其特有的优势,如无需监测井、成本低、实用性强和监测范围广等,使得它具有更加广泛的应用前景。然而,地面检波器采集到的微震数据通常具有有效信号能量弱、信噪比低等缺点,这给整个微震数据的处理和解释工作带来巨大的困难。因此,需要通过各种噪声压制技术来提高地面微地震监测数据的信噪比和分辨率,这对微地震资料事件拾取和裂缝解释等后续工作具有重要的意义和价值。Shearlet变换是近几年发展起来的一种新型多尺度几何分析技术,它同时也是多维函数的一种接近最优的稀疏表示方法。此外Shearlet变换可以和多分辨率分析关联起来,从而建立了连续和离散形式的统一。基于其优秀的性质,采用Shearlet变换对地面微地震数据进行噪声压制。Shearlet变换可以将微地震数据分解到不同的尺度层,随着尺度的增加Shearlet系数逐渐由Coarse尺度层系数变为Fine尺度层系数。Coarse尺度层对应原始数据的低频概貌信息,而信号的高频细节信息主要集中在Fine尺度层,随机噪声主要集中在Fine尺度层中。然而,地面微地震数据常常表现为极低的信噪比,含噪的地面微地震数据经Shearlet变换后,Fine尺度层中部分随机噪声对应的系数与微震有效事件对应的系数在数值上近似相等。此时,基于阈值的Shearlet变换去噪算法已经不能有效地移除和噪声相关的系数,影响最终的噪声压制效果。为了更好分离Shearlet变换域中有效信号相关的系数和噪声相关的系数,我们提出了基于谱多流行聚类的Shearlet变换去噪算法。基于谱多流行聚类的Shearlet变换去噪算法充分考虑了Shearlet变换域中有效信号系数的流行结构特征,并采用谱多流行聚类算法检测或分组有效信号系数,从而达到系数分离的目的。地面微地震有效事件与噪声分别位于各自所在的低维流行上,即二者来自不同的流行。有效事件同相轴具有明显的流行结构,而随机噪声没有明显的流行结构信息,并且经过Shearlet变换后,它们的流行结构特征不会发生变化。基于这个差异,谱多流行聚类可以很容易地检测到微地震有效事件对应的系数,从而避免了阈值方案的缺陷。为了验证该算法的有效性和可行性,将其应用于模拟和实际地面微地震数据处理中。模拟和实际微地面微地震数据的处理结果表明,基于谱多流行聚类的Shearlet变换去噪算法在压制强随机噪声的同时能更好地保留信号特征。
[Abstract]:Micro-seismic monitoring technology is an advanced hydraulic fracturing fracture monitoring technology, it has become a very important means to reconstruct the reservoir structure of low permeability reservoirs. Micro-seismic monitoring in wells and ground microseismic monitoring are two different ways to realize micro-seismic monitoring technology. Among them, ground micro-seismic monitoring technology has more extensive application prospects because of its unique advantages, such as no monitoring wells, low cost, strong practicability and wide monitoring scope. However, the microseismic data collected by the ground geophone usually have the disadvantages of weak signal energy and low signal-to-noise ratio, which bring great difficulties to the processing and interpretation of the whole microseismic data. Therefore, a variety of noise suppression techniques are needed to improve the signal-to-noise ratio and resolution of ground microseismic monitoring data. This is of great significance and value to the follow-up work of micro-seismic data gathering and fracture interpretation. The Shearlet transform is a new multi-scale geometric analysis technique developed in recent years. It is also a near optimal sparse representation of multidimensional functions. In addition, the Shearlet transform can be associated with multi-resolution analysis, thus establishing the unity of continuous and discrete forms. Based on its excellent properties, the Shearlet transform can decompose the microseismic data into different scale layers by using Shearlet transform to suppress the noise of ground microseismic data. With the increase of scale, the Shearlet coefficient gradually changes from Coarse scale layer coefficient to Fine scale coefficient .Coarse scale layer corresponds to the low frequency profile information of the original data, while the high-frequency detail information of the signal is mainly concentrated in the Fine scale layer. The random noise is mainly concentrated in the fine scale layer. However, ground microseismic data often exhibit very low signal-to-noise ratio (SNR). After Shearlet transform, the coefficients corresponding to some random noises in the fine scale layer are approximately equal to those corresponding to the effective microseismic events. In this case, the threshold based Shearlet transform de-noising algorithm can not effectively remove the noise-related coefficients and affect the final noise suppression effect. In order to better separate effective signal correlation coefficients from noise correlation coefficients in the Shearlet transform domain, we propose a Shearlet transform de-noising algorithm based on spectral multi-popular clustering. The Shearlet transform denoising algorithm based on spectral multi-popular clustering takes full account of the popular structural features of effective signal coefficients in Shearlet transform domain, and uses spectral multi-popular clustering algorithm to detect or group effective signal coefficients, so as to achieve the purpose of separating coefficients. The effective events and noises of ground microearthquakes are located on their respective low dimensional epidemics, that is to say, they come from different epidemics. The effective event cophase axis has obvious popular structure, but the random noise has no obvious popular structure information, and after Shearlet transform, their popular structure characteristics will not change. Based on this difference, spectral multi-popular clustering can easily detect the coefficients corresponding to micro-seismic effective events, thus avoiding the defect of threshold scheme. In order to verify the validity and feasibility of the algorithm, it is applied to the simulation and practical ground microseismic data processing. The results of simulation and practical microseismic data processing show that the Shearlet transform denoising algorithm based on spectral multi-popular clustering can suppress the strong random noise and retain the signal features better.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P631.4

【相似文献】