基于循环平移Shearlet变换自适应阈值消减微震勘探随机噪声

发布时间：2018-12-08 18:02

【摘要】：随着世界油气需求的持续增长,油气作为不可再生资源逐渐被消耗。复杂的非常规油气藏开采成为储量增长的热点,微震作为非常规水力压裂观测信号,具有能量弱、信噪比很低的特点,常规的滤波方法具有局限性。严重影响微震信号初至拾取及反演定位的精度,对地质构造解释及油气藏开发带来困难。因此,提高监测记录信噪比是微震资料处理中的重要环节,我们需要在强随机噪声背景下提取有用信号,有效消减微震资料中的随机噪声。本文重点研究微震信号随机噪声压制方法,结合Shearlet变换和循环平移变换,构建自适应阈值估计方案。仿真实验和实际资料处理结果证明该方法可以有效提取微震信号,较大程度去除随机噪声且能保持有效信号幅度,提高微震资料信噪比。Shearlet变换是一种新的多尺度时频分析工具,具备多尺度、多方向及最佳稀疏逼近性质,即通过稀疏矩阵重构有用信号,计算效率高。近年来,该变换方法已逐步应用于地震勘探资料去噪领域,取得了一定成效。然而基于阈值的Shearlet变换方法在微震消噪过程中存在一定局限性:1、井中微震信号能量弱、主频较高,信号和噪声在高频段重合,阈值法很难分离信号和噪声。2、对于地面微震信号,目前基于Shearlet变换阈值降噪方法往往假设噪声分布在高频段,而没考虑到低频段噪声对信号的干扰。3、传统的基于阈值的Shearlet变换消噪方法在变换域使用统一阈值,因而容易丢失部分有效信号,造成能量损失。4、传统的Shearlet变换存在下采样过程,因此变换缺乏移不变性。本文从Shearlet变换本质出发,研究基于Shearlet变换的微地震噪声压制技术。分析微震数据经过Shearlet变换后的系数分布特征,考虑信号的方向和空间相关性,对待处理数据进行循环平移,利用Shearlet变换的多尺度特性并增强其平移不变性,对信号进行多尺度多方向分解并根据微震信号与随机噪声在Shearlet域表征差异建立基于分块主成分分析的权重阈值收缩方案,最后对处理后的空间排列数据进行叠加平均从而达到增强有效信号的目的,在降噪的同时更好的恢复有效信号的幅度,实现低SNR条件下的微震信号识别。通过人工模拟仿真和实际记录处理,验证了本文提出的自适应方向阈值Shearlet变换算法在幅度保持和噪声压制的能力均优于基于传统阈值的Shearlet变换算法。
[Abstract]:With the continuous growth of world oil and gas demand, oil and gas as a non-renewable resource is gradually consumed. The complex unconventional oil and gas reservoir exploitation has become the hot spot of reserve growth. As an unconventional hydraulic fracturing observation signal, microseismic is characterized by weak energy and low signal-to-noise ratio (SNR), and the conventional filtering method is limited. The accuracy of initial arrival pickup and inversion positioning of microseismic signals is seriously affected, and it is difficult to interpret geological structure and develop oil and gas reservoirs. Therefore, improving the signal-to-noise ratio (SNR) of monitoring records is an important step in microseismic data processing. We need to extract useful signals in strong random noise background and effectively reduce random noise in microseismic data. In this paper, the random noise suppression method of microseismic signal is studied, and the adaptive threshold estimation scheme is constructed by combining Shearlet transform and cyclic translation transform. Simulation experiments and actual data processing results show that this method can effectively extract microseismic signals, remove random noise to a large extent and maintain the effective signal amplitude. Shearlet transform is a new multi-scale time-frequency analysis tool, which has the properties of multi-scale, multi-direction and best sparse approximation, that is, reconstruction of useful signals by sparse matrix has high computational efficiency. In recent years, the transform method has been gradually applied in the field of seismic exploration data denoising, and achieved certain results. However, the Shearlet transform method based on threshold has some limitations in the process of microseismic de-noising. 1. The signal energy is weak, the main frequency is high, the signal and noise coincide in the high frequency band, the threshold method is difficult to separate signal and noise. For ground microseismic signals, the current threshold noise reduction methods based on Shearlet transform often assume that the noise is distributed in high frequency band without considering the interference of low frequency band noise to the signal. The traditional threshold-based Shearlet transform denoising method uses a unified threshold in the transform domain, so it is easy to lose part of the effective signal, resulting in energy loss. 4. The traditional Shearlet transform has the process of downsampling, so the transformation lacks shift invariance. Based on the essence of Shearlet transform, the technique of microseismic noise suppression based on Shearlet transform is studied in this paper. This paper analyzes the coefficient distribution characteristics of microseismic data after Shearlet transform, considers the direction and spatial correlation of the signal, makes cyclic translation of the processed data, and makes use of the multi-scale characteristics of Shearlet transform to enhance its translation invariance. The signal is decomposed in multi-scale and multi-direction. According to the difference between microseismic signal and random noise in Shearlet domain, the weight threshold shrinkage scheme based on block principal component analysis (PCA) is established. Finally, the processed spatial array data are superimposed and averaged to enhance the effective signal, to restore the amplitude of the effective signal while reducing the noise, and to realize the recognition of the microseismic signal under the condition of low SNR. Through artificial simulation and actual record processing, it is verified that the proposed adaptive directional threshold Shearlet transform algorithm is superior to the traditional Shearlet transform algorithm in amplitude retention and noise suppression.
【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】：TE937

【相似文献】