测井曲线相似度算法的研究

发布时间：2018-01-17 12:29

本文关键词：测井曲线相似度算法的研究　出处：《东北石油大学》2015年硕士论文　论文类型：学位论文

【摘要】：校深工作是套管射孔前的重要环节,对精确度的要求很高。校深环节不准确会让后续的射孔工作出现偏差,减少石油的产量,直接影响到对这个区块的评价,进而影响到其未来发展前景。校深的核心是对齐套前、套后测井曲线的相似点,所以相似度算法的准确度是研究的重点问题。本文主要研究适合测井曲线的相似度算法及如何高效准确地将相似度算法融入到校深环节中。为了使相似度算法精确度更高、耗时更少,对曲线进行预处理就成为了不可缺少的环节。曲线自身的毛刺噪声影响到精确度,庞大的数据量影响到计算复杂度,也就影响到了计算时间。对现有的剔除曲线毛刺方法的研究,发现小波领域的阈值法很适合对测井曲线处理。在分析了现存阈值函数及阈值的基础上,引出了一种改进的自适应小波域值法并应用到了测井曲线中,通过仿真实验,发现该方法不仅可以剔除噪声还能很好地保留住曲线的重要信息。在对几种线性分段法的讨论比较后,选取重要特征值线性分段法处理去噪后的曲线,这样可以达到压缩数据量的作用。通过重要特征值提取后的点组成的曲线段即为线性分段曲线。经仿真实验对处理后的曲线形态和数据量的观查,发现这种方法在保留局部重要信息的同时还可以压缩数据量。相似度算法有很多,结合测井曲线的自身特点,决定选用动态时间弯曲距离作为曲线的相似度核心算法。虽然松弛点动态时间弯曲方法有一定的局限性,但很适用于测井曲线校正。对原有的松弛点算法进行改进,可以进一步缩短计算时间。人工校深和编程校深的思路是完全不同的,本文提出了两种方案,将原有改进的时间弯曲方法与改进的松弛点方法分别融入适合它们各自的方案中,通过仿真实验得出结果与人工校正做比对,发现两种方案均可用于校深。经讨论比较后,发现“方案二”精确度更高、耗时更少,可以用于测井曲线校深,对减少生产力、节约成本有着重要意义。
[Abstract]:The depth adjustment is an important link before casing perforation, and the requirement of accuracy is very high. The inaccuracy of depth correction will make the subsequent perforation work deviate, reduce the oil production, and directly affect the evaluation of this block. The core of the depth calibration is the similarity of the log curves before and after the alignment. Therefore, the accuracy of similarity algorithm is the key issue. This paper mainly studies the similarity algorithm suitable for logging curve and how to integrate the similarity algorithm into the depth link efficiently and accurately. In order to make the similarity algorithm fine. More accurate. It takes less time to preprocess the curve. The burr noise of the curve itself affects the accuracy and the huge amount of data affects the computational complexity. On the basis of analyzing the existing threshold function and threshold value, it is found that the threshold method in wavelet domain is very suitable for processing log curves. An improved adaptive wavelet domain value method is introduced and applied to the logging curve. It is found that this method can not only eliminate the noise but also preserve the important information of the curve. After the discussion and comparison of several linear segmentation methods, the important eigenvalue linear segmentation method is selected to deal with the denoised curve. In this way, the function of data compression can be achieved. The curve segment composed of points extracted by important eigenvalues is a linear piecewise curve. Through simulation experiments, the shape of the processed curve and the observation of the amount of data are observed. It is found that this method can compress the data while preserving the local important information. There are many similarity algorithms, combined with the characteristics of logging curves. It is decided to choose the dynamic time bending distance as the core algorithm of curve similarity, although the relaxation point dynamic time bending method has some limitations. But it is very suitable for log curve correction. The calculation time can be further shortened by improving the original relaxation point algorithm. The ideas of artificial depth correction and programming depth correction are completely different. Two schemes are proposed in this paper. The original improved time bending method and the improved relaxation point method are incorporated into their respective schemes, and the simulation results are compared with manual correction. It is found that the two schemes can be used to calibrate the depth. After discussion and comparison, it is found that the "scheme two" has higher accuracy and less time, and can be used to calibrate logging curves, which is of great significance to reduce productivity and save cost.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P631.81

【参考文献】