基于云模型的EMD相控测井储层预测研究

发布时间：2018-07-10 01:30

本文选题：经验模态分解 + 测井相　；参考：《成都理工大学》2016年博士论文

【摘要】：在岩性地层油气藏勘探中,由于低渗透储层物性相对较差,导致测井曲线信噪比低,储层流体识别较为困难,存在多解性。基于云模型的EMD相控测井储层预测研究是在确立优势测井相序集的基础上,探索经验模态分解(简称EMD)提高信噪比的方法,达到提高测井数据分辨率之目的;以层点为样本,研究时频分析技术在测井层序地层学中应用的新方法;在对主力油层精细分层的基础上,提出了基于EMD分解的储层流体识别的新算法和新参数;针对储层产能评价中的多解性问题,建立了储层产能评估的二维正态云模型判定标准,实现了测井产能评价过程的可视化;通过对低渗透致密砂岩储层预测实例的分析,从科学上,客观的给出测井解释结果,用于指导油气藏勘探的进程。具体来说,主要研究内容及成果有下几个方面:(1)从多角度、多方面探讨用于测井数据处理的算法理论。在时频分析技术方面,分析了传统傅里叶变换、小波变换及EMD分解算法的优缺点及局限性,指出:傅里叶变换适合于分析频率固定的平稳信号;小波变换和EMD分解更适合于分析非线性、非平稳的测井序列。在空间数据挖掘理论研究方面,介绍了不确定性人工智能中用于定性概念与其定量数据之间相互转换的“云模型”理论,讨论了描述“云”的三个基本数字特征的意义,给出了“正态云”云模型发生器的实现算法,为解决测井储层产能评价中的多解性问题奠定了理论基础。(2)深入探讨了多参数雷达图在相控测井数据处理中的应用技巧,编写了雷达图绘图程序,为用户提供便利的使用环境。以四川盆地大邑构造须家河组为例,详细介绍了雷达图用于测井数据处理的三种典型方法与操作步骤;建立了该地区岩性和流体识别的雷达图标准模板;利用雷达图绘图程序有效地帮助数据解释人员较为精确的实现了分选敏感测井指标、识别未知井剖面地层岩性以及初步判定储集层的流体性质。(3)深入分析了三种利用测井资料进行层序界面划分和旋回识别的方法,给出了新的认识。以鄂尔多斯盆地子北采油厂L井区为例,通过对比传统的测井曲线形态识别法、小波变换精细分层法及经验模态分解旋回识别法的理论基础、应用方法与技巧,指出:传统的测井曲线形态识别法所投入的工作量大,难度高,并且受人为主观性因素影响较大;小波变换是一种时频局部分析方法,它通过两个互补的滤波器产生低频的近似信号和高频的细节信号,分别对应于信号中的稳定变化趋势和突变点,利用这一特性,可以较为准确的实现小层的精细划分;而emd分解函数中,每一个分量都包含了信号的完整信息,反映出信号的瞬时频率特征,更适合于识别地层不同级次的沉积旋回。(4)提出了基于emd分解的测井地层旋回识别的新方法。概括出五种emd分解曲线响应特征在各种沉积环境中的基本形态;以利萨如图(lissajousfigures)理论为基础,指出了在深度域上作imfs分量的散点图时出现的光滑闭合曲线正是地层沉积旋回特征的表现;并以imfs瞬时频率主值区间推算可识别旋回级次,证明了新方法的科学性。结合四川盆地川西坳陷中段孝泉-丰谷构造带层序地层研究的实例,详细介绍了emd分解用于测井地层旋回识别的新方法和操作步骤,深入探讨了不同尺度下本征固有模态函数的地质意义,指出:对地区敏感参数作emd分解,依据相关性原理优化组合imfs分量可以强化反映地层旋回特征的较为稳定的信息;在层序地层学研究中,地层的旋回是以分布时限为标准划分的,由瞬时频率的主值区间推算出测井数据在深度域的可分辨深度范围,据此可以很好的识别不同级次的地层旋回特征。(5)提出了基于emd的储层流体识别的新算法和新参数。将经验模态分析方法用于低渗透致密砂岩储层的天然气识别当中,通过对适合地区特征的天然气敏感参数作经验模态分解,展开多尺度的相关性对比分析,选择能较好反映流体性质变化的imfs函数分量作交会图,揭示出非平稳的测井信号在微观上隐藏的局部曲线相关和局部线性相关的规律,找出了新的水层指示参数pw、气层指示参数pα及优质气层指示参数hgr,并给出依据新参数进行储层流体判别的方法和计算公式。(6)发现了一个适用于致密砂岩储层流体判别的新规律。以川西凹陷须家河组气藏识别为例,揭示了优势测井相序集信号中隐藏的局部相关性规律,验证了水层和气层的新指示参数和新算法在低渗透性致密砂岩储层中流体识别的可行性。详细介绍了在常规方法无法识别气藏的情况下,应用emd分解法进行气、水界面识别的新算法与步骤,证实了新算法和新参数的有效性,指出:在依据相关性原理,对优选的天然气敏感参数ac和cnl的经验模态分解函数作交会图时,天然气储层呈现出局部曲线相关的特征,表现为明显的弧线,并且储层物性越好,弧线的曲率越大;而水层和干层则呈现出局部线性相关或可近似为线性相关的特征。(7)提出了基于云模型的测井储层产能评价的新方法。针对油气储层产能评估过程中广泛存在的不确定性因素,以及测井数据与专家意见之间存在的定量与定性变量转换和映射的需求,深入分析云模型的数字特征,以鄂尔多斯盆地子北采油厂X井区延长组地层为例,结合试油数据,建立储层不同产能标准的二维正态云图,分别从定性和定量上给出了判定储层产能的方法,实现了数据挖掘的可视化解释。该方法从多角度研究储层数据的分布特征,最大限度的保留了评估过程中概念描述的不确定性因素,建立了专家定性意见的定量可比性,客观的对储层产能作出评价,从而保证了测井解释的可信度。
[Abstract]:In the exploration of lithostratigraphic reservoir, due to the relatively poor physical property of low permeability reservoir, it leads to low signal-to-noise ratio of logging curves, and the reservoir fluid recognition is difficult and multi solvable. The research of EMD phase controlled logging reservoir prediction based on cloud model is based on the establishment of the predominant log phase sequence set, exploring the empirical mode decomposition (EMD) to improve the signal to noise ratio. In order to improve the resolution of log data, a new method of applying time frequency analysis technique to log sequence stratigraphy is studied with stratified sampling. On the basis of fine stratification of main oil layers, a new algorithm and new parameters for reservoir fluid identification based on EMD decomposition are put forward. The two dimensional normal cloud model criterion for reservoir productivity evaluation has been established, and the visualization of logging productivity evaluation process is realized. Through the analysis of the low permeability and dense sandstone reservoir prediction example, the results of logging interpretation are objectively given and used to guide the process of oil and gas exploration. The following aspects: (1) the algorithm theory of logging data processing is discussed from many angles and many aspects. In the aspect of time-frequency analysis, the advantages and disadvantages of traditional Fourier transform, wavelet transform and EMD decomposition are analyzed. It is pointed out that Fourier transform is suitable for stationary signals with fixed frequency, and the wavelet transform and EMD decomposition are more suitable. In the analysis of nonlinear and non-stationary log sequences, in the field of spatial data mining theory, the "cloud model" theory used for the conversion between qualitative concepts and quantitative data in uncertain artificial intelligence is introduced, and the significance of describing the three basic digital characteristics of "cloud" is discussed, and the occurrence of "normal cloud" cloud model is given. The realization algorithm of the device has laid the theoretical foundation for solving the multi solvability problem in the logging reservoir productivity evaluation. (2) the application techniques of the multi parameter radar map in the phase controlled logging data processing are deeply discussed, and the radar drawing program is written to provide the convenient use environment for the users. The Xujiahe Formation in Dayi of the Sichuan basin is taken as an example. This paper introduces three typical methods and operation steps of radar map for logging data processing, establishes a standard template for radar map of lithology and fluid identification in this area, and effectively helps the data interpreters to achieve more accurate identification of sensitive logging indicators, identify the lithology of unknown well profile and preliminary identification. The fluid properties of the reservoir are determined. (3) three methods of sequence boundary division and cycle identification using logging data are deeply analyzed. A new understanding is given. Taking the L well area of the Erdos Basin subarea oil production plant as an example, the traditional logging curve shape recognition method, the fine stratification method of wavelet transform and the empirical mode decomposition cycle are compared. The theoretical basis, the application method and the technique of the recognition method, point out that the traditional logging curve shape recognition method has a large amount of work, high difficulty and influenced by the subjective subjective factors. The wavelet transform is a time-frequency local analysis method, which produces low frequency approximate signal and high frequency detail signal through two complementary filters. In the EMD decomposition function, each component contains the complete information of the signal, reflects the instantaneous frequency characteristic of the signal, and is more suitable for identifying the sedimentary cycles of different gradation. (4) proposed A new method of identification of log stratigraphic cycles based on EMD decomposition. The basic morphology of the response characteristics of five EMD decomposition curves in various sedimentary environments is summarized. Based on the theory of lissajousfigures, the smooth and closed curve of the formation of the stratigraphic sedimentary cycles is pointed out when the scatter plot of the IMFs component in the depth domain is the characteristic of the sedimentary cycle. It is proved that the new method is scientific, and the new method is proved to be scientific by using the main interval of IMFs instantaneous frequency. The new method and operation steps of EMD decomposition for identification of log stratigraphic cycle are introduced in detail, combined with the example of the sequence stratigraphy of the middle section of the Sichuan basin in the middle section of the West Sichuan depression. The geological significance of the intrinsic intrinsic modal function indicates that the EMD decomposition of sensitive parameters of the region and the optimization of the combined IMFs component according to the principle of correlation can strengthen the more stable information reflecting the characteristics of the stratigraphic cycle. In the sequence stratigraphic study, the formation cycle is divided by the time limit of distribution and is pushed by the main interval of the instantaneous frequency. The distinguishable depth range of the log data in the depth domain can be calculated and the stratigraphic cycle characteristics of different grades can be identified well. (5) a new algorithm and new parameters for reservoir fluid identification based on EMD are proposed. The empirical mode analysis method is applied to the identification of the sky gas in the low permeability and dense sandstone reservoir, and the characteristics are suitable for the region. The sensitive parameters of natural gas are decomposed by empirical mode, and multiscale correlation analysis is carried out. The IMFs function component, which can better reflect the change of fluid properties, is selected as the rendezvous graph, and the laws of local curve correlation and local linear correlation hidden by non-stationary logging signals are revealed, and the new water layer indication parameter PW, gas layer is found. The indicator parameter p alpha and the indicator parameter HGR of high quality gas reservoir are given, and the method and calculation formula of reservoir fluid discrimination according to the new parameters are given. (6) a new law for identifying the fluid in the tight sandstone reservoir is discovered. The local correlation hidden in the phase sequence set signal of the best potential well log is revealed with the identification of the Xujiahe gas reservoir in the West Sichuan depression as an example. The feasibility of the new indicator parameters of water layer and gas reservoir and the new algorithm in the low permeability tight sandstone reservoir are verified. The new algorithm and steps of gas and water interface identification by the EMD decomposition method are introduced in detail, and the validity of the new algorithm and the new parameters are confirmed. On the basis of the principle of correlation, when the selected natural gas sensitive parameters AC and CNL are rendezvous with the empirical mode decomposition function, the natural gas reservoir shows the characteristics related to the local curve, showing the obvious arc, and the better the physical property of the reservoir, the greater the curvature of the arc line, while the water layer and the dry layer show a local linear correlation or can be approximated as a line. (7) a new method of logging reservoir productivity evaluation based on cloud model is proposed. The uncertain factors which exist widely in the process of oil and gas reservoir productivity evaluation, and the requirement of quantitative and qualitative variable transformation and mapping between logging data and expert opinion are needed, and the digital characteristics of the cloud model are analyzed in depth, and the ERL is analyzed. As an example of the Yanchang Formation in the X well area of the duo basin sub oil production plant, combining with the oil test data, the two-dimensional normal cloud map of the reservoir with different productivity standards is established. The method of determining the reservoir productivity is given from the qualitative and quantitative terms, and the visual interpretation of the data mining is realized. The method is used to study the distribution characteristics of the reservoir data from multiple angles and maximize the distribution characteristics. The uncertainty factors of the conceptual description in the evaluation process are retained, the quantitative comparability of the expert qualitative opinion is established, and the reservoir productivity is evaluated objectively, thus ensuring the credibility of the log interpretation.
【学位授予单位】：成都理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：P618.13;P631.81

【参考文献】