【作者】 洪德成；

【导师】 吴式枢； 杨善德；

【作者基本信息】 吉林大学 ， 理论物理， 2009， 博士

【摘要】 近十年发展起来的三轴感应测井由于能探测地层的电各向异性,提供关于地层水平和垂直电导率以及地层倾角和走向的信息,这对砂—泥岩薄交互油储层组这类“低阻产层”的识别和开发有着重要意义,倍受人们的重视。Schlumberger公司推出的完全三轴感应仪器通过在仪器体内安装经过精心设计的电导元件而显著地减少了井眼环境包括仪器偏心等对测井响应的影响,使仪器的实用性得到极大提高。三轴感应测井响应与地层水平和垂直电导率σh和σv、井眼(或地层)斜角α以及方位角?同时有关,而且关系复杂,再加上邻层和井眼效应等,目前普遍认为很难对三轴感应测井资料进行直观解释,有意义的资料解释结果只能通过多参数的非线性反演才能达到。然而,由于三轴感应仪器在含井眼和泥浆侵入的水平层状TI倾斜地层中的测井问题是个3D问题,其正演模拟已无解析解,目前普遍采用3D有限差分法求数值解,它的计算速度慢、占用计算机内存大,还难以适应做资料反演的要求,因而一般都采取先对实测资料进行井眼效应校正,利用校正后的数据在忽略泥浆侵入的条件下来做水平层状TI倾斜地层的纵向一维反演。本文应用横电(TE)和横磁(TM)波分解法推导的在水平层状TI倾斜地层中三轴感应测井响应的解析解,可为这种资料反演提供一种高效的正演模拟手段。即便是做纵向一维反演,由于无法对资料做直观解释,要给资料反演提供一个较好的初始地层模型也有很大困难,不仅这种资料反演的难度大,也无法满足现场对资料实时解释的需要,从而制约了它的广泛使用和进一步发展。因此三轴感应测井资料处理和解释问题至今仍是一个具有挑战性的问题,不能对它进行直观解释是个关键因素。现在普遍采用首先确定方位角? ,将实测到的9条测井曲线变换成井眼坐标系的5条响应曲线再来做资料解释。我们发现现有的确定?的方法存在可相差π的不确定性,且转换得到的5条响应曲线的精度与?的提取精度有关。本文给出了一种与?的提取精度无关的由实测资料直接获取井眼坐标系的5条响应曲线的方法。由于井眼坐标系的每一条响应曲线都和σh、σv、α同时相关,再加上邻层和井眼效应等复杂因素,使得很难对这些资料进行直观解释。本文最有意义的结果是:我们找到这样一条组合曲线,它通过与井眼斜角α有关的权重系数把井眼坐标系磁场张量的主分量和交叉分量组合成一条新的曲线,我们称它为水平电导率视值的组合曲线,并用符号σha表示。σha能以准方框图的形式直观再现地层水平电导率的纵向分布,通过σha对井眼斜角α的扫描还可提供一种确定井眼斜角视值αa的方法。组合曲线σha的纵向分辨率可用L cosα表征,在大井眼斜角和短源距时,σha曲线的纵向分辨率极高,用它不仅可直观定准地层的纵向边界位置,包括薄交互层组这样的薄层边界位置,对厚度大于L cosα的层σha曲线在每层中央区的标志值也相当接近于该层的水平电导率真值。σha组合曲线使三轴感应测井仪器的纵向分辨能力得到了充分的直观显示,显著地提高了它的资料直观解释能力。本文给出了一种以σha组合曲线为核心的三轴感应测井资料直观解释的方法和流程,此直观解释结果还可为资料的多参数的非线性反演提供一种较好的初始地层模型,对几个典型的水平层状TI倾斜地层模型的模拟测井资料所做的纵向1D反演得到了令人满意的结果,它迭代收敛快,反演精度高,这在很大程度上得益于σha有很好的纵向分辨率,为反演提供了好的初始模型。本文给出的以组合曲线σha为核心的三轴感应测井资料的处理方法不仅显著提高了资料的直观解释能力和效果,也使其资料的实时解释成为可能,这不仅有助于仪器的推广利用,还可为仪器进一步向随钻测井和阵列测井方向发展提供资料处理和解释方法上的支撑。因为随钻测井的地质导向必然要求能对资料进行实时解释,以便能够及时做出钻井位置和方向的安排与调整,本文3.8节的结果初步显示了将它用于地质导向的可能性。另外,在相当一段时间内,要大量对三轴感应测井资料进行3D反演还是不现实的,采取与现在阵列感应类似的资料处理和解释方法是较为现实的途径,只要三轴感应仪器有几个不同探测深度的探头,对每个探头的实测资料做了井眼效应校正后,可按本文提出的方法分别对它们进行一维反演,以实现它们在纵向分辨率上的匹配,几个不同探测深度的这种反演结果的σh和σv的差别就可用来解释它们在径向上的不均匀性,即泥浆侵入情况和原状地层的电参数。本文的工作有些还需要充实和深入。首先,本文对存在井眼和泥浆侵入时的组合曲线的研究还只做了一个地层模型的实例,其读图数据的精度也不够高,还需要用精度高的数据做更多的地层模型实例来充实和完善它。其次,本文考察用的是无噪声数据,由于交叉耦合分量在层中央,特别是各向同性厚层的中央附近,其信噪比比主分量的低,而我们所用的σha组合曲线包含了交叉耦合分量,因此它的抗噪声能力如何是个值得重视并且最后还要通过实测资料来加以检验的问题。总之本文的结果还是初步的,要做的工作还很多,我们将继续努力。更多还原

【Abstract】 In recently 10 years, people pay much attention to a new induction logging technology called fully triaxial induction logging. It can detect and characterize resistivity anisotropy and provide the information of formation dips and azimuths that are essential in charactering and developing low-resistivity reservoirs such as thinly laminated sand-shale sequences. A newly developed fully triaxial induction tool by Schlumberger is designed to great minimize the effects of borehole environments, deep invasion and tool eccentricity by using conductive elements in tool body and great improve the practical of the instrument.The fully triaxial induction logging responses simultaneously depend on formation horizontal conductivityσh, vertical conductivityσv, borehole (or formation) dipαand the tool azimuth? and are highly non-linear functions, in addition to the strong effects by borehole and adjacent layers that make the responses more complex, until now, people generally agree that it is hard to process the logging data with visual explanation and the reliable results can only be obtained by nonlinear multi-parameter inversion.However, the fully triaxial induction logging is a 3D problem and no analytic forward modeling while the tool traverses a dipping TI layered formation with borehole and invasion. The 3D finite-difference modeling is generally used to solve the numerical solution that is now difficult to adapt to the requirements of data inversion for slow computing speed and very large workload. Hence, people always remove the borehole effect from the actual field logs first and then inverse the borehole-effect-corrected data with one-dimensional horizontal layered TI forward modeling.In this paper, we derive an analytic solution in horizontal layered anisotropic formation with no borehole and invasion by using TE and TM wave decomposing method to simulate the response of a fully triaxial induction tool in dipping well, which provides a high efficient forward modeling for logging data inversion. However, it is difficult to provide a reasonable initial model for inversion when we can’t make the visual explanation of logging data. It increases the difficulty even for vertical 1D inversion and also can not meet the real-time interpretation in scene that restricts the extensive application and further development of fully triaxial induction tool. So, it is still a challenging problem for fully triaxial induction logging data processing and interpretation, and the key problem is the visual interpretation of the logging data.Now, people commonly determine the azimuth? first, and then transform the nine actual field logging curves in the tool coordinate system to the five logging curves in the borehole coordinate system, and make logging data processing and interpretation with them at last. We found that there exists the uncertainty to determine the azimuth? or( ? +π)using the existing method and the precisions of obtained five response curves depend on the precision of the extracted? . In this paper, we develop a fine algorithm to directly obtain five response curves independent of the precision of the extracted ? from nine actual field logging curves in the tool coordinate system.Because of each response curve simultaneously depending onσh,σvandαin addition to some complex factors such as borehole and invasion effects, it is hard to make the visual explanation of logging data. The most meaningful result in this paper is that we found a fine combination curve with the principle components and cross-component via different weight coefficient relevant toαin borehole coordinate system calledσha, which has high vertical resolution and can reconstruct the distribution of horizontal conductivityσhwith a quasi-block response curve. We also provide an algorithm to determine the apparent dipαavia the scanning of dipαusing the combinationσha.The vertical resolution of the combination curveαacan be characterized by L cosα, where, L means the distance between transmitter and receiver. In highly deviated well, the combination curveσhahas high vertical resolution to visually determine the location of the bed boundary including thinly laminated sand-shale sequences with short transmitter-to-receiver structure. The value ofσhaon electric midpoint each layer is also very close to the true horizontal conductivity in the layers with thickness greater then L cosα.The combination curveσha fully reflects the vertical resolving power of the triaxial induction tool with visual display, significantly improves its ability to data visual explanation. In this paper, we develop a data visual explanation method and process with the combination curveσhaas the core for trixial induction logging. This visual explanation results are also considered as the initial value for nonlinear multi-parameter inversion. Some vertical 1D inversions of typical horizontal layered TI dipping bed have been done using simulating logging data and obtained satisfactory results. The iteration of inversion is fast and high precision. This is largely due to a very good vertical resolution ofσhathat provides a good initial value.The fully triaxial induction logging data processing and interpretation method with the core of a combination curveσhain this paper not only significantly improve the capacity and effectiveness of the logging data visual interpretation, but also makes the real-time logging data interpretation possible. That helps promote the extension and utilization of the equipment, and provide the support of data processing and interpretation methods for the further development of the triaxial induction tool to logging-while-drilling (LWD) and array induction logging. For geosteering, logging data interpretation in real-time is extremely important in order to arrange and adjust the location and direction of the borehole. The results in the section 3.8 show the preliminary possibility for geosteering of a fully triaxial induction tool. In addition, during a certain period, a large number data inversion of triaxial induction logging using 3D numerical modeling is unrealistic, however, making a similar data processing and interpretation as array induction tool is a more realistic way. As long as there are several different probes to detect the different formation depth and correcting the borehole effect of each measured data from corresponding probe, the vertical 1D inversion can be implemented first using our method mentioned in this paper in order to meet the match of the vertical resolution for different logging curve. The difference of the electric parametersσhandσvbetween the results of inversion of different detecting depth helps to explain the inhomogeneity of formation in radial, that is, borehole, invasion and raw formation.Some work in this paper needed to improvement and further study. First of all, we only study the combination curveσhawith one formation model including borehole and invasion, and the accuracy of its data is also not high enough. There still need higher precision simulating logging data and more formation model to improve it. Secondly, the logging data studied in this article is noise-free data, as a result of combination curveσhastudied in this article including the cross–coupling component that has lower signal to noise ratio than the main component in the middle point of bed, especially the near the central of isotropic thick bed, its anti-noise ability derive our attention and also need to be tested in the final through the actual logging data. In short, the results of this paper are still preliminary and lots of work remains to be done. We will continue to work hard.更多还原