【作者】 王婧慈；

【导师】 郭海敏；

【作者基本信息】 长江大学 ， 地球探测与信息技术， 2013， 博士

【摘要】 在油田水驱开发过程中,准确估算剩余油饱和度及其分布规律,对于进一步提高二次采油开发效果、有针对性地实施三次采油具有十分重要的意义。因此,本文针对动、静态测井评价及剩余油综合预测方法展开研究,为清污混注水淹层剩余油精确定位工作提供了一些新的研究思路。本次研究中进行的主要工作及取得的认识如下。(1)首先,在两种假设条件下对不同注入水条件下混合地层水矿化度变化规律进行研究。进而,利用岩电实验对水淹开发过程中饱和度评价模型中的参数加以确定。结合尕斯NI-N2’油藏水淹层岩心实验资料来看,水驱开发对储层m值影响不大。但注水开发各个阶段中饱和度指数n呈现阶段性变化,随着注入水矿化度的不同,n的变化特征也不尽相同。最终,综合模拟研究结论和岩心实验资料建立出混合地层水动态分析模型,该模型可对注入水与原始地层水之间的离子交换现象做出更为细致的考虑,从而为后续油水饱和度的精细计算打下基础。(2)在利用大量裸眼井测井资料进行水淹层剩余油饱和度的程序化计算时,为解决混合地层水电阻率不易确定和无法有效考虑饱和度计算模型中的参数随地层含水饱和度的增加而发生的变化这两项难题,进行了以下研究。首先对研究区块水淹特征进行分析,进而建立了研究区块孔隙度、渗透率、原始含油饱和度、原始地层水矿化度、混合地层水电阻率等相应模型,最后基于水驱油藏混合地层水矿化度的变化特征,利用变参数阿尔奇公式构建了含水饱和度与注入水矿化度关系矩阵。在注入水矿化度资料不全的情况下,可利用原始含油饱和度确定出注入水矿化度。最终,针对解释层中的每个采样点确定出混合地层水矿化度和流体饱和度。该方法可回避传统方法的缺陷,植入解释程序后批量应用效果良好,给水淹层定量评价开辟了新的思路,并可移植到过套管剩余油饱和度测井评价中去。(3)利用PNN测井资料进行剩余油评价时,为了获得在平面上、垂向上、时间推移上具有针对性的解释参数,首先选择出未经开采及注入水未波及的相对封闭层段作为标准层。进而把经过必要改进的自适应遗传算法编入解释程序,并设定好各未知参数的进化范围。单井解释时,利用改进的算法对单井标准层样本点进行处理,实现测井资料二次校正并最终确定出针对单井、针对油组的饱和度模型参数。应用效果表明该方法更贴近测量环境的非均质性,有效弥补了井间、层间差异所带来的解释误差,因而解释结果与实际生产动态更加相符。(4)清污水交替注入型水淹层混合地层水矿化度变化较大。在进行脉冲中子类过套管剩余油饱和度测井解释时,地层水矿化度的多变性将导致地层水宏观俘获截面这一重要参数不易确定。对研究区块水淹层混合地层水的水型和矿化度进行了分析,分析结果表明不同小层间混合地层水矿化度变化较大,因而逐层计算混合地层水宏观俘获截面十分必要。以脉冲中子-中子测井评价为例,首先构建原始含油饱和度解释模型,进而对地层流体宏观俘获截面进行信息提取,最后提出了利用动静态测井资料确定混合地层水宏观俘获截面的方法。植入解释程序后,该方法还可对同一解释层内水淹程度不一致的地方分别计算出混合地层水宏观俘获截面。由于计算过程中较好地考虑了层间差异性,因而计算结果准确性更高。(5)论文中分析了单因素评价剩余油饱和度的几项缺陷,并提出了基于多因素水淹指数的剩余油综合预测方法。在计算多因素水淹指数的过程中,有必要将各水淹强度评价指标反映水淹程度的能力纳入考虑范围之内。因此,在分析研究区块多种动、静态资料的基础上,建立了一种基于椭圆基函数(Ellipse Basis Function)的模糊神经网络水淹指数预测系统。该预测系统可根据学习样本自行创建或删减模糊规则,并考虑输入变量的动态权值。测井资料信息量庞大,因此这种具有自学习机制的预测系统更有利于有效信息的提取和利用,特别对于复杂储层而言,减轻了预测过程中对先验信息的依赖程度,因而效率和精度更高。更多还原

【Abstract】 During the exploitation process of an oil field, it is of great importance to make an accurate estimate of the saturation and distribution of remaining oil, so as to further enhance the secondary, or even the third oil recovery. Therefor, dynamic and static logging evaluation methods and remaining oil prediction methods were discussed in this dissertation, which provides some new research ideas for accurate positioning of remaining oil in water flooded interval injected by fresh water and sewage.The major research and results of the dissertation are as follows.(1) First of all, under two conditions assumed, the change law of the mixed formation water salinity was researched under the conditions of variable injection water salinity. Moreover, the parameters in the saturation assessment model during waterflooding were confirmed based on rock electricity mechanism experiments data. Integrating the core experiment data on the water-flooding layer in Gasikule oil field N1-N21reservior, the water-flooding development has no great influence on porosity index m. However, the saturation index n in the various stages of water-flooding development shows stage-based changes, with the change of injected water salinity, the change characteristics of n are all different. Finally, a mixed formation water dynamic analysis model was hereby established integrating the simulative researches and experimental data, such model could carry out more careful considerations on the ion exchange phenomenon between injected water and original formation water, so as to set up foundations for the elaborative calculation of remaining oil and water saturation.(2) When a programmable calculation was made on the remaining oil saturation in the water-flooded layer by a large amount of open-hole log information, in order to resolve such two difficulties as uncertainty of the mixed formation water resistivity and incapably-effective consideration of parameter changes in saturation calculation models with the increase of formation water saturation, the following research was done in this dissertation.First of all, the flooding characteristics of the study area was analyzed, and then relative models were established, including porosity,permeability,original oil saturation, original salinity of formation water, mixed formation water resistivity and so on, so as to establish a rational matrix between water saturation and injected water salinity by the Ariche formula of variable element based on the change characteristics of the mixed formation water salinity in the water-drive reservoir. Under the conditions of no complete injected water salinity data, the injected water salinity was confirmed by the original oil saturation so as to finally determine the mixed formation water salinity and fluid saturation at each sampling point of the interpretative layer. This method can evade the defects of traditional methods and batches of applications can have favorable effects after insertion of the interpretative procedures, which open up new thoughts for the quantitative evaluation on the water-flooded layer. This method can also be applied to the through casing saturation logging evaluation.(3) The distribution of remaining oil and the changes of reservoir properties can be dynamically reflected by PNN logging. However, the accuracy is affected by many factors.In different regions,and even in different wells in the same area,the parameters of models are not identical. In time dimension, in order to obtain more accurate parameters in horizontal and vertical directions, a method has been discussed in this dissertation. Firstly, the relatively sealed layers which are not affected by exploitation and waterflood development were selected as marker bed. The vertical distance between marker bed and major reservoirs should be within a certain range. Then necessarily improved adaptive genetic algorithm with evolution range had been properly compiled into the interpretation program, so that the second correction of log data could be considered and the most suitable parameters could be determined for each well.The practical application shows that this method accurately reflects the heterogeneity of the measured environments. The interpretation process effectively compensates for the errors caused by difference between wells and layers,and the evaluation results are more consistent with the practical production performance.(4) The salinity of mixed water in water flooded zones injected by fresh water and sewage is variable. Therefore, as an important parameter in remaining oil saturation calculation model in cased hole, the macroscopic capture cross section of mixed formation water is difficult to determine. The distribution regulation of salinity and solutes of mixed formation water in the water flooded layer in study area shows that it is very necessary to calculate the macroscopic capture cross section of mixed formation water layer by layer.Pulsed neutron-neutron well logging evaluation in water-flooded zone was taken as an example in this dissertation. First of all, original oil saturation model was established. And then the macroscopic capture cross section of formation fluid was obtained. Finally, macroscopic capture cross section of mixed formation water was calculated by static and dynamic logging information. In view of the water flooding degree of some sampling points is inhomogeneous in the same layer, the method can also calculate the corresponding macroscopic capture cross section of mixed formation water.The practical application shows that this method accurately reflects the variability of the salinity of formation water. The calculation accuracy has been improved because the difference between layers has been taken into consideration.This method also provides a new idea for the quantitative assessment of water flooded zones injected by fresh water and sewage by using some comprehensive methods which take dynamic and static logging data into consideration.(5) Several defects of the calculation of remaining oil saturation based on single factor were pointed out and remaining oil prediction method based on multi-factor were discussed in this dissertation. It is very necessary to consider the weight of each flooding strength indicator in calculation of multifactorial flooding index. Therefore, a fuzzy neural network prediction system of multifactorial flooding index based on ellipse basis function was established on the basis of the analysis of a variety of static and dynamic data of Gasikule oil field N1-N21reservior. This prediction system can create or delete fuzzy rules by analyzing samples and take the dynamic weight values of the input variables into consideration. The information contained in the log data is enormous. By using this prediction system with self-learning mechanism, the extraction and utilization of information is more effective. Practical application shows that the accuracy of identification is high. Especially for complex reservoirs, the application of this Fuzzy Neural Networks on reservoir characteristic parameters prediction improves the precision of prediction results and reduces the dependency on prior informations.更多还原