【作者】 杨宇；

【导师】 冯文光；

【作者基本信息】 成都理工大学 ， 油气田开发工程， 2004， 博士

【摘要】 以洛带气田为代表的岩性浅气藏，开发早期阶段由于静态资料少、动态资料缺乏，加上严重的层内和层间非均质性，要对气藏进行全面研究和评价，具有较大的难度。开发早期气藏描述的目的是尽可能地充分利用已有的静动态资料，从三维空间定量表征气藏静态特征，并且对单井、气藏的动态特征进行归纳和分析，在此基础上提出气田的初步开发方案。通过本次研究，取得了以下进展： 1．根据有取心资料的关键井，提出了一种基于模糊聚类和人工筛选样本相结合划分储层类型的新方法，建立储层的单井分类标准。在此基础上，提出了一种基于岩心分析的模糊综合判别方法，用于储层分类的定量判识。 2．在储层四性特征研究的基础上，针对BP神经网络的收敛速度慢、计算误差大的缺陷，首次提出了遗传-自适应神经网络算法，用于样本学习和样本预测，并且首次在洛带气田的孔隙度、渗透率和饱和度的测井二次解释中得到应用。与传统的多元统计算法和基本BP算法的结果进行的对比后，证明本文采用的遗传-自适应神经网络算法在计算速度和精度方面都有显著提高。 3．本文对洛带气田气井产能和影响因素进行了分析，在此基础上，首次提出了气井压后产能的测井预测解释方法，对洛带气田的压后产能的测井预测解释作了可行性讨论，并且采用神经网络方法进行了样本训练和预测，从而为洛带气田开发决策及时提供了一定的依据。 4．现代试井解释是开发早期从动态资料获取地层参数的重要手段，本文按照“模型判定—分段建立目标函数—最优化求解”的思路，阐述了最优化试井的计算步骤。首次提出了基于模拟退火原理的改进复合形全局最优算法，以气井的实际试井解释过程为例，具体阐明了这种全局最优化算法在实际试井分析中的应用步骤，并证明了该算法具有较强的稳定性和收敛性。 5．首次提出了开发早期气井单井的分类动态分析方法：①对产量较稳定的气井，采用最优化算法联立求解气藏物质平衡方程式和气井二项式采气方程。在此基础上进行了定产量／压力条件下的动态预测；②对产量递减井，采用分形理论中的R／S方法，对气井产量变化作了连续趋势性论证，在此基础上采用GM(1，1)模型对递减气井的生产动态进行快速拟合和预测。采用提出的开发早期气井单井的分类动态分析方法进行拟合、分析与预测，解决了气井开发早期由于资料缺乏，难以进行单井动态分析的难题。 6．针对低渗岩性气藏具有复杂含气边界，气井采用压裂开采等特点，采用了最新的径向+PEBI网格系统，将整个研究区域分成两个区域：气井区和气藏主体区。在气井区采用径向网格，可以充分刻画气井附近流动状态；在气藏主体区采用PEBI网格，可以很好地逼近复杂油藏边界形态，同时减少网格数。按上述网格系统把储层物性模型转化为气藏数值模型后，对气田的初步开发方案进行了论证，并且和多种其它方法相对照，寻求气田开采的最佳方式。更多还原

【Abstract】 During the early stage of development of LuoDai gas field,which is typical of lithologic shallow-depth gas pool, both of the static and dynamic information is scarce to describe the heterogeneous character of 3-D geologic model, so it’s quite difficult to evaluate the gas pool thoroughly. The main purpose of description in the early stage of development is to make use of all those available information, for the sake of characterizing the 3-D static model ,meanwhile, sum up and analyze the dynamic character on the scale of single well and gas field , which aims to bring up the elementary development project. There are several major progresses obtained in this research as below:1. On the basis of coring data from key-well, put forward a new method to classify reservoir ,which consists of fuzzy clustering and manual filtration. Furthermore ,taking use of coring data ,a new fuzzy differentiation is used to partition the reservoir quantificationally.2. According to the research on litho-electricity correlation of reservoir ,a new kind of GA-adaptive BP neural network is brought forward to train and forecast the samples in the secondary logging interpretation of porosity ,permeability and saturation in LuoDai gas field. The result is compared with traditional BP and linear multi-variant analysis to testify that iterative speed and accuracy is much better.3. Based on the analysis of productivity and it’s influencing factor ,the feasibility and procedure of interpreting the productivity after artificial fracturing by logging method is demonstrated in the first time in LuoDai gas field. The anticipation from sample training can be useful proof for the early judgment of development.4. The modern transitional pressure test is a major method to determine the reservoir parameter via dynamic information. The interpretation procedure is explained as "firstly ,model judgment; secondly, multi-segment object function; thirdly ,global optimization solution", during which a new covering complex optimization method improved by simulated annealing is introduced .At last an actual gas well is interpreted following this procedure as an example, which proves the strong robust character.5. The dynamic analysis based classification method in the early stage of development is brought forward in the first time ,which consists of:firstly ,the material balance equation and the binomial production equation is solved together for those wells with stable production., to yield the anticipation under constant production ,and constant THP afterwards; Secondly, the R/S fractal method is used to demonstrate the declining tendency for those withdegressive production, and the GM(1,1) model is adopted for simulation and anticipation. As a result ,the hard-nut in single well prediction because of deficiency of information is tackled better.6. Aiming at the complexity of tight shallow lithologic gas pool, which consists of irregular outer reservoir boundary, the fracturing of wellbore and so on , the brand new compound PEBI grid model is adopted . For the purpose of decreasing grid number and incarnate the reservoir ,the object area is divided into two parts: the near well bore area, which is described by radial grid, and the bulk area, which is described by PEBI grid. This type of compound grid is used in the simulation of LuoDai gas field to demonstrate the best recovery scheme , with the aiding of other methods.更多还原