【作者】 赵启蒙；

【导师】 高美娟；

【作者基本信息】 大庆石油学院 ， 地球探测与信息技术， 2004， 硕士

【摘要】 现代油藏描述技术在勘探开发中的应用成效关键在于对油藏认识是否符合客观实际，十年来的油藏描述实践表明，要建立符合客观实际的三维油藏地质模型，关键在于不同的油藏类型，不同的勘探开发阶段，不同资料拥有程度和精度，应采用不同的油藏描述技术路线。 本文在前人研究的基础上，主要对序贯指示随机模拟方法和退火随机模拟方法作了深入的探讨。随机模拟方法越来越多的适用于储层非均质性建模中，各种随机模拟方法在其基本原理、复杂程度和应用条件诸方面均有不同，每一种方法都有它的适用条件、优点及缺点。 序贯指示随机模拟方法首先将地质信息进行离散编码，通常编码成0与1两值的指示变量，然后将克里金的基本思想用于指示变量，最终得到指示变量的克里金估计，即未知位置变量的概率分布的估计。定义一个经过所有网格节点的随机路径，在给定n个条件数据的情况下，在第一个网格节点处从随机变量的条件分布中抽取一个值，将这个新值加入到条件数据集中，在给定的n+1个条件数据的情况下，在节点处从随机变量的条件分布中抽取一个值，重新进行，直到所有节点被模拟完为止。在模拟过程中，克里金方法本身不与模型发生关系，它仅利用协方差或变异函数这个矩信息，这样就克服了克里金方法对参数的光滑效应，使得其适用于象渗透率这样变化幅度较大的地质参数的数值模拟，对地质专家和油藏工程师最关心的渗透率的特高值或特低值的分布情况预测更为全面、准确。 模拟退火方法是一种启发式蒙特卡罗法，在解决优化的问题中，它克服了常规蒙特卡罗方法盲目的随机搜索机制，而是在一定的理论指导下搜索，故能保证搜索成功。而且在搜索的过程中，不仅接受使目标函数变好的解，而且还能以一定的概率接受坏解，这样将尽量避免陷入局部极值解而达到全局最优解。同时，它几乎可以满足任意给定的统计量，正是模拟退火法的这一系列优点，本文将其应用于储层建模的优化问题中，以更好的搜寻满足建模条件的优化解。更多还原

【Abstract】 How to evaluate the effect of modern technology of reservoir description in exploration and development, this mainly based on whether what we know about reservoir match what it is. Ten-year practice of reservoir description indicate: if we want to establish objective 3-dimension geological model of reservoir, the different reservoir type, the different exploration and development phase, the different data quantity and quality decide what technical route of reservoir description we should apply.Doing the deep discussion about the method of sequential indicator stochastic simulation and annealing stochastic simulation on the foundation of study of predecessor in this paper. More and more people applied the method of stochastic simulation to heterogeneous modeling of reservoir. And each method is different from others such as basal principle, extent of complex, applied condition and so on. They all have their own applicability, advantages and disadvantages.The method of sequential indicator stochastic simulation firstly make the geological information discretization code, normally two indicator variables of 0 and 1. Then make the Kriging theory act on the variables to get the Kriging estimation of indicator variables, namely estimation of probability distribution of the variables in a unknown position. Define a random route, which pass through all the grid node, on the condition of given n conditional datum, get a value on the first grid node from the conditional distribution of stochastic variable, Add the new value into the conditional datum as a new conditional data. On the condition of current n+1 conditional datum, get a new value from conditional distribution of stochastic variable on the next node again. Then continue until all the nodes gets own value. In the course of simulation, the model has no relation to Kriging method, it only use the matrix information of covariance or variogram. So it can conquer the disadvantage that Kriging method smooth the geological parameters, and fit the numerical simulation for the geological parameters such as permeability whose values change quickly and largely. It can estimate the distribution of extra high value and extra low value of permeability more perfectly and more accurately, which the geologist and field engineer mainly care for.Simulated annealing algorithm is a kind of heuristic Monte Carlo method. In the course of solve the optimum problem, it conquer the blindfold searching mechanism of normal Monte Carlo method, and based on a appointed theory to guides search, so it can ensure a successful search. In the course of searching the optimum solution, it can accept a value make objective function good, but also a bad one. In this way it will avoid falling into a local extremum and get a global optimum value. At the same time, it can almost fulfill any statistical request. Just about its advantages, try to solve the optimum problem of reservoir modeling by it to searching better solution.更多还原