多点地质统计学建模参数敏感性分析

Sensitivity Analyses of Parameters for Multiple-point Geostatistics Modeling

与基于像元的两点模拟和基于目标的模拟相比,多点地质统计学能较好地忠实于硬数据(井数据)和再现复杂的地质体形态。在介绍多点模拟算法(Snesim)实现流程的基础上,对Snesim算法中重要的输入参数进行了敏感性分析,结果表明:目标比率越接近训练图像的边缘相概率,模拟效果越好;目标比率一定时,提高伺服参数可使模拟相的比率更接近目标体,但以损失相结构信息为代价;搜索邻域的设计、网格级数的选择取决于训练图像的大小以及需重现的结构信息;此外,在一个较小的数据事件重复数下,随着最大条件数据的增加,其结构信息的再现效果越好,所需机时则呈线性增加。可见,参数设置对多点地质建模中模拟效果的好坏至关重要。

Compared with pixel-based two-point simulation and object-based simulation,multi-point geostatistics method can agree with the data of wells and reproduce the morphology of complex geologic bodies better. The sensitivity of the important input parameters in Snesim algrithm is analyzed based on the introduction of the implementation process of multi-point simulation algorithm. The result shows that,the closer the target ratio is to the facies probability in training image edge,the better the simulation result is; when the target ratio is constant,the improvement of servo parameters can make the ratio of simulation facies closer to the target,but this is at the expense of the loss of the structural information of facies; the design of the search neighborhood and the choice of the grid series depend on the size of the training image and the structure information to be reproduced. In addition,under the smaller repeat number of data events,the increase of the largest conditional data will make the reproduction effect of the structural information better,but the required computing time will linearly increase. It is shown that the reasonable setting of parameters is very important to the simulation effect of the multiple-point geology modeling.