多点地质统计学算法Snesim参数敏感性分析

Parameter sensitivity analysis of multiple-point geostatistical Snesim program

 王红 李少华

 （长江大学地球科学学院，湖北 武汉 430100）

 (College of Geosciences, Yangtze University, Wuhan 430100, China)

 Snesim是第一个被应用到油田储层建模中的多点地质统计学算法。文中以弯曲河道作训练图像，对该算法建模过程中常涉及到的参数（如目标边际分布与伺服系统参数、搜索半径与搜索角、旋转角与仿射率，以及多重网格层数等）进行敏感性分析。分析结果表明：当目标相比例值与训练图像边际概率值接近时，模拟结果较合理；增大伺服系统参数可使模拟结果的边际概率接近目标相比例，但伺服系统参数设置过高将降低河道连续性；搜索邻域应与模拟维度相适应；局部旋转角和仿射率可将训练图像中的平稳与非平稳信息融合起来，用于趋势模拟；多重网格模拟中，网格层数取决于最大尺度结构信息的大小。以上认识可为利用Snesim算法进行储层建模提供一定的指导。

 Snesim is a multiple-point geostatistical program which was applied to reservoir modeling for the first time. This paper makes the first attempt. Using sinuous channels as training image, this paper studies the sensitivities of important input parameters such as target proportion, servosystem parameter, search radii and angles for search ellipsoid, rotation angles and affinity, number of multiple grids. The analysis shows that: the target proportion should not be too different from that of the training image; the simulated channel proportion can approach the target proportion by using high servosystem parameters, whereas an excessive high servosystem parameter is not recommended as it may reduce the channel continuity; the search neighborhood should be adapted to the dimensions of the structures to be reproduced; the non-stationary information typically found for facies distribution can be incorporated into stationary information using rotation and affinity ratio parameters; multiple-grid simulation is always recommended, the number of grid levels depends on the size of the largest scale structures. These understandings can be used as references and guidance for reservoir modeling with Snesim.

10.6056/dkyqt201704012