基于连通性函数的多点地质统计建模参数敏感性分析方法  

Parameter sensitivity analysis method of multi-point geostatistical modeling based on connected function

作　　者：杜川 喻思羽[1] 李少华[1] 方红 DU Chuan;YU Siyu;LI Shaohua;FANG Hong(School of Geosciences Yangtze University,Wuhan 430100,China;Exploration and Development Research Institute of Liaohe Oilfield Branch of CNPC,Panjin 124010,China)

机构地区：[1]长江大学地球科学学院,武汉430100 [2]中国石油天然气辽河油田分公司勘探开发研究院,盘锦124010

出　　处：《物探化探计算技术》2022年第2期195-202,共8页Computing Techniques For Geophysical and Geochemical Exploration

基　　金：国家自然科学基金(42002147,41872129);2019年度地质资源与地质工程一流学科开放基金(2019KFJJ0818021)。

摘　　要：不同建模参数不仅影响建模质量,也决定了油气藏预测开发风险,找到一个合适的建模参数设置是获得良好建模的前提,对降低油藏开发风险具有积极意义。这里以多点地质统计学经典算法Snesim为例,通过连通性函数定量化地分析Snesim建模过程中参数两个关键参数(搜索节点数、多重网格层数)对建模效果的影响。针对以上参数利用连通性函数和余弦相似度CosSim函数,评价建模参数集的多点地质统计随机模型与训练图像的空间相关性及结构特征相似性,进而建立基于连通性函数的空间相关性评价指标与建模参数的关系曲线,选取评价指标开始平稳、进入平台区域时的拐点所对应的参数值作为最优参数。大于最优值参数将变得不敏感,模型效果不会随着参数值增加而得到明显改善,相比传统人工视觉判别方法,本方法可以客观定量化地描述建模参数的敏感性。Different modeling parameters not only affect the modeling quality,but also determine the prediction and development risks of oil and gas reservoirs.To find a suitable setting of modeling parameters is the premise of obtaining good modeling,which has positive significance for reducing the development risks of oil reservoirs.Taking Snesim,a classical algorithm of multi-point geostatistics,as an example,this paper quantitatively analyzes the influence of two key parameters(the number of search nodes and the number of multi-grid layers)on the modeling effect in Snesim modeling process through connectivity function.According to the above parameters,the connectivity function and cosine similarity CosSim function are used to evaluate the spatial correlation and structural feature similarity between the multi-point geostatistical random model of the modeling parameter set and the training image,and then the relationship curve between the spatial correlation evaluation index and the modeling parameters based on the connectivity function is established.The parameter value corresponding to the inflection point when the evaluation index starts to be stable and enters the platform area is selected as the optimal parameter.Parameters larger than the optimal value will become insensitive,and the model effect will not be significantly improved with the increase of parameter value.Compared with the traditional artificial vision discrimination method,the proposed method can objectively and quantitatively describe the sensitivity of modeling parameters.

关 键 词：连通性概率 多点地质统计学 储层建模 参数优选 参数敏感性 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]