利用马尔科夫链模型对多点地质统计建模影响因素分析及优化

Multi-point Geostatistics Modeling Factors Analysis and Optimization by Using Markov Chain Model

多点地质统计方法作为一种随机建模方法,在很多领域得到了广泛的应用,取得了很好的效果.但模拟过程中随机性强,难以控制其模拟效果.以地震尺度的精细离散模型(地震相)为训练图像,通过大量的实验探索,分析多点模式的选取对多点地质统计方法模拟效果的影响,提取模式大小、模式形态(各向异性)以及多级网格约束三个因素对建模结果影响.模式大小应根据模拟目标的尺度进行选择,模式的形态应与模拟地质目标的各向异性一致,多级网格控制的多少应权衡模拟目标的复杂性和计算速度.成功应用马尔科夫链模型,评价模型的空间结构恢复效果,对影响参数遍历设置和组合优化,探索了训练图像的最优多点模式参数设置组合.

Multiple-point Geostatistics methods acts ,as a stochastic modeling approach has been widely used in many areas and achieved good results. Because of the strong randomness in simulation process, it is difficult to control the simulation results. Based on Multiple-point Geostatistics knowledge and training image made by precision earthquake discrete modeling （seismic face modeling）, through a large number of experiments to explore, the article analyzed the effect of pattern selection method for Multiple-point Geostatistics simulation. Studles have shown that model size, model shape （Anisotropy） and muitigrid are the three significantly affected factors on the results of modeling. The model size should be selected by measuring the target scale. The model shape should be consistent with the anisotropy of simulation geologic targets. The selection of number of multigrid should weight the complexitY of simulated target and computation speed. And through the Markov chain model to evalUate recovery effect of spatial structure. Traverse setting and optimize the affecting parameters, Arrive at the optimal combination of parameter settings of this training image.