MCMC方法在多孔介质流体预测中的应用

Flow Prediction in Porous Media Using Markov Chains Monte Carlo Approach

孔隙度和渗透率作为油气储层的重要参数,对石油产量的预测有至关重要的作用。在多孔介质流体流动过程中,孔隙度和渗透率的概率密度分布函数结构复杂,难以用经典分布予以描述,该文介绍了应用马尔可夫链蒙特卡罗方法(Markov chain Monte Carlo method)对孔隙度和渗透率进行贝叶斯估计,然后在其后验概率分布中采样,得到部分已知流量数据并计算流量的似然分布,最终得到生产曲线并用该方法成功预测了生产曲线的走势。同时在文章的最后,基于现存方法中存在的问题,提出了相关的改进方向。

Permeability and porosity, which significantly describing of subsurface properties, are essential factors for the predicting gas production. But given the fact that in porous media flow process, the probability density functions for permeability and porosity are usually too complex for direct sampling, using classical statistical ways to describe the process is troublesome. This article introduced a relative Markov chain Monte Carlo method to solve this kind of problems. In this article, we demonstrated the process consisting Bayes estimation of permeability and porosity, sampling from their posterior distribution, finding the likelihood of the flows and prediction for the production given limited training data. Also, at the end of this article also listed the problems existing in current methods and provided several potential ways for improving.