协方差局地化方法在自动历史拟合中的应用

Application of covariance localization method in automatic history matching

 张黎明¹ 周建人¹ 张凯¹ 董振华² 朱孟高³

中国石油大学（华东），山东 青岛 266580 中国科学院声学研究所，北京 100190 中国石化胜利油田分公司滨南采油厂，山东 滨州　２５６６００）

China University of Petroleum, Qingdao 266580, China Institute of Acoustics, Chinese Academy of Sciences, Beijing 100190, China Binnan Oil Production Plant, Shengli Oilfield Company, SINOPEC, Ｂinzhou 256600, China)

 集合卡尔曼滤波（EnKF）是自动历史拟合领域应用较为广泛的智能算法。为了解决该算法应用过程中出现的滤波发散问题，文中运用协方差局地化方法，综合考虑先验地质模型的相关半径和观测数据的观测影响半径，计算临界半径长度，并在油藏模型模型中的水平和垂直方向引入局地化相关函数，滤除远距离观测数据产生的相关噪音，降低协方差矩阵计算过程中的伪相关。将改进的算法编程实现并运用理论实例进行验证，并对比反演得到的渗透率场与真实渗透率场，结果表明，改进后的理论在渗透率反演精度方面提高了28%，数据拟合速度提高了16%。反演的渗透率场能够清晰刻画出大孔道，对于优势通道识别、精细油藏描述具有重要意义。

 Ensemble Kalman Filter (EnKF) is one kind of intelligent algorithms and it is widely used in the field of automatic history matching. In this paper, a local covariance method is proposed to solve the problem of filtering divergence in the process of the algorithm application. This method firstly calculates a reliable critical radius which is based on the relevant radius of the prior geological models and the relevant radius of the observation data. Then the local correlation functions are added to the assimilation system in the horizontal and vertical directions on the reference of the critical radius. These local correlation functions could filter out the relevant noise of the remote observation data and improve the calculation accuracy of the covariance matrix. The improved algorithm is then programmed and examined by the theoretical examples. By comparing the inversed permeability field with the real permeability field, it indicates that the improved algorithm can increase the permeability inversion precision by 28% and shorten the running time by 16%. The inversed permeability field can clearly distinguish the large pores, which is significant for the predominant pathway identification and fine reservoir description.

10.6056/dkyqt201701011