协方差局地化方法在自动历史拟合中的应用

发布时间：2018-06-20 01:00

本文选题：历史拟合 + 集合卡尔曼滤波　；参考：《断块油气田》2017年01期

【摘要】：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。
[Abstract]:Ensemble Kalman filter (EKF) is a widely used intelligent algorithm in the field of automatic history fitting. In order to solve the problem of filtering divergence in the application of the algorithm, a covariance localization method is used to calculate the critical radius length by taking into account the correlation radius of the prior geological model and the observed influence radius of the observed data. The local correlation function is introduced in the horizontal and vertical direction of the reservoir model to filter the correlation noise generated by the remote observation data and to reduce the pseudo-correlation in the calculation of covariance matrix. The improved algorithm is programmed and verified by a theoretical example, and the permeability field obtained by inversion is compared with the true permeability field. The results show that the improved theory improves the accuracy of permeability inversion by 28% and the speed of data fitting increases by 16%. The inverse permeability field can clearly depict the large pore path, which is of great significance for the identification of dominant channels and fine reservoir description.
【作者单位】：中国石油大学(华东);中国科学院声学研究所;中国石化胜利油田分公司滨南采油厂;
【基金】：国家自然科学基金项目“基于嵌入式离散裂缝模拟的裂缝性油藏缝网分布反演”(61573018) 国家科技重大专项课题“海上高含水期油藏大井距井网加密矢量优化与注采结构调整技术”(2016ZX05025001-006) 山东省自然科学基金项目“基于嵌入式离散裂缝模拟的裂缝性油藏缝网分布反演”(ZR2015EL014)
【分类号】：P618.13;TE319

【相似文献】