石油学报 ›› 2019, Vol. 40 ›› Issue (6): 716-725.DOI: 10.7623/syxb201906007

• 油田开发 • 上一篇    下一篇

基于单井敏感性局域化EnKF的油藏辅助历史拟合方法

刘伟1, 赵辉1, 雷占祥2, 陈增顺3, 曹琳1, 张凯4   

  1. 1. 长江大学石油工程学院 湖北武汉 430100;
    2. 中国石油勘探开发研究院 北京 100083;
    3. 江苏油田矿业开发总公司 江苏扬州 225000;
    4. 中国石油大学(华东)石油工程学院 山东青岛 266580
  • 收稿日期:2018-12-13 修回日期:2019-02-27 出版日期:2019-06-25 发布日期:2019-07-02
  • 通讯作者: 赵辉,男,1984年3月生,2006年获中国石油大学(华东)学士学位,2011年获中国石油大学(华东)博士学位,现为长江大学石油工程学院教授、博士生导师,主要从事油气田开发、油藏工程及优化控制工程方面的研究。Email:zhaohui-712@163.com
  • 作者简介:刘伟,男,1994年10月生,2016年获长江大学学士学位,现为长江大学石油工程学院硕士研究生,主要从事油藏数值模拟历史拟合方面的研究。Email:liuwei9346@163.com
  • 基金资助:

    国家自然科学基金项目(No.51674039,No.51604035,No.51874044)和国家科技重大专项(2016ZX05014)资助。

Reservoir assisted history matching method using a local ensemble Kalman filter based on single-well sensitivity region

Liu Wei1, Zhao Hui1, Lei Zhanxiang2, Chen Zengshun3, Cao Lin1, Zhang Kai4   

  1. 1. School of Petroleum Engineering, Yangtze University, Hubei Wuhan 430100, China;
    2. PetroChina Research Institute of Petroleum Exploration and Development, Beijing 100083, China;
    3. Jiangsu Oilfield Mining Development Corporation, Jiangsu Yangzhou 225000, China;
    4. School of Petroleum Engineering, China University of Petroleum, Shandong Qingdao 266580, China
  • Received:2018-12-13 Revised:2019-02-27 Online:2019-06-25 Published:2019-07-02

摘要:

常用的油藏自动历史拟合算法多数存在梯度计算不准确、产生伪相关性等问题,导致参数修正错误、模型反演失真。通过建立一种基于单井动态敏感性的局域化集合卡尔曼滤波(FMM-CL-EnKF)历史拟合方法,解决了传统距离截断方法处理伪相关性时与实际地层状况不匹配的问题。基于程函方程,根据地质模型静态参数场信息,快速追踪压力波从井点到地层网格的传播时间确定各单井动态最大敏感性区域,从而构建局域化矩阵。同时,结合集合卡尔曼滤波(EnKF)方法,实现数据同化方法梯度的矫正,减弱伪相关,通过逐步拟合生产动态达到更新油藏模型和获取最优估计的目的。概念算例和矿场算例的计算结果表明,FMM-CL-EnKF方法在模型集合生产动态拟合效果及反演模型参数场准确性等方面均优于标准EnKF方法。

关键词: 历史拟合, 伪相关, 局域化方法, 程函方程, 集合卡尔曼滤波

Abstract:

At present, the commonly used reservoir automatic history matching algorithms have problems such as inaccurate gradient calculation and pseudo-correlation, resulting in incorrect modification of geological parameter and model inversion distortion. The history matching method of ensemble Kalman filter with localization based on single-well dynamic sensitivity region (FMM-CL-EnKF)is newly established to solve the problem that the pseudo-correlation treated using the traditional distance truncation method may not conform to real geological characteristics. Based on Eikonal equation, according to the static parameter information of geological model, the propagation time of pressure wave from well point to a grid is rapidly calculated to determine the maximum sensitivity region of single well, so as to construct the localization matrix. In combination with EnKF, this study realizes the correction of spurious gradient information using data assimilation method, weakens the pseudo-correlation, and achieves the purposes of updating reservoir model and acquiring optimal estimate by gradually matching dynamic production data.The applications of synthetic and real examples show that the proposed FMM-CL-EnKF method is superior to the standard EnKF method in the matching effect of dynamic production data and the inversion accuracy of model parameter.

Key words: history matching, pseudo-correlation, localization method, Eikonal equation, ensemble Kalman filter

中图分类号: