1.China University of Petroleum (East China), Qingdao 266580, China 2.Department of Oilfield Exploration and Development, Sinopec, Beijing 100728, China
Fund Project:Project supported by the National Science and Technology Major Project of the Ministry of Science and Technology of China (Grant No. 2016ZX05060-010), the Fundamental Research Fund for the Central Universities, China (Grant No. 17CX06007), and the National Natural Science Foundation of China (Grant No. 51404292).
Received Date:31 August 2018
Accepted Date:21 January 2019
Available Online:01 March 2019
Published Online:20 March 2019
Abstract:Numerical simulation of a fractured-vuggy porous medium is a challenging problem. One reason is the coexistence of matrix, fractures and vugs on multiple scales that need to be coupled, and the other reason is that the high-resolution fractured-vuggy model may contain up to several millions of gridcells in applications, which brings severe computational challenges into the numerical methods. Therefore, the requirement for accurate and efficient technique is widely increasing. Fractured-vuggy porous medium is generally represented by triple-continuum model in which the matrix system, fracture system and vug system each are treated as a parallel continuous system. Although triple-continuum model is widely used because of its easy-implementation and high efficiency, it fails to capture the detailed flow patterns of reservoir with disconnected long fractures. Discrete fracture-vug network (DFVN) model can precisely model the fluid flow in fractures and vugs. However, the simulation of this model is deemed intractable even with the advent of supercomputers because of the large amount of calculation. In view of the fact that the multigrid method is now well known as one of the fastest method of solving elliptic problems, in this paper we introduce a nearly linear complexity multiresolution decomposition method for fluid flow in a fractured-vuggy reservoir. The detailed flow patterns are described by combing the advantages of continuum model and discrete model. That is, the homogenization theory is used to construct an equivalent permeability in each coarse grid block in which the vugs and small-scale fractures are represented by discrete fracture-vug network model. We decompose the solution space into several subspaces and then we compute the corresponding solutions of heterogeneous discrete fracture network model in each subspace. Gamblets are constructed and they are elementary solutions of hierarchical information games associated with the process of computing with partial information and limited resources. These gamblets have a natural Bayesian interpretation under the mixed strategy emerging from the game theoretic formulation. This method could realize its fast simulation by decomposing the solution space into a direct sum of linear subspaces that are orthogonal to each other. Finally, the pressure difference distribution of fractured-vuggy porous medium is obtained by combing the DFVN solutions of all subspaces. Numerical results are presented to demonstrate the accuracy and efficiency of the proposed multiresolution decomposition method. The results show that this method is a promising method of numerically simulating the fractured-vuggy porous medium. Keywords:fractured-vuggy porous media/ multiresolution decomposition/ discrete fracture-vug network model/ numerical simulation
如图2所示, $\varOmega $被分为若干${2^{ - k}} \times {2^{ - k}}$的子单元${\left( {\tau _i^{\left( k \right)}} \right)_{i \in {I^{\left( k \right)}}}}$. 令$\phi _i^{\left( k \right)} = 1_{\tau _i^{\left( k \right)}}^{}/\sqrt {\left| {\tau _i^{\left( k \right)}} \right|} $, 其中$\left| {\tau _i^{\left( k \right)}} \right|$为$\tau _i^{\left( k \right)}$的体积, $1_{\tau _i^{\left( k \right)}}^{}$为$\tau _i^{\left( k \right)}$的指示函数. 对于边界条件问题, 使用非零边界的测试函数. 通过假设A和B在进行一个博弈游戏来定义基函数, 具体理论证明见文献[24]. 根据博弈理论[27,28]并结合流动方程, B的最佳选择为 图 2 区域$\varOmega $的网格剖分示意图 Figure2. Schematic of grid partition of solution space.
$p ^{\left( k \right)}\left( x \right){{ = }}\sum\limits_{i \in {{\cal{I}}^{\left( k \right)}}} {\psi _i^{\left( k \right)}\left( x \right)} \int\limits_\varOmega {p\left( y \right)\phi _i^{\left( k \right)}\left( y \right)} \;{\rm{d}}y, $
其中$\psi _i^{\left( k \right)}$为基函数, 定义为
$\psi _i^{\left( k \right)} = \mathbb{E}\left[\!{v|\int_\varOmega \!{v\left( y \right)\phi _j^{\left( k \right)}\left( y \right){\rm{d}}y = {\delta _{i, j}}, \;\;i \!\in\! {{\cal{I}}^{\left( k \right)}}} } \!\right]\!.$
式中${p_{\rm{f}}}$为参考压力解, ${p_{{\rm{mg}}}}$ 为多尺度分解法所计算的压力解. 图7给出了参考解和多重网格解的对比. 可以看出, 当k = 3时多重网格法所得压力分布与参考解几乎一致, 因此多重网格解能够精确地反映非均质性, 从而捕捉到小尺度裂缝和溶洞的影响. 表2列出了不同k时的计算误差, 其中当k = 4时为多层分解法的精确解, 当k = 3和k = 2时为均化计算, 计算误差均很小, 体现了该方法的精确性. 小尺度缝洞模型的计算结果表明, 此方法可以有效地进行非均质地层以及小尺度缝洞介质的流动模拟. 图 7 对于小尺度缝洞模型, 参考解和多重网格解对比 (a) 参考解; (b) k = 3时的多重网格解 Figure7. Comparison of reference solution and gamblets solution for a small-scale-fractured vuggy porous medium: (a) Reference solution; (b) gamblets solution with k = 3.
k
2
3
4
计算误差
0.0823
0.0132
3.2549 × 10–15
表2对于小尺度缝洞模型, 不同k时的计算误差 Table2.Relative error in different k for a small-scale-fractured vuggy porous medium.
24.2.长裂缝模型算例 -->
4.2.长裂缝模型算例
算例2考虑存在长裂缝的情况. 考虑如图8所示的裂缝型介质, 在小裂缝和溶洞的背景下包含一条长裂缝. 裂缝开度df = 1 × 10–3 m, 裂缝渗透率为$d_{\rm f}^2/12 $${\text{μ}}{\rm m}^2$, 基岩孔隙度$\phi =0.2$. 边界均为不流动边界, 流体从上往下流动. 采用离散裂缝模型表征裂缝. 图 8 长裂缝介质模型 Figure8. Geometrical model of a fractured-vuggy porous medium with a long fracture.
本算例考虑长裂缝与小裂缝溶洞共存的情况. 压力分布图9展示了不同k时多尺度分解法与参考解的对比. 当k = 3时计算误差为0.0131, 当k = 2时计算误差为0.0741, 在进行均化的前提下依然能够反映裂缝的存在, 保证较高的计算精度. 参考解与多重网格解的对比结果说明, 本文构建的多尺度分解法能够有效地处理长裂缝的情况, 并能通过均化大幅减少计算量, 同时保持较高的计算精度. 图 9 对于长裂缝模型, 参考解和多重网格解对比 (a)参考解; (b) k = 3时的多重网络解; (c) k = 2时的多重网格解 Figure9. Comparison of reference solution and gamblets solution for a fractured-vuggy porous medium with a long fracture: (a) Reference solution; (b) gamblets solution with k = 3; (c) gamblets solution with k = 2.
24.3.大尺度缝洞模型算例 -->
4.3.大尺度缝洞模型算例
自然介质中, 除了小裂缝, 往往伴随着压裂等增产措施产生大裂缝, 该算例检验本文提出的方法对大裂缝的模拟能力. 图10所示为10 m × 10 m缝洞型介质, 其中包含的小尺度裂缝和溶洞的参数与算例1相同. 在包含小尺度溶洞的同时, 还包含着尺度较大的裂缝网络系统. 裂缝网络由6条长裂缝组成, 长裂缝开度df = 1 × 10–3 m, 裂缝渗透率为$ d_{\rm f}^2/12$${\text{μ}}{\rm m}^2 $, 基岩孔隙度$\phi = 0.2$. 流体从上至下流动. 图 10 大尺度缝洞介质几何模型 Figure10. Geometrical model of a large-scale-fractured vuggy porous medium.
多孔介质中的长裂缝作为导流通道, 对压力分布有显著的影响, 并且裂缝之间也会相互影响. 该方法在每一个子空间内求解离散裂缝模型, 图11为离散裂缝模型在各层上的解, 可以很明显地看出, 各个层次上的解都能反映出裂缝的存在. 然后将k个层次上的解加起来, 即得到最终的解. 图 11 大尺度缝洞介质模型在各层上的解 Figure11. Solutions in different levels for a large-scale-fractured vuggy porous medium.
图12给出了参考解和多重网格解的对比, 可以看出, 本文构建的数值算法在捕捉小尺度裂缝和溶洞的同时, 能够精确模拟长裂缝对压力场的影响, 并反映裂缝间的相互作用. 结合表3可知, 当均化程度较大(k = 1)时存在较大误差. 图13给出了沿x = 5 m的参考解与不同k时多重网格解的对比. 裂缝在流动过程中可视为一个等势体, 对应于图11中较平缓的部分, 与实际符合, 随着均化程度加大误差变大. 该算例展示了本文构建的数值方法对长裂缝的精确模拟能力. 图 12 对于大尺度缝洞介质模型, 参考解和多重网格解对比 (a) 参考解; (b) k = 3时的多重网格解; (c) k = 2时的多重网格解; (d) k = 1时的多重网格解 Figure12. Comparison of reference solution and gamblets solution for a large-scale-fractured vuggy porous medium: (a) Reference solution; (b) gamblets solution with k = 3; (c) gamblets solution with k = 2; (d) gamblets solution with k = 1.