Modelling saline intrusion using dynamic mesh optimization with parallel processing
Modelling saline intrusion using dynamic mesh optimization with parallel processing
复制标题
DOI：
10.1016/j.advwatres.2022.104189
复制
发表时间：
2022-04
影响因子：
4.7
通讯作者：
A. Hamzehloo;M. Bahlali;P. Salinas;C. Jacquemyn;C. Pain;A. Butler;M. Jackson
中科院分区：
文献类型：
--
作者：
A. Hamzehloo;M. Bahlali;P. Salinas;C. Jacquemyn;C. Pain;A. Butler;M. Jackson
沿海含水层中的盐水入侵（SI）是一个全球性问题，有可能污染超过10亿人使用的地下水。沿海含水层SI的数值模拟是风险评估、含水层管理和资源调节的关键工具，但极具挑战性，因为盐锋上的混合区通常非常狭窄，延伸超过米或10米，而盐锋本身可能横向延伸超过一个大的（即许多公里）三维（3D）域。此外，含水层可能具有很大的异质性，使前缘的运动和几何形状更加复杂。我们在这里测试使用动态网格优化（DMO）在一个并行计算框架，以更高的精度和更低的计算成本相比，固定网格的方法来模拟SI。该框架使用双控制体积有限元（DCVFE）方法，并在开源帝国理工学院有限元油藏模拟器（IC-FERST）中实现，但也可以在其他基于有限元的模拟器中实现。我们确认的准确性和收敛性，使用测试案例的基础上经典的“亨利”SI问题，证明使用DMO获得的解决方案需要显着更少的元素，因此具有更低的计算成本相比，等效的固定网格解决方案。我们将该框架应用于模拟非均匀白垩含水层中盐水入侵的现实3D案例研究，模拟速度超过120倍。我们建议，并行DMO提供了显着的优势，现有的方法来模拟SI。
Saline intrusion (SI) in coastal aquifers is a global problem with the potential to contaminate groundwater used by over a billion people. Numerical modelling of SI in coastal aquifers is a key tool for risk assessment, aquifer management and resource regulation, but is extremely challenging because the mixing zone across the saline front is often very narrow, extending over metres or 10’s metres, yet the saline front itself may extend laterally over a large (ie many km) three-dimensional (3D) domain. Moreover, the aquifer may be highly heterogeneous, further complicating the movement and geometry of the front. We test here the use of dynamic mesh optimization (DMO) in a parallel computational framework to simulate SI with higher accuracy and lower computational cost compared to fixed-mesh approaches. The framework uses a double control-volume-finite-element (DCVFE) method and is implemented in the open-source Imperial College Finite Element Reservoir SimulaTor (IC-FERST), but could be implemented in other FE-based simulators. We confirm accuracy and convergence using test cases based on the classic’Henry’SI problem, demonstrating that solutions obtained using DMO require significantly fewer elements and therefore have much lower computational cost compared to equivalent fixed mesh solutions. We apply the framework to a realistic 3D case study simulating saline intrusion in a heterogeneous chalk aquifer, demonstrating simulation speed-up in excess of 120×. We suggest that parallelized DMO offers significant advantages over existing methods to simulate SI.