Introduction
As global urban centers transition toward renewable energy, shallow geothermal systems—specifically ground-source heat pumps—are increasingly deployed for building climate control. While efficient, the large-scale implementation of these systems can significantly alter the subsurface thermal environment. This phenomenon can lead to the formation of "underground heat islands," potentially compromising the efficiency of neighboring geothermal installations and impacting local ecosystems.
Simulating these effects at a "community scale"—covering tens of kilometers and involving thousands of thermal foundations—presents a massive computational challenge. Existing commercial finite element (FE) tools often lack the necessary flexibility for custom physics or the raw processing speed required to handle high-resolution regional models with millions of degrees of freedom.
Methodology: An HPC-Driven Approach
To address these limitations, researchers developed a specialized hydrothermal finite element simulator leveraging deal.II, an open-source C++ library designed for solving partial differential equations. The simulator is built to handle coupled physics, accounting for both groundwater flow (fluid pressure) and heat transfer (temperature).
The technical framework utilizes a "one-way" coupling where groundwater velocity influences heat transport. For temporal stability, the model employs a backward Euler integral scheme. A core feature of the system is its distributed memory parallelization. By partitioning the mesh and Degrees of Freedom (DOFs) across thousands of processor cores using MPI (Message Passing Interface), the simulator avoids the memory bottlenecks typical of single-machine commercial software.
Validation and Case Studies
The simulator's accuracy was rigorously benchmarked against COMSOL Multiphysics using an Aquifer Thermal Energy Storage (ATES) case study. This benchmark involved a three-layer geological model (caprock, aquifer, and basement rock) simulating the injection of hot water.
Beyond benchmarking, the team applied the tool to a real-world scenario: the Treasure Island redevelopment project in San Francisco. This complex model integrated geological data for five distinct soil layers, historical climate data for surface boundaries, and a dense network of 1,130 energy piles integrated into the island's footprint.
Key Insights and Performance Results
The study revealed several critical findings regarding both computational performance and geothermal physics:
- The Assembly Bottleneck: As models scale to millions of elements, the primary computational cost shifts from solving linear equations to the "assembly" phase—the process of traversing the grid to build massive system matrices.
- Diffusion Dominance: In the Treasure Island study, slow groundwater flow meant that heat migration was primarily diffusion-dominated. Consequently, thermal impacts remained localized near the energy piles, highlighting the importance of precise urban planning and pile spacing.
- Residual Heating: The simulations showed that temperatures at specific observation points could continue to rise even after the system entered a cooling phase, due to the slow diffusion of residual heat from the piles.
- Speed and Scalability: On a standard PC, the new simulator outperformed COMSOL by approximately 3.6x. When deployed on a high-performance computing (HPC) node, it achieved a 10x speedup over the PC version. The system demonstrated linear scalability, with computation time decreasing proportionally as more processors were added.
Conclusion
The developed simulator successfully managed models with over 1.3 million degrees of freedom, enabling the simulation of a full year of cyclic thermal loading for an entire island community. While the current version uses simplified homogeneous soil layers and excludes tidal effects, it provides a powerful proof-of-concept for regional-scale geothermal modeling. Future work will focus on incorporating 3D heterogeneous geometries and more complex boundary conditions to further refine urban thermal sustainability planning.
Summary updated