CONTEXT AND PROJECT GOAL
The OL-STAR project, OLkiluoto STate-of-the-Art Reactive hydrogeochemical modelling, aims at ensuring the safe long-term deep disposal of spent nuclear fuel at the Olkiluoto site in Finland.
Through advanced computer modelling, the project simulates the evolution on time scales spanning over several millennia of groundwater chemistry located in the fractured crystalline bedrock that will store the waste repository known as the ONKALO®. The project seeks to consider factors like density-driven flow, microbiological reactions in fractured crystalline rocks coupled with various chemical and environmental events. This project is probably one of the most ambitious modelling projects of groundwater chemistry to date globally. The results will be crucial to confirm the safety features of this site.
The long-term safety of deep geological repositories for nuclear waste depends on demonstrating that suitable hydrogeochemical conditions in the host rock can be maintained over timescales lasting thousands of years, and they must also be maintained during future ice age and subsequent submerged and temperate phases of planetary activity.
High computational power is needed to calculate the length of the modelled time span, and the large size of about 70 cubic kilometres of the modelled area.
COMPUTATIONAL METHODS
The project team confronted a substantial challenge: understanding how water and various substances navigate through the geological setting prevailing at the nuclear waste storage site at Olkiluoto. The inclusion of sophisticated processes like electromigration and matrix diffusion have been included which, until now, have not been simultaneously considered in this type of calculations.
To address this, the project team employs PFLOTRAN, an advanced parallel code for modeling flow and transport phenomena in porous and fractured media. This upscaling approach relies on a parameterization of flow, transport, and geochemical parameters (hydraulic conductivity, kinematic porosity, mineral surface area, and mineral abundance) that is consistent with the underlying Discrete Fracture Network (DFN) and other field data of Olkiluoto. To ensure the accuracy of the computer models, the project team subjected them to a thorough verification, calibration, and validation process in upscaling simulations.
RESEARCH TEAM
The OL-STAR project brings together a team of experts from Posiva Oy and Amphos21. Posiva Oy is a leading company in nuclear waste management and provided its extensive scientific knowledge of the disposal site, industrial experience, and sustainable management of radioactive materials to the project. Amphos21 is a scientific and strategic environmental consulting company and brings extensive knowledge on the entire nuclear fuel cycle and the presence of radioactive material in the environment.
TECHNICAL CHALLENGES AND COMPUTATIONAL METHODS
The OL-STAR project initially faced challenges in compiling PFLOTRAN within the LUMI computing environment. The modelling problem introduced a complex, heterogeneous system involving various processes (groundwater flow and transport, matrix diffusion, chemical reactions, and microbial activity) operating across diverse temporal and spatial scales. The high non-linearity and complexity of kinetic reactions and multicomponent diffusion causes the numerical model to be stiff, thus requiring very small time steps to simulate the paleo evolution of Olkiluoto. Therefore, the challenge is to derive an optimal parallelization, consisting of subdividing the model domain in several sub-domains where equations are easier to be solved, which would help to overcome the inherent stiffness of the system.
To handle dynamic boundary conditions and streamline a multitude of simulations, the project team is using a PFLOTRAN manager coded in Python. Calibration of PFLOTRAN models will be achieved through Optuna, a tool vital for computational models to simulate and reconstruct past geological and environmental conditions in the LUMI system.
Integrating these numerical tools into the LUMI supercomputing environment presents a significant challenge but is paramount for optimising simulation execution and handling extensive datasets efficiently. Overcoming these technical challenges ensures the seamless operation of the project’s advanced numerical tools in the LUMI supercomputing environment, contributing to the project's success in tackling intricate hydrogeochemisty dynamics.
POTENTIAL IMPACT
The safe disposal of highly radioactive spent nuclear fuel arising from nuclear power generation requires that it is isolated from the geo-biosphere for extended periods in order to protect humans and the environment against the dangers of radiation.
Deep geological disposal of spent nuclear fuel is, for now, the preferred solution for the final disposal of the most hazardous nuclear waste.
The advancements in computational technologies provide opportunities for ever more realistic modelling, allowing analysing the evolution of groundwater chemistry caused by repository construction, geological and climatic changes. This project integrates all the modelling knowledge developed over the years in a holistic numerical model to simulate the hydrogeological and hydrogeochemical evolution of Olkiluoto in the past, present, and future.
Furthermore, the groundbreaking nature of this project holds the potential to obtain the scientific-technical interest and recognition of other spent nuclear fuel storage projects globally. Currently, all spent nuclear fuel is stored in interim facilities, vulnerable to natural disasters or national emergencies. The project's innovative work opens opportunities for collaboration with stakeholders worldwide, contributing to a collective effort in advancing safe and sustainable solutions for nuclear waste disposal.
ACHIEVEMENTS LEVERAGING EUROHPC SUPERCOMPUTERS
The use of LUMI supercomputer and HPC in general was decisive in tackling the underlying high complexity modelling environment and improving the safety assessment of the spent nuclear fuel repository at Olkiluoto.
The project aims at embracing Olkiluoto's hydrogeological flow and transport model through a comprehensive/holistic hydrogeochemical reactive transport modelling methodology, in which supercomputing is the key tool to overcome hurdles arising from model complexity coupled with the fabric of the site’s hydrogeology. Using LUMI or other EuroHPC supercomputers is helping to add new important concepts or processes (e.g., cation exchange, cement leachates) while allowing advances in other processes in existing models (e.g., anion exclusion, matrix diffusion, surface hydrology). It also gives the possibility to explore novel numerical developments for a better representation of flow and transport through fractured rock media.
Supercomputing is planned to be leveraged to reduce model uncertainty or evaluating different repository designs and means of control by performing as many variant cases as needed. Thus, HPC at LUMI is decisive in tackling the underlying high complexity modelling environment and improving the safety assessment of the spent nuclear fuel repository at Olkiluoto.
Leveraging LUMI or other EuroHPC supercomputers became imperative in the above pursuits.
FUTURE UTILISATION OF EUROHPC RESOURCES
The timescale for these simulations extends from hundreds of thousands to a million years. Given the multitude of stochastic realisations and the diverse scenarios requiring evaluation, the future utilisation of EuroHPC resources is deemed crucial for the successful completion of the project.