In the fast-paced and highly competitive energy market, accurate and timely power production planning is crucial for success. The Nord Pool power market, a leading electricity trading platform in Europe, requires sophisticated simulation tools to help participants make informed decisions within tight time constraints.
Traditional simulation systems, however, often struggle to meet these demands due to limitations in computational power and efficiency. This success story highlights how SINTEF Energy Research leveraged high-performance computing (HPC) to revolutionise power production, transmission and trading planning.
SINTEF Energy Research is an applied research institute dedicated to creating innovative energy solutions. The institute specialises in research-based knowledge and infrastructure both in Norway and internationally to provide its clients with solutions and services that increase value and strengthen their competitive ability. SINTEF Energy Research is a part of the SINTEF group, one of Europe's largest independent research institutes. SINTEF is a non-profit research foundation where financial profits are invested in scientific equipment and expertise.
The development of simulation models and tools for power production and planning have been a core business for SINTEF Energy Research for more than 40 years. Most of the power-marked actors in the Nordics and Baltics use the tools offered for production planning and engaging in the Nord Pool day-ahead and intraday trading system.
The power marked and the evolution of power simulation systems
The history of power simulation systems has evolved over time. Initially, these systems were designed to work with single machines and have transitioned from older computers (like VAX) to modern ones using Windows and Linux.
Today, the simulation system uses a sophisticated model that combines randomness and dynamic changes. It involves running a large, complex program multiple times to predict different production outcomes and price scenarios.
Challenges in the current simulation framework
Due to the nature of the Nord Pool power market, participants have a two-hour window to run simulations and prepare their bids. This time constraint limits the complexity of the simulations that can be performed for decision support. Although there has been a recent focus on using the Message Passing Interface (MPI) to facilitate multi-node simulations, the full potential of high-performance computing (HPC) has not been explored.
Additionally, the platform and compiler independence, build system, and library structure were not designed to be used in a typical HPC setting, which includes a module system (Lmod) and queue management (Slurm).
Proof-of-concept: Refactoring for HPC
We established a proof-of-concept by refactoring parts of the simulation system using CMake as a build automation tool. Adaptations were made to integrate with the module system and job queue. This allowed us to test and benchmark day-ahead production planning simulations for typical market scenarios on the Sigma2 supercomputer, Betzy.
It also enabled us to investigate how to scale up for more complex simulations with significantly higher granularity—a complexity currently prohibitive in business operations due to time and computational resource constraints.
Preliminary results and insights
Preliminary results show that simulations that typically require two hours within the "two-hour window", can be reduced to less than two minutes on an HPC system, allowing the traders to run more simulations with higher granularity and precision within the trading window, allowing more informed decision-making.
The CPU efficiency when scaling to multi-node simulations for higher granularity and model complexity can reach 60-80% utilisation. This scaling efficiency depends on the model complexity; more complex models benefit from a higher number of nodes.
Transforming power system simulation tools to use HPC can streamline production planning in the industry. It also enables the creation of simulation models that provide more accurate results with higher granularity, all within the time constraints of the Nord Pool power market bidding processes.
Key benefits and wins:
- Standardisation of simulation code and build tools
- Running more complex simulations with higher granularity
- Off-loading computation to existing HPC resources
- Reduced wall time (2h -> 2min) and increased scalability
- Ability to study multiple scenarios and lower risk