Compared to today’s standard methods, Olivia reduces calculation times from minutes to seconds. For example, a typical hydropower optimisation problem with 3-hour resolution that used to take 100 seconds can now be run in just over 2 seconds. At a finer 1-hour resolution, runtimes drop from around 17 minutes to less than 5 seconds.
“This is a game changer,” says Hans Ivar Skjelbred, research manager at SINTEF Energy Research. “We can now explore hundreds of possible scenarios in the time it used to take to run just one. That means better planning, lower risk and more robust decisions for the power system.”
With Olivia, running more detailed simulations doesn’t just become possible — it actually works better, giving faster and more reliable results for hydropower planning.
The reason is simple: larger and more detailed simulations give each graphics processor more work to do, which makes better use of the machine’s massive parallel power. For coarser runs, the processor finishes the job so quickly that the overhead of splitting the job between many processors becomes the bottleneck. But at higher resolutions the balance shifts: the extra work fills the processors efficiently, and communication between them becomes less of a constraint. This means that aiming for more granular simulations is not only possible with Olivia, it is actually the most efficient way to use the machine. In addition, Olivia’s architecture enables both traditional processors and graphics processors to share the same memory, which increases efficiency by removing the need to copy data from one place to the other.
Olivia is a national infrastructure and will be used by research groups across Norway on a wide range of challenges. SINTEF Energy Research’s work focuses on hydropower, the backbone of the Norwegian energy system, where improved planning allows us to get the full benefit of Norway’s energy resources for the community.
