CoE RAISE Technical Workshop - AI for Wind Farm Layout Optimization
From 15 to 18 November, the Barcelona Supercomputing Center (BSC) in Spain hosted a four day in-person hands-on workshop on Machine Learning (ML) supported methodology development, testing, and tuning for wind turbine simulations. The workshop took place in the context of CoE RAISE Task 3.2 “AI for Wind Farm Layout Optimization”. Researchers from BSC and Riga Technical University (RTU) (Latvia) gathered to discuss and experimentally explore ML methods to model the wind turbine flows and integrate these models into wind turbine simulations.
The first day session of the workshop focused on coupling and integration strategies for intertwining PyTorch-based ML implementations and large-eddy simulations using the Alya code (developed at BSC). In addition to the research work, BSC colleagues organized a visit to the supercomputer MareNostrum 4, known not only for its computing power, but also for its aesthetics style.
Arnis, Ilze and Guillaume in the computer hall of MareNostrum 4 (left picture) and in front of the Universitat Politècnica de Catalunya.
On the next days, the focus was devoted to intensive experimental work to develop a new Alya dataset format support, as well as to implement, test, and tune convolutional neural network models for low- and high-resolution modeling needs.
The RTU team would like to thank the Barcelona colleagues for their warm welcome. A similar in-person hands-on workshop for CoE RAISE Task 3.2 is planned in Riga in the spring of 2023.