Purdue University

A researcher from the Massachusetts Institute of Technology (MIT) used Purdue’s Anvil supercomputer to help create more accurate Numerical Weather Prediction (NWP) models for offshore wind farms. Better model predictions can lead to a slew of benefits, such as determining ideal site selection and optimizing turbine placement within a farm. These benefits can save businesses and the government both time and money as they seek to provide clean energy for the nation.

Dr. Sara Porchetta is an Assistant Professor at the Delft University of Technology in the Netherlands and a recent postdoc alumni of Howland Lab at MIT. While at Howland Lab, she used the Anvil supercomputer to determine which wind farm parameterization (WFP) is most reliable for offshore wind conditions in mesoscale NWP models. The study looked at the results of six different WFPs and compared them to recorded power and velocity measurements from wind farms in the North Sea, with the goal of improving the future power yield and wake prediction of offshore wind farms.

“I'm investigating offshore wind turbines in the North Sea because they released their power data, which is typically quite difficult to obtain,” says Porchetta. “With this data, I was able to review the production of the actual turbines and compare it with the different model predictions to see which is most accurate.”

Offshore wind farms offer the promise of clean, renewable energy and are more efficient and reliable than wind farms located on land. Higher and more consistent wind speeds over the ocean mean fewer turbines are needed to produce the same amount of power as land-based wind farms. However, building an offshore wind farm is very expensive, and relocating one after it has been constructed is not an option. Companies and investors need to know the best location to place their offshore wind farms to ensure maximum efficiency and power production. Part of this process is predicting the performance of the wind farm year-round instead of just one season or a few days at a time. This is where Porchetta’s research comes into play.

Historically, research studies looking at WFPs for power prediction have only considered a few days at most for their simulations. But because of the abundance of variation in weather systems, simulating only a few days does not provide robust validation. Porchetta decided to address this deficiency in the literature by conducting much longer simulations.

“There has been quite some development recently of weather prediction models for predicting wind farm power production, but there were no significant validations or comparisons of these models. This is why we decided to extend the study over as long a timeframe as we possibly could. With Anvil, I was able to do a six-month simulation, which represents quite a lot of different weather phenomena.”

Porchetta’s simulations not only covered a massive time scale, but also used a very high spatial resolution of 1km by 1km. This allowed her to look at the output and effects of individual turbines versus only viewing the wind farm as a whole. These parameter decisions provided Porchetta with a tremendous amount of useful data, but running such extensive simulations is computationally costly. Without access to a supercomputer like Anvil, Porchetta would never have been able to take on a project of this magnitude.

To run these simulations, Porchetta used WRF, an open-sourced code that is very well-known in weather prediction. She had no issues with getting the software up and running on Anvil. Once it was compiled, she was able to begin her simulations.

“To start the simulation, I used reanalysis data. This is data made by global models with a very coarse resolution and observations that are readily available. This was all assimilated into one big database, where I then did some dynamical downscaling. So basically, I took a global model, and then I ran my simulation on a much higher temporal and spatial resolution to update my results.”

Those results so far are promising, with each of the six simulations showing impressive accuracy, even though there was quite a bit of variation in the assumptions that each model made. This brings to light a new avenue of discovery for Porchetta to follow—how to take these models and create an even better one.

“The power prediction simulations, they come quite close to what is actually produced. So now it's scientifically interesting to check which assumptions are more relevant, what should we focus on, where should we develop new things, and so on.”

As for her experience with the Anvil supercomputer, Porchetta says she couldn’t have been happier. She praised Anvil for its low queue times and the ability to run multiple jobs in parallel, enabling her to make quick progress on her project. She was also pleased with how easy it was to compile and run WRF on Anvil.

“I’m a big user of the Anvil supercomputer,” says Porchetta. “Yeah, I really love it. It's the best supercomputer I’ve had so far. I’ve been on other ones, and this was by far the easiest.”

To learn more about High-Performance Computing and how it can help you, please visit our “Why HPC?” page.

Anvil is one of Purdue University’s most powerful supercomputers, providing researchers from diverse backgrounds with advanced computing capabilities. Built through a $10 million system acquisition grant from the National Science Foundation (NSF), Anvil supports scientific discovery by providing resources through the NSF’s Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS), a program that serves tens of thousands of researchers across the United States.

Researchers may request access to Anvil via the ACCESS allocations process. More information about Anvil is available on Purdue’s Anvil website. Anyone with questions should contact anvil@purdue.edu. Anvil is funded under NSF award No. 2005632.

Written by: Jonathan Poole, poole43@purdue.edu