Calculating noise with precision
Ken Mattsson and Gustav Eriksson are developing a new method for accurately simulating sound that they hope will be useful, for example, when planning and establishing new wind farms. Photo: Mikael Wallerstedt
How much noise does a wind farm, a busy road or a shooting range actually make? We will soon know for sure, according to Ken Mattsson at Uppsala University. Together with fellow researcher Gustav Eriksson, he is developing a new calculation tool that leaves no room for chance or arbitrary decisions, but is instead driven by physics.
“What we are now developing is a calculation model that will be able to accurately simulate how sound propagates in large areas, and to do that we need to take physics into account,” says Ken Mattsson, Professor of Scientific Computing at the Department of Information Technology.
It is a matter of taking careful account of conditions in the atmosphere, the terrain and, not least, low-frequency sounds, including infrasound.
“If you’re interested in how sound propagates over a large area of several square kilometres, only low-frequency sound, particularly infrasound, will remain. It passes through walls and noise barriers, and can be felt without being heard. High-frequency sound, on the other hand, is effectively attenuated in the atmosphere,” he says.
The calculation programs used today are very simplified and take almost no account at all of the low-frequency sounds, according to Ken Mattsson. The fact that the calculations can therefore be wildly inaccurate is something that he and several colleagues had demonstrated in their research as far back as ten years ago.
And as today’s wind turbines get bigger, they make more noise, especially in the low frequencies.
Dependence on good measurements
“To calculate wind power noise levels, we therefore need to start by measuring how much noise a wind turbine actually makes. It’s not enough to simply rely on specifications from the wind turbine supplier, in part because they tend to exclude most of the low-frequency sound,” says Ken Mattsson.
At the end of May, the team carried out measurements at the Lervik Wind Farm in Västervik with the help of colleagues from the University of Gävle. The measurement results are being used to accurately simulate the sound from the entire wind farm, every day of the year.
So, to make their calculation model as accurate as possible, he and his colleague Gustav Eriksson are incorporating on-site measurements from wind farms. Together with measurement experts, they recently visited the Lervik Wind Farm in Västervik to take measurements. They then feed this measurement data into their model, along with information about the weather, wind and terrain conditions.
Ken Mattsson explains that sound is greatly influenced by conditions in the atmosphere.
“So you need to know in detail what the weather is like on the day you’re measuring, and what the land is like at and around the site. This enables us to find out how high the sound levels will be, whether it’s a cold day in January or a windy early summer day in June.”
The time is ripe
The team has already been able to optimise good positions for wind turbines – good in the sense of minimising the actual sound level, including low-frequency sound, in areas where people live, for example.
“We can simply do something that current methods can’t handle, and the measurements in Lervik help us demonstrate that. If you’ve measured correctly on one occasion, with all relevant noise sources, that’s enough to be able to make accurate predictions with our model in other cases,” says Ken Mattsson.
The physical conditions at the site of a noise source, such as a tunnel or wind turbine, are key to the team's calculation model so that the sound can be simulated as accurately as possible over large distances. Photo: Mikael Wallerstedt
He has been researching advanced numerical algorithms for 25 years, and the new model represents a synthesis of this knowledge. When combined with modern graphics cards, great possibilities open up. A calculation that would have taken three days and a supercomputer ten years ago can now be handled in ten minutes.
“It's still a very complicated problem that we’re solving, but we’re making the process go quickly. This opens the door for anyone to use what we’ve developed,” he says.
Choosing a path
The technology is not linked specifically to wind turbines. There are many other potential applications where a lot of infrasound is generated, such as traffic noise from aircraft, railways and roads. Their model can also handle impulse noise from shooting ranges and explosions, and of course indoor noise.
“Wind power is something we are a bit specialised in, but there is a lot we could do. The tricky part is deciding which problems others are most interested in solving,” says Ken Mattsson.
Now that it is time to decide how to package and commercialise the technology, he has the support of UU Innovation. Among other things, they have helped him find a mentor, who founded their own company to take algorithm-based products to market.
“It was a perfect match. My mentor knows how to make money, and we’re also looking at the possibility of doing something together business-wise,” says Ken Mattsson.
Sara Gredemark