Machine learning looks to unlock Swiss solar energy

February 03, 2020 //By Nick Flaherty
Switzerland has tapped into only a tenth of the potential of photovoltaic panels, and a study using machine learning shows another 90 percent remains to be unlocked.
Switzerland has tapped into only a tenth of the potential of photovoltaic solar panels, and a study using machine learning shows another 90 percent of the 24TWh capacity remains to be unlocked.

Researchers at EPFL in Switzerland have shown that photovoltaic (PV) solar panels could be installed on more than half of the country’s 9.6 million rooftops to provide over 40 percent of the annual electricity demand.

The overall photovoltaic potential of Swiss rooftops has not been estimated accurately owing to a lack of data about buildings and their environments, along with wide margins of error arising from existing calculation methods. The researchers at EPFL’s Solar Energy and Building Physics Laboratory (LESO-PB) developed a methodology combining machine learning algorithms with geographical information systems and physical models to estimate PV potential. And, for the first time, they estimated hourly profiles of PV potential. Their results have been published in Applied Energy.

“We’re not just looking at solar radiation, but also at the space available on rooftops. Some rooftops have an unusual shape or contain superstructures such as chimneys that prevent photovoltaic panels from being installed,” said Alina Walch, who led the second phase of the study. The algorithm takes into account parameters such as the size of the roof, its orientation and whether the building is in a city centre or a more isolated location. The results show that PV panels could be fitted to 55 percent of Switzerland’s total rooftop area. Even if panels were only installed on mainly south-facing rooftops, this could cover more than 40 percent of Switzerland’s electricity demand

A previous study explored the use of artificial intelligence to quantify the potential for the large-scale installation of photovoltaic panels on building rooftops. “Using new high-resolution data, we have now improved the estimation method and increased the spatio-temporal resolution of the results. This will enable us to model future energy systems that are 100% renewable,” said Jean-Louis Scartezzini, the head of LESO-PB.

Using the national SIG-Énergie geographic information system, the Swiss Federal Office of Energy has created a highly accurate model of Switzerland’s buildings. Using machine learning, estimates were made of the total


Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.