Contact us
Contact us

 

 

Machine Learning Safeguards Solar Panels From Extreme Wind

PVH is a global leader in solar tracker manufacturing, a fundamental element for solar, or photovoltaic, parks. Their projects generate potential energy of more than 7GW in 130 plants and counting around the world– enough electricity to power more than 4.9 million homes. The photovoltaic industry has experienced incredible growth due to energy transformation initiatives across the world. PVH invests heavily in R&D, constantly evolving their edge in a highly competitive global marketplace where efficiency improvements are crucial for success.

The Challenge

Solar panels are angled towards the sun to maximize power generation; however, the panels become vulnerable to violent wind gusts, which can lead to severe damage. Solar trackers switch the panels to a defensive position when the wind is rough. The defense position is not optimal for generation, but ensures the panels’ safety. PVH began investigating advanced algorithms that improve prediction of wind speed with certainty, improving generation times and safeguarding panels. The solution had to work autonomously while accurately predicting future wind, a highly complex problem that required the use of novel approaches, such as machine learning.

ben-soyka-gDULrt23rXM-unsplash

The Approach

Based on extensive knowledge in predictive systems, Avvale developed a machine learning algorithm capable of anticipating wind gusts and alerting photovoltaic panels to activate the defense position. The developed algorithms are based on the application of deep neural networks, or deep learning, and data obtained in each of the locations, which achieves accurate forecasts of wind behavior. 

Using machine learning, a wind forecast is made from the historical time series generated at the physical location of the generation facility and nearby weather stations. The solution requires wind speed data supported by various meteorological data such as temperature and humidity. The system generates wind alarms that trigger the installation to move to a defense or security position, depending on the wind direction. The resulting algorithm can be installed in any facility and run in isolation. It can be continuously retrained as more local data information becomes available from the facility.

mariana-proenca-GXiHwHkIdVs-unsplash

The Impact

PVH’s machine learning algorithm has created a competitive edge by enhancing detection of wind intensity and direction, increasing installation efficiency by optimizing generation time, and introducing artificial intelligence algorithms in installations. With more local information to feed machine learning algorithms, PVH has increased the value of its technological proposals to the industry and customers. PVH has ambitious plans to continue developing and integrating machine and deep learning technologies in its systems, including predictive maintenance through proactive interpretation of data, improved generation forecasting based on historical time series, and the detection of technical losses in facilities through data analysis.

andreas-gucklhorn-Ilpf2eUPpUE-unsplash