École Polytechnique Fédérale de Lausanne (EPFL) researchers have enhanced efficiency as well as durability of vertical-axis wind turbines (VAWTs) by using machine learning. The innovation is learned to have increased the efficiency by 200% in turbine efficiency and a 77% reduction in damaging vibrations. This is being considered as a potential revolution in wind energy technology.
VAWTs have several advantages over traditional horizontal-axis wind turbines (HAWTs) with respect to being quieter and requiring less space. It is simultaneously more wildlife-friendly. VAWT has been less used until now due vulnerability to strong gusts of wind that causes structural damage. The damage usually occurs when the angle between the airflow and the blade changes rapidly. This thereafter creates vortices that can impose heavy loads on the turbine structure.
EPFL researchers led by Sébastien Le Fouest from the School of Engineering’s Unsteady Flow Diagnostics Lab (UNFOLD) have addressed the challenges. They have utilized sensor technology and machine learning to overcome the issue. The team mounted sensors on the turbine blades to measure the forces acting on them. They experimented with various pitch angles, speeds and amplitudes. They performed more than 3500 iterations to identify the most efficient and robust pitch profiles by using machine learning. The algorithm mimics natural selection and continuously improves the blade pitch.
The research is basically a set of optimal pitch profiles that mitigates the effects of dynamic stall and simultaneously utilize it to enhance power production. The researchers adjusted the blade pitch to form smaller vortices and thereafter redirect the aerodynamic forces. They were also able to turn a major disadvantage of VAWTs into strength. The method boosts energy output of the turbines significantly while reducing the risk of structural damage.
The research was published in Nature Communications journal and it highlights the potential for VAWTs to play a significant role in wind energy production.
