Improving the efficiency of wind energy through early damage detection

Wind energy has proven to be a highly valuable sustainable energy source, but maintenance issues with turbine blades have presented persistent and costly challenges. Blade damage can be divided into two categories, early and severe. Early-stage damage consists of coating defects and cracks. If left unattended, these may progress to severe damage, characterized by fractures and breakage.

While drones equipped with optical cameras can collect images from all angles of the blades, manual interpretation of the images, along with judgment of potential damage, is required and introduces a level of human error. Gao et al. proposed an improved end-to-end object detection model, called WTB-DINO, which combines deep learning technology with drone-collected images.

“Developing a detection method that can accurately and conveniently identify early-stage damage to wind turbine blades is essential to the future of wind energy,” said author Yongfei Ma.

The authors collected daily inspection data from four wind farms in central China, and combined them into a training and testing dataset consisting of 2,577 images of damaged turbine blades. The resulting machine-learning model accurately located and effectively classified damaged areas in real-time.

“By addressing the shortcomings of existing algorithms and proposing the WTB-DINO model, we achieve a high detection precision and recall rate of up to 93.2% and 93.6%, respectively, while maintaining a high frame rate of 27 frames per second,” said Ma.

While the experimental model was proven effective, the researchers plan to collect more comprehensive drone image data to further enhance the application in other types of blade surface damage.

Source: “Early-stage damage detection of wind turbine blades based on UAV images and deep learning,” by Ruxin Gao, Yongfei Ma, and Tengfei Wang, Journal of Renewable and Sustainable Energy (2023). The article can be accessed at https://doi.org/10.1063/5.0157624 .