Résumé: Wind energy is one of the fastest-growing renewable energy resources. The blades are regarded as one of the most critical components in a wind turbine. The appropriate detection scheme to ensure the safety of the blade is crucial. Although there are many ways to detect blade damage and distinguish the types of them, a real-time online blade alert is important to ensure that potential wind turbine problems can be corrected in a timely manner. In this paper, a wind turbine blade damage early warning system was designed and developed based on sound signal fusion. Firstly, a wind turbine blade early warning method based on wavelet packet decomposition is proposed, which mainly includes data processing, feature extraction and early warning mechanism. Specifically, the beamforming technology of minimum variance distortion-less response(MVDR) is applied to enhance the weak signal and suppress the interference signal in the data processing. In the feature extraction, four-layer wavelet packet decomposition is applied to fully retain the information in the original signal. To improve the robustness of early warning, the two early warning strategies are introduced into in early warning mechanism. Then, a wind turbine blade damage early warning system was developed based on specific hardware. Finally, the system is tested on active wind farms and can achieve good early warning results.

Langue: Anglais