Abstract
Wind power has become one of the most important clean energy sources today, and the complex environment and physical structures lead to high operation and maintenance costs. Achieving rapid and precise wind turbine fault diagnosis can potentially decrease wind power’s operational and maintenance expenses. The conventional fault detection methods of the wind turbine system usually focus on one specific component, with multiple limitations, high computational complexity and low prediction efficiency. In this paper, a kind of feature extraction method integrated heatmap with MMSVM-RFE is applied on XGBoost model to diagnose overall wind turbine faults. The heatmap-MMSVM-RFE model presented in this paper fuses the advantages of the physical interpretability of the heatmap model with the mathematical statistics of the SVM model. A new criterion Ψ(α) is defined to assess whether the multi-faults model is comprehensively diagnosed. In the case study, the heatmap-MMSVM-RFE-XGBoost model has achieved the highest classification accuracy rate of 91%, which can broadly categorize faults. Compared to Random Forest and LightGBM, the integrated fault diagnosis model we propose improves Ψ(α) by 0.4% and 0.3%.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call