Research On Fatigue Damage Assessment And Blade Icing Condition Diagnosis Of Wind Turbine

China’s wind power has entered a new stage of grid parity and auction,and extending the service life of wind turbines is a crucial way to reduce the lifetime levelized cost of energy of wind turbines.Fatigue damage assessment technology can accurately determine the remaining service life of critical components of wind turbines,providing a fundamental basis for wind farm life extension decisions and technological upgrading.Condition diagnosis can predict the abnormal condition of essential components of wind turbines in advance,extending the wind turbine’s service life.Therefore,this paper focuses on wind turbine fatigue damage assessment and blade icing condition diagnosis.The main contributions include:1.The influence of power curtailment control on the fatigue damage of wind turbine tower bolts is revealed,which can provide theoretical support for the fatigue life assessment of wind turbines with frequent power curtailment.Firstly,the time series load of wind turbines tower sections under various power curtailment levels is obtained by GH Bladed wind turbine load simulation software.Then,the Schmidt-Neuper engineering stress algorithm,rain flow counting method,and linear cumulative fatigue damage theory are used to calculate the fatigue damage of tower bolts.Finally,the influence of turbulence intensity,wind speed,and power curtailment level on the fatigue damage of wind turbine tower bolts is quantitatively analyzed.The influence of power curtailment control on fatigue life is compared based on the assumption of power curtailment scenarios.Taking a 2 MW wind turbine as an example to verify,the results show that when the wind speed of the wind turbine is low,the fatigue damage of bolts in power curtailment control is not much different from that in normal power control.When the wind speed of wind turbine is high,the fatigue damage of bolts in power curtailment control is much more minor than that in normal power control.2.The influence of second-order vortex-induced vibration on the fatigue damage of the high-flexible tower weld of the wind turbine is revealed,which can provide a theoretical basis for the fatigue life assessment of the high-flexible tower wind turbine.Firstly,the wind turbine model containing high-flexible tower is established using GH Bladed wind turbine load simulation software.The high-flexible tower’s natural frequencies and mode shapes are obtained through modal analysis.Then,the inertial force at each height subject to the excitation of the second-order vortexinduced vibration is calculated according to the relevant standards.Finally,the influence trend of the second-order vortex-induced vibration on the fatigue damage of the wind turbine tower is quantitatively analyzed and compared with the firstorder vortex-induced vibration.Taking a 2 MW high-flexible tower wind turbine as an example to verify,the results show that the second-order vortex-induced vibration may easily lead to insufficient tower strength and accelerate fatigue failure,which will shorten the fatigue life of the wind turbine and increase the risk of tower collapse.3.The research on the fatigue damage assessment method of wind turbine nacelle chassis based on machine learning is carried out,which can provide efficient and reliable fatigue life assessment results for mountain wind turbines with significant differences in wind resources.Firstly,the key environmental influencing factors of wind turbine load are permutation and combination to obtain different inflow wind conditions.Calculating the hub center load of wind turbine corresponding to different inflow wind conditions and form an inflow wind conditions-load database.Secondly,the stress time series at the dangerous node of the wind turbine nacelle chassis is obtained based on the quasi-static method and Signed Von Mises equivalent stress calculation method.Thirdly,transform the stress time series into equivalent fatigue stress based on the principle of equal fatigue damage.Finally,the nonlinear mapping relationship between inflow wind conditions and equivalent fatigue stress is established based on SVR,LASSO,RF,XGBoost,and DNN algorithms.Taking the nacelle chassis of a 2.5 MW wind turbine as an example to verify,the results show that for the lifetime equivalent fatigue stress prediction,the machine learning algorithm with the slightest prediction error is the SVR algorithm,and the value of each error evaluation index is maintained at a low level,about 0.383%.The machine learning algorithm with the most significant prediction error is the LASSO algorithm,and each error evaluation index is maintained at a high level,about 10.037%.The RF,XGBoost,and DNN algorithms can all achieve good prediction results,and the prediction error is about 1%.4.The research on wind turbine blade icing diagnosis method based on hybrid features and Stacked-XGBoost algorithm is carried out,which significantly improves the diagnostic accuracy and generalization ability of wind turbine blade icing.Firstly,hybrid features that consider both short-term and long-term icing influence are extracted based on the underlying icing physics and sliding window algorithm.Then,the Stacking-XGBoost wind turbine blade icing diagnosis model is proposed based on the Stacking ensemble learning and XGBoost machine learning algorithms.Finally,the accuracy and generalization ability of the hybrid features and blade icing diagnosis model are analyzed and compared with RF,SVM,and XGBoost algorithms.Taking the SCADA data of 6 wind turbines from 2 wind farms during the winter icing period as an example to verify the improvement of the proposed method,the results show that the icing diagnosis accuracy of the four algorithms is improved by 19.51%on average after applying the hybrid features compared with original features.The icing diagnostic accuracy of the proposed Stacked-XGBoost icing diagnosis model is improved by 0.55%to 6.91%compared with the single-model algorithm.The revealed fatigue damage influence law and the machine learning-based fatigue damage assessment method can realize rapid and accurate fatigue damage assessment of in-service wind turbines,providing theoretical support and methodological basis for wind farm life extension and technological upgrading.The proposed wind turbine blade icing diagnosis method based on hybrid features and Stacked XGBoost algorithm can accurately identify the blade icing condition,which is essential to eliminate the negative impact of fatigue life-shortening and output power reduction caused by blade icing in time.