Research On Diagnosis And Prediction Method Of Large Wind Turbine Blade Ice Coating

With the deepening of modern industrialization,the impact of global warming is becoming more and more obvious.Looking for renewable energy alternatives to fossil energy and changing the existing energy structure has become an inevitable way to achieve sustainable energy development.Wind energy,as an inexhaustible new energy source,has abundant resources in developed cities in the south and coastal areas.Vigorously promoting the development of the wind power industry will promote the adjustment of my country’s energy structure and the construction of the entire ecological civilization.However,in order to better obtain wind energy,wind farms are often built in high mountains and hilly areas.In the cold winter climate,the low temperature and high humidity environment makes the wind turbine blades prone to icing,which greatly affects the economy and efficiency of wind farms.safety.Therefore,the use of big data and artificial intelligence to predict the icing of wind turbine blades can reduce wind power costs for wind farms,improve the power generation efficiency of wind turbines,and avoid potential safety hazards caused by wind turbine blades falling after icing and melting.This thesis mainly studies the diagnosis and prediction of icing on wind turbine blades.Based on the data acquisition system(SCADA)commonly installed in wind farms,deep learning algorithms are used to achieve preprocessing,feature selection and feature construction of the wind turbine data collected by SCADA.The processed data is imported into the fan blade icing model for training,and the model is used to diagnose and predict the icing state of the fan blade.The main contents of this thesis are as follows:1)A diagnostic method for fan icing based on Pearson correlation coefficient(PCC)and eXtreme Gradient Boosting(XGBoost)algorithm is designed.First of all,this method labels the icing state of the input wind turbine data,and uses the Pearson correlation coefficient to calculate the correlation between each characteristic variable in the data and the icing state,and selects the one with a high degree of correlation with the icing state of the wind turbine.Then use these features as the input features of the icing diagnosis model,and use the corresponding icing state as the label of the icing diagnosis model to establish a fan icing diagnosis model based on the limit gradient boosting tree algorithm;finally,the wind collected in real time The unit monitoring data is imported into the model,and the model will automatically identify the operating status of the wind turbine unit.Experimental results show that the accuracy of this method for the diagnosis of wind turbine icing can reach more than 90%,which meets the actual needs of wind farms.2)Combining the characteristics of time continuity,a wind turbine icing prediction algorithm based on Long Short-term Memory Networks(LSTM)algorithm is designed.The algorithm is based on the known time information in the original wind turbine data,and uses the Mutual Information(MI)method to find features that are highly correlated with the wind turbine data collected by SCADA as time interval related features;then use these features As the input vector of LSTM,the icing state prediction model of normal wind turbine is established.The experimental results show that the model has smaller prediction errors than the model directly based on the original data prediction,and the convergence speed is faster during training.3)In order to improve the performance of the fan icing diagnosis model,a multi-model fusion algorithm based on the stacking combination strategy is designed.The algorithm first fused the CatBoost,LightBoost and XGBoost models,and then used the fused model to train the data set after feature selection.Finally,the experiment found that the fusion model using the Stacking algorithm has significantly improved classification accuracy and recall rate than the single model,which can effectively realize the diagnosis of the icing state of the wind turbine,and has good generalization,and is popularized and applied to the wind farm.It is of great significance.