Research On Wind Turbine Transmission Chain Fault Diagnosis Based On Time-Frequency Domain Features

As the mounted capacity of wind turbines continues to increase,the stable and efficient operation of wind turbines becomes particularly important,which requires accurate diagnosis and identification of the operational status of wind turbines.The drive train components of wind turbines,which have a high frequency of failures,are directly involved in the energy conversion process of the wind turbine.Diagnosis of drive train faults helps wind farm personnel to maintain the wind turbine and reduce all losses due to faults.In order to enhance the efficiency of fault diagnosis,several aspects are studied in terms of feature extraction,feature selection,model building and system development,as follows:(1)In order to explore the deep features of the vibration signal,CEEMDAN decomposes the vibration signal into a series of modal components,and the vibration signal and its modal components form a new data set.The time domain characteristics of the original signal and each modal component,including statistical parameters and information entropy,are calculated.The signal is transformed into the frequency domain using FFT,and the statistical parameters and frequencies corresponding to the maximum amplitude in the frequency domain are calculated,and the above features are fused to construct a candidate feature set.(2)A comprehensive feature extraction method is designed by combining the advantages of RFE and XGBoost.The two feature selection algorithms were used to rank the importance of the feature variables separately,and the features ranked 20 in importance by the two algorithms were fused to form a model building requisition set,which was input to a classification model for wind turbine drive chain fault diagnosis.(3)In order to carry out accurate fault identification of the drive chain,a fault diagnosis model was established for each component of the drive chain separately.The classifier is a DNN,and the optimization algorithm JAYA is used to optimize the hyperparameters of DNN.The model achieved good classification performance on the wind turbine vibration dataset,and its classification accuracy all exceeded 0.985.The model has certain reference value for transmission chain fault diagnosis.(4)A wind turbine drive chain fault diagnosis system based on time-frequency domain features was developed based on C# and python language.The system contains five modules: user login,data connection,feature selection,model building and user management.After entering the system,the user can select the data set according to actual needs,extract the modelling features and perform dimensionality reduction.Appropriate hyperparameters are set to complete the establishment of the transmission chain fault diagnosis model,and the performance of the model is evaluated by means of evaluation indicators.