Research On Intelligent Fault Diagnosis Of Wind Turbine Drive Chain Based On Time-Frequency Characteristics

The operation of transmission chain in wind turbine directly affects the performance and life of wind turbine.The vibration signal obtained by the vibration sensor can effectively reflect the running state of the drive chain,and the fault diagnosis of the drive chain based on the vibration signal can provide help for the safe operation of the wind turbine.At present,the fault diagnosis method of transmission chain using vibration signal mainly relies on a lot of manual analysis,which takes a long time and is inefficient,so it is difficult to meet the requirement of real-time state detection of wind power production.Therefore,how to realize intelligent and fast fault diagnosis of wind turbine drive chain has become an urgent problem to be solved.In order to improve the intelligence of fault diagnosis of wind turbine drive chain and reduce the diagnosis time,the research is mainly carried out from the following aspects:（1）Carry out envelope spectrum analysis and spectrum analysis on the vibration signals collected from the public data set and the blades,bearings and gearboxes in the actual wind turbine drive chain,analyze the natural frequency of each fault type,determine the signal fault type,and specify the fault category labels for all kinds of samples to form the data set for research.（2）Because vibration signals contain abundant fault features and are easy to handle,most intelligent methods such as neural network can’t directly extract their intrinsic features,and it is difficult to fit the input and output,so that ideal classification accuracy can’t be obtained.Therefore,the research uses feature extraction methods based on variational modal decomposition,fast Fourier transform and statistics to mine the hidden features of vibration signals.On this basis,the random forest is used to select the features of the obtained time-frequency domain feature data,which ensures that the recognition accuracy does not decrease and reduces the training time of the subsequent neural network.（3）In order to realize fast fault diagnosis modeling and identification speed,based on the above-mentioned optimal features,an intelligent fault classification model based on generative extreme learning machine is established according to different transmission chain components.The model has good classification performance in the mentioned data set,and according to its high-speed and accurate characteristics,an intelligent fault diagnosis system software which can be used for real-time fault diagnosis is designed and developed.Through the research of this topic,the proposed intelligent diagnosis method of wind turbine drive chain fault has an average recognition accuracy of 98.03% in actual data,an average modeling time of 0.026 seconds,and an identification time of 0.674 seconds,which can provide certain reference significance for fault diagnosis of wind turbine drive chain.