Research On Fault Diagnosis Method Of Wind Turbine Main Bearing Based On Extreme Learning Machine

Wind energy is a kind of new energy which is non-polluting and renewable,so wind power has been getting more and more research in recent years.However,most wind turbines work in the relatively bad environment.As the vital transmission part of wind turbine,main bearing plays the role of supporting and guiding.If the main bearing fails,the unit will stop running,bringing huge economic losses to the wind farm.Therefore,it is an effective measure to improve the utilization rate of the wind turbine and improve the economic benefits of the wind farm by quickly and effectively diagnosing the faults of the wind turbine main bearing.The main contents of the thesis are as follows:The wavelet packet energy feature extraction method is given to extract the vibration signal eigenvectors of wind turbine main bearing.By comparing the noise reduction effect of different combinations of wavelet basis functions and threshold functions,the optimal noise reduction combination is chosen,and the analysis shows that the noise reduction effect of soft threshold processing method is better.The energy vectors of vibration signal of the wind turbine main bearing are extracted by the wavelet packet energy feature extraction method,and the similarities of the feature vectors of different fault types are analyzed,which lays the foundation for the fault identification below.The fault diagnosis method for wind turbine main bearing based on Extreme Learning Machine(ELM)is given.By comparing the influence of different activation functions on the diagnostic effect of ELM,the optimal activation function is selected.The influence of the ELM parameters on the diagnostic effect of ELM is analyzed,and the concrete realization process of diagnosis is given.Compared with the Least Squares Support Vector Machine algorithm,experiment shows the fault diagnosis method of wind turbine main bearing based on the ELM algorithm has better diagnostic effect.The fault diagnosis method for wind turbine main bearing based on Kernel-based Extreme Learning Machine(KELM)is given.Genetic Algorithm is used to optimize the parameters of KELM to further improve the diagnostic accuracy.The concrete realization process of diagnosis is given.By contrasting the confounding matrixes of fault diagnosis based on ELM and KELM,it is concluded that the diagnostic effect of KELM is better.The fault diagnosis method for wind turbine main bearing based on Multi-Kernel Extreme Learning Machine(MKELM)is given.The paper constructs the multi-kernel function by using the linear combination of global kernel function and local kernel function so that the multi-kernel function has both global and local characteristics.The parameter optimization method of the MKELM is employed which combines the Genetic Algorithm and the Cross Validation.The concrete realization process of diagnosis is given.Experiments show that the fault diagnosis method of wind turbine main bearing based on MKELM optimized by Genetic Algorithm and Cross Validation has better diagnostic effect.