Research On Multi-domain Feature Fault Diagnosis Of Wind Turbine Based On Complex Empirical Mode Decomposition And Random Forest Theory

With the global warming problem becoming more and more prominent,the energy security problem is becoming more and more serious,the production of fossil energy is unsustainable,and the development and utilization of renewable energy has become more and more important internationally.Wind power generation,as an important renewable energy,has the characteristics of safety,cleanliness and abundant reserves.It is considered as an important measure to safeguard energy security and cope with climate change.Now it has become an important area of research and development in various countries.Wind turbines work in harsh environment for a long time,and there is a strong coupling between the various equipment of the unit,so that the vibration signals of the unit show strong non-stationary,non-linear and coupling characteristics.So the fault diagnosis of wind turbine is more difficult than that of traditional power equipment.In this paper,the fault diagnosis of rolling bearing vibration signals of wind turbines is taken as the goal,and the time-frequency feature extraction and fault diagnosis model of wind turbines are studied.The key problems of weak fault feature extraction,multi-domain eigenvector fault diagnosis and fault information redundancy of wind turbines vibration signals are solved emphatically.Finally,the multi-domain characteristics vector fault diagnosis of wind turbines vibration signals are realized.Firstly,the background,purpose and significance of the topic selection for fault diagnosis of wind turbines are expounded.Then,the research status of signal de-noising technology,feature extraction method and fault identification method for wind turbines at home and abroad are introduced in detail.Aiming at the existing problems,the research contents and ideas are presented.Secondly,introduce the basic structure of wind turbine,including impeller system,transmission system,yaw system,generator-converter system and control system,and each part is introduced in detail.Then the typical faults of wind turbines are analyzed,including the common faults and causes of gearboxes,the main faults and causes of generators in wind turbines,and the main faults and causes of impellers,which lay a theoretical foundation for feature extraction and fault diagnosis of vibration signals of wind turbines.Thirdly,aiming at the difficulty of detecting non-stationary vibration signals of wind turbines and the mode aliasing problem of EMD decomposition,this paper applies empirical mode decomposition of complex data to the fault diagnosis of wind turbines transmission system,adds white noise as imaginary part,thus constitutes complex signal,and influences the selection of extremum points by projection of white noise in all directions,at the same time,uses the influence of noise projection.Then the characteristics of the envelope center of mass are eliminated to suppress the mode aliasing.The validity of the method is verified by simulation signals.Finally,on the basis of empirical mode decomposition of complex data,a multi-domain feature fault diagnosis method for wind turbines based on empirical mode decomposition of complex data(CEMD)and random forest theory(RF)is proposed.The collected vibration signals are decomposed into multiple intrinsic mode functions(IMF)by using CEMD.Then the energy and energy entropy of IMF components are calculated as eigenvectors in time-frequency domain.Eleven time-domain eigenvectors and thirteen frequency-domain eigenvectors of vibration signals are calculated.All the eigenvectors are constituted into multi-domain eigenvectors of wind turbines.Finally,the weight of each eigenvector is calculated.Redundant feature vectors should be removed to a certain extent,and the removed feature vectors should be input into the random forest pattern classification model to realize the fault diagnosis of wind turbines.Taking the rolling bearing of wind turbine as an example,the case pattern recognition is carried out from three aspects:different fault types,different fault degrees of the same fault and different operation states of the same fault.The simulation and experimental results show that the method can effectively extract the fault features in the signal and realize the fault diagnosis of wind turbine.Compared with the traditional classification method,it has higher accuracy and recognition rate.