Research On Fault Diagnosis Method Of Wind Turbine Pitch System Based On MIMM-KEnPF

With the development of wind power technology,wind turbines designed and manufactured at home and abroad are becoming more and more large-volume,largecapacity,and the issues of stability and safety are becoming increasingly prominent.As one of the important components of the variable-pitch wind turbine,the pitch system plays a decisive role in the safe and efficient operation of the wind turbine.Affected by random wind speed and wind direction changes,the pitch system needs to be frequently operated,and due to the year-round working in harsh conditions,the damage and failure rate of various components of the pitch system are high,which ultimately affects the stability of the power output.It is important to study the fault diagnosis method of the pitch system to reduce the damage to the system and ensure the safe operation of the system.Particle filtering,as a state estimation algorithm suitable for non-linear model and non-Gaussian noise,has been widely used in fault diagnosis of non-linear systems.However,the algorithm itself has disadvantages such as particle degradation,impoverishment and low real-time performance.For this reason,the fault diagnosis method of wind turbine pitch system based on particle filter is studied in this paper.The main research contents are as follows:(1)The working principle of the wind turbine is analyzed,and simplified models of the main subsystems of the wind turbine are given,including wind speed,aerodynamic system,pitch system,transmission system,generator and converter,and control system.The failure of the pitch system is analyzed and the fault model set is established.(2)Aiming at the problems of particle filtering,including the problem of low estimation accuracy caused by particle degradation and impoverishment,and the problem of poor real-time estimation caused by large number of particles,this paper combines Ensemble Kalman Filter(EnKF)and Kullback-Leibler Distance(KLD)sampling method with particle filtering to obtain KEn PF.First,the EnKF is used to obtain the suggestion distribution that is closer to the actual posterior distribution to alleviate the problem of particle degradation and improve the estimation accuracy.Then,KLD is used as a criterion in each iteration period to adaptively reduce the number of particles and improve the filter calculation efficiency,thereby improving real-time.(3)Aiming at the problem of multi-fault diagnosis of wind turbine pitch system,an interacting multiple model(IMM)method based on adaptive estimation for fault diagnosis is introduced,which is combined with particle filter to design an IMM-PF based fault diagnosis algorithm.Finally,the diagnosis algorithm is verified on the established simulation platform of wind turbine.(4)Aiming at the problems of IMM-PF fault diagnosis method,including low state estimation and diagnosis accuracy,the existence of diagnosis delay,and poor real-time performance,in the non-mode change stage,a designed correction function based on the model probability gradient is used to adaptively modify the IMM model transition probability to improve the model probability accuracy.In the mode change stage,the likelihood function is used as a criterion to reverse the model probability to improve the diagnosis speed.At the same time,the improved particle filter(KEn PF)is combined with the modified IMM to improve the performance of estimation,diagnosis and the real-time.Finally,a fault diagnosis algorithm based on MIMM-KEn PF(Modified IMMKEn PF)is obtained,and the fault diagnosis problem of the pitch system is solved by using this method.