Research On Fault Diagnosis For DFIG Based On Wavelet Neural Network

Along with the strong recovery of wind power,32.97 GW of new wind capacity was installed across China in 2015.New installed capacity makes a high record again.DFIG is the key equipment of wind turbine.The effectiveness of the wind turbine is determined by DFIG in the process of operation.Therefore,research on fault diagnosis for DFIG is very vital to ensuring the long-term safe and stable operation of DFIG.The main researching contents of the paper are as follows.(1)The fault types and fault mechanism of DFIG are analyzed.The higher fault rate of stator faults and bearing faults are chosen as the researching object.A 3.5kw of DFIG faults simulation experimental platform is designed.The platform can realize the stator windings and bearing faults.The stator windings faults include the inter-turn short circuit fault,phase to phase circuit fault and single phase to ground.Stator voltage and bearing vibration signal are objects of the data acquisition.(2)An improved wavelet threshold function and threshold method are presented.The improved threshold function achieves better de-noising and retains more effective signal.SNR and RSME are better than traditional wavelet threshold function.The stator voltage and bearing vibration signal are de-noised respectively as the faults example.The signal after de-noised is compared with the original signal.The effectiveness of the improved wavelet threshold de-noising method is verified.(3)The wavelet packet is used to decompose the stator voltage and bearing vibration signal after de-noised.After each frequency band energy value is calculated,they are normalized.The fault frequency band energy value is compared with no fault frequency band energy value.The accuracy of the fault feature extraction of wavelet packet is verified.(4)An improved BP neural network hidden layer node algorithm is presented.It can determine the better number of hidden layer node and make the network performance optimal.The sample data of the stator windings and bearing faults are used to respectively train the improved BP neural network.The trained network is tested by testing data.The experimental results show that the recognition accuracy rate of DFIG fault is higher by the improved BP neural network.(5)The fault diagnosis system of DFIG is designed by Virtual LabVIEW.The system includes user login,data acquisition,data display,data processing and fault diagnosis module.The fault diagnosis system and the internet realize the combination by the Web server function of LabVIEW.