Analysis of electromagnetic characteristics of typical faults in permanent magnet wind generators

Due to the harsh actual operating environment of the permanent magnet wind turbine,it is easy to break down and difficult to monitor.Therefore,the electromagnetic characteristics identification of major fault types of large-scale permanent magnet wind turbines is studied in this paper.The typical faults of rotor eccentricity,stator winding short circuit and permanent magnet demagnetization of permanent magnet wind turbines are analyzed theoretically.The wavelet analysis algorithm is used to decompose and reconstruct the abnormal electromagnetic signal waveform band,and the characteristic frequency of the electromagnetic signal is obtained when the fault occurs.In order to verify the effectiveness of the proposed method,a 3.680MW permanent magnet wind turbine was taken as the research object.Its physical simulation model was established,and an external circuit was built to carry out field co-simulation.The results show that the motor fault type can be determined by detecting the change rule of fault characteristic frequency in the spectrum diagram,and the electromagnetic characteristic analysis can be applied to the early monitoring of the permanent magnet wind turbine fault.

Artificial Monitoring of Eccentric Synchronous Reluctance Motors Using Neural Networks

This paper proposes an artificial neural network for monitoring and detecting the eccentric error of synchronous reluctance motors.Firstly,a 15 kWsynchronous reluctance motor is introduced and took as a case study to investigate the effects of eccentric rotor.Then,the equivalent magnetic circuits of the studied motor are analyzed and developed,in cases of dynamic eccentric rotor and static eccentric rotor condition,respectively.After that,the analytical equations of the studied motor are derived,in terms of its air-gap flux density,electromagnetic torque,and electromagnetic force,followed by the electromagnetic finite element analyses.Then,the modal analyses of the stator and the whole motor are performed,respectively,to explore the natural frequency and the modal shape of the motor,by which the further vibrational analysis is possible to be conducted.The vibration level of the housing is furtherly studied to investigate its relationship with the rotor eccentricity,which is validated by the prototype test.Furthermore,an artificial neural network,which has 3 layers,is proposed.By taking the air-gap flux density,the electromagnetic force,and the vibrational level as inputs,and taking the eccentric distance as output,the proposed neural network is trained till the error smaller than 5%.Therefore,this neural network is obtaining the input parameters of the tested motor,based on which it is automatically monitoring and reporting the eccentric error to the upper-level control center.