Research On Sensorless Control Technology Of PMSM Based On Pseudo-Random Square Wave Injection

Permanent magnet synchronous motors(PMSMs)are widely used in industrial manufacturing and new energy vehicles because of their high power density and fast dynamic response,etc.The normal operation of PMSM depends on the acquisition of rotor position,and the addition of position sensors increases the complexity of the motor structure and reduces the reliability of the PMSM control system.Therefore,sensorless control schemes are drawing a wide spread concern for current research.The high-frequency voltage injection(HFVI)method is an effective sensorless control method for PMSM at low speed state,but its application is prone to cause large harmonic noise.In this thesis,a sensorless control method based on pseudo-random square wave voltage injection(HFSVI)is proposed to reduce the audible noise and harmonic vibration during PMSM control.Firstly,this thesis analyzes the principle of HFVI method.Based on constructing the basic mathematical model and HF mathematical model of PMSM in different coordinate systems,the response current acquisition and information processing methods of the commonly used HFVI method are analyzed.A simulation was made to compare the pulse vibration sinusoidal injection method and the HFVI method.The simulation results indicate that under the same conditions,the high-frequency square wave injection method possesses better rotor position estimation accuracy and speed following effect.Secondly,in order to fully exploit the advantages of the HFSVI method,and to solve the problems of high sampling frequency requirements and large hardware memory usage of traditional position extraction methods,this thesis proposes a Fourier decomposition-based rotor position estimation error extraction method.The method converts discrete square wave signals into multiple continuous sinusoidal signals for processing,reducing the sampling frequency and the hardware burden.In addition,a third-order non-linear state dilation observer for HFSVI is designed to improve the rotor position estimation accuracy,and a principle for parameter selection is given.The feasibility and effectiveness of the proposed method are verified by the simulation analysis.Furthermore,in order to address the problem of large harmonic noise during HFSVI.This thesis analyses the causes of noise generation during fixed frequency injection based on power spectral density(PSD),and gives the best frequency selection principle with reference to the PSD analysis results by comparing the noise suppression effects of different multi-frequency random injection methods.In addition,to compensate for the shortcomings of random frequency injection,Bayesian sequential decision making is introduced to replace the random number generator for pseudo-random frequency generation.Then we simulated and compared various frequency combinations and pseudo-random injection methods to verify the theoretical correctness in this thesis.Finally,an experimental platform for the sensorless control system of PMSMs based on TMS320F28335 was designed.The multi-frequency pseudo-random square wave injection method proposed in this thesis is deeply experimentally analyzed.The experiment shows that the proposed method has good results in rotor position estimation,speed following and noise reduction.