Online Parameter Identification Of PMSM Based On Collaborative Particle Swarm Optimization

Permanent Magnet Synchronous Motor(PMSM) is the execution unit of servo system.In order to control PMSM effectively, proper control parameters of the servo systemshould be obtained. As the control parameters is directly bound up with parameters ofPMSM, and the parameters of PMSM change all the time, it is necessary to identifyPMSM parameters online.This thesis does a deep analysis of the online identification methods which are nowwidely investigated, elaborate merits and flaws of respective method, and figure out theexisting common problems of these methods, including large amount of calculation andinter-coupling problem of multi-parameter identification. Aiming at the two problems, thisthesis proposes an identification method using an improved particle swarm optimizationalgorithm. First, the reason for the early maturity of standard particle swarmoptimization(SPSO) is deeply analyzed, and an collaborative particle swarm optimizationalgorithm(CPSO) is proposed as an improvement to solve the problem. With comparisonof CPSO and SPSO via simulation on MATLAB, it proves that the CPSO algorithm solvesearly maturity problem existed in SPSO, and can achieve higher convergence precision.Next, the proposed CPSO algorithm is combined with the model of PMSM vector controlsystem in SIMULINK, and an online parameter identification method using CPSO isrealized to identify parameters of PMSM in MATLAB. The simulation results testify theeffectiveness of the proposed online identification method.Since the PSO algorithm has a nature characteristic of concurrency, which is similarto the hardware concurrency of FPGA, the experiment is done via FPGA to insure thereal-time requirement of the identification method. As the on-chip resource of the FPGAused is limited, and the resource used to realize CPSO is much more than that to realizeCPSO, the experiment in this thesis is done in two steps. The first step is to realize theparameter identification method based on SPSO via FPGA to identify the mechanical andelectrical parameters of PMSM. The effectiveness of using PSO to identify motorparameters online is testified through the precision of the identified results and the time spent on the identification. The second step is to realize CPSO algorithm via FPGA, and tocompare the performance of hardware-realized CPSO and SPSO through typical fitnessfunction. The results prove that the time cost of finishing CPSO is much less than that tofinish SPSO, and its optimal precision is much higher than that of SPSO, but CPSOspends much more FPGA resources.The simulation and experiment results indicate that the parameter identificationmethod based on CPSO is easy to realize and achieves the real-time requirement. Thismethod solves the early maturity problem of SPSO and the inter-coupling problem ofmulti-parameter identification, and realizes the high-precision concurrent onlineidentification of two mechanical parameters and three electrical parameters.