【作者】 金海；

【导师】 黄进；

【作者基本信息】 浙江大学 ， 电气工程， 2006， 博士

【摘要】 交流传动已经占领变速传动领域的主导地位,各种高性能的交流电机变频调速系统也广泛应用于工农业生产。磁链估计和参数辨识是高性能变频调速传动的关键技术,已成为各国学者研究的热点。交流三相异步电动机是多变量、非线性和强耦合的高阶系统,基于空间矢量分析的交流电机复数模型具有物理概念清晰、形式简洁的特点,论文系统总结了基于空间矢量分析的交流电机复数模型。提出了以Luenberger状态观测器理论为基础的定、转子磁链观测器的一般形式,并通过反馈矩阵的选择,可以使得磁链观测模型不出现某些特定的参数,从而使得磁链观测不受某些参数变化的影响,提高磁链观测的鲁棒性。证明了各种常见磁链模型都只是Luenberger观测器的特例。 提出了基于模型参考方法的磁链自适应观测与参数辨识方法,不但可以准确观测电动机的磁链,当满足充分激励的条件时还可以同时辨识电动机的所有四个参数,即使不满足充分激励的条件,当部分参数已知时也可辨识其余参数。分别研究了在定子坐标系下和转子坐标系下基于模型参考方法的磁链观测和参数辨识,推导了参数的自适应律,从理论上证明了一定条件下的磁链和参数的收敛性,基于Matlab/Simulink进行了仿真,并对实际电动机的采样数据进行了实验,仿真和实验结果表明这种方法正确而有效。 遗传算法是一种成熟的具有极高鲁棒性和广泛适用性的全局优化方法。由于遗传算法具有不受问题性质的限制,在解决电机参数辨识方面的较大潜力。本文提出利用电动机的启动过程进行基于遗传算法的参数辨识。结果表明,在定子坐标系下辨识时,定子电阻和总漏感的辨识精度较高,转子时间常数和定子电感的辨识精度较低;而在转子坐标系下辨识时,漏感、转子时间常数和定子电感的辨识精度较高,定子电阻的辨识精度却有明显下降。如果将两种坐标系下的辨识结合起来,以在定子坐标系下辨识得到的定子电阻为已知参数,再在转子坐标系下进行其它参数辨识,全部参数辨识精度大大提高。 对按转子磁场定向的矢量控制系统来说,由于参数变化影响系统运行性能的程度最大的是转子时间常数的变化。本文提出一种新的异步电动机转子时间常数的在线校准方法,这种方法在励磁电流中间歇地叠加窄的负脉冲信号,根据此叠更多还原

【Abstract】 The AC drives have played a dominant place in the field of variable speed drives and various high-performance variable frequency AC drives are widely applied in industrial and agricultural production. Flux estimation and parameters identification are the critical technologies in high-performance AC drives, and have turned to be the hotspot of research. AC three-phase induction motors are strongly coupling high-order systems of multiple variables and nonlinearity. This paper summarized the complex number models of AC motors based on space vector analysis, which have the features of clear physical concepts and concise forms. In this paper, it proposed the universal form of flux observer of stator and rotor based on Luenberger state-observer theory, and it improved the robustness of flux observer by the selection of feedback matrix, which makes certain parameters not appear in models of flux observer so that it are not influenced by the parameters. It is verified that the common flux models are just the special cases of Luenberger flux observer.One kind of model reference approach for adaptive flux estimation and parameter identification are presented, which can not only observe the stator or rotor flux accurately, but also identify all the four parameters rapidly in the case of persistent excitation. When some parameters are known, the rest of the parameters could also be identified even though the excitation is not persistent. The paper has studied this approach in stator and rotor reference frame respectively, deducted the adaptive laws of parameters and theoretically proved the astringency of flux and parameters in certain conditions. Simulation by Matlab/Simulink and then experiments with sampled data from an induction motors have been done, and the results of simulation and experiments show that the method are accurate and effective.Genetic algorithm is a kind of well-rounded global optimization method that owns the features of strong robustness and broad applicability. As the genetic algorithm is not limited by character of the problems, it has deep potential in dealing with parameters identification of motors. This paper proposed the parameters identification based on genetic algorithm using data of induction motors during starting process. It is showed that in the stator reference frame, the stator resistance and whole leaking inductance have relatively high identification accuracy while therotor time constant and stator inductance have low identification accuracy, and in rotor reference frame, the leaking inductance, rotor time constant and stator inductance have high identification accuracy, but the identification accuracy of stator resistance has obvious decrease. If the identification in two reference frames are combined together, the stator resistance is identified in stator reference frame in advance, and then other parameters are gained in rotor reference frame;the identification accuracy of all the parameters will be improved a lot.With regard to the rotor-field-orientated vector control system, the change of rotor time constant has the greatest influence on the running performance of system compared to other parameters. So in this paper, it proposed another new method for adjusting the time constant of induction motors online, which intermittently overlap narrow negative pulse signal into the flux reference current, and then find out whether vector control was rotor-flux-oriented accurately according to the change of torque caused by overlapped pulse, and hereby adjust the time constant correspondingly. The simulation results showed this method could track the variance of rotor time constant very well.更多还原