【作者】 刘蜀宁；

【导师】 张代润；

【作者基本信息】 四川大学 ， 电力电子与电力传动， 2001， 硕士

【摘要】 异步电机的参数辨识既是基础理论课题，又是应用课题。目前有两类辨识方法。一为频域响应法，该法在计算方法上已较为成熟，算法稳定性较好且具有一定的滤波功能；但其缺点是对输入的信号要求严格。频响分析是建立在线性系统基础上，它不能反映动态过程中的非线性特征。二为时域辨识法，其最大优点是能够自动计及工况影响，一旦辨识成功，很多因素都自然包含在参数的估计值中，简单易行、计算简便、不需附加任何其他条件。但时域辨识法用于在线辨识时也存在一些问题：如周围环境的影响；辨识信号不能太大否则会影响电机的正常工作；在线辨识还有可能遇到不同实验或同一实验所识别的参数不稳定的问题。目前时域辨识方法主要有卡尔曼滤波法和最小二乘法，这两种方法一是存在局部最小或收敛性问题，二是往往采取线性化而带来误差。 本文分析了上述辨识方法，结合实例对比分析了其应用特点；并提出了在电机T参数模型下将转子漏抗集中折算到定子侧的电机参数识别方案。通过理论分析，阐述了方案的可行性，并通过仿真试验验证了其可靠性和准确性。在可接受测量精度的条件下，该方案使辨识方法更简单，运算速度更快，易实现，对驱动系统本身不须做大的改动。更多还原

【Abstract】 There are two kinds of methods for parameter identification at present. The one is making use of frequency response of system, the other is identification in time domain. The former has the well-developed algorithm, which is stable and can own some function of filter. However, too strictly demand about inputs and analysis of frequency response based on linear system stop it developing deeply. The latter is a simple approach, without any limits. The greatest advantage is that the results include many natural factors so long as identification is successful. But on-line identification also meets some difficulty such as disadvantage, unstable parameter and extra input which will cause motor work abnormally. There two kinds of algorithm in identification in time domain, the extended Kalman filter and recursive least square estimation, which also have problems of convergence and linearization errors.This paper explains the theory of approaches of identification above and discusses their application with some example, then sets forth a plan that real-time measures parameter of induction by transform the leakage inductance on rotor to stator in T topol configuration. The paperexpatiates the feasibility of the plan in theory and proves its veracity by simulation, and finally show it is a good plan that use less time, cause less changes to the drive, but results is satisfied.更多还原