【作者】 陆可；

【导师】 肖建；

【作者基本信息】 西南交通大学 ， 电气系统控制与信息技术， 2009， 博士

【摘要】 交流电机以其经济和技术优势占据了电力传动领域的主导地位,各种高性能的交流调速技术也得到了广泛的研究和应用。转子磁场定向控制使交流调速系统的性能产生了质的飞跃,感应电机无速度传感器控制更是增加了系统的简易性和鲁棒性。感应电机无速度传感器控制系统需要解决的关键问题是电机转速估计和转子磁链观测。扩展Kalman滤波(EKF)是一种有效的感应电机状态估计算法,但其存在两大缺陷:(1)对电机参数变化的鲁棒性欠佳;(2)对突变状态的跟踪能力较弱。本文利用强跟踪滤波(STF),提出了基于STF的感应电机状态估计算法,有效提高了对于突变状态的估计性能和参数变化的鲁棒性。此外,传统基于EKF的感应电机状态估计算法将电机转速作为常量处理,导致算法在极低速和零速时的估计精度不佳。本文引入电机的机械和转矩方程,将转速作为变量处理,并增加对负载转矩的估计,从而避免零速附近激励信号不足和摩擦阻力影响,提高状态估计精度。上述建立在感应电机全阶模型基础上的状态估计方法存在高阶矩阵运算,计算量偏大。为此,导出感应电机的降阶模型,此模型的观测量为状态的一阶延迟,无法直接利用EKF进行状态估计,引入延迟扩展Kalman滤波算法(SEKF)实现电机的状态估计。由于SEKF是在EKF的基础上得到的,因此存在与EKF同样的缺陷,利用STF的思想对其进行改进,提出了强跟踪延迟滤波(STSF)算法,并将其应用于转速估计和磁链观测。仿真和实验研究表明,基于STSF的感应电机状态估计算法具有满意的动、静态估计性能,同时计算量也有显著降低。前面提出的状态估计方法均假设电机参数保持不变,然而感应电机在运行过程中,参数随着工况和环境的变化表现出时变性。仿真研究表明,电机参数变化对EKF和STF的估计精度均会产生不良影响。为了在实践中获得高性能的状态估计,必须对电机参数进行在线辨识。对于定、转子电阻,提出了基于STF的辨识方法,得到了满意的估计精度;对于励磁电感,由于非线性程度较高,直接利用STF估计会增加算法的复杂度,因此提出了基于无轨迹Kalman滤波(UKF)和基于双重EKF(Dual EKF)估计的辨识方法。仿真和实验研究表明,上述方法均能实现对电机参数的准确辨识,从而避免状态估计受电机参数变化的影响。前面通过在线辨识解决了状态估计方法中的参数自适应问题,然而辨识本身需要一个过程,即当前周期得到的参数辨识结果到下一周期才被更新,因此本质上算法对系统模型的跟踪存在滞后,从而影响算法的动态性能。为此,提出利用多模型(MM)算法对电机状态和参数进行估计。为了降低传统MM算法的计算量,提高估计精度,提出了单滤波器多模型(SFMM)算法,并引入变结构思想,提出了单滤波器变结构多模型(SFVSMM)算法,将其应用于感应电机的状态估计和参数辨识中。仿真和实验研究表明,SFVSMM算法具有满意的状态和参数估计性能,并且计算量适中,为感应电机的无速度传感器控制提供了一种新的解决方案。最后,利用交流传动互馈实验台对本文提出的感应电机状态估计和参数辨识方法进行了实验研究。利用联合矢量控制对互馈双电机的同轴转速和转矩进行调节,相应的提出了基于STSF的联合状态估计方法,实现对互馈双电机状态的同时估计,能有效提高同轴转速和转矩的估计精度。针对定、转子电阻和励磁电感辨识方法的实验结果同样具有满意的精度,与实际值相当接近,从而满足感应电机高性能无速度传感器控制的要求。更多还原

【Abstract】 The AC motors have occupied a dominant position in the electrical power drive field based on the advantages of economy and technology. Various high performance AC speed adjustment technologies are widely researched and applied. The rotor field oriented control has brought essential advances in AC speed adjustment system. Speed sensorless control of induction motor (IM) promotes the simplicity and robustness further. There are two problems must be solved in this system: the speed estimation and rotor flux observation.Extended Kalman Filter (EKF) is an effective state estimation algorithm of IM. But it has two major defects: (1) bad robustness to the variation of motor parameters; (2) bad tracking ability to the abrupt change of states. To overcome these disadvantages, the Strong Track Filter (STF) is introduced to estimate the motor states, which can improve the estimation performance of abrupt change states and the robustness of variable parameters. Besides, the speed is considered as a constant in the traditional EKF-based estimator, which results to the bad estimation precision at very low and zero speed. In this paper, the mechanism and torque equations are introduced into the model of IM. Additionally, the speed is regarded as a variable, and the load torque is added to the state vector. It can improve the speed estimation precision and avoid the effects of lacking signal or friction at very low and zero speed to estimate the load torque.The state estimators based on full-order model of IM need high-order matrix operations, which has large computational burden. Therefor, the reduced-order model of IM is derived, while its observation equation is the first-order state delay, so its states can’t be estimated by EKF directly. Therefore, the Schmidt Extended Kalman Filter (SEKF) is introduced. Since SEKF inherits the basic algorithm of EKF, it has the same disadvantages as EKF. Using the concept of STF to improve it, the Strong Track Schmidt Filter (STSF) is proposed, and is applied to speed estimation and flux observation of IM. The simulation and experiment results illustrate that the STSF-based state estimator of IM has satisfactory dynamic and static estimation performance, and it also has lower computational complexity. The parameters of IM are supposed as constant in the above proposed state estimation algorithms, but in fact they are time-varying along with the changes of working condition during the operation. Simulation results illuminate that the estimation precisions of EKF and STF are affected by the changes of parameters. In order to obtain the high performance of state estimation, the parameters must be online identified. Therefore, an STF-based identification method is proposed to estimate the stator and rotor resistance, which has satisfactory estimation precision. Since magnetic inductance is highly nonlinear, the algorithm will be complex using STF. Two novel magnetic inductance identification approaches are proposed, one is based on Unscented Kalman Filter (UKF), and the other is based on Dual Extended Kalman Filter (Dual EKF). Simulation and experiment results show that the proposed algorithms can identify the parameters of IM exactly, and avoid the state estimation impacted by parameter variety.In the above state estimation methods, the problem of parameter self-adaptive is solved by online identification. However, identification needs a process, the result of this cycle will be used at the next cycle, so the track of system model has delay essentially, and the dynamic performance is affected. Therefore, Multiple Model (MM) algorithm is introduced to estimate the states and parameters of IM. In order to increase estimation precision and decrease computational burden, a Single Filter Multiple Model (SFMM) algorithm is proposed. Combined with the variable structure method, a Single Filter Variable Structure Multiple Model (SFVSMM) algorithm is obtained. Simulation and experiment results illustrate that it has satisfactory estimation performance and proper computational burden.The proposed state and parameter estimation methods in this paper are experimented using the reciprocal power-fed AC drive test-bed. A combined state estimation method is proposed based on STSF, which can estimate the coaxial speed and load torque of two motors simultaneously and increase the precision efficiently. The parameter identification methods also have high precision, which can satisfy the request of high performance speed sensorless control of IM.更多还原