基于卡尔曼滤波的感应电机无速度传感器控制

发布时间：2018-06-24 04:49

本文选题：感应电机 + 直接转矩控制　；参考：《昆明理工大学》2016年硕士论文

【摘要】：感应电机价格低廉、可靠性强,近年来在工业、农业、家用等场合得到了广泛了应用。为了实现感应电机的闭环调速,通常采用的方法是在感应电机的传动轴上安装机械传感器。但安装机械传感器也带来造成增加安装体积,增加成本,降低系统可靠性等问题。对于这种问题,无速度传感器是一种良好的解决方案,能够解决以上出现的问题。目前已有多种无速度传感器算法被提出,其中无迹卡尔曼滤波算法UKF具有良好的转速预测性能而得到广大学者的研究,但应用中发现UKF算法存在运算时间过长，容易滤波发散,不适用于3维以上的高维非线性系统等问题。为了克服这些问题,本文对此进行深入研究,具体研究工作如下：首先,对感应电机的数学模型,逆变器和电压空间矢量原理,直接转矩控制系统组成进行详细介绍,为建立基于卡尔曼滤波的感应电机无速度传感器直接转矩控制仿真模型奠定理论基础。其次,介绍无迹卡尔曼滤波算法UKF,针对其运行时间过长,容易滤波发散的问题,对其进行改进,提出一种可克服这些问题的超球体平方根UKF算法。在此基础上,建立UKF,超球体平方根UKF这两种算法的感应电机无速度传感器Matlab仿真模型。通过仿真证明超球体平方根UKF算法能够有效降低运行时间和保证系统滤波不发散。但发现超球体平方根UKF算法和UKF算法一样依然无法克服在高维非线性系统上滤波精度下降的问题。为此,本文采用一种可用于高维非线性系统的卡尔曼滤波算法,即容积卡尔曼滤波算法CKF。然后,对CKF和UKF算法在预测感应电机转速性能方面进行Matlab仿真研究。通过仿真对比得知,CKF运行时间比UKF短,滤波精度比UKF高,在高维非线性系统具有良好的转速预测性能。虽然CKF比UKF运行时间短,但只是稍微短,在对实时性能要求高的电机控制系统中依然显得过长。为此,本文研究了一种可进一步降低运算时间的降维CKF算法。与此同时,针对CKF存在滤波发散的问题,提出一种可解决滤波发散问题的平方根CKF算法SRCKF.最后,通过Matlab仿真证明,SRCKF和CKF算法在转速预测性能方面具有相同的滤波精度,有效克服滤波发散问题,在实际工程更具实用性。降维CKF算法滤波精度与CKF一样,运行时间比CKF短,比CKF算法更具优越性。
[Abstract]:Induction motor has been widely used in industrial, agricultural, household and other fields in recent years because of its low price and high reliability. In order to realize the closed-loop speed regulation of induction motor, mechanical sensors are usually installed on the drive shaft of induction motor. But the installation of mechanical sensors also brings problems such as increasing installation volume, increasing cost and reducing system reliability. For this problem, the speed sensor is a good solution, can solve the above problems. At present, many speed sensorless algorithms have been proposed, among which the unscented Kalman filter (UKF) algorithm has been studied by many scholars for its good speed prediction performance. However, it is found that the UKF algorithm has long operation time and is easy to filter divergence. It is not suitable for high dimensional nonlinear systems with more than three dimensions. In order to overcome these problems, this paper carries on the thorough research, the concrete research work is as follows: first, introduces the induction motor mathematical model, the inverter and the voltage space vector principle, the direct torque control system constitution in detail, It lays a theoretical foundation for establishing the speed sensorless direct torque control simulation model of induction motor based on Kalman filter. Secondly, the unscented Kalman filter algorithm UKF is introduced. Aiming at the problem of long running time and easy filtering divergence, a hypersphere square-root UKF algorithm is proposed to overcome these problems. On this basis, the speed sensorless Matlab simulation model of induction motor based on UKF algorithm and hypersphere square root UKF algorithm is established. The simulation results show that the hypersphere square root UKF algorithm can effectively reduce the running time and ensure that the filtering of the system does not diverge. However, it is found that the hypersphere square root UKF algorithm and the UKF algorithm can not overcome the problem of filtering accuracy decline in high dimensional nonlinear systems. In this paper, a novel Kalman filtering algorithm, called volumetric Kalman filter, is proposed for high dimensional nonlinear systems. Then, the CKF and UKF algorithms are simulated by Matlab to predict the speed performance of induction motor. The simulation results show that CKF has shorter running time and higher filtering accuracy than UKF and has good speed prediction performance in high dimensional nonlinear systems. Although the running time of CKF is shorter than that of UKF, it is only a little shorter, and it is still too long in the motor control system which requires high real-time performance. Therefore, this paper studies a reduced dimension C KF algorithm which can further reduce the computational time. At the same time, aiming at the problem of filtering divergence in CKF, a square root CKF algorithm SRCKF, which can solve the problem of filtering divergence, is proposed. Finally, it is proved by Matlab simulation that SRCKF and CKF have the same filtering accuracy in speed prediction, which can effectively overcome the problem of filtering divergence, and are more practical in practical engineering. The filtering accuracy of the reduced dimension C KF algorithm is the same as that of the CKF algorithm, and its running time is shorter than that of the CKF algorithm, and it is superior to the CKF algorithm.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TM346

【相似文献】