Based On The Kalman Filter Algorithm Of Induction Motor Speed Sensorless Control

So far most problems on AC drives have been solved in modern power electronics. The relevant mature technologies supply the most acceptable solution to the industry field and have achieved many successful applications. Now the another important task in induction machine drivers development is that how to combine the machines with drivers organically, which means the more low-cost, more reliable and more capable Driver Module to be developed. Based on this thought, developers have done many deep researches on how to eliminate the shaft speed sensors and motor current sensors. Especially the realization of high performance speed sensorless vector control has appealed the abroad attention by researchers from all around the world and has become focus on AC drivers' development in recent years. Based on the background introduced above, this thesis realized the observation of speed and decoupling control of flux by sufficiently consideration of electromagnetism characteristics and mechanism characteristics of induction machine. The robustness against the disturbance of environment noises and the variation of induction motor parameters are the problems solved essentially for application on speed sensorless vector control in this paper.The thesis introduced the application on family of kalman filters in observation of variables in induction motors by the numbers. The following sensorless control strategies are researched in deep: sensorless vector control on induction motor with extended kalman filter;The original kalman filter is designed for estimation on linear systems. In chapter 4 the extended kalman filter is designed for nonlinear induction motor system. The way of dealing with nonlinear system is appealed to the truncation of Taylor series expansion, which is the most used method in industry application. In induction motor model state equations, the system noises and measurement noises are introduced to reflect the uncertainty of modeling and disturbance of environment including the variation of loads and wheel inertia, which can make the estimation more reliable and accurate. The EKF can estimate the 5 variables at the same time and supply the speed information to the sensorless vector control system. Simulation results show that the EKF has better low speed observation capability, well dynamic response and immunity to the noises.Because of the complexity of high-level structure and excessive-variables in induction motor model used in EKF, the implementation is very difficult due to the large amount of real-time calculation and requirement of very short sample period. In chapter 5 it presents and proposes a new approach to resolve these problems based on a reduced-order EKF. With this model structure, only the rotor flux components are estimated, besides the rotor speed itself. The new EKF predicts the rotor speed in the every sample time. So it eliminates the need for short sample period. Simulation results show that reduced-order EKF enables us to reduce the execution time without difficulties related to the tuning of covariance matrices and achieve robust speed estimation with elimination of stator resistance.The weighted sigma-points linearization is also called UT transformation which is based on statistical auto-recursive technology. Other than traditional local linearization it is a linearization for whole system. In chapter 6 an unscented kalman filter is designed with UT transformation replaced with truncation of Taylor series expansion. UKF uses a deterministic sampling approach choosing a set of sigma points by statistic knowledge which can completely capture the mean and covariance of the Gaussian random variables of the nonlinear system. Simulation results show that the posterior mean and covariance of the sigma points can achieve the accuracy of the third-order Taylor series expansion after having propagated through the true nonlinear system. EKF only can achieve the first-order accuracy. The UKF is more adapt to the nonlinear process than EKF in industry application with more reliable results and more accurate tracking ability.