 |
|
Theses
Application des outils du traitement du signal a la commande des machines tournantes
Application des outils du traitement du signal a la commande des machines tournantes
Abstract : The main objectives of our research was the application of Kalman filters and frequency estimators for the sensorless speed estimation of electrical machines with a restrained computational cost.
Extended Kalman filters are frequently used to estimate the flux and rotor velocity of an induction machine. Generally, these filters are implemented through their conventional equations, although their large computational burden. Therefore, we present a modification of the Kalman filter equations that we call « extended Kalman filter with virtual state». The significant decrease of the computional burden justifies its application on low power variable speed drives where electronic hardware represents an important part of the cost price. We also experienced the use of optimal two stages Kalman filters for the estimation of the rotor velocity.
We propose a procedure for the optimization of the Kalman filter parameters, according to the dynamics of the controlled variables, to the sensor noise level and to parameters uncertainties. This procedure allows the estimator to be robust against parameters variations and to avoid instability when they are included in a feedback loop.
To avoid parameter dependencies, the stator currents can be analysed with frequency estimators with low computational cost. The experimental results obtained with both a DC motor and a induction motor showed that the speed estimation is accurate at steady state. However, the results at low speed and the tracking capability during transients are slightly lower than those obtained by observers based on a state space model of the machine. The interconnection of these complementary estimators, a frequency estimator and a parameter-dependent observer is proposed to compensate for the drawbacks of each one. This approach was successfully experimented for the sensorless control of a DC motor.
Extended Kalman filters are frequently used to estimate the flux and rotor velocity of an induction machine. Generally, these filters are implemented through their conventional equations, although their large computational burden. Therefore, we present a modification of the Kalman filter equations that we call « extended Kalman filter with virtual state». The significant decrease of the computional burden justifies its application on low power variable speed drives where electronic hardware represents an important part of the cost price. We also experienced the use of optimal two stages Kalman filters for the estimation of the rotor velocity.
We propose a procedure for the optimization of the Kalman filter parameters, according to the dynamics of the controlled variables, to the sensor noise level and to parameters uncertainties. This procedure allows the estimator to be robust against parameters variations and to avoid instability when they are included in a feedback loop.
To avoid parameter dependencies, the stator currents can be analysed with frequency estimators with low computational cost. The experimental results obtained with both a DC motor and a induction motor showed that the speed estimation is accurate at steady state. However, the results at low speed and the tracking capability during transients are slightly lower than those obtained by observers based on a state space model of the machine. The interconnection of these complementary estimators, a frequency estimator and a parameter-dependent observer is proposed to compensate for the drawbacks of each one. This approach was successfully experimented for the sensorless control of a DC motor.
https://tel.archives-ouvertes.fr/tel-00179108
Contributor : Mickaël Hilairet <>
Submitted on : Friday, October 12, 2007 - 7:21:50 PM
Last modification on : Wednesday, February 13, 2019 - 3:20:03 PM
Long-term archiving on: : Thursday, September 27, 2012 - 1:05:11 PM
Contributor : Mickaël Hilairet <>
Submitted on : Friday, October 12, 2007 - 7:21:50 PM
Last modification on : Wednesday, February 13, 2019 - 3:20:03 PM
Long-term archiving on: : Thursday, September 27, 2012 - 1:05:11 PM