Adaptive, predictive and neural approaches in drive automation and control of power converters

Abstract

It is well known that the parameters of the load, of the motor and of the converter are not known exactly and may vary during the drive system operation. Therefore, adaptive control seems to be a natural choice for drive automation. Unfortunately, the complexity of adaptive controllers forces us to constantly seek balance between the universal theoretical soundness and the ability of practical implementation of the designed control algorithm. A nonlinear approach to adaptive motion control is presented in [5]. It is one of the first contributions using a nonlinear state transformation and backstepping. Its practical importance results from the fact that it allows us to impose hard constraints on the state variables directly and to achieve asymptotic tracking of any reference trajectory satisfying the constraints, despite unknown parameters of a drive. An interesting cooperation of adaptive and algebraic speed estimation laws is proposed in [6] for induction machines. This approach improves stable operation in the considered range of changes of rotor speed and load torque. Although the permanent magnet synchronous motor drive presented in [7] is not an adaptive control system literally, it possesses an important ability to reduce velocity ripples due to periodic disturbances. The disturbances are estimated by an extended Kalman filter and a state feedback controller with feedforward action regulates and compensates for disturbance. Adaptive control is not the only methodology to obtain proper drive operation in the presence of parameter variations. A similar result may be achieved ensuring the system robustness by proper tuning of the state controller parameters. This technique is demonstrated in [8], where coefficients of the controller of PMSM servo-drive are tuned by an artificial bee colony optimization algorithm. Hence, the readers of the Special Section get an interesting opportunity to compare adaptive and robust approaches to the same type of drive with a permanent magnet synchronous motor. The same type of motor is considered in [9]. Smart modification of a classical linear, model-reference adaptive control is presented with the Widrow-Hoff rule used to adjust controller’s coefficients. Originally, the Widrow-Hoff rule was proposed in