ELECTROMOTION 15 (2008) 141-120 141 A particle swarm optimization-based deadbeat on-line speed control for sensorless induction motor drives E.H.E. Bayoumi and H.M. Soliman Abstract: Particle swarm optimization (PSO) is utilized to derive a deadbeat speed control for sensorless induction motors (IM) drive system. The accuracy of the rotor speed estimation is sensitive to motor parameters variations. To alleviate the computation burden of such parameters, they are divided into two groups: on-line and off-line. Rotor flux angular speed is estimated using Proportional Integral (PI) flux observer (estimator). The observer is on-line designed to cope with the updated parameters. A PI current controller is similarly designed. Both controller and observer are tuned to achieve a deadbeat performance. The method guarantees accurate and precise steady-state speed estimation in addition to high dynamic performance. A comparative study is done between the proposed design and the conventional PI current controller and flux observer with/without parameters update. The results show the superiority of the proposed design under zero and very low speed operations. 1. Introduction In high performance vector-controlled induction motor drives, the knowledge of the rotor speed is required. Since speed sensors decrease the reliability, increase the hardware complexity and can be affected by noise, it is valuable to eliminate the speed sensor from the control system. Thus, during the past decades, there were serious research works to control induction machines without speed sensors [1-3]. A common expansion progress trend in induction motor control is to use different type of observer (estimator) to calculate the speed. A large variety of different schemes for speed sensorless ac drives is proposed in [1–17]. Depending on the respective approach, very good dynamic performance can be achieved over a fairly large speed range. In [4] the direct calculation method is introduced. This method uses the motor equations directly to calculate the rotor speed which leads to numerical errors and steady state errors. Model Reference Adaptive (MRA) observers to estimate stator or rotor quantities (e.g flux, back emf) are given in [5-6]. An extended Kalman filter (EKF) method is proposed in [7] for rotor speed estimation. Sliding mode observers [8–10] use the estimated speed to correct a flux-current observer; the correction is based on a sliding mode surface that combines the current error with flux estimation. Chattering is a big problem has to be solved when using sliding mode observers. All well- known rotor speed estimators depend on the induction motor model. Correct knowledge of the model parameters is crucial for speed estimation particularly in low speed range. Considerable research has been done for induction motor parameters estimations [11]. Offline methods [12-13] aim for estimation of all induction motor parameters with high accuracy. Self-commissioning methods [14-15] focus on algorithms that can be implemented directly on the motor controller and run as an initialization routine. Online parameter estimation [16-17] usually focuses on one or two parameters only (rotor or stator resistance); its main purpose is the tracking of the parameters that change while the motor is operating. There is quite few research conducted in detuned the modeling the induction machine as a function of stray load losses [11]. In this paper, particle swarm optimization (PSO)-based PI current controllers are used in © 2008 – Mediamira Science Publisher. All rights reserved.