Sensorless nonlinear control of asynchronous machines.

An alternative to the existing technique of controlling induction machines - vector control - was investigated. In the new technique, sensorless nonlinear control, the mathematical model of the machine was transformed to a linear form and those states which were not measured (sensed) were reconstructed. It was shown possible to maintain stability despite the errors which the use of observers introduces into the control loop. The investigation is founded primarily on a study of the theory. Demonstration of the proposed scheme has been done mainly by means of simulation. The stability of the computer code for simulation was verified by a study of the stiffness and modes (eigenvalues) of the overall system. Furthermore, modifications to the simulation such as the introduction of noise, were made to increase its validity. The main findings were as follows: 1. In the new technique the decoupling of the speed (torque) and flux control is perfect. 2. The linearised model separates into two subsystems, one mechanical and one electromagnetic. 3. The control scheme is robust against disturbances (step change of load). 4. An asymptotic observer can successfully reconstruct both flux and speed states when the whole system is under the linearisation approach. Therefore sensorless nonlinear control is made possible and effective. 5. By selecting initial value(s) for the estimated state(s) close to the command(s) the transient demand on the power supply is reduced. Thus it may be possible to use smaller and cheaper power converters in a drive system. 6. This technique is also able to counteract some of the model uncertainties resulting from temperature variations. The significance of these findings is as follows: 1. Maximum torque and efficiency are achievable only by the new technique. In contrast, in field-weakening under vector control a near optimum is achievable only at the price of degrading the transient behaviour. 2. Direct measurement of flux and speed is avoided, which has, in practice economic benefit. 3. Although the proposed control technique is more complex than field-oriented control, its stability has been analytically proven and, in addition, the rate of convergence (for both the control and the estimator) is adjustable. 4. A sensitivity investigation shows the robustness of the proposed method when the most critical parameter i.e. the rotor resistance, varies slightly from its nominal value. 5. The simulation approach was developed to the point where it was believed possible to investigate the capability of the proposed technique and to identify potential practical problems with the nonlinear control strategy. Moreover the same code can easily be modified for compiling into a microprocessor, so allowing easy implementation of the scheme. 6.The proposed control is easily implementable in commercial drives due to the modest voltage and current demand, even in the initial transient