Control and observation of electric machines by sliding modes.

摘要

The objective of this dissertation is to develop control and estimation methods for electric machines based on sliding mode control theory. Major attention is paid to two types of AC machines, i.e. the induction machine (IM) and the synchronous machine, including the permanent magnet synchronous machine (PMSM). This choice may be explained by the fact that AC drives are gradually superseding DC ones for many dynamic plants in modern industrial applications. The method proposed in this dissertation for both control and observation is the so-called sliding mode approach chosen because of its robustness and ability to reduce the order of the motion models. A further advantage is that the average values of discontinuous inputs (i.e. the so-called equivalent control) in sliding modes are algebraic functions of unknown state components and parameters. These equivalent control values can be easily obtained by using low pass filters and they are useful in calculation and estimation.; As real-time computation costs continually decline, both mechanical robustness and economic considerations increasingly stimulate the replacement of mechanical sensors by software-based observation methods. These so-called sensorless systems are free of maintenance and exhibit high reliability and low cost. Elimination of encoders or resolvers on induction machine drives is a prime example.; Due to the above reasons, many sensorless control schemes have been developed and described in literature. High order models of AC machines, nonlinearities in motion equations, uncertainties in model parameters and disturbances are the main obstacles hindering the development and rigorous mathematical analysis of such systems. However, their efficiency has been demonstrated by experiments and real applications. In contrast to conventional approaches, where control and observation are handled independently, the core idea of the approach proposed in this dissertation implies that they are treated as one interconnected system. This approach facilitates control system analysis and design since the speed is not an arbitrary time function any more but the solution to the known differential equations. As a result, the new structure of the observer is offered and the convergence of the observation is proven.; There is one very important issue in the framework of the studies: varying of the model parameters in a wide range, in particular the rotor resistance, which may be within 30--40% because of heating. New approach is developed to identify speed, flux and rotor resistance simultaneously under the common assumption that the electromagnetic processes are faster than the mechanical ones.; The developed control and estimation algorithms are tested experimentally for different types of induction machines. The sensorless systems demonstrate high accuracy of tracking reference inputs for speed and torque.