Despite its simplicity and the development of computing and digital processing, the classical proportional-integral (PI) controller is the best choice in industrial power electronics and electric motor applications. Implementing PI controllers on microcomputer systems is very simple and the operation is fast and robust. Fast operation is particularly important for real-time control of electromechanical converters, with time constants on the order of tens of microseconds. The PI controller also has limitations that need special attention, one of which is the integral windup, which can have a crucial effect on the controller and the quality of the control.
The thesis examines the phenomenon of integral runaway, presents several solutions to prevent the phenomenon, and compares in detail the two most common solutions using the example of electric motor control, and then the effect of sampling time on the control with discrete AW implementation. The theory of the integral runaway phenomenon and its solutions is based on the existing literature. The comparison of the most common solutions (conditional integration and back-calculation) based on a computer simulation in the Matlab/Simulink software environment.
With the controller set according to the Ziegler - Nichols set-up rules, we ran a system excited by a step function. On the unipolar-limited system, both methods improved all the tested integral performance indices compared to the basic controller alone by between 2.1% and 11.8%. Similar results are also seen for the bipolar-limited system, where the controller is kept in saturation for a longer period of time, resulting in even larger differences. In the case of the cascade-regulated unipolar-limited system, the values of the indices also improved, but slightly less significantly, between 1.1% and 5.3% depending on the performance index. When implementing AW on a discrete system, the best ratio between the sample time and the minimum time constant that we still want to take into account turned out to be at least 1:2 or even better, 1:10, where we increase the value of the performance indices by 2 to 3% compared to a continuous controller.
|