Predictive deviation filter for deadbeat control

Abstract

Deadbeat control calculates excitation by a differential equation model that takes measurements of state variables as initial values and instructions as terminal values. Ideally, this excitation can force state variables to track instructions in the succeeding control interval. However, in a practical system, noises in measured signals can mislead the control system and degrade its performance with increased ripples and harmonics. Therefore, some filters are often used in measurements, but their inherent delay can lead to inaccurate initial values and cause severe overshoots in dynamic cases. In this study, a new type filtering algorithm called predictive deviation filter is proposed for deadbeat control, which filters only the predicted deviation obtained by removing the expected values from measured ones, and then the real values of state variables are recovered. With this method, the influence of the filter delay can be limited only on the deviation which is a small part of the signal, significantly improving the accuracy of the initial values. Furthermore, a comprehensive frequency-domain analysis of the predictive deviation filter is provided. Finally, the new filtering algorithm is applied to drive control of permanent magnet synchronous motor. The simulation and experimental results verify the effectiveness and performance of this approach.