An Adaptive PPI Controller

Sammanfattning: Processes with long dead times are among the most difficult to control. This has urged the development of controllers that cope with this kind of delays. A possible approach is to use a model predictive controller, i.e. a controller with an internal model of the process and its dead time. In ABB’s new control system, Control IT, there exists such a controller, namely the PPI controller. PPI stands for Predictive PI, and as the name indicates it shares some of its features with the common PI controller. The system contains an auto-tuner that is able to detect long dead times and to design a PPI controller. Another feature of Control IT is adaptive control. An adaptive controller has the ability to change its parameters in order to optimize its performance. Today, there is no adaptation mechanism for the PPI controller. This thesis investigates the possibilities of an adaptive version of the PPI controller. A test version has been implemented and simulation results show that there are processes that would benefit from an adaptive PPI. The big flaw is however that adaptation is possible only at set point changes, an event that might not be that common in industry. This implies that industrial studies must be performed before the solution is included in the final product. Other problems that have been encountered are sensitivity to noise and ramped set point changes. It has also been shown that the existing auto-tuner sometimes chooses a PPI design in cases were a PID design would be more useful. The thesis further discusses the subject of switching controller structure while in adaptation mode