Comparing filtered PID and smith predictor control of a thermal plant

The paper compares filtered PID control with two filtered Smith predictor modifications and experimentally points out different impact of the derivative terms in the considered controllers on the loop robustness and performance. All structures applied to a thermal plant control are based on its approximation by the first order time-delayed model. They include an nth order binomial filter for measurement noise attenuation. In order to stress the time delay impact, the intrinsic plant time delay is increased by an additional delay in Simulink. The evaluation shows a central role of an appropriate filtration in the controller design. It enables a significant control effort reduction by keeping nearly the same speed of controlled processes. To avoid different tuning scenarios for particular controllers, choice of an optimal solution is based on noise characteristics confronting the speed of transients with the additional control effort required.