Fuzzy predictive PID controllers: a heating control application

The authors describe a new fuzzy predictive proportional-integral-derivative (PID) algorithm that is derived from the generalized predictive control theory. The key idea is the introduction of fuzzy models for the prediction of process output and of perturbations that affect the process. The control policy is obtained by a simple PID-like correction from the last control sequence as provided by a receding control strategy. The resulting algorithm proves to be a versatile tool, easy to design and to implement with reduced computational power. An interesting feature of the proposed algorithm is the looking-ahead property that is useful in applications where the setpoint is known in advance. Another advantage is that the algorithm needs only a rough model of the process, and behaves robustly even when the fuzzy models are only coarse approximations of the process. A self tuning version is outlined using an online fuzzy model identification of the process. An example of an application to thermal control of a building is discussed