This article proposes an option to execute conveniently the traditional Model Predictive Control (GPC), called the Alternative Generalized Predictive Control (AGPC). In this AGPC the disturbed discrete-time input-output mapping of controlled plant is utilized directly for the prediction of plant outputs, instead of its transfer function as being done in conventional approach. Hence, the solving Diophantine equations will be avoided. Within this AGPC all recorded values of plant inputs/outputs in the last time-horizon are matched into separate vectors for computing predictive control signals at the next control step, which helps therefore that its implementation becomes more manageable. To verify via virtually real simulation the control performance of this proposed AGPC a Simscape water tank model, which is chosen as the controlled plant, had been created among Thermal Fluids Toolbox. The simulation is carried out for two different circumstances, one by using AGPC and the other by applying conventional PID, for comparison purpose. The simulation demonstrates also how to realize this AGPC in practise.
Keyword
MPC, alternative GPC, PID, optimization, virtually real tank model.
