ROBUST IMPLEMENTATION OF THE MODEL PREDICTIVE CONTROL FOR DCDC POWER CONVERTERS AND INDUSTRIAL CONTROL APPLICATIONS

This dissertation addresses the development of the robustness and strength of the Model Predictive Control algorithm subjected to input constraints for a plant system with and without parameters' uncertainties. In the beginning, the MPC control system was implemented for systems with no types of parameters' uncertainties. The proposed system models were stable and linear and all of its parameters were fully known. They were formulated in model state-space system format. The main objective of this control system design was to maintain a smooth and constant output signal that could easily track theassigned desired output signal. The technical process of this control design was to calculate the optimal solutions for the proposed plant system by optimizing the Quadratic programming problem (QP) subjected to linear inequality constraints. Therefore, the proposed control system successfully forced the outputs of the proposed systems to track the output reference signals in a fast response and with very small steady-state errors, even with the change in prediction horizon values. The second control system approach was for a system model which assumed its parameters’ uncertainties. In the other words, the parameters were not precisely known, but they were bounded in a minimum and maximum range. The parameters' uncertainties and the converter's switching behaviors made it act as a highly nonlinear system. Therefore, the Adaptive Model Predictive controller (AMPC) with the Linear Parameter Varying (LPV) control algorithm was implemented to address these issues and to secure a sustainable output signal with on types of noise or degradations. In this algorithm, the LPV model was created out of a set of Linear Time-Invariant (LTI) models, which are used to update the AMPC controller based on the feedback signals that come from the plant model and the change in the system parameters. Due to the changes in the plant model parameters over time, the AMPC was the perfect control approach due to its capability to update its prediction model and the operating condition over a prediction horizon interval. Since the AMPC is an online optimization-based approach, the QP variables and parameters can be tuned based on changes of the system measurements in real-time, and the LPV scheduling parameters. The proposed AMPC and LPV algorithm was compared against different control system approaches. Also, the proposed AMPC and LPV algorithm was implemented using an Arduino Mega 2560 microcontroller to show its performance in a real-time environment. In summary, from the outputs and the results, the proposed AMPC and LPV control system showed higher levels of the performance interims of the purified output, faster responses, and the computational time in both the simulation and real time results. MATLAB, SIMULINK, and ARDUINO support packages were used for the system design and implementations.