Date of Award


Document Type

Open Access Dissertation


Electrical Engineering

First Advisor

Enrico Santi


Model Predictive Control (MPC) with a finite control set has been successfully applied to several power converter topologies as reported in the scientific literature and research activity on predictive control techniques has increased over the last few years. MPC uses a discrete-time model of the system to predict future values of control variables for all possible control actions and computes a cost function related to control objectives to find the optimal control action. The control action which minimizes the cost function is selected and applied to the system for the next time interval. Different control objectives can be introduced in the user-defined cost function and controlled simultaneously by solving the multi-objective optimization problem. This approach is particularly advantageous for certain power converter topologies, such as Direct Matrix Converter (DMC) and dual-output power converters, for which conventional control techniques require complicated Pulse Width Modulation (PWM) schemes and multi-loop control, incurring high computational burden and complexity. Conversely, since MPC does not need a modulator to generate switching signals, implementation of the MPC technique is simple and intuitive. However, the MPC method also has several drawbacks:
1. Real-time implementation of MPC incurs high computational burden
2. There is no analytical procedure to adjust the weighting factors for multi-objective optimization problem
3. A complete system model must be derived since MPC method uses this model to predict control variables
4. MPC implementation is not straightforward for several power converter topologies, such as dual-output power converters.

In this dissertation four specific contributions are reported that address these drawbacks.

First, a fully FPGA-based real-time implementation of model predictive controller is proposed for direct matrix converter. In conventional real-time implementation of model predictive control method, Digital Signal Processors (DSPs) and Field-Programmable Gate Arrays (FPGA) are both used to ensure fast processing operation and preserve performance of the predictive controller. For the proposed, real-time implementation method, all control calculations and the safe commutation scheme for DMC are fully implemented in the FPGA and the need for a DSP is eliminated. Advantages of the proposed approach are simplicity and the ability to exploit the parallel computation capability of the FPGA to calculate in parallel the predictive state for all switch combination. This translates in a significant reduction of required computation time and potentially in reduced control hardware cost.

Second, a novel model predictive control scheme for the three-phase direct matrix converter based on switching state elimination is proposed. The conventional MPC solves a multi-objective optimization problem by minimizing a multi-objective cost function over a one-step horizon. The control performance is strongly affected by the weighting factors used in the cost function and this is problematic. The proposed method solves this difficulty by eliminating the weighting factors and using a state elimination method based on error constraints that have a clear physical interpretation.

Third, the model predictive control scheme is proposed for Nine-Switch Inverter (NSI) under an unknown load condition. Nine-switch inverter is a dual-output inverter and the proposed method can control two three-phase load simultaneously by solving single optimization problem. In power electronics applications, control of the power converter must work well under all load conditions and the control method should provide clean power no matter what the load is. In this work, two ac load currents are estimated using full-order observers and converter is controlled by using model predictive control method.

Fourth, the model predictive control scheme is proposed for dual-output Indirect Matrix Converter (IMC). Modulation method for this topology is complicated and conventional linear control techniques require tuning of the controller parameters. In conventional control technique, multi-loop control is required to independently adjust the two ac outputs. The usage of multi-loop control techniques increases the complexity of implementation of the controller. On the other hand, proposed method can achieve several control goals by using single control loop and provide good system performance.