Synthesis And Implementation Of Model Predictive Control For Power Converters

This work focuses on the development of Model Predictive Control(MPC)for multilevel converters to control and optimize the output currents and voltage.The controller objective is to improve the converters performance in term of fewer harmonics and good voltage balancing.The proposed model predictive algorithm is developed through multiple steps to solve and tackle the common issues with multilevel converters.The model was developed for single-phase and three-phase seven-level hybrid cascaded multilevel converter based on Active Neutral Point Clamped(ANPC)converter and an H-bridge with a flying capacitor.The proposed control techniques are all based on Finite Set Model Predictive Control(FS-MPC)which takes advantage of the discrete nature of power electronics to set finite prediction set of the system behavior.On the basis of the hybrid seven-level converters and the FS-MPC the following research is conducted.(1)Conventional FS-MPCs for single-phase multilevel converters are based on cost function that contains multiple objectives in a single formulation such as current tracking and capacitor charge balancing for the ANPC hybrid seven-level converter.These objectives are balanced using a weighting factor that may have a negative impact on the optimization priorities.A satisfactory MPC was developed to introduce an improved control strategy to balance the capacitors voltage separately from the current tracking subject to soft constraints.The degree of membership of the capacitor charge to a nominal voltage range defines the degree of freedom(soft constraints)in which the current tracking can be prioritized for certain voltage levels.Compared with conventional methods the proposed method can maintain low harmonics distortion in output current without compromising the capacitor charge for all output voltage levels.(2)A Multilevel-MPC(MMPC)architecture is proposed for solving the control issues faced when upgrading a single-phase converter into a three-phase converter.It exploits decomposition and aggregation to divide the optimization process into two stages.A primary stage that contains three autonomous sub-MPCs each for a single-phase current tracking complemented with a satisfactory capacitor optimization.The second optimization stage is concerned with capacitors balancing only.It defines the exact control sequence subject to the load current direction and the capacitor charge as well as it introduces swapping between redundant voltage levels to ensure equal distribution of the switching efforts.The satisfactory controller is also adopted and used in the multilevel MPC to minimize the Common Mode Voltage(CMV)between the three phases and the ground.The influence of the proposed methods on the performance is analyzed in steady state and transient operational conditions.The results show the effectiveness of the MMPC and the advantages of the satisfactory optimization in transient operation.(3)Since the performance of MPC for multilevel converters is often subject to high switching frequencies and therefore high switching efforts and short life cycle of the switching devices.A multistep MPC(also known as long horizon MPC)is proposed for the three-phase seven-level converter.It adopts a Sphere Decoding Algorithm(SDA)with a modified formulation to act as an Evolutionary Optimization(EO)algorithm.Multistep MPC enables better performance of the converter even with a reduced switching frequency but it is associated with large number of possible control inputs,thus,it is an NP-hard optimization problem.The Proposed optimization reduces the number of control entries to be processed as it determines a smaller set of possible elite solutions.Then it constructs a new set of possible sub-optimal solutions from the elite solutions vector.These solutions are evaluated sequentially through multiple iterations until the set of solutions is shortened to contain only one single optimal control vector.The performance of the proposed optimization is evaluated through experimental tests and statistical analysis.Results have shown its effectiveness in term of current quality,switching frequency and computations burden.