Design and development of advanced model-based robust multi-objective controllers for multilevel converters

Abstract

Considering the essential role of multilevel converters (MLC) in the future of distributed renewable energy resources, electric vehicles, and automation, this thesis contributes to developing advanced multi-objective controllers and topologies for MLCs to enhance their performance, stability, and reliability in grid-tied and stand-alone modes of operation. Accordingly, in the first research work of this dissertation, an optimized sliding mode controller (OSMC) is introduced to address the voltage balancing issue (recognized as a benchmark) of the seven-level packed U-cell (PUC) converter without involving an outer voltage loop controller. Further stability analysis is also accomplished to guarantee the stability and robustness of the sensorless OSMC algorithm under dynamic conditions and parametric uncertainties. Concerning the lack of an effective analytical method to tune the weighing factors in model predictive control (MPC) strategies, an offline artificial intelligent-based (AI) training technique is developed in the second work using the artificial bee colony (ABC) algorithm to enhance the multi-objective control performance of predictive controllers, which are dedicated to MLCs. The implementation results of applying the supervised learning MPC (SLMPC) to a three-phase neutral point clamped (NPC) converter demonstrate that AI is quite effective in dealing with the tedious tuning of the MPC-based algorithms. Even so, offline training cannot guarantee the optimal control performance of the MPC under dynamic conditions. Regarding this fact, an online tuning strategy is developed in the following research work based on Artificial Neural Network (ANN) to adapt the weighing factors for an MPC controller, which is applied to a seven-level modified PUC (MPUC) active rectifier. A novel data-free selftraining strategy is also established using the particle swarm optimization (PSO) algorithm to train the ANN-based regulator. The proposed training strategy significantly improves the contribution of ANNs in power electronics control problems. Despite the astonishing multiobjective performance offered by finite control set model predictive controllers (FCSMPC), Lyapunov stability analysis is not supported due to the discrete control structure. To address this issue, a novel predictive Lyapunov function is developed in the same research work to guarantee the stability of multi-objective predictive controllers. Since the proposed stability objective is independent, it causes zero impact on the optimal multi-objective control performance of the predictive controllers. To reduce the sensitivity of the self-training method to the initial parameters, an advanced multi-core fast self-training strategy (FSTS) is constructed by the imperialist competitive algorithm (ICA) in the next research work. Using the new generalized FSTS, an intelligent predictive multi-objective controller (IPMOC) is developed for MLCs, which can track over seven control objectives simultaneously. In addition, a novel selective predictive harmonic mitigation (SPHM) objective is formulated in the same work for MPC-based controllers to suppress any harmonic order directly. The proposed SPHM is generalized and functional for various MLCs. Finally, boost back E-Cell (BPEC) topology as the modified version of the conventional packed E-Cell (PEC) is introduced to provide a cost-effective compact MLC (CMLC) for power quality ancillary services. Despite the original PEC, the proposed BPEC dominates the point of common coupling (PCC) using lower DC link voltages. As a result, the volume and cost of the converter remarkably reduce as lower power components are required. As a case study, a compact active power filter (CAPF) is designed using the proposed BPEC to verify its feasibility. The corresponding test results show that using an MPC-based multi-objective controller for the BPEC leads to an advantageous CAPF, which can meet all the expectations for the ancillary services. The performance of the developed multi-objective controllers, as well as the BPEC topology, has been extensively evaluated using various test scenarios applied to specific testbeds designed based on dSPACE 1104, Microlabox 1202, OPAL-RT OP8662, Chroma 61086, and the power board of the converters, including the seven-level PUC, seven-level MPUC, threephase NPC, and 11-level BPEC.

Date	21 Nov 2023
Original language	American English
Awarding Institution	École de technologie supérieure
Supervisor	Kamal Al-Haddad (Supervisor)