Description: In the context of renewable energy penetration as production source, high power centralized electromechanical generators are progressively replaced by distributed static converters. Power quality of these converters therefore becomes a concern of growing importance. A significant part of power quality is related to harmonic generation and damping inside the static converter. A lot of theories exist to improve power quality of AC grid connected converters. The level of complexity of these algorithms varies from simple control loops to complex self-tuning systems. From an industrial perspective, developing and improving these algorithms is a very difficult task. On the one hand, the grid connection requirements become always more stringent. On the other hand, grid parameters can change in a very large range depending on the particular installation location.
The proposed master thesis will consist in investigating the applicability of stateof-the-art deep learning algorithms to achieve better static converters power quality and robustness to parameter variations than deterministic approaches. Such learning algorithms appear promising given the possibility to ingest data acquired over one or several grid periods and leverage them over the next period. They may also be applicable to DC micro-grids, a possible extension of the thesis.
After a short review of existing current control and harmonic reduction algorithms, the student will analyze possible algorithms that could be applied to this specific task. He will carry out simulations to validate first concepts and evaluate hardware resources required to perform real time calculations. He will choose – and justify the choice of – relevant test environments among the numerous ones available at CE+T Power, e.g., Hardware-In-the-Loop (HIL), FPGA-based control boards, etc. This master thesis lies at the intersection of the control, power electronics and deep learning algorithms fields, a challenging combination. It constitutes a great opportunity to advance knowledge in areas of major importance for the ongoing energy transition.
Labo Entreprise: CE+T Power, rue du Charbonnage, B-4020 Wandre, Belgique
CE+T Power offer:
Nom du contact industriel: Benoît Bidaine
email du contact industriel: b dot bidaine at cet-power dot com
Adresse Web: https://www.cet-power.com
Contact FSA: Fabrice Frebel - fabrice dot frebel at uliege dot be
Supervision: This master thesis is co-promoted by Professor Louppe.