04 Décembre – Thesis defense - Samuel Jupin

10 h Full videoconferencing

Advanced control of power converters for weak grid applications.

With the progressive rise of the micro-grids incorporating renewable energy sources, a new electricity distribution paradigm is emerging. These new architectures interface uncontrolled consumers with intermittent energy sources, therefore imposing more stress on the conversion, storage and management of the energy.
Power converters are adapting accordingly, in particular, with the development of multi-level converters, which allow higher power rates and better power quality than their predecessors with similar components, but whose control is becoming increasingly complex.
Due to their hybrid nature, the control of power converters is traditionally split into two parts: on the one side, the continuous objectives related to the main interfacing function of the power converters, and, on the other side, the driving of their quantized power switches, known as the modulation strategy.
In this context, the growing demands in efficiency, reliability, versatility and performance require a high level of intelligence of the complete control structure. To meet these requirements, the objectives of this research work are to address both the interfacing objectives and the inner driving of the converter into a single controller. This decision implies incorporating the non-linearity of power converters into the controller, equivalent to suppressing the traditional modulation block. Modulation is the traditional solution to linearize the inner operation of the converters. The Model Predictive Control (MPC) approach was chosen to handle the non-linearity and the diversity of control objectives that accompany power converters.
The developed control algorithm combines graph theory, with Dijkstra, A* and other algorithms, with a special state-space model designed for switching systems to form a powerful universal tool capable of simultaneously manipulating the discrete and continuous nature of the converter and its environment. Switched state-space models are studied, leading to interesting results on stability and controllability concerning their application on power converters.
The obtained controller is then tested in simulation, with various case studies: grid-connected and standalone inverter, rectifier and bidirectional operation. These situations are studied for three common multi-level topologies: Neutral Point-Clamped, Flying Capacitor and Cascaded H-Bridge. The exact same MPC structure is used for each and every one of the case studies, with adaptations of its internal behavior. This behavior is agglomerated in two functions: the prediction, containing the model of the converter, and the cost function, which translates the control requirements into the optimal problem solved by the algorithm. Changing the topology implies adjusting the model, without impacting the cost function, while modifying this function is sufficient to adapt to the different applications.
The results show that the controller manages to directly drive the power switches according to the application, demonstrating a large variety of considerations and objectives. The overall performance of this unique structure is comparable to that of the multiple structures used for each of the studied cases, with the notable exception of rectifier operation mode, where the speed and range of possibilities are particularly interesting.
In conclusion, the developed controller manages miscellaneous applications, topologies, objectives and constraints. While the traditional linear control structures have to change, often deeply, for different operation modes and control requirements, such modifications do not affect the control architecture of the designed MPC controller. This shows the versatility of the proposed solution and its universality, further demonstrated by its ability to adapt to different power converters without modifications. Finally, the complexity of the modulation is fully included in the structure, offering simplicity and flexibility to the control design.

Event localization