| Περίληψη: | The rapid integration of renewables into the daily production schedule has greatly increased the complexity of today’s contemporary power systems. These sources suffer significantly from variability in terms of production due to their large dependance on ambient conditions. As the main purpose of a power system is to ceaselessly supply the demand, in a supervised manner, renewables are usually installed in systems in ways that facilitate cooperation between them and conventional power sources. These systems are known as microgrids and they combine distinct renewables and conventional power sources to either supply local demand or inject power to the grid. The introduction of such complex systems has also led to the need of proper control algorithms that can accommodate their performance.
The general notion of a microgrid is not a novel term, but had been first introduced in the late 1950s. Nevertheless, the conventional control systems that have been developed, mainly dismiss dynamic response and unfortunately these assumptions may lead to a system failure due to unaccounted parameters. In this frame, researchers mainly focused on model-based analysis and development of model-based control algorithms, that regulate these contemporary power systems dynamically, while only making mild assumptions that do not significantly restrict the operation bounds of the final system. Furthermore, the results of the control design are also exploited to appropriately select system parameters that can facilitate its operation and improve the response of the applied control scheme. The proportional and integral controllers constitute the main blueprint of these controllers and have been repeatedly proven to be quite efficient when regulating power systems, but without proper analysis, they may fail to control more complex installations such as those of the microgrids. Thus, a proper stability analysis is important when applying any type of control scheme that aims to regulate specific states of a larger system. Even though PI-based control schemes are usually able to achieve the control design goals, they unfortunately suffer from lack of adaptation and when they are applied in more complex systems, where strict transfer functions cannot be derived, it is almost impossible to control parameters such as overshoots, oscillations or even settling time. Solution to this problem is provided with the introduction of Artificial intelligence. More specifically, intelligent controllers such as neuro-fuzzy techniques and Neural Networks are able to adapt their output in a way that minimizes specific cost functions. Unfortunately, though, they lack vital stability foundations and that’s the main reason why researchers have not yet introduced effective AI-based control schemes.
In this frame, main purpose of this thesis is to design, control and implement a novel microgrid that combines distinct renewable power sources that are regulated by an elaborate energy storage system. The microgrid consists a Permanent Magnet Synchronous Generator-based wind-turbine and a PV panel farm. The microgrid operates in two unique modes known as grid-following and grid-forming. In the former, the microgrid control design goal is to inject set amounts of active power to the grid, while the power sources are regulated to operate in maximum power point mode. In grid forming mode, the microgrid’s sole purpose is to supply local loads under nominal voltage and frequency conditions, whereas the power sources proportionally share the load power through a novel droop technique. Details on the modeling and design of the interface of the power system that provides the ability to control and interconnect the renewable power sources are also provided. Furthermore, based on the equivalent average model, a rigorous control system is developed that is based on the highly-efficient cascaded-PI scheme and relies on the time separation principle. Finally, to prove the efficiency of AI-based controllers, suitable Neural Networks were properly trained, under various circumstances, to improve the transient response of the conventional control technique. Hence, simulation results pertaining the response of the microgrid that took place on Matlab’s Simulink virtual environment, are provided for both the cascaded-PI scheme-based system and the Neural network one, to facilitate their comparison, where it is found that the later significantly, improves the system transient response and therefore the potential of AI-control algorithms on power system applications is established.
|
|---|
