Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/E1A7AF8B-BFA5-455F-9B3D-8EE65A0DB67C | ||
| Year | 2025 | ||
| Type of Item | Master Thesis | ||
| License |
|
||
| Bibliographic Citation | Emmanouil Lioudakis, "Regulation of the power production of photovoltaic DC/AC converters", Master Thesis, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2025 https://doi.org/10.26233/heallink.tuc.102040 | ||
| Appears in Collections |
The continuously increasing demand of electric energy in modern societies, along with the increasing cost of fossil fuels are two of the main factors that led to a rapid increase on the share of photovoltaic (PV) systems in modern electricity generation and transmission grids. This may affect the grid frequency stability, since the energy production of PV systems varies over time, following the stochastic variation of solar irradiance. Grid frequency stability can be improved by regulating the output power of grid-connected PV systems to a specific reference value with the application of a Flexible Power Point Tracking (FPPT) algorithm. In cases where some of the modules of a PV array are shaded (e.g., due to neighboring buildings), the application of a Global FPPT (GFPPT) algorithm ensures efficient regulation of the PV system’s power production. In this thesis a novel reinforcement learning-based GFPPT algorithm is presented. After examining its performance on a standalone PV battery charging system, it was modified to be applicable on a grid-connected PV inverter with Internet of Things (IoT) connectivity, which is remotely controlled via a Wi-Fi interface. In this case the algorithm had been trained before being applied to the real system, to ensure satisfying grid-injected power quality and quick convergence to the reference power from the first operation hours of the system. The experimental results of applying the proposed GFPPT algorithm on both systems showed that it is more efficient than the existing GFPPT algorithms that have been presented in the literature. A neural network-based GFPPT algorithm was also developed and compared with the aforementioned GFPPT method. The experimental results showed that while both algorithms had been trained on the same dataset, the neural network-based algorithm cannot guarantee the minimization of the PV inverter power losses during the process of regulating its power production.
Copyright © DIAS 2013
