Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/13982849-CB6E-45A2-93DD-2C28B6FD694A | ||
| Year | 2025 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Δημήτριος Λύτης, "Διάγνωση βλαβών φωτοβολταϊκών μετατροπέων ισχύος με τεχνικές τεχνητής νοημοσύνης", Διπλωματική Εργασία, Πολυτεχνείο Κρήτης:Σχολή Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών, Χανιά, Ελλάς, 2025 https://doi.org/10.26233/heallink.tuc.104154 | ||
| Appears in Collections |
The growing global demand for sustainable electricity production has led to the widespread adoption of Renewable Energy Sources (RES), with photovoltaic (PV) systems playing a central role. However, the reliability and performance of PV installations heavily depend on the proper operation of power converters, which are responsible for energy conversion and management. These converters are prone to component faults and external disturbances, making timely fault detection essential to ensure energy efficiency and system continuity.This thesis focuses on the development and evaluation of fault detection methodology for PV power converters using Artificial Intelligence (AI) techniques. Specifically, it employs Gaussian Process Regression (GPR) to estimate the normal operating range of a converter and Genetic Algorithms (GA) to extract the extreme values of seven statistical features of the system’s output (range, mean, standard deviation, skewness, kurtosis, entropy, and centroid). By comparing actual output signals with the limits defined by GPR and GA, operational anomalies can be accurately and promptly identified.The proposed methodology was applied and assessed on different types of converters (Buck converter, Boost converter, and full-bridge inverter), demonstrating high fault detection accuracy even in cases of subtle deviations. Results indicate that integrating AI methods into PV converter monitoring systems can significantly improve the predictive maintenance capabilities and reduce operational costs, thus enhancing the long-term reliability and efficiency of RES systems.
Copyright © DIAS 2013
