Democritus University of Thrace (DUTH)
Μοντελοποίηση της αξιοποίησης των ανανεώσιμων πηγών ενέργειας με την χρήση μηχανικής μάθησης και μαθηματικών εργαλείων
Abstract
dc:descriptionThe production of electricity through renewable energy sources has increased greatly in recent years. Countries have primarily turned to this form of energy for environmental, economic and security reasons. Optimizing the efficiency of renewable energy sources and minimizing their impacts are influenced by many factors and depend mainly on the characteristics of the installation site. The literature, in particular, divides the criteria that affect the location of wind farms into three categories: environmental, economic and social. Most of the research that has been done on the particular topic aims to establish individual criteria that correspond to a weighting factor in order to achieve balance among these three categories. Such an approach can lead to incorrect conclusions or to conclusions that cannot be generalized and applied to areas with different characteristics. In this doctoral thesis, a wind farm siting methodology is proposed, according to which it is not necessary to define weighting factors for each criterion. This is achieved through machine learning techniques and methods such as clustering and neural networks. The Greek Territory as a whole is the study area, as its geomorphology and the special meteorological conditions that prevail make it one of the most suitable areas for the installation of renewable energy sources worldwide. The locations of the already installed wind turbines were retrieved and for each one nine characteristics were calculated which, according to the extensive literature review carried out, are considered the most important criteria. Initially, clustering was performed on the locations of the wind turbines, which is based on the objects of computational geometry and graph theory. The results of this process lead to important conclusions about the characteristics of the areas where wind turbines are installed, as well as how these areas were affected by the said investments. Then it was designed a neural network of radial basis functions, which is trained using the characteristics of the areas with installed wind turbines. This allows the network to assess potential new locations for wind farm installation and operation. The neural network exploits the clustering of wind turbines and classifies the new locations into the resulting clusters. Finally, the landscape fragmentation index is added in order to exclude areas that have not been affected by external interventions. Also, protected areas (wildlife refuges, wetlands, Natura areas, etc.) have been excluded from the possible locations, for environmental reasons. Through these two additions, the increase in the acceptance of public opinion regarding the installation and operation of wind turbines is achieved. According to the results, it was found that the locations of wind turbines in Greece can be separated into four groups with different characteristics. New sites were assessed for suitability and classified according to these four categories. Subsequently, all sites presenting a “Low” or “Very Low” landscape fragmentation index were removed. Therefore, potential locations for the installation of wind farms in the Greek Territory have been evaluated, drawing information from the characteristics of the already installed wind turbines. A key feature of the methodology is the ability to redefine the findings in the event of new data or new features becoming available. Finally, it is possible to apply it to different types of renewable sources, such as photovoltaic or offshore wind farms.
Degree
thesis:*- Grantor dc:publisher
- Democritus University of Thrace (DUTH)
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Xenitidis, Kleanthis
- Ξενιτίδης, Κλεάνθης
Subjects
dc:subject × 24- Ανανεώσιμες πηγές ενέργειας
- Αιολικά πάρκα
- Συσταδοποίηση
- Διαίρεση συμπλεγμάτων / τριγωνοποίηση Delaunay
- Τεχνητά νευρωνικά δίκτυα
- Μηχανική μάθηση
- Χωροθέτηση
- Renewable energy sources (RES)
- Wind farms
- Clustering
- Delaunay triangulation
- Artificial neural networks
- Machine learning
- Spatial planning
- Γεωεπιστήμες και Επιστήμες Περιβάλλοντος
- Φυσικές Επιστήμες
- Επιστήμες περιβάλλοντος
- Μαθηματικά
- Διεπιστημονικές εφαρμογές των μαθηματικών
- Earth and Related Environmental Sciences
- Natural Sciences
- Environmental Sciences
- Mathematics
- Mathematics, Interdisciplinary Applications
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/56109