National Technical University of Athens (NTUA)
Συμβολή στις πιθανοτικές προβλέψεις παραγωγής ΑΠΕ με χρήση προηγμένων τεχνικών μηχανικής μάθησης
Abstract
dc:descriptionThe PhD thesis focuses on the development of advanced machine learning techniques for the probabilistic prediction of renewable energy production, with emphasis on wind and solar energy. The integration of RES into existing power systems presents significant difficulties for system operators and market players, mainly due to their dependence on weather conditions. To successfully integrate RES and reduce power imbalances in the market, accurate forecasting is essential along with the development of decision support tools. The first chapter introduces the scope of the thesis and describes the research objectives and the context in which the thesis is developed. The second chapter focuses on weather forecasting and discusses weather forecasting methods and their importance for accurate solar and wind energy forecasting. The third chapter presents the physical and statistical modelling techniques used to forecast renewable energy production. The forecasting methods and their importance for the integration of RES into existing power systems are analysed. Chapter four discusses the processing of numerical weather forecasts for wind speed prediction. A new method based on artificial neural networks is presented, which approximates the probability distribution functions. The fifth chapter presents the use of hybrid neural networks to estimate probability distributions in forecasting renewable energy production. The new parametric method based on artificial neural networks is analyzed, which combines the cluster forecasting and probability density estimation processes into a single multi-component architecture. In chapter six, the challenge of forecasting RES generation from geographically distributed plants over large areas is addressed. The dimensionality problem created by the large number of input data based on weather forecasts is solved by a dimensionality reduction method. This method divides the numerical weather forecasts into sub-regions and selects only the most relevant ones, which are used as inputs to the forecast models. Chapter seven investigates the accuracy, consistency and value of the probabilistic RES generation forecasts. The basic principles of energy markets are analysed. This analysis includes the use of probabilistic models to forecast total generation and manage the risks associated with price fluctuations and volatility of energy production. The final chapter summarizes the main findings of the thesis and suggests directions for future research. The main contributions of the thesis to the development of reliable probabilistic models for predicting wind speed and renewable generation, as well as the prospects for further refinement and application of these models in real-world conditions are discussed.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Konstantinou, Theodoros
- Κωνσταντίνου, Θεόδωρος
Subjects
dc:subject × 12- Ανανεώσιμες πηγές ενέργειας
- Αλγόριθμοι τεχνητής νοημοσύνης
- Προβλέψεις
- Renewable energy sources (RES)
- Artificial intelligence
- Forecast
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Ηλεκτρική και Ηλεκτρονική μηχανική
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Electrical and Electronic Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/57569