Advanced deep learning methodologies for renewable energy sources

Artificial Intelligence and Renewable Energy Sources represent two fields with immense potential to impact the upcoming years: the former enhances data-driven tasks through advanced learning and automation, while Renewable Energy Sources offer sustainable and clean alternatives to traditional fossil fuels. Integrating renewable sources into the global energy landscape has become essential for achieving decarbonization targets; however, the inherent variability and unpredictability of renewable resources, such as solar and wind, pose substantial challenges to their efficient deployment and integration into existing energy grids. Deep learning has already proven to be an effective tool in solving complex tasks across various domains, continuously evolving and improving through advances in architectures, training techniques, and computational resources. Its ability to leverage large amounts of data enables the extraction of intricate patterns and representations, with ongoing developments further enhancing its adaptability, efficiency, and generalization capabilities in increasingly challenging applications. This doctoral thesis investigates the potential of advanced deep learning methodologies in the renewable energy sector, examining their effectiveness across multiple tasks to address key challenges in the field. The research focuses on three core areas: forecasting renewable energy production, detecting anomalies in time series data, and enhancing explainability within energy networks, providing a comprehensive analysis of their capabilities and impact. Furthermore, this doctoral research expanded its focus beyond the primary domain, exploring methodologies that complement Artificial Intelligence, such as quantum computing, and validating advanced architectures like Graph Neural Networks and Data Attribution techniques in entirely different contexts, e.g. high-energy physics or telecommunication. By integrating Artificial Intelligence with cutting-edge technologies across diverse application domains, this work highlights the power of interdisciplinary research in driving innovation and unlocking new opportunities for scientific advancement. These advancements not only address immediate practical issues but also contribute novel theoretical frameworks that push the boundaries of current understanding in the field. In fact, the dual approach of this research, balancing practical application with theoretical innovation, ensures that the findings are both impactful in real-world scenarios and instrumental in advancing academic discourse. By resolving existing challenges and simultaneously laying the groundwork for future theoretical advancements, this work offers a comprehensive and multifaceted contribution to the field of Deep Learning and renewable energy.