Introduction & Objectives

• Ioannis Vlahavas (AUTH)

The introductory presentation will briefly present the necessity of and possibilities for usingAI and in particular ML for forecasting time series that are influenced by many parameters. It will also discuss the value of reliable forecasting by renewable energy providers for the Energy Exchange, consumers and the environment. Moreover it will highlight the importance of abundant and high quality data, as well as the continuous improvement of a forecasting system will also be addressed.

Concepts of Machine Learning & Deep Learning

• Vasileios Kochliaridis (AUTH)

This presentation introduces concepts of Machine Learning and Deep Learning, with a focus on time-series forecasting. The talk explains how systems can learn patterns from data and make predictions without relying on explicit programming. It then focuses on Deep Learning, which can model complex relationships in large datasets, with a special emphasis on forecasting models for time-series data. The webinar covers modern deep learning methods, including LSTMs, GRUs, Temporal Convolutional Networks and Transformers. Participants will gain a clear overview of how these models are used in real-world applications such as finance and energy.

Forecasting Renewable Energy Production Using Machine Learning

• Anestis Ampatzidis (EY & AUTH)

This presentation focuses on a practical application of Machine Learning and Deep Learning in renewable energy forecasting. Through Day-Ahead forecasting of solar energy production, the discussion covers the complete data pipeline, detailing how meteorological and historical records are processed to train predictive ML models. Particular emphasis is placed onDeep Learning approaches such as Neural Networks. Additionally, the role of Aggregators (FOSE) in the modern energy market is examined to emphasize the real-world value of accurate forecasting. Finally, a first version of an intuitive User Interface, currently under development to manage the entire application pipeline, is presented, offering a complete end-to-end perspective from raw data to the final user experience. The presentation also includes a hands-on demonstration, showcasing how a pre-trained Deep Learning model is used to perform inference and generate solar energy production forecasts.