Short, medium, and long-term electricity generation forecasting with machine learning in wind energy systems

The increasing demand for energy and the necessity to utilize renewable energy sources more effectively and sustainably have made accurate forecasting of energy production a critical challenge. Wind energy, as a clean and infinite resource, plays a crucial role in meeting this demand. However, optimizing energy production processes and ensuring their sustainability require accurate predictions of wind speed and energy output.In this thesis, wind speed and energy production data were analyzed using time series analysis and machine learning approaches. The study aimed to develop models that can effectively predict energy production based on historical data, thereby contributing to sustainable energy management and efficient use of renewable energy resources. The research began with the preprocessing of energy production data and exploratory data analysis, focusing on uncovering trends, seasonal patterns, and anomalies in wind speed. In the main part of the study, wind speed and energy production forecasting were performed under different scenarios using Long Short-Term Memory (LSTM) algorithms. Models were developed during the training processes using different time windows (e.g., 7, 14, and 30 days of historical data) and different forecasting intervals (e.g., predicting 7, 14, and 30 days into the future).The models were evaluated based on their predictive performance and ability to generalize to unseen data.Findings indicate that applying advanced machine learning techniques to time series data significantly improves the accuracy of energy production forecasts. The developed models demonstrated strong potential in supporting decision-making processes for energy sector planning and operations, ensuring efficient and sustainable management of renewable energy sources.