Date of Award
2026
Degree Type
Thesis
Degree Name
Master of Science in Mechanical Engineering and Applied Mechanics
Department
Mechanical, Industrial and Systems Engineering
First Advisor
Sumanta Das
Abstract
This study explores the use of machine learning (ML) to forecast solar and wind energy generation using meteorological and temporal data. Due to the inherent variability and intermittency of renewable energy sources, accurate prediction remains a significant challenge, emphasizing the need for robust data-driven approaches. A supervised learning framework was developed using key input variables, including irradiance, temperature, humidity, wind speed, and time-based features. A multilayer perceptron (MLP) neural network was implemented and systematically optimized through extensive hyperparameter tuning to improve predictive performance. Model accuracy was evaluated using Mean Squared Error (MSE) and the coefficient of determination (R²). The optimized model demonstrated strong predictive capability, highlighting the critical role of hyperparameter tuning in capturing complex nonlinear relationships. To enhance model transparency, SHapley Additive exPlanations (SHAP) analysis was applied to identify the most influential input variables. The results provide insight into the key drivers of renewable energy generation and improve model interpretability. Overall, this work presents a structured and reproducible framework for accurate and interpretable ML-based forecasting of renewable energy.
Recommended Citation
Massa, Eleonora, "INTERPRETABLE MACHINE LEARNING FOR RENEWABLE ENERGY FORECASTING" (2026). Open Access Master's Theses. Paper 2748.
https://digitalcommons.uri.edu/theses/2748