Publicação
Forecasting Renewable Energy Production in Madeira Island
| Resumo: | The global shift toward sustainability has become increasingly urgent as reliance on fossil fuels accelerates climate change, pollutes ecosystems, and depletes finite resources. Yet, modern society depends on these energy sources for essential needs, given their stable nature. Balancing the demand for reliable power with the need for sustainable solutions is one of today's greatest challenges. Renewable energy offers a cleaner alternative, as it neither harms the environment nor relies on finite resources. However, its primary limitation is weather related intermittency, which poses a significant challenge for Madeira Island, a Portuguese region that remains more dependent on fossil fuels than its mainland. This increased reliance emerges from the island's status as an isolated electrical system, where the need for a stable and reliable energy supply is critical. In response, this research aimed to develop a practical and effective forecasting framework to support accurate renewable energy predictions and contribute to Madeira’s decarbonization efforts. This case study leveraged solar, wind, hydropower, and biomass energy data, provided by Empresa de Eletricidade da Madeira, the region's main utility company, and complemented by weather data retrieved from Visual Crossing Weather API, to develop a clear and practical forecasting framework. The proposed solution consisted of day-ahead forecasts of XGBoost for wind and hydropower energy sources, and Random Forest for solar and biomass energy sources, all with grid search defined parameters, delivering the most accurate day-ahead forecasts of this research. To assess this solution’s performance, error metrics and visualization methods (line chart and SHAP values) were employed. Although performance varied across energy sources, the model achieved strong results for hydropower, acceptable outcomes for wind, moderate but cautious results for biomass, and underwhelming performance for solar. Nonetheless, this research delivered a comprehensive and applicable forecasting framework, not only for Madeira Island but also for other regions aiming to apply time series techniques to enhance their reliance on renewable energy. In doing so, it contributed to the United Nations’ 7th Sustainable Development Goal: ensuring access to affordable and clean energy for all. |
|---|---|
| Autores principais: | Ferreira, Gonçalo Lagos |
| Assunto: | Renewable Energy Production Energy Analysis Machine Learning Time Series Analysis SDG 7 - Affordable and clean energy |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | The global shift toward sustainability has become increasingly urgent as reliance on fossil fuels accelerates climate change, pollutes ecosystems, and depletes finite resources. Yet, modern society depends on these energy sources for essential needs, given their stable nature. Balancing the demand for reliable power with the need for sustainable solutions is one of today's greatest challenges. Renewable energy offers a cleaner alternative, as it neither harms the environment nor relies on finite resources. However, its primary limitation is weather related intermittency, which poses a significant challenge for Madeira Island, a Portuguese region that remains more dependent on fossil fuels than its mainland. This increased reliance emerges from the island's status as an isolated electrical system, where the need for a stable and reliable energy supply is critical. In response, this research aimed to develop a practical and effective forecasting framework to support accurate renewable energy predictions and contribute to Madeira’s decarbonization efforts. This case study leveraged solar, wind, hydropower, and biomass energy data, provided by Empresa de Eletricidade da Madeira, the region's main utility company, and complemented by weather data retrieved from Visual Crossing Weather API, to develop a clear and practical forecasting framework. The proposed solution consisted of day-ahead forecasts of XGBoost for wind and hydropower energy sources, and Random Forest for solar and biomass energy sources, all with grid search defined parameters, delivering the most accurate day-ahead forecasts of this research. To assess this solution’s performance, error metrics and visualization methods (line chart and SHAP values) were employed. Although performance varied across energy sources, the model achieved strong results for hydropower, acceptable outcomes for wind, moderate but cautious results for biomass, and underwhelming performance for solar. Nonetheless, this research delivered a comprehensive and applicable forecasting framework, not only for Madeira Island but also for other regions aiming to apply time series techniques to enhance their reliance on renewable energy. In doing so, it contributed to the United Nations’ 7th Sustainable Development Goal: ensuring access to affordable and clean energy for all. |
|---|