Machine learning models to predict electricity consumption and the impacts of COVID-19 in Portugal

• Autor(a) principal: Sucena, Ana Sofia Lobato
• Data de Publicação: 2023
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/29599
• topic: Electricity consumption; Machine learning; COVID-19 pandemic; Predictive models; Consumo de eletricidade; Pandemia COVID-19; Modelos preditivos

This thesis analyzes how data from public data sources and machine learning models can be used to forecast electricity consumption in Portugal. Accurate forecasts are crucial for efficient energy management, given the rising global demand for energy. Portugal presents a compelling case for consumption projections since it significantly relies on energy imports and suffers from poverty. The study uses a data-driven methodology to analyze twelve years of consumption patterns and examine how the COVID-19 pandemic, weather patterns, and GDP affect electricity use. Five predictive models were studied - SARIMA, SARIMAX, VAR, SVR and LSTM - and their performance indicators in two different periods (one for the twelve years of analysis, including during Covid-19, and the other only for data before Covid-19). Thus, this study makes it possible to evaluate the performance of machine learning models in stable and non-stable periods. The study acknowledges its limitations, such as the lack of data in the post-COVID era, while providing valuable insights for developing and managing energy policies.