A carregar...
Publicação
Participação da geração renovável no mercado de reservas de um sistema eléctrico
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.88 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
A crescente penetração de fontes de energia renováveis variáveis no tempo, vRES, no sistema eléctrico, como a solar ou eólica, está a transformar significativamente os mercados de eletricidade, devido
à sua natureza intermitente e imprevisível. Isso torna as previsões de produção e consumo de energia
mais desafiantes, especialmente porque os mercados fecham entre 1 e 37 horas antes da entrega real da
energia, podendo originar discrepâncias entre as energias contratadas e necessárias. Manter o equilíbrio
entre a oferta e a procura em tempo real é vital para a segurança e estabilidade da rede, função que recai
principalmente sobre os operadores de redes de transporte (TSO).
Os TSO utilizam mercados de reserva de energia, onde adquirem de forma simétrica potência secundária ascendente e descendente, com base em previsões de procura para as horas subsequentes. No
entanto, essa abordagem é ineficaz face às flutuações das renováveis, levando à necessidade de ajustes
mais dinâmicos e precisos.
Este trabalho propõe um estudo de parâmetros fórmula do TSO português para a previsão de reserva
necessária (ρ), onde, usando os dados históricos horários no período de 2008 a 2023, é calculado o ρ que
apresente menor erro na previsão, atingindo erros inferiores a 5%.
O presente trabalho propõe também um modelo machine learning para calcular dinamicamente as
reservas de potência secundária, utilizando dados operacionais abertos do TSO espanhol. O modelo foi
treinado com dados no período de 2014 a 2023, e validado com dados de referência de 2024. A metodologia proposta demonstra uma melhoria significativa na utilização das reservas de potência secundária, com
um aumento de aproximadamente 15% na eficiência das reservas ascendentes e cerca de 12% nas reservas descendentes. Este avanço contribui para uma gestão mais eficiente e equilibrada do sistema elétrico,
especialmente em cenários com elevada penetração de vRES.
The growing penetration of variable Renewable Energy Systems (vRES) into the electricity system, such as solar or wind, is significantly transforming the electricity markets due to their intermittent and unpredictable nature. This makes forecasting energy production and consumption more challenging, especially as markets close between 1 and 37 hours before the actual delivery of energy, which can lead to discrepancies between contracted and required energy. Maintaining the balance between supply and demand in real time is vital for the security and stability of the network, a task that falls mainly to the Transmission System Operators (TSO). TSOs use power reserve markets, where they symmetrically purchase upward and downward secondary power based on demand forecasts for subsequent hours. However, this approach is ineffective in the face of renewable fluctuations, leading to the need for more dynamic and precise adjustments. This work proposes a study of the parameters of the Portuguese TSO formula for forecasting the reserve requirement (ρ), where, using historical hourly data for the period 2008 to 2023, the ρ that presents the smallest error in the forecast is calculated, reaching errors of less than 5%. This work also proposes a machine learning model to dynamically calculate secondary power reserves using open operational data from the Spanish TSO. The model was trained with data from 2014 to 2023 and validated with reference data from 2024. The proposed methodology demonstrates a significant improvement in the utilisation of secondary power reserves, with an increase of approximately 15% in the efficiency of upward reserves and around 12% in downward reserves. This advance contributes to more efficient and balanced management of the electricity system, especially in scenarios with high vRES penetration
The growing penetration of variable Renewable Energy Systems (vRES) into the electricity system, such as solar or wind, is significantly transforming the electricity markets due to their intermittent and unpredictable nature. This makes forecasting energy production and consumption more challenging, especially as markets close between 1 and 37 hours before the actual delivery of energy, which can lead to discrepancies between contracted and required energy. Maintaining the balance between supply and demand in real time is vital for the security and stability of the network, a task that falls mainly to the Transmission System Operators (TSO). TSOs use power reserve markets, where they symmetrically purchase upward and downward secondary power based on demand forecasts for subsequent hours. However, this approach is ineffective in the face of renewable fluctuations, leading to the need for more dynamic and precise adjustments. This work proposes a study of the parameters of the Portuguese TSO formula for forecasting the reserve requirement (ρ), where, using historical hourly data for the period 2008 to 2023, the ρ that presents the smallest error in the forecast is calculated, reaching errors of less than 5%. This work also proposes a machine learning model to dynamically calculate secondary power reserves using open operational data from the Spanish TSO. The model was trained with data from 2014 to 2023 and validated with reference data from 2024. The proposed methodology demonstrates a significant improvement in the utilisation of secondary power reserves, with an increase of approximately 15% in the efficiency of upward reserves and around 12% in downward reserves. This advance contributes to more efficient and balanced management of the electricity system, especially in scenarios with high vRES penetration
Descrição
Tese de mestrado, Engenharia da Energia e Ambiente , 2024, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
sistemas de reserva mercados de energia redes neuronais previsões Teses de mestrado - 2024