Previsão probabilística de vento via modelos dinâmicos lineares em um contexto de produção de hidrogênio

• Ano de defesa: 2024
• Autor(a) principal: Leal, Jairon Isaias
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Energia eólica; Previsão dos ventos; Modelos lineares (Estatística); Hidrogênio como combustível; Wind energy; Probabilistic wind speed forecasting; Linear models (Statistics); Hydrogen as fuel; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA
• Link de acesso: http://repositorio.ufc.br/handle/riufc/78828

The consolidation of renewable hydrogen as one of the mainstays of the energy transition depends on overcoming some operational challenges, such as the predictability of the primary energy source and the impacts caused by intermittent operation. Building on this premise, the present research aims to evaluate the application of Dynamic Linear Models for the probabilistic forecasting of wind speed in the context of hydrogen production. To this end, wind measurements of wind power plants located in the states of Bahia (BA), Ceará (CE) and Rio Grande do Sul (RS) were used as input for a rolling forecasting procedure. The 329 daily forecasts resulting from this procedure were used to estimate hydrogen production from three conversion methods differentiated from each other by calibrating the efficiency of the electrolytic process. The results show that CE obtained the best results regarding point adjustment and probability of coverage but BA had relatively more accurate prediction intervals. The normalized Root Mean Squared Error (nRMSE) median values for CE, RS, and BA are 0.1501, 0.2855, and 0.3272. The Prediction Interval Coverage Probability (PICP) median values for CE, BA, and RS are 97.92%, 79.17%, and 70.83%. The Prediction Interval Normalized Average Width (PINAW) median values are 65.66%, 69.15%, and 73.94% for BA, CE, and RS. Among the electricity-hydrogen conversion methods, C2, which is theoretically calibrated, resulted in smaller differences between the observed and predicted values over the monthly horizon. From a daily perspective, BA’s sample contained the highest daily amount of critical periods due to fluctuations greater than ±50%.