Modelos de Séries Temporais em Cascata para Previsão da Velocidade do Vento

Wallace P. P. Porto; Ângelo M. O. Sant’Anna; Daniel Barbosa; Nyegirton Barreiros dos Santos Costa

doi:10.20906/sbse.v2i1.3051

Wallace P. P. Porto Departamento de Engenharia Elétrica e da Computação, Universidade Federal da Bahia, BA
Ângelo M. O. Sant’Anna Departamento de Engenharia Mecânica, Universidade Federal da Bahia, BA
Daniel Barbosa Departamento de Engenharia Elétrica e da Computação, Universidade Federal da Bahia, BA
Nyegirton Barreiros dos Santos Costa Instituto Federal do Sertão Pernambucano, PE

DOI: https://doi.org/10.20906/sbse.v2i1.3051

Keywords: Forecasting, SARIMA, ARIMA, Prophet, Time Series, Wind Speed, Performance

Abstract

This paper proposes developing cascate time series models for forecasting of wind speed. It was developed the Seasonal Autoregressive Integrated Moving Average (SARIMA), Prophet, Cascaded Autoregressive Integrated Moving Average (ARIMAc) and Cascaded Prophet . The called cascaded times series models are an approach in which, unlike the others, the forecast was made using data from each time to distinct time series. From predicted data by time series, its was reassembled in the original order. The analysis and comparison of the performances of each model was carried out based on different evaluation criteria to determine the most accurated forecasting. Finally, the Bergmann-Hommel post-hoc test was applied to assess the level of difference between developed times series models.