Improving current forecast by Leveraging Measured Data and numerical models via LiESNs

Felipe M. Moreno, Marcel R. de Barros, Artur Jordão, Marlon S. Mathias, Marcelo Dottori, Anna Helena Reali Costa, Edson S. Gomi, Fábio G. Cozman, Eduardo Aoun Tannuri

Abstract: Highlights•In this work, currents in a port channel access are forecasted by Echo State Networks.•Irregular time series from sensors and numerical model forecasts are used as input.•Three architectures are evaluated with different combinations of those inputs.•Those architectures are evaluated in scenarios that take into account input blackout.•Inclusion of numerical model data improves robustness in case of input blackout.