AI identification of new Hydro-Climate models

This work introduces a new AI identification model, applied to the forecasting of hydro-climate series. Unlike most currently used models, based solely on the study of the historical data, this approach proposes the use of delayed "explaining" variables feeding an unknown nonlinear system. An optimal combination of two different models (neural networks and vector quantization) "explains" the desired variables, yielding the prediction output. The method is applied to the prediction of a Brazilian river inflow, with immediate use into future energy availability forecasts.