Forecast modeling a time series of water reservoir levels using exponential smoothing method

Abstract

Exponential smoothing is a time series forecasting method that presents the forecast based on trend and seasonality components. In this work, we study the behavior of two time series that describe the level of the water reservoirs of the Descoberto and Santa Maria dams. We trained the fifteen models present in the Pregels taxonomy, the criterion for choosing the model consists of the model with the lowest Akaike information criterion. The results indicate that the exponential smoothing model with damped additive trend and additive seasonality best describes both time series.