Thermometer Data Forecast for a Buttress Block of the Itaipu Dam

Resumo

The article presents the temperature forecast of thermometers installed in a buttress block of the Itaipu
dam, more specifically the D-38 block. The models used were ARIMA (Autoregressive Integrated Moving
Averages) and Holt-Winters (simple exponential smoothing), additive and multiplicative. Temperature data from
three thermometers were used: TI-D-001, TS-D-003 and TS-D-004. The analyzes were carried out with the help
of free software, R. The series with quarterly periodicity from 2008 to 2017 were considered for the modeling,
with forecasts for the quarters of 2018, which were compared with the real data. The modeling that presented the
best result, for the TI-D-001 and TS-D-004 instruments, was the additive Holt-Winters whose information criterion
AICc obtained the lowest value and, in addition, presented the lowest MAPE in relation to the ARIMA model ,
which makes it the most suitable model. As for the TS-D-003 instrument, the model that best fitted, according to
the AICc criterion, was the multiplicative Holt-Winters, but for the 2018 quarters, it presents MAPE superior to
those of the ARIMA model.