TEMPERATURE FORECAST IN CONCRETE DAM BY HOLT-WINTERS MODELS AND THE FINITE ELEMENT METHOD

Resumo

The article describes a method of forecasting the temperature fields of a buttress-type concrete
block from the Itaipu Hydroelectric Dam located in a region with high thermal gradients, approximately
7 to 40°C. The method involves the application of Holt-Winters exponential smoothing models and a
heat conduction model that follows Fourier's Law. Initially, the temperature series from 2000 to 2016
of two surface-to-block thermometers, whose read periodicity is not well defined, were adjusted by cubic splines and the now monthly series were used as input for Holt- Winters to produce temperature
forecasts as outputs for 2017. Monthly (interpolated) and 2017 (predicted) monthly temperatures were
used as boundary conditions for the transient thermal block model and it was solved by the Finite
Element Method (FEM) in Python language. The predicted temperature fields of the block for 2017 are
thus obtained. The numerical results at a specific point and interior of the structure were compared with
the observed data of an internal thermometer installed in the block by means of the MAPE error, and
this was around 7%. The proposed method has an innovative approach to the thermal analysis of
structures, particularly in concrete dams, regarding thermal field prediction and a first application of the
Python language to this problem.