For comparison purposes, an uncalibrated model of the TxAIRE
House #1 was developed using the engineering software BEopt [59] that uses EnergyPlus [60] as
simulation engine. Information for the geometry was obtained from blueprints and other known
parameters were obtained from a visit to the house. All unknown parameters required by the software
were left as default values as given by the software for a new standard construction. The weather
file used for the simulation was a modified file of the USA_TX_Tyler-Pounds.Field.722448_TMY3.epw
available from the EnergyPlus website [60]. The outdoor temperature and global horizontal radiation
recorded onsite were used in the modified weather file. Since the simulations are performed in time
steps of 10 min and the internal algorithm within EnergyPlus is used to handle the weather
variables during the computation, the outdoor temperature and global horizontal radiation data
points do not match with the data in the modified weather file. Therefore, the same data of the two
variables from the software simulations are also used in the statistical models to obtain results
for model comparison. Table 4 shows the quality parameters obtained for the statistical models and
the BEopt software. It can be noted that the coefficient of determination and the RMSE for the
hourly statistical models are better than the ones for the engineering model. However, all models
are fairly similar for the daily data. Since statistical models avoid the burden associated with
the collection of information needed to develop engineering models, the com- parison illustrates
that the statistical models can be a cost- effective approach to forecast energy
consumption in residential
buildings in a reasonable and accurate manner.