Civil Engineering Reference
In-Depth Information
Fig. 4.30
Results of validation set 1.
a
Temperature model.
b
Relative humidity model.
c
CO
2
concentration model
series methods are modelling, prediction and characterisation. More specifically,
prediction by means of time-series methods requires, first, to identify the pattern
observed in the data, and secondly, to propagate it in time with obtained trends and
integrate it with other data.
To do this, in Pawlowski et al. (
2010
) four different well-known time-series meth-
ods are analysed and compared: discrete Kalman Filter (KF) (Sarkka et al.
2004
),
discrete Kalman Filter with Data Fusion (KFDF) (Olfati-Saber
2005
), Exponentially
Weighted Moving Average (EWMA) (Vadakkoot et al.
2009
) and Double Exponen-
tial Smoothing (DES) (NIST/SEMATECH
2012
). More specifically, the are used to
estimate short-term forecasts of the solar radiation from real data of a meteorological
