II. LOAD SERIES ANALYSIS
A previous load series analysis allows us to infer many
aspects of its behavior, such as periodicity and trends. This
procedure is fundamental to the Neural Network design.
Figure 1 illustrates a. piece of the load series (hourly load
data from Jan/01 to Jan/14 of 1993) and Figure 2 shows its
auto correlation function.
As we can see, the load series has a very regular
behavior. The auto correlation function of the load series
shows peaks at lags multiples of 24, what indicates that the
load series has a high periodic component in the period of
24 hours. It can also be seen that the actual load value is
highly correlated with the last load values, as denoted by the
high values of the auto correlation function in lags 1 and 2!.