We considered four different features of the hourly electricity
consumption data: daily average, minimum, maximum, and
maximum-minus-minimum (also called “range”). Daily minimum,
maximum, and average values for each user were calculated for
each day of the experiment (238 values for each user). Then, these
238 values for each user were averaged over the summer and
winter (partial) periods, resulting in two daily minimum, two daily
maximum, two daily average, and two daily maximum-minus minimum
consumption values for each user (eight values in total
for each user). Then separate regression models were fitted to each
load feature for each season (eight regression models). Each
regression model consists of 952 response variables (one value for
each user).