| dc.description.abstract | The aim of this study is to find out from what point in time and with what amount and type of data you
can detect with a certain amount of certainty a significant decrease of the gas consumption for an individual
household. Data points for the summed gas consumption for the average temperature differences between
indoor and outdoor temperature for each day for annual periods between September and April from 2015 till
2020 were taken. To be able to make the earliest possible detection of a valid decrease of gas consumption,
three consecutive heating periods are needed.
Afterwards, the slopes were compared with the following period slopes to identify an increase or decrease. If
there is a significant change that was determined differently in three different approaches, you can assume that
a possible reason is a newly add insulation of that household. Those household where a significant decrease
has been detected by the different approaches linear regression, Support Vector Regression and Random For-
est, were afterwards filtered out to have a final dataset with houses where an insulation has possibly been
added.
The findings of the study showed that with two linear models, linear regression and support vector regression,
significant decreases in gas consumption can be detected in the data.
These results lead to the assumption that the gas consumption and the average temperature difference per day
alone show a change in gas consumption, but this cannot be attributed to a newly added insulation, as this can
also have many other reasons. | |
