Electricité de Strasbourg features 1000 load curve electric meters. This means that the average load of each customer is calculated with a period of 30 minutes and the resulting curve was recorded for winter 2013. Further information is available such as outside temperature during the same period, habitat type and heating mode, type of customer (individual or professional). The problem is: for marketing purposes on one hand and on the other network planning, it is desired to achieve a non-supervised classification of this population.
The following characteristics should be taken into account:
- shape of the curve
- periodicity
- correlation of load curves of the different days of the week between them
- dependence of the curve to the current and past outside temperature
Objective
Can we define a set of classes which would express the characteristics metric?
Recent Posts
Collaborators
- Danier Wagner (Électricité de Strasbourg)
- Myriam Maumy-Bertrand & Frédéric Bertrand (Cemosis, UNISTRA)