Artificial aggregation of domestic appliance currents for NILM evaluation
Résumé
Dedicated Nonintrusive Load Monitoring (NILM) datasets are needed to test any new NILM process. This, however, introduces limitation, since datasets can be lacking in terms of load diversity and multi-load signals. The first problem can be solved by acquiring data from other datasets. The second one requires real measurements of said signals. With the intention of solving the second limitation, we present a new method to create artificial multi-load signals using real individual load signals. The method is able to place the signals at any position relative to one another. Its performance is tested by comparing the resulting artificial signal with the real signal. The results shows strong similarities between the real and artificial signals. An application in ON-OFF transition counts is introduced showing the usefulness of the proposed artificial aggregation approach for any deep evaluation of event detectors in NILM.
Origine | Fichiers produits par l'(les) auteur(s) |
---|