| Paper ID | MLSP-19.6 |
| Paper Title |
MULTIVARIATE NON-NEGATIVE MATRIX FACTORIZATION WITH APPLICATION TO ENERGY DISAGGREGATION |
| Authors |
Pascal Alexander Schirmer, Iosif Mporas, University of Hertfordhshire, United Kingdom |
| Session | MLSP-19: Non-Negative Matrix Factorization |
| Location | Gather.Town |
| Session Time: | Wednesday, 09 June, 14:00 - 14:45 |
| Presentation Time: | Wednesday, 09 June, 14:00 - 14:45 |
| Presentation |
Poster
|
| Topic |
Machine Learning for Signal Processing: [MLR-MFC] Matrix factorizations/completion |
| IEEE Xplore Open Preview |
Click here to view in IEEE Xplore |
| Virtual Presentation |
Click here to watch in the Virtual Conference |
| Abstract |
Non-Intrusive Load Monitoring aims to extract multiple unknown variables, i.e. device signatures, from a single observation, thus it can be considered as a single-channel source separation problem. Source separation methods and especially Non-negative Matrix Factorization have been utilized to solve the Non-Intrusive Load Monitoring problem. However, due to the restrictions of Non-negative Matrix Factorization being only applicable for utilization of one feature only active power signals have been used as features and multivariate signals, i.e. active and reactive power have not been used. In the proposed architecture the baseline Non-negative Matrix Factorization approach is expanded utilizing multivariate signals. The proposed approach was evaluated on the AMPds2 dataset reporting performance improvements up to 21.7% when being compared with other source separation approaches reported in the literature. |
