The large-scale introduction of renewable energy resources will cause instability in the power supply. Residential energy management systems will be even more important in the near future. An important function of such systems is visualization of appliance-wise energy consumption; residents will be able to consciously avoid unnecessary consumption behavior. However, visualization requires sensors to measure appliance-wise energy consumption and is generally a costly task. In this paper, an unsupervised method for non-intrusive appliance load monitoring based on a semi-binary non-negative matrix factorization model is proposed. This framework utilizes the total power consumption patterns measured at the circuit breaker panel in a house, and derives disaggregated appliance-wise energy consumption. In the proposed approach, the energy consumption of individual appliances is estimated by considering the appliance-specific variances based on an aggregated energy consumption data set. The authors implement the proposed method and evaluate disaggregation accuracy using real world data sets.
[1]
Paul Van Dooren,et al.
Weighted Nonnegative Matrix Factorization and Face Feature Extraction
,
2007
.
[2]
D. Hunter,et al.
Optimization Transfer Using Surrogate Objective Functions
,
2000
.
[3]
Rafal Zdunek,et al.
Data Clustering with Semi-binary Nonnegative Matrix Factorization
,
2006,
ICAISC.
[4]
H. Sebastian Seung,et al.
Algorithms for Non-negative Matrix Factorization
,
2000,
NIPS.
[5]
G. W. Hart,et al.
Nonintrusive appliance load monitoring
,
1992,
Proc. IEEE.
[6]
Chris H. Q. Ding,et al.
Binary Matrix Factorization with Applications
,
2007,
Seventh IEEE International Conference on Data Mining (ICDM 2007).
[7]
Andrew Y. Ng,et al.
Energy Disaggregation via Discriminative Sparse Coding
,
2010,
NIPS.
[8]
Manish Marwah,et al.
Unsupervised Disaggregation of Low Frequency Power Measurements
,
2011,
SDM.