Non-Intrusive Load Monitoring Using Current Harmonic Vectors and Adaptive Feature Selection

The non-intrusive load monitoring method presented in this paper uses changes in current harmonic vectors to identify the operational state of appliances. The algorithm based on this feature has low complexity, but it may suffer from an information loss caused by a random fluctuation of the current harmonic vectors. In order to deal with this problem, we propose the algorithm which includes a stage which identifies and select a subset of relevant features in the set of available appliance features. The proposed load disaggregation algorithm is demonstrated through experiments on a representative set of household appliances.