Odd Harmonic Distortion Contribution on a Support Vector Machine NILM Approach

In this paper, the harmonic distortion (HD) of the 3 rd , 5 th and 7 th harmonic currents is utilized in order to improve the load identification performance in addition to common electrical features that are usually applied in NILM applications. The evaluation of the proposed feature employment is achieved under a multi-label consideration through a Support Vector Machine (SVM) learning approach. For the needs of this work, a private dataset was developed with measurements extracted from a low-cost smart energy meter under high sampling rate, installed in a household's main feeding panel. The load identification concerns four different load clusters of the examined residence. The results indicate that the introduction of the HD of the first three odd harmonic order currents could enhance the robustness of the formed clusters during the training phase and subsequently improves the efficiency of the load identification scheme.