A new approach for event detection and feature extraction for NILM

Nabil Amirach (1) , Bernard Xerri (1) , Bruno Borloz (1) , Claude Jauffret (1)

1 IM2NP - Institut des Matériaux, de Microélectronique et des Nanosciences de Provence

—In this paper is proposed a method, working on the overall current and voltage measurements, in order to detect events, estimate the features of the electric loads underlying these events, and eventually discriminate the loads. This method is a non-intrusive approach: it extracts information from the electric consumption measured only at the electrical switchboard. The chosen approach works in the frequency domain for two reasons: select the frequencies of interest and the reduction of the influence of possible disturbances or some nonlinearities. The searched features for each event (transition) are: transition instant, duration of the transition, admittance modulus and phase of each frequency of interest. To reach the objectives, a modelization of the measurements is proposed, which takes into account the transitions. Index Terms— Estimation; event detection; feature extraction; least squares method; non-intrusive load monitoring (NILM).

Mots clés: – Non-intrusive load monitoring (NILM); event detection; feature extraction; estimation

Domaines: Traitement du signal et de l'image [eess.SP]

Nabil Amirach, Bernard Xerri, Bruno Borloz, Claude Jauffret. A new approach for event detection and feature extraction for NILM. 2014 21st IEEE International Conference on Electronics, Circuits and Systems (ICECS), Dec 2014, Marseille, France. ⟨10.1109/ICECS.2014.7049978⟩. ⟨hal-01823740⟩