Non-intrusive appliance recognition

Energy conservation becomes more important nowadays. The use of smart meters and, in the near future, smart appliances, are the key to achieve reduction in energy consumption. This research proposes a non-intrusive appliance monitor and recognition system for implementation on an embedded system. The research focuses on computational efficiency of such a system as embedded systems usually have limited computation power. In this paper, research has been done on the effects on the accuracy of the sample frequency. The algorithm first detects an event in which an appliance is turned either on or off. Subsequently its profile is extracted. A hierarchical support vector machine (HSVM) is used to classify the appliance. The result is a complete algorithm that recognizes individual appliances within a household. Tests on this appliance recognizer show that the proposed algorithm can correctly detect appliances with reasonable accuracy.