DC appliance classification and identification using k-Nearest Neighbours technique on features extracted within the 1st second of current waveforms

The commonly used identification techniques for appliances in a household are usually performed on the AC power supply side. However, as more household appliances and gadgets are now being DC powered, it is more accurate to perform the measurement and identification on the DC demand side. In addition, the AC identification method is not applicable for the DC household-grid. This paper discusses the application of a computational intelligence technique, k-Nearest Neighbours, to classify and identify DC appliances in a low voltage DC household through their 1st second of DC demand-side waveforms, sampled at 500Hz. Voltage and current waveforms were collected from an experiment conducted using this technique and it has been observed from the data collected that DC appliances generate unique current waveforms, similar to signatures, during the 1st second of operation. This time window can be spilt further into an inrush current stage and a steady-state stage. Two primary features and three secondary features of the waveforms were extracted and employed as attributes in the kNN technique, which was successfully used to classify and identify three appliances: a Peltier technology fridge, LED lights and a DC motor fan.