Disordered charging of scaled electric vehicles (EVs) can cause many power quality problems, so it is necessary to grasp the charging behavior of EVs to cope with their disordered impacts. To this end, this paper proposes a non-intrusive load extraction method applicable to 3min low-frequency power data. Firstly, the load waveforms of small power appliances are removed from the aggregated power signals and the onset moments of each segment are detected; then the remaining power segments are categorized based on the gradient of the counting function and the EV charging power amplitude is calculated; finally, the EV charging loads are extracted from overlapping segments by combining with the charging event onset moments. In this paper, the effectiveness of the algorithm is verified using the Pecan Street Dataport data set, and the recognition accuracy of the proposed algorithm is 94.98% under 3min sampling data.
 

electric vehicles; non-intrusive;load extraction; low frequency data; count function gradient