Title:Incorporating Coincidental Water Data into Non-intrusive Load Monitoring

Abstract:Non-intrusive load monitoring (NILM) as the process of extracting the usage pattern of appliances from the aggregated power signal is among successful approaches aiding residential energy management. In recent years, high volume datasets on power profiles have become available, which has helped make classification methods employed for the NILM purpose more effective and more accurate. However, the presence of multi-mode appliances and appliances with close power values have remained influential in worsening the computational complexity and diminishing the accuracy of these algorithms. To tackle these challenges, we propose an event-based classification process, in the first phase of which the $K$-nearest neighbors method, as a fast classification technique, is employed to extract power signals of appliances with exclusive non-overlapping power values. Then, two deep learning models, which consider the water consumption of some appliances as a novel signature in the network, are utilized to distinguish between appliances with overlapping power values. In addition to power disaggregation, the proposed process as well extracts the water consumption profiles of specific appliances. To illustrate the proposed process and validate its efficiency, seven appliances of the AMPds are considered, with the numerical classification results showing marked improvement with respect to the existing classification-based NILM techniques.