On the Impact of Temporal Data Resolution on the Accuracy of Non-Intrusive Load Monitoring

Many approaches to perform Non-Intrusive Load Monitoring, i.e., to disaggregate electrical load curves collected at a single measurement point, have been presented in literature. The largely different characteristics of the datasets used to evaluate newly proposed disaggregation algorithms, however, complicate an objective and comparable assessment of their capabilities. Different temporal resolutions of the input data (i.e., different sampling rates) are a major impediment to the comparative evaluation of load disaggregation methods in particular. We hence investigate the impact of the temporal data resolution on the disaggregation results of three state-of-the-art algorithms in this work. Our study not only confirms that temporal resolution has an impact on load disaggregation accuracy, but also highlights that a favourable low-frequency sampling rate exists for the appliances under consideration, and generally falls within the range from 1 Hz to 1/30 Hz.