A Multi-Scale Framework for Nonintrusive Load Identification

Nonintrusive load monitoring, i.e., the process of identifying individual load information from aggregate electrical measurements, is useful for a variety of smart grid applications including energy scorekeeping, condition monitoring, and activity tracking. Numerous load disaggregation algorithms have been used for nonintrusive monitoring. Many of these perform well only on certain datasets or load types, because transient electrical events can occur on vastly different time-scales and operating schedules with significantly different regularities. This paper presents a nonintrusive load monitoring framework that allows multiple algorithms to be used across multiple time-scales, with their outputs combined to enhance load recognition. Results are demonstrated with power system data from a United States Coast Guard Cutter (USCGC), demonstrating the utility of the framework for developing applications for condition-based maintenance, among other applications.