Abstract
Non-Intrusive Load Monitoring (NILM) researchers have always assumed the switch continuity principle (SCP), which assumes that only one appliance ever changes state at any given point in time. However, SCP cannot be relied upon 100% of the time, especially when unsupervised NILM is used to guess what appliances might be in a house. This principle breaks under certain conditions based on the data capture choices made: number of occupants, number of appliances, measurement unit, measurement precision, and sampling frequency. This paper identifies and explores the conditions under which SCP can and cannot be assumed. This is done through empirical tests performed on two of the most popular datasets used for NILM.

This publication has 10 references indexed in Scilit: