Unsupervised Learning Algorithm using multiple Electrical Low and High Frequency Features for the task of Load Disaggregation

Device specific power consumption information leads to a high potential for energy savings. Smart meters are currently deployed in several countries, but they are only able to track the overall consumption in domestic and commercial buildings. One promising option to gain device specific information is called Nonintrusive Load Monitoring (NILM), which can be of great use in combination with smart metering. In NILM, device specific information is achieved by disaggregating the overall load profile from a single-point measurement using device fingerprints and machine learning techniques. In this paper we focus on unsupervised learning methods to minimize the learning phase of device fingerprints. To increase the algorithm accuracy a range of several electrical features are taken into account. This research is part of a public funded project, which aims to increase the energy efficiency in industrial applications. Keywords—Nonintrusive Load Monitoring (NILM); load disaggregation; unsupervised learning; machine learning; electrical disaggregation features; smart metering; energy efficiency; energy management systems