The Use of Canopy Clustering within Non-Intrusive Load Monitoring (NILM)

Daniel Carr and Stephen Gardner


Canopy clustering, NILM, K-means clustering, Data mining


Many environmental issues are affecting the way we look at and think about energy, rising fuel costs and the dwindling supplies of oil are forcing people to look at their energy consumption. By using methods such as Non-Intrusive Loan Monitoring (NILM), the ability to monitor individual loads within a property allows consumers to make informed decisions on how and where energy savings can be made. The domestic environment provides a challenging environment for NILM with a large amount of loads being consumed at random throughout the day, with the combination of loads that are being consumed becoming exponentially large. By conducting NILM on groups of loads instead of individual loads, profiles are more easily created, and groups of loads can more readily be identified. NILM by traditional methods is resource intensive and still in its infancy. By using canopy clustering to initially segregate data into a set of over-lapping canopies of data, the process of NILM becomes faster, without the loss of accuracy in the clustering process. Using k-means clustering within the canopies provides the output clusters, and as k-means is only carried out within its relevant canopy, the distance measurements to all other points outside of this canopy are negated, thus providing speedy, efficient and accurate clustering.

Important Links:

Go Back