Large Scope Device Recognition by Power Usage for Crownstones

Abstract

Electrical appliance classification has great potential. It has potential uses in analysis of power usage within households, automation of households, and detection of hazards and electrical decay. The field is well researched, but has yet to see successful mass deployment in the modern world. A new device called the Crownstone may be the solution, as it is capable of intrusive load monitoring and being developed to be distributed into many households. In this research, an easily implementable established method for classification of electrical appliances by intrusive load monitoring is tested on a new, more challenging dataset recorded using Crownstones. An analysis is made of the achievable accuracy, as well as the effects of the noise and larger number of classes. It is found that the method continues to perform surprisingly well under these more demanding conditions, especially with the help of simple preprocessing steps.