Appliance signature identification solution using K-means clustering

Sustainability, energy conservation and demand response have become an inevitable concern around the world. In the light of electricity companies' demand about what electric appliances that the end-users are switching on, appliance signature is suggested to increase the performance of demand response. The main idea behind appliance signature is that it utilizes the characteristic that same types of electric appliances should have similar features like current, power and harmonic distortion. Utility can get not only the energy profile of households with current metering system but also acquires evidence in energy management according to the energy usage pattern. In this paper, K-means clustering is used for the classification of eight types of common household electric appliances which is an appliance signature identification solution for appliance signature. A digital Butterworth filter has been firstly introduced to remove noisy data before data analyzing by K-means clustering. The performance is evaluated by 10-fold cross validation. Three indexes, CH index, DB index and SH index have been calculated to determine the optimal number of clusters used in K-means clustering. These indexes achieve accuracy of 55.5%, 42.1% and 67.7% respectively.