Rules induction-supported random forest for the non-intrusive electrical appliances identification

Piotr Bilski, Wieslaw Winiecki

Abstract

The paper presents the application of computational intelligence methods to the nonintrusive identification of electrical appliances in the household. The difficulty of the problem calls for implementation more than one algorithm. The presented two-staged methodology monitors the aggregated current level and includes the detection of the change in the state of appliances’ configuration with the subsequent identification of the device responsible for it. The latter uses the random forest as the main classifier and the rules induction algorithm as the supporting method to increase the appliance identification accuracy. The tests conducted on the configuration of two-state appliances show the supremacy of such a combination of classifiers.

Download

IMEKO-TC10-2016-045.pdf

DOI

-

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings