Non-Intrusive Load Monitoring Method for Appliance Identification Using Random Forest Algorithm

Abstract

The Non-Intrusive Load Monitoring (NILM) method in energy disaggregation is an effective way to disaggregate overall power consumption and obtain information on electricity usage for each load. The load identification is determined by the signature of each appliance. As the main contribution in this research, implementing the Random Forest algorithm in the application of the NILM method to identify the type of appliance and compare it with the supervised algorithms that are often used in NILM, such as Support Vector Machine, Multi-Layer Perceptron, K-Nearest Neighbors, and Naïve Bayes. The proposed algorithm was tested using data on household appliances collected using a single-phase power metering system with five electrical appliances tested, i.e., fans, lamps, rice cookers, televisions, and telephone chargers. The effectiveness of the proposed algorithm on the tested appliances is also validated using the WHITED public dataset under current and power features. The proposed method identifies appliance types correctly above 90% of the total events in the private and WHITE datasets. The results of a series of experiments show that the proposed algorithm is more optimal than the other algorithms tested.