Automatic Appliance Classification for Non-Intrusive Load Monitoring

This paper is based on non-intrusive load monitoring (NILM), which uses low-frequency sensor in power circuit. Traditional process must establish a database with features before identifying what the circuit is. If the system wants to add new feature of appliances into database, it must relearn electrical data. Therefore, this paper proposes a method, which can identify appliances status and whether new appliances exist or not. It can also learn feature of appliances automatically at the same time. The proposed method combines statistics with classification techniques to simplify the feature extraction. The consequent is quite valid in the economy, accuracy and feasibility. In addition, if NILM system does not identify successfully, it might contain the unknown appliances. The unknown appliances can thus be identified. The system will be able to expand its appliances amount in the database automatically. Experiment performed with a variety of single or multiple classifications which include the unknown appliances.