Deep Learning Based Active Power Disaggregation of Appliances for Personalized Energy Feedback System

Energy consumption feedback with an appliance-level system can reduce by a maximum of 12%. In this study, we proposed data acquisition and training framework for configuring deep learning based on Non-Intrusive Load Monitoring (NILM) personalized energy system. To construct dataset, aggregation active power from four types home appliances (refrigerator, induction, TV, washing machine) was performed approximately three weeks. LSTNet applied to extract recognize the features state each appliance. With accuracy metric more than 90% disaggregation result, applicability verified.