A Non-Intrusive Appliance Recognition System

• Main Authors: Bugnot, Reinelle Jan C, Macabebe, Erees Queen B
• Format: text
• Published: Archīum Ateneo 2020
• Subjects: Load Monitoring; Machine Learning; Energy Management; Appliance Recognition; Electrical and Computer Engineering
• Online Access: https://archium.ateneo.edu/ecce-faculty-pubs/15; https://ieeexplore.ieee.org/abstract/document/8980438
• Institution: Ateneo De Manila University

Depleting energy resources and the unstable supply of raw materials call for innovations in the energy industry, such as in energy generation, distribution, and management. Moreover, increasing electricity prices take a toll on consumers that operates within a strict budget. This study in particular focuses on the proper management and utilization of energy from the consumer's perspective. The objective is to develop a non-intrusive appliance recognition system that can identify the appliances that are being used and calculate how much each of these appliances contribute to the total electricity consumption. Installation of the monitoring system, which utilizes a single sensor clamped to the main power line and used to measure the total energy consumption, does not alter the electrical system, thus, non-intrusive. With this system, the homeowner can monitor which appliances are in use and how much energy they consume. Also, this translates to savings for the household when data provided by the system lead to smarter energy-management choices. For this study to be deployable in households, a data-acquisition system to streamline the data-gathering procedure was needed. Also, a machine learning algorithm was trained and implemented to perform the appliance recognition task given input features from the frequency domain of the measured aggregate data from the main power line. Lastly, the system was tested for prediction accuracy and characterized; and then necessary optimizations were implemented.