An In Depth Study into Using EMI Signatures for Appliance Identification

TL;DR

This paper investigates the use of electromagnetic interference (EMI) signatures for non-intrusive appliance identification, providing empirical evaluations, simulation methods, and insights into factors affecting EMI-based load disaggregation.

Contribution

It offers a comprehensive empirical analysis of EMI signatures across appliances and sensing configurations, introduces a simulation approach, and releases a new EMI dataset for research.

Findings

EMI signatures vary with external parameters like noise and impedance.

Simulation models can replicate real-world EMI behavior effectively.

Empirical data supports feasibility of EMI for appliance identification.

Abstract

Energy conservation is a key factor towards long term energy sustainability. Real-time end user energy feedback, using disaggregated electric load composition, can play a pivotal role in motivating consumers towards energy conservation. Recent works have explored using high frequency conducted electromagnetic interference (EMI) on power lines as a single point sensing parameter for monitoring common home appliances. However, key questions regarding the reliability and feasibility of using EMI signatures for non-intrusive load monitoring over multiple appliances across different sensing paradigms remain unanswered. This work presents some of the key challenges towards using EMI as a unique and time invariant feature for load disaggregation. In-depth empirical evaluations of a large number of appliances in different sensing configurations are carried out, in both laboratory and real world settings. Insights into the effects of external parameters such as line impedance, background noise and appliance coupling on the EMI behavior of an appliance are realized through simulations and measurements. A generic approach for simulating the EMI behavior of an appliance that can then be used to do a detailed analysis of real world phenomenology is presented. The simulation approach is validated with EMI data from a router. Our EMI dataset - High Frequency EMI Dataset (HFED) is also released.