Non-Intrusive Load Monitoring: an Architecture and its evaluation for Power Electronics loads

Abstract

NILM (Non-Intrusive Load Monitoring) may well become a widespread solution for diagnostic of Electrical Energy consumption available to every end user. Such a diagnostic may identify waste and improper use; it is also an important tool for energy management both by the residential users and by commercial/industrial users. An architecture for a NILM solution is proposed and evaluated. A comparison is performed among common NILM event detection algorithms and the algorithms proposed in this work. Of particular interest in this study is the detection and classification of power electronics loads, as they impose specific challenges in their detection and correct disaggregation (classification). Our proposed algorithm achieved 100% detection of on/off events for the loads in the COOLL dataset.