Study on the Influence of voltage variations for Non-Intrusive Load Identifications

In a smart grid, electricity energy demands requested from down-stream sectors continuously increase. One way to meet the energy demands is utilizing a Non-Intrusive Load Monitoring (NILM)-style system to monitor and manage residential and commercial electrical appliances in respond to demand response programs. NILM is as an electricity energy audit to energy-saving issues.This paper aims at developing an NILM identification considering the influence of voltage variations. An NILM system normally consists of "Data Acquisition", "Event Detection and Feature Extraction", and "Load Classification." The goal of load classification in NILM is to identify the operation ON/OFF status of individual household appliances. The k-Nearest Neighbors Classifier is used as the load classifier of the NILM. In NILM, measured physical phenomena, voltage signals/power profiles, to load classification for long-term load monitoring vary. The variations affect the identification performance. In this paper, a cross-validation strategy is conducted and used to deal with sampled data with systematic errors from classified loads. Different types of loads are used to verify the influence of the voltage variations to load identifications. As the experimentation reported in this paper shows, a satisfactory recognition rate of 97.60% to the NILM load identification addressed in this paper is achieved. The proposed NILM system is able to classify loads with proper robustness.