Implementation of a robust real-time non-intrusive load monitoring solution

Shirantha Welikala, Neelanga Thelasingha, Muhammed Akram, Parakrama B. Ekanayake, Roshan I. Godaliyadda, Janaka B. Ekanayake

Research output: Contribution to journalArticlepeer-review


This paper presents the formulation and practical implementation of a spectral decomposition based, Real-Time Non-Intrusive Load Monitoring (RT-NILM) solution. Many of the NILM techniques reported in the literature have been validated on environments with non-varying supply voltages, while relying on multiple measurements taken at high sampling rates. In contrast, the RT-NILM solution proposed in this paper has addressed the issue of supply voltage variability, which is a common practical problem prevalent in many developing countries and is anticipated to emerge globally with the increased penetration of renewable energy sources. Therefore, the proposed RT-NILM algorithm was implemented to maintain high accuracy levels even under severe supply voltage fluctuations. An iterative implementation of the Karhunen-Loève expansion was introduced to improve the spectrum decomposition resolution. Further, a fast deconvolution based technique was introduced for the disaggregation of individual power levels of active appliances in an computationally efficient manner. The proposed solution has been validated on a real voltage varying environment, at a real house, in real-time, using active power and voltage measurements taken at a low sampling rate of 1 Hz.

Original languageEnglish
Pages (from-to)1519-1529
Number of pages11
JournalApplied Energy
StatePublished - Mar 15 2019

ASJC Scopus subject areas

  • Mechanical Engineering
  • General Energy
  • Management, Monitoring, Policy and Law
  • Building and Construction
  • Renewable Energy, Sustainability and the Environment


  • Demand side management
  • Non-intrusive load monitoring
  • Real-time load monitoring
  • Smart grid
  • Subspace techniques
  • Supply voltage variation

Cite this