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

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-Loeve 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.