A scalable and practical method for disaggregating heating and cooling electrical usage using smart thermostat and smart metre data

Abstract

We present a scalable and practical method for disaggregating electrical usage for heat pump heating and cooling (HC) that uses low-resolution data from existing smart energy metres and smart thermostats. The disaggregation model is based on a Bayesian approach to account for the skewed characteristics of HC and non-HC energy consumption and adopts sequential Bayesian update to enable reliable predictions without long-term data. The modelling approach is demonstrated using disaggregated electricity consumption and thermostat operation signal data in two multi-family residential communities located in two different cities in Indiana, U.S. The results show that the model successfully disaggregated HC electricity consumption for various housing units by using 15-minute interval data with less than 12% error for a weekly time interval. Finally, seasonal parameters of the model were updated when a new HC operation signal was observed resulting in good predictions for different seasons.

KEYWORDS:

• Energy disaggregation
• residential heating and cooling energy
• nonintrusive load monitoring
• NILM
• smart metre
• smart thermostat

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was funded by the National Science Foundation [grant number 1737591] and supported by the Big Ideas Challenge programme at Purdue University.