Reducing energy consumption in commercial buildings, such as offices, is an important step towards lowering greenhouse gas emissions and achieving carbon-neutral goals. This requires determining the individual energy use of all the operational electrical equipment in the building. However, for large commercial spaces where it is impractical and expensive to install power measurement devices for every electrical facility, this can be difficult task. Instead, using a technology called "disaggregation," the aggregated power readings of the entire building are broken down to estimate the power consumption of individual electrical facilities.

In most buildings, only the most demanding electrical facilities ― such as air conditioners― are typically monitored. Other electrical facilities, such as lighting and electrical appliances, are usually left unmonitored and their operational statuses are estimated assuming a single periodic usage pattern. However, this is a largely oversimplified approach as most buildings have different usage patterns for workdays and holidays. This, in turn, reduces the disaggregation accuracy, and while it can be improved, it involves lengthy procedures and increases management cost.

Now, in a study by researchers from Tokyo University of Science, Japan, and Mitsubishi Electric Corporation, Japan, a disaggregation strategy for providing accurate power estimations with a limited number of monitoring points has been proposed. "The method can accurately estimate and specify the energy consumption of each monitored and unmonitored facility by using the status of the monitored facilities," say the authors of the study. This paper was made available online on 01 November 2021 and was published in Volume 255 of the journal Energy and Buildings on 15 January 2022.

How exactly does the strategy accomplish this? The authors explain, "The consumption of the unmonitored facilities is estimated using the linear regression residuals of the monitored facilities and clustered with a daily routine. Next, linear basis function models with different diurnal periods are assigned to each cluster to estimate the daily variation in energy consumption of the unmonitored facilities."

Put simply, the method involves making initial estimations of the unmonitored power consumption based on the total power consumption of the building and the monitored facilities. Based on the estimated power of the unmonitored facilities, these values are clustered. The purpose of the clusters is to reflect possible energy usage scenarios such as weekdays, holidays, and weekends. For each cluster, the daily variations in power consumption are then taken into account and used to estimate the energy consumption patterns for the electrical facilities.

The method was used to disaggregate the power consumption of an eight-story office building consisting of 94 monitored air-conditioners and unmonitored facilities including lightning, office automation equipment, and outlets. The power consumption of the unmonitored facilities was spread between three clusters. Compared to the conventional method that uses a consistent daily routine, the proposed method could apply different daily routines to match the power consumption scenarios of the three clusters. This, in turn, provided more accurate estimations for workdays and holidays than the conventional method.

Notably, the method only requires energy logs collected from the building, allowing accurate power consumption estimations to be made quickly. "The technology will enable commercial building energy-managers to make low-cost as well as highly accurate energy-saving proposals. This could become an indispensable technology for a carbon-neutral society in 2050," the team adds.

The authors are now working to factor in seasonal trends into their estimations to make them even more accurate and reflect real-world usage. We certainly hope their efforts to build a greener society come to fruition soon.