CaltechAUTHORS
  A Caltech Library Service

Unsupervised Non-Intrusive Energy Disaggregation for Commercial Buildings

Hossain, Mohammad A. and Khalilnejad, Arash and Swanson, Rachel A. and Mousseau, Jack and Pickering, Ethan M. and French, Roger H. and Abramson, Alexis R. (2017) Unsupervised Non-Intrusive Energy Disaggregation for Commercial Buildings. In: 2017 ASHRAE Annual Conference. American Society of Heating, Refrigerating and Air Conditioning Engineers (ASHRAE) , Atlanta, GA. http://resolver.caltech.edu/CaltechAUTHORS:20180314-125829362

Full text is not posted in this repository.

Use this Persistent URL to link to this item: http://resolver.caltech.edu/CaltechAUTHORS:20180314-125829362

Abstract

Commercial buildings alone are responsible for 36% of the total United States electricity consumption, and 30% of this electricity consumption is wasted on average. One of the greatest challenges in improving building energy efficiency lies in the ability to do simple and non-intrusive load disaggregation. In this paper, we propose a novel unsupervised, non-intrusive building energy disaggregation technique using only 15-minute interval whole-building energy consumption and weather data. The proposed disaggregation technique consists of an analysis loop with three steps. Step 1 analyzes the change in energy consumption in 15-minute intervals for every 3 degrees C temperature range to identify statistically significant temperature dependent and independent the equipment turn on-off states. In step 2, a probabilistic approach is developed using a Factorial Hidden Markov Model (FHMM) that combines equipment state transition matrices from the equipment turn on-off states of step 1 and an emission matrix representing all possible combinations of equipment energy consumption in each state. In step 3, FHMM identifies the best possible combinations of the equipment state and consumption for every point in time using an Expectation Maximization (EM) algorithm. Ultimately, specific equipment such as HVAC, plug load, and lighting, can be identified virtually without setting foot in the building. The proposed disaggregation technique is validated by comparing real-world, sub-metered building equipment energy consumption data to the whole building data. The results show that the method achieves at least 80% accuracy. Using the disaggregation technique to identify the breakdown of HVAC, lighting, and plug load, we compare the results from building population studies to identify potential areas of energy waste and generate an automated virtual energy audit report. The report provides personalized building recommendations such as improving HVAC operations, lighting, and equipment scheduling, including a measure of potential return-on-investment.


Item Type:Book Section
Additional Information:© 2017 ASHRAE.
Funders:
Funding AgencyGrant Number
Department of Energy (DOE)DE-AR0000668
Record Number:CaltechAUTHORS:20180314-125829362
Persistent URL:http://resolver.caltech.edu/CaltechAUTHORS:20180314-125829362
Usage Policy:No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:85309
Collection:CaltechAUTHORS
Deposited By: Tony Diaz
Deposited On:26 Mar 2018 22:52
Last Modified:26 Mar 2018 22:52

Repository Staff Only: item control page