Measurement & verification
The M&V Challenge
When electricity is used is as important as how much.
This puts a premium on measurement & verification (M&V) methods that can prove hourly performance of energy efficiency measures by establishing an hourly counter-factual or baseline. That is, what electricity would the building have used in each hour absent the efficiency measure?
Proving EMeister MPC hourly performance is necessary for carbon reduction (e.g. using hourly marginal emission rates) as well as energy/expense savings and financial risk management (e.g. using locational marginal prices). Important going forward, M&V methods need also take into account the now more dynamic seasonal/daily/hourly building occupancy.
QCo uses off-the-shelf machine learning algorithm, XGBoost, to prove EMeister MPC performance.
This baseline example reflects all occupied weekday hours (8am to 6pm) over an entire June-September cooling season.
This application is undergoing peer review and industry testing on additional buildings in anticipation of journal publication and public dissemination for broad industry use prior to the Summer 2023 cooling season.
This application far exceeds the accuracy required by ASHRAE Guideline 14 and IPMVP.
How to read this graph?
By definition, the features add to 100% … so the percentages are a measure of relative importance.
Observations
Simple data required -- hourly electric use, hourly weather, and hourly occupancy.
Note the importance of occupancy! People do not generate much heat, but their activities do. For this building, "Friday" simply reflects that the building did not have hourly occupancy data to distinguish lightly occupied Fridays from other days of the week.
Definitions
Prior weekend CDD: prior Saturday and Sunday CDD, applied only to Monday and Tuesday hours.
Other: less significant features, including overnight weather, solar radiation, and day of week.
