GREEN is a novel building energy performance grading system based on machine learning and city benchmarking data. The grading process is detailed below:
GREEN uses XGBoost to model energy performance and K-Means to cluster the model errors into interpretable grades.
The power of GREEN lies mainly on three axes:
- Advanced modeling techniques that capture non-linear relationships in the data and assign grades in an intelligent way.
- Volume of data. GREEN is trained and validated on more than 7,500 residential buildings in New York City.
- Contextualization. GREEN is a market-specific index, meaning the model is specifically trained on and for New York City buildings. This allows to capture energy dynamics that describe a particular building stock.
Note: Although GREEN grades are currently available for New York City, the system is designed to be scalable and reproducible to any city with available energy benchmarking data.
In the Sankey plot below we see how GREEN compares with a grading system based on EnergyStar that was recently adopted in New York City. Notice that there is more than 40% on the grade classification between the two approaches. See GREEN grading method.ipynb here to find out why and how GREEN outperforms EnergyStar, as well as step-by-step implementation details.
More details in the GREEN method, its applications, and policy implications for climate action in cities can be found in this paper.
Please cite as: Papadopoulos, S. and Kontokosta, C.E., 2019. Grading buildings on energy performance using city benchmarking data. Applied Energy, 233, pp.244-253.