This is like a highly accurate “digital twin” of a building’s energy use. You feed it information about the building and how it’s used, and it predicts how much energy the building will consume over time under different conditions.
Manual or rules-based approaches to estimating building energy use are slow, inaccurate, and can’t easily adapt to changing occupancy, weather, or control strategies. This work aims to provide a dynamic, data-driven model that can forecast and optimize energy consumption at the building level.
Potential moat comes from high-quality longitudinal building data, calibrated models for specific building types, and integration into existing design/BIM and building-management workflows.
Classical-ML (Scikit/XGBoost)
Time-Series DB
High (Custom Models/Infra)
Data availability and quality for each building (sensor coverage, historical logs), plus model retraining and calibration costs across many heterogeneous buildings.
Early Majority
Focus on dynamic, data-driven energy-consumption modeling at the building level, likely capturing transient behavior (e.g., occupancy, weather, controls) more accurately than static benchmark or steady-state models commonly used in design tools.