Sensitivity analysis of US housing stock building characteristics using ResStock

Traditional building energy model calibrations require time-intensive parameter adjustments for individual buildings, making it impractical for large building portfolios. This paper presents a novel framework combining ResStock’s statistical sampling with random forest sensitivity analysis to identify key calibration parameters across different housing categories.

The method analyzes 1.2 million simulations across 18 housing category bins, classified by combinations of climate zone, floor area, and vintage.Results from a case study of small, pre-1990 homes in hot climates revealed that HVAC heating and cooling efficiency, heating setpoint, water heater efficiency, duct leakage, and infiltration dominate total energy consumption (81% combined influence).

This framework allows practitioners to streamline large-scale building calibration by prioritizing the most influential parameters for each housing category.