Data-Driven Estimation of Design Heating Loads for HVAC Equipment Sizing

Oversized heating and cooling equipment entails unnecessarily high up-front costs, energy costs, pollutant emissions, rates of wear and tear due to equipment cycling, and strain on power grid infrastructure. This paper develops two data-driven methods for estimating heating loads at design conditions to improve equipment sizing. The methods are tested on a new dataset that were gathered as a part of this study. The dataset of 74 detached single-family houses in five United States climate zones includes smart thermostat time-series data, monthly utility bills, weather data, existing equipment specifications, and Manual J design load calculations (the United States industry standard) procured from practicing professionals. One data-driven design load estimation method uses smart thermostat data; the other, utility bills. We find that both methods have strong goodness-of-fit statistics individually and agree reasonably well with each other. On average over the 74 houses, existing heating equipment is 150\% larger (95\% confidence interval: 134--167\%) than the central estimate from the two data-driven methods. As implemented by practicing professionals, Manual J overestimates design heating loads by 116\% (96--135\%) on average. We discuss prospects for implementing the data-driven methods at scale and for incorporating data-driven sizing into industry standards.