Structural-behavioral determinants of residential energy use: Summer electricity use in Davis

Engineering models of the physical processes of energy use in individual houses are quite complex. We investigate simple statistical models of summer electricity use and compare them to engineering models. Our data include interviews, energy audits, and utility billings for a random sample of residences in Davis, California. We predict summer kWh using appliance and cooling-load models. The appliance model is based on manufacturers' or published data on average annual kWh used by major appliances; refinements for appliance location, seasonality and frequency of use have mixed success. The cooling-load model includes the major variables used in the DOE 2.1A simulation; coefficients estimated by a multiple regression model closely resemble interpolation parameters derived from DOE 2.1A. The appliance and cooling-load models explain over 50% of the variation in summer kWh in single-family detached houses. Using the appliance model and only two variables from the cooling-load model, house area and self-reported frequency of air-conditioner use, we explain nearly 60% of summer kWh in houses. The simple interview question on frequency of air-conditioner use captures most of the effects of structural features such as insulation and glazing. Finally, the appliance and cooling-load models are applied successfully to other house types (common-wall houses and apartments).