Estimates of Refrigerator Loads in Public Housing Based on Metered Consumption Data

Abstract

The New York Power Authority (NYPA), the New York City Housing Authority (NYCHA), and the U.S. Departments of Housing and Urban Development (HUD) and Energy (DOE) have joined in a project to replace refrigerators in New York City public housing with new, highly energy-efficient models. This project laid the ground work for the Consortium for Energy Efficiency (CEE) and DOE to enable housing authorities throughout the United States to bulk-purchase energy-efficient appliances. DOE helped develop and plan the program through the ENERGY STAR@ Partnerships program conducted by its Pacific Nofiwest National Laboratory (PNNL). PNNL was subsequently asked to conduct the savings evahations for 1996 and 1997. PNNL designed the metering protocol and occupant survey, supplied and calibrated the metering equipment, and managed and analyzed the data. The 1996 metering study of refrigerator energy usage in New York City public housing (Pratt and Miller 1997) established the need and justification for a regression-model-based approach to an energy savings estimate. The need originated in logistical difficulties associated with sampling the population and pen?orming a stratified analysis. Commonly, refrigerators[a) with high representation in the population were missed in the sampling schedule, leaving significant holes in the sample and difficulties for the stratified anrdysis. The just{jfcation was found in the fact that strata (distinct groups of identical refrigerators) were not statistically distinct in terms of their label ratio (ratio of metered consumption to label rating). This finding suggested a general regression model could be used to represent the consumption of all refrigerators in the population. In 1996 a simple two-coefficient regression model, a function of only the refrigerator label rating, was developed and used to represent the existing population of refrigerators. A key concept used in the 1997 study grew from findings in a small number of apartments metered in 1996 with a detailed protocol. Fifteen-minute time-series data of ambient and compartment temperatures and refrigerator power were analyzed and demonstrated the potential for reducing power records into three components. This motivated the development of an analysis process to divide the metered consumption into baseline load, occupant-associated load, and defrosting load. The baseline load is the consumption that would occur if the refrigerator were on but had no occupant usage load (no door-opening events) and the defrosting mechanism was disabled. The motivation behind this component reduction process was the hope that components could be more effectively modeled than the total. We reasoned that the components would lead to abetter (more general and more significant) understanding of the relationships between consumption, the characteristics of the refrigerator, and its operating environment.