A Novel Modeling Approach to Assess the Electricity Consumption of LEED-Certified Research Buildings Using Big Data Predictive Methods

Given the growing interest in improving the energy efficiency of buildings, researchers have generated many predictive models of energy consumption. Most studies use physical and statistical tools, and regression models were shown to be the most practical and accurate. This paper compares the electricity consumption performance of leadership in energy and environmental design (LEED) certified buildings to similar non-LEED buildings in climate zone 2B. To test the hypothesis that electricity consumption is similar for these two types of facilities, the research method is based on developing a performance model for non-LEED buildings, and investigating whether this same model predicts the performance of LEED certified buildings. Electricity data was collected from the buildings, in addition to data for multiple variables including weather, time, and several building factors. The data is used to generate multiple regression models that predict electricity consumption with an accuracy of 90% or more. The results show that the differences between LEED and non-LEED residuals are not as large as anticipated. After discussing the possible reasons behind the observed results, the authors commend the U.S. Green Building Council (USGBC) in their latest move toward considering the actual performance of the building during the occupation phase, as opposed to just the intended performance as set during the design and construction stages, in the newest LEED rating system.