Benchmarking Building Energy Performance Using Data Envelopment Analysis with Normalized Metrics—A Residential Case Study

Abstract

Data envelopment analysis (DEA) is emerging as an important benchmarking tool for building energy performance with its unique capability of discriminating scale efficiency and management efficiency. Generally, the impacts of climate factors are desired to be neutralized to obtain a climate-adjusted energy parameter when using DEA for building energy performance benchmarking. Multiple linear regression is often adopted for this neutralization. While very useful, this approach rarely considers the multicollinearity trap referring to the statistical issue that the strong correlations among explanatory variables can lead to non-robust building energy regression models. This paper presents a simple alternative normalizing approach to avoid the multicollinearity deficiency in DEA benchmarking application through neutralizing energy input with degree day and floor area. First, the annual energy consumption of each building is normalized by its floor area and local degree day to acquire the degree day normalized energy use intensity (DEUI). Second, with each building as one decision making unit, DEA model is constructed with the number of occupants and the floor area as DEA outputs and DEUI as input. Finally, DEA is calculated to obtain overall efficiency, scale efficiency and management efficiency. The energy performance of 31 one-story residential buildings is benchmarked using the developed approach based on historical data. The case study reveals that the low energy performance of the targeted buildings is mainly due to the inefficiency of management factors but further verification is desired.