Automated Data Mining Methods for Identifying Energy Efficiency Opportunities Using Whole-Building Electricity Data

Automated detection of schedule- and operation-related energy savings opportunities in commercial buildings can help building owners lower operating expenses while also reducing adverse societal impacts such as global greenhouse gas emissions. We propose automated methods of identifying certain energy-efficiency opportunities (EEOs) in commercial buildings using only whole-building electricity consumption and local climate data. Our two-step approach uses piecewise linear regression and density-based robust regression model residual clustering to detect both schedule- and operation-related electricity consumption faults. This paper discusses results obtained from applying this approach to two all-electric office buildings meant to demonstrate our model's effectiveness in identifying such EEOs. Ways by which the analysis results can be conveniently and succinctly presented to building managers and operators are also suggested.