Improving A Probabilistic Approach to Assessing Building Performance

Buildings and their associated electrical and mechanical systems are responsible for 35 percent of U.S. carbon emissions, and it is estimated that up to 30 percent of the energy used in commercial buildings is wasted due to inefficient operation and poorly performing control systems. Improving building operation typically involves an expert manually sifting through data from the control system to identify issues. We have previously proposed a novel standardized “control score” concept to simplify control performance assessment by quantifying three aspects of system control in terms of probabilities. This work focuses on improvements to these probability measures and presents an improved method for assessing control stability using a combination of the Hurst exponent and measures of the variance and periodicity in time series data. We also present a data-driven approach for assessing the supervisory aspect of control performance based on a multivariate analysis in a feature space determined via clustering. Results will be shown from application of these methods to over 70 buildings across three organizations.