Data-driven building load profiling and energy management

Abstract

Commercial buildings consume a lot of energy and contribute a significant part of greenhouse gas emission. Many energy-saving or green-building initiatives were compromised by equipment and human-related faults under the umbrella of poor facility management. Data-driven building energy management is a cost-effective approach to improve energy efficiency of commercial buildings, and gains more and more popularity worldwide with the deployment of smart metering systems. This paper developed a systematic process of using smart metering data to quantify building daily load profiles (i.e. energy consumption patterns) with a set of statistics, e.g. base load, peak load, rising time and so on. Then prediction models of these building load statistics are constructed from historical training data consisting of energy consumption, environment and holiday information. At last residuals of the prediction models are analyzed to form statistical control charts. As a result anomaly energy consumption could be detected by comparing the predicted statistics and observed ones, which will help building managers to locate problems just in time. The effectiveness of the proposed solution is verified through real-world data analysis and computational studies.