Data-quality detection and recovery for building energy management and control systems: Case study on submetering

作者：Fu, Yangyang; Li, Zhengwei; Feng, Fan; Xu, Peng^*

来源：Science and Technology for the Built Environment, 2016, 22(6): 798-809.

DOI：10.1080/23744731.2016.1195658

• FAULT-TOLERANT CONTROL
• DIAGNOSIS

Most modern buildings are equipped with building energy management and control systems. These systems can store tremendous amounts of data on buildings' performance and energy usage. A significant amount of data on buildings' mechanical devices, particularly electricity-consumption data, is now available for analysis. However, the quality of the collected data is questionable. Some data are mislabeled, and others contain gaps and errors. In this article, a methodology based on a correlation coefficient and a wavelet-based support vector machine predictor is proposed to detect and recover the proportional deviation data faults and faults caused by network communication. After testing this methodology with electricity data collected from a large commercial building, it is found that a high accuracy of faulty data alerts and automated data recovery can be achieved. Considering the wide use of building energy management and control system data for performance monitoring, fault detection and diagnostics, and demand responsive control, this method is useful and practical in many engineering situations.

• 出版日期2016
• 单位同济大学