Using Weather and Schedule-Based Pattern Matching and Feature-Based Principal Component Analysis for Whole Building Fault Detection—Part I Development of the Method
- 9 November 2021
- journal article
- research article
- Published by ASME International in ASME Journal of Engineering for Sustainable Buildings and Cities
- Vol. 3 (1), 1-23
- https://doi.org/10.1115/1.4052729
Abstract
A whole building fault (WBF) refers to a fault occurring in one component, but may cause impacts on other components or subsystems, or arise significant impacts on energy consumption and thermal comfort. Conventional methods (such as component level rule-based method or physical model-based method) which targeted at component level fault detection cannot be successfully used to detect a WBF because of the fault propagation among the closely coupled equipment or subsystems. Therefore, a novel data-driven method named weather and schedule-based pattern matching (WPM) and feature-based principal component analysis (FPCA) method for WBF detection is developed. Three processes are established in the WPM-FPCA method to address three main issues in WBF detection. First, a feature selection process is used to pre-select data measurements which represent a whole building’s operation performance under a satisfied status, namely, baseline status. Second, a WPM process is used to locate weather and schedule patterns in the historical baseline database, which are similar to that from the current/incoming operation data, and to generate a WPM baseline. Lastly, real-time PCA models are generated for both the WPM baseline data and the current operation data. Statistic thresholds used to differentiate normal and abnormal (faulty) operations are automatically generated in this PCA modeling process. The PCA models and thresholds are used to detect the WBF. This paper is the first of a two-part study. Performance evaluation of the developed method is conducted using data collected from a real campus building and will be described in the second part of this paper.Keywords
Funding Information
- Building Technologies Program (DE-FOA-0001167)
This publication has 47 references indexed in Scilit:
- A GA-based feature selection and parameters optimizationfor support vector machinesExpert Systems with Applications, 2006
- Review Article : Methods for Fault Detection, Diagnostics, and Prognostics for Building Systems—A Review, Part IIHVAC&R Research, 2005
- Review Article : Methods for Fault Detection, Diagnostics, and Prognostics for Building Systems—A Review, Part IHVAC&R Research, 2005
- Detection and diagnosis of AHU sensor faults using principal component analysis methodEnergy Conversion and Management, 2004
- Partial least squares for discriminationJournal of Chemometrics, 2003
- A Critique of the Bayesian Information Criterion for Model SelectionSociological Methods & Research, 1999
- Wrappers for feature subset selectionArtificial Intelligence, 1997
- GA Strategy for Variable Selection in QSAR Studies: GA-Based PLS Analysis of Calcium Channel AntagonistsJournal of Chemical Information and Computer Sciences, 1997
- Multi-objective genetic algorithm and its applications to flowshop schedulingComputers & Industrial Engineering, 1996
- A note on genetic algorithms for large-scale feature selectionPattern Recognition Letters, 1989