skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Hybrid Model-Based and Data-Driven Fault Detection and Diagnostics for Commercial Buildings

Abstract

Commercial buildings often experience faults that produce undesirable behavior in building systems. Building faults waste energy, decrease occupants' comfort, and increase operating costs. Automated fault detection and diagnosis (FDD) tools for buildings help building owners discover and identify the root causes of faults in building systems, equipment, and controls. Proper implementation of FDD has the potential to simultaneously improve comfort, reduce energy use, and narrow the gap between actual and optimal building performance. However, conventional rule-based FDD requires expensive instrumentation and valuable engineering labor, which limit deployment opportunities. This paper presents a hybrid, automated FDD approach that combines building energy models and statistical learning tools to detect and diagnose faults noninvasively, using minimal sensors, with little customization. We compare and contrast the performance of several hybrid FDD algorithms for a small security building. Our results indicate that the algorithms can detect and diagnose several common faults, but more work is required to reduce false positive rates and improve diagnosis accuracy.

Authors:
; ; ; ; ; ;
Publication Date:
Research Org.:
National Renewable Energy Lab. (NREL), Golden, CO (United States)
Sponsoring Org.:
USDOE Office of Energy Efficiency and Renewable Energy (EERE), Building Technologies Office (EE-5B)
OSTI Identifier:
1324382
Report Number(s):
NREL/CP-5500-67077
DOE Contract Number:
AC36-08GO28308
Resource Type:
Conference
Resource Relation:
Conference: Presented at the 2016 ACEEE Summer Study on Energy Efficiency in Buildings, 21-26 August 2016, Pacific Grove, California
Country of Publication:
United States
Language:
English
Subject:
32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATION; commercial buildings; fault detection and diagnosis; FDD; building energy models

Citation Formats

Frank, Stephen, Heaney, Michael, Jin, Xin, Robertson, Joseph, Cheung, Howard, Elmore, Ryan, and Henze, Gregor. Hybrid Model-Based and Data-Driven Fault Detection and Diagnostics for Commercial Buildings. United States: N. p., 2016. Web.
Frank, Stephen, Heaney, Michael, Jin, Xin, Robertson, Joseph, Cheung, Howard, Elmore, Ryan, & Henze, Gregor. Hybrid Model-Based and Data-Driven Fault Detection and Diagnostics for Commercial Buildings. United States.
Frank, Stephen, Heaney, Michael, Jin, Xin, Robertson, Joseph, Cheung, Howard, Elmore, Ryan, and Henze, Gregor. 2016. "Hybrid Model-Based and Data-Driven Fault Detection and Diagnostics for Commercial Buildings". United States. doi:.
@article{osti_1324382,
title = {Hybrid Model-Based and Data-Driven Fault Detection and Diagnostics for Commercial Buildings},
author = {Frank, Stephen and Heaney, Michael and Jin, Xin and Robertson, Joseph and Cheung, Howard and Elmore, Ryan and Henze, Gregor},
abstractNote = {Commercial buildings often experience faults that produce undesirable behavior in building systems. Building faults waste energy, decrease occupants' comfort, and increase operating costs. Automated fault detection and diagnosis (FDD) tools for buildings help building owners discover and identify the root causes of faults in building systems, equipment, and controls. Proper implementation of FDD has the potential to simultaneously improve comfort, reduce energy use, and narrow the gap between actual and optimal building performance. However, conventional rule-based FDD requires expensive instrumentation and valuable engineering labor, which limit deployment opportunities. This paper presents a hybrid, automated FDD approach that combines building energy models and statistical learning tools to detect and diagnose faults noninvasively, using minimal sensors, with little customization. We compare and contrast the performance of several hybrid FDD algorithms for a small security building. Our results indicate that the algorithms can detect and diagnose several common faults, but more work is required to reduce false positive rates and improve diagnosis accuracy.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = 2016,
month = 8
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share:
  • Commercial buildings often experience faults that produce undesirable behavior in building systems. Building faults waste energy, decrease occupants' comfort, and increase operating costs. Automated fault detection and diagnosis (FDD) tools for buildings help building owners discover and identify the root causes of faults in building systems, equipment, and controls. Proper implementation of FDD has the potential to simultaneously improve comfort, reduce energy use, and narrow the gap between actual and optimal building performance. However, conventional rule-based FDD requires expensive instrumentation and valuable engineering labor, which limit deployment opportunities. This paper presents a hybrid, automated FDD approach that combines building energymore » models and statistical learning tools to detect and diagnose faults noninvasively, using minimal sensors, with little customization. We compare and contrast the performance of several hybrid FDD algorithms for a small security building. Our results indicate that the algorithms can detect and diagnose several common faults, but more work is required to reduce false positive rates and improve diagnosis accuracy.« less
  • Fault Detection and Diagnostics (FDD) is a technology that has a great potential for improving performance and reducing energy consumed in commercial buildings, and is rapidly becoming feasible for the buildings sector. Scientists have developed algorithms for FDD, and are making plans for field-testing and demonstration of these methods in real buildings. These efforts will provide a sound technical basis for FDD product offerings. FDD has the potential to dramatically improve the quality of operation of buildings. However, progress on technical issues is only one step towards implementing FDD in the market. FDD cannot be expected to have a majormore » impact on buildings unless market issues are addressed. Many questions will have to be answered regarding the users of FDD systems, the usability of the product, the market for FDD, and the nature of possible FDD offerings. It is crucial to consider marketing issues in parallel with the more technical issues. Constraints and opportunities that will be faced in marketing the products must be recognized early in technology development, and addressed and integrated into designs to ensure an appropriate system design. This paper identified a number of key questions that will arise in addressing marketability issues. These questions will have to be answered individually by technology developers and entities intending to market FDD. This paper presents some of the considerations that must go into the answering the questions, and provides a framework for analyzing the market requirements.« less
  • This paper presents and compares model-based and data-driven fault detection approaches for coal mill systems. The first approach detects faults with an optimal unknown input observer developed from a simplified energy balance model. Due to the time-consuming effort in developing a first principles model with motor power as the controlled variable, data-driven methods for fault detection are also investigated. Regression models that represent normal operating conditions (NOCs) are developed with both static and dynamic principal component analysis and partial least squares methods. The residual between process measurement and the NOC model prediction is used for fault detection. A hybrid approach,more » where a data-driven model is employed to derive an optimal unknown input observer, is also implemented. The three methods are evaluated with case studies on coal mill data, which includes a fault caused by a blocked inlet pipe. All three approaches detect the fault as it emerges. The optimal unknown input observer approach is most robust, in that, it has no false positives. On the other hand, the data-driven approaches are more straightforward to implement, since they just require the selection of appropriate confidence limit to avoid false detection. The proposed hybrid approach is promising for systems where a first principles model is cumbersome to obtain.« less
  • Commercial buildings consume significant amount of energy. Facility managers are increasingly grappling with the problem of reducing their buildings’ peak power, overall energy consumption and energy bills. In this paper, we first develop an optimization framework – based on a gray box model for zone thermal dynamics – to determine a pre-cooling strategy that simultaneously shifts the peak power to low energy tariff regimes, and reduces both the peak power and overall energy consumption by exploiting the flexibility in a building’s thermal comfort range. We then evaluate the efficacy of the pre-cooling optimization framework by applying it to building managementmore » system data, spanning several days, obtained from a large commercial building located in a tropical region of the world. The results from simulations show that optimal pre-cooling reduces peak power by over 50%, energy consumption by up to 30% and energy bills by up to 37%. Next, to enable ease of use of our framework, we also propose a shortest path based heuristic algorithmfor solving the optimization problemand show that it has comparable erformance with the optimal solution. Finally, we describe an application of the proposed optimization framework for developing countries to reduce the dependency on expensive fossil fuels, which are often used as a source for energy backup.We conclude by highlighting our real world deployment of the optimal pre-cooling framework via a software service on the cloud platform of a major provider. Our pre-cooling methodology, based on the gray box optimization framework, incurs no capital expense and relies on data readily available from a building management system, thus enabling facility managers to take informed decisions for improving the energy and cost footprints of their buildings« less
  • There are two indicated torque input observers presented in paper. The input estimation problem is transformed into the control tracking problem. The sliding mode control with integrator and PI with feedforward control are used for tracking control blocks. The indicated torque estimations are applied to detect engine firing faults. Cylinder pressure centroid and cylinder pressure difference approaches for engine diagnostics are also discussed in this paper. Simulation results for a six-cylinder engine with abnormal combustion are plotted to show engine combustion problems.