1011-RP Utility/EMCS Communication Protocol Requirements

Seem reviewed progress on RP-1043 (Fault Detection and Diagnostic Requirements and Evaluation Tools for Chillers), for which Purdue University (Braun, PI) is the contractor. The contractor has performed a detailed literature review for fault detection and diagnosis methods and will extend the literature review to cover chiller models. Purdue could not obtain the chiller originally sought, a 50-Ton screw machine, but is now installing a 90-Ton centrifugal chiller. The PMS (at least a subset) plans to meet in a few months with the contractor, before the Seattle meeting. The PMS requests an hour-long review period at the Seattle meeting.

Norford reviewed progress on RP-1011 (Utility/EMCS Communication Protocol Requirements). This project requires that the contractor identify existing services and communications protocols, identify possible new services, develop models for the exchange of information for identified services, and make suggestions for ASHRAE's involvement in this area. The PMS is reviewing a substantial draft final report, not yet complete, that was prepared by the contractors, PNNL and Hypertek. The contractors plan to have a complete version to the PMS before the Seattle meeting. The scheduled contract completion date is June 30, 1999.

ASHRAE American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.

Programs for Chicago, Seattle, Dallas, Minneapolis and were discussed at the general
TC 4.11 meeting. TC 4.11 reviewed numerous candidates for future programs. We voted on the Seattle program and priorities. See below in Section II for the approved list.

Carol Lomonaco will recommend the program for Dallas to TC 4.11 at the Seattle Annual Meeting-June 1999, and request acceptance and prioritization. A list of suggested program for Minneapolis will also be prepared for the Seattle meeting.

1. Sponsoring Committee: TC 4.6 & Co-Sponsoring Committee: TC 4.11

Fault Detection and Diagnostics – Learning from Building Operations

Symposium CH-99-05, Osman Ahmed, Sunday 1/24/99 @10:15am, Program SEQ #1196

2. Sponsoring Committee: TC 4.11 & Co-Sponsoring Committee:

Customer Experience with Real-Time-Pricing Electric Rates

Seminar 02, Michael Kintner-Meyer, Sunday 1/24/99 @8:00am

3. Sponsoring Committee: TC 4.11 & Co-Sponsoring Committee:

Fault Detection and Diagnostics – Learning from Building Operations

Symposium CH-99-18, Natascha Castro, Wednesday 1/27/99 @8:00am

4. Sponsoring Committee: TC 1.4 & Co-Sponsoring Committee: TC 4.11

How Do We Achieve Multi-Vendor Interoperability In Existing Facilities?

Forum 12, Jim Coogan, Sunday 01/24/99 @2:00pm

5. Sponsoring Committee: TC 1.4 and TC 9.9 & Co-Sponsoring Committee: TC 4.11

Public Session, Jim Gartner, Monday 01/25/99 @3:00pm, Program SEQ #1195

A. FAULT DETECTION AND DIAGNOSTICS

1. Sponsoring Committee: TC 4.11 & Co-Sponsoring Committee: TC 1.4

2. B. APPLICATIONS
   

1. Sponsoring Committee: TC 1.4 & Co-Sponsoring Committee: TC 4.11

C. UTILITY/BUILDING INTERFACE

1. Sponsoring Committee: TC 4.11 & Co-Sponsoring Committee:

III. COMMENTS

1. TC 4.11’s web site (www.ecw.org/tc411/program_summary.htm) should reflect the voted programs and the priority.
2. Future meetings and deadlines are listed below:

1. TC 4.11 UTILITY/BUILDING INTERFACE

1. DALLAS
   

1. Sponsoring Committee: TC 4.11 & Co-Sponsoring Committee: TC 1.4

Case Studies Of Energy Procurement And Aggregation

Seminar, Les Norford, Steve Blanc & Doug Nordham, Track1_____ & Track2_____


2. MINNEAPOLIS

1. Sponsoring Committee: TC 4.11 & Co-Sponsoring Committee: TC 1.4

1. Sponsoring Committee: TC 4.11 & Co-Sponsoring Committee:

1. TC 4.11 FAULT DETECTION AND DIAGNOSTICS

1. Sponsoring Committee: TC 4.11 & Co-Sponsoring Committee:

Fault Detection and Diagnostics (FDD)

Seminar, Jim Braun, Michael Kintner-Meyer, Ian McIntosh, Track1_____ & Track2_____

2. Sponsoring Committee: TC 4.6 & Co-Sponsoring Committee: TC 4.11

Symposium, Barry Bridges, Track1_____ & Track2_____

1. TC 4.11 APPLICATIONS

1. DALLAS

1. Sponsoring Committee: TC 1.4 & Co-Sponsoring Committee: TC 4.11

Should We Consider Expanding The Typical Control Applications From Fan Coil Units To…

Forum, Clay Nesler

2. Sponsoring Committee: TC 1.4 & Co-Sponsoring Committee: TC 4.11

Adding New Life To Old Systems. Retrofit Case Studies.

Symposium, Gaylen Atkinson

3. Sponsoring Committee: TC 1.4 & Co-Sponsoring Committee: TC 4.11

A Paper on Research Project RP-1011

As the cost of hardware (e.g., sensors, microprocessors) goes down, interest in automated fault detection and diagnostics (FDD) for vapor compression systems grows. For the most part, the methods that have been developed for on-line FDD involve the use of thermodynamic measurements to detect and diagnosis faults that degrade system cooling capacity and efficiency and impact the life of equipment. Methods based upon thermodynamic measurements have been documented by McKellar (1987), Stallard (1989), Yoshimua and Noboru (1989), Kumamaru et al. (1991), Wagner and Shoureshi (1992), Hiroshi et al. (1992), Grimmelius et al. (1995), Stylianou and Nikanpour (1996), and Rossi and Braun (1997). The faults considered include compressor valve leakage, heat exchanger fan failures, evaporator frosting, condenser fouling, evaporator air filter fouling, liquid line restriction, and refrigerant leakage. Typically, temperature and pressure measurements have been investigated because of their relatively low cost.

Most of these FDD methods rely on a mathematical model to predict the fault-free thermodynamic states as a function of the driving conditions that affect the operation of the vapor compression system. For example, on a simple rooftop air conditioning unit, the driving conditions might be the ambient air temperature, indoor air temperature, and indoor air relative humidity. These independent variables uniquely determine the system’s dependent variables such as the compressor discharge temperature and suction superheat during fault-free operation. When the thermodynamic states differ from those predicted by the model, residuals are generated. Classification of these residuals is then performed to detect and diagnose faults. An alternative approach would be to compare current model parameters to model parameters generated for normal conditions. The differences in the model parameters could be analyzed by classification techniques to detect and diagnose faults.

The importance of developing an accurate and robust model is crucial to the success of the FDD technique. However, most of research work performed in this field has focused more on classifying the differences between measurements and model predictions (residuals) for both fault detection and diagnostics. Little work has been done in developing and evaluating effective models for fault-free operation. Some of the references mentioned above and along with some additional references including Norford and Little (1993), Haves et al. (1996), Breuker (1997a, 1997b), and Breuker and Braun (1998) have dealt with the issue of developing a model for normal system operation. However, most of the techniques have assumed that a robust experimental data set is available for training the model. If a FDD device is to be added to a vapor compression system that is already installed and operating in the field, there is no simple way to obtain training data at a robust set of driving conditions. In this case a method that learns on-line while the system is operating would be more practical for training a model.

A major contribution to the field of FDD for vapor compression equipment could be made by a study that simply focused on developing a simple, robust, on-line training method of learning a mathematical model for normal system operating states. The development of a mathematical model and the specification of the training technique are viewed as a crucial step in the effort to apply FDD methods to existing vapor compression equipment. The results of this work could be used by a number of researchers in the field to improve their existing FDD methods and push this technology closer to widespread commercialization and application.

The objective of this work is to develop and compare mathematical models that could be used to predict the fault-free operation of a vapor compression system and the on-line training techniques that allow this model to be trained during day-to-day operation. The findings of this study will be used as the basis for a follow-up study that will focus on demonstrating the most promising model and training technique at one or more field sites.

In order to meet the objectives outlined above, a number of stages of work need to be completed. A suggested order for the project is outlined below. Important information that should be included in proposals for this work statement is included in the section Other Information for Bidders.

As a first step in the project, the current methods available for FDD on vapor compression equipment should be reviewed. In particular, the reviews should note the measurements used by the FDD techniques and the type of mathematical models that were used to predict normal system behavior. This information will be necessary to determine what measurements on the vapor compression equipment should be modeled for this project. A list of the measurements that will be required shall be submitted to the project monitoring subcommittee (PMS) for approval prior to the commencement of Task 3.

There are a number of fields of study that have attempted to apply on-line training techniques to learn mathematical models for physical systems. The successful contractor will be expected to examine and produce an interim report on the different classes of mathematical models and training techniques that could be utilized to develop a model of a vapor compression system using day-to-day operational data. Examples of different classes of models include physical models, regression models, neural network models, fuzzy models, time-series models, state estimation models, and look-up tables. The interim report should describe the models and training techniques, assess their strengths and weaknesses, and recommend modeling and training techniques from at least four separate classes that will be further developed and customized to the vapor compression problem. Emphasis should be placed on selecting modeling techniques that can be learned without operator supervision. Models that require an expert to tweak parameters are not considered appropriate candidates for study. The report shall be submitted to the PMS for approval prior to the commencement of Task 4.

In order to begin developing and testing the modeling techniques selected in Task 2, some operating data from a working vapor compression system will be required. To obtain this data, a data acquisition system should be installed on a vapor compression system to record the desired system measurements as the system operates in a normal mode. This task can take place as soon as the measurements and equipment are selected (and approval from the PMS obtained) in Task 1. Care should be taken to ensure that no faults are developing while the data acquisition is in progress. There are two types of data that may be valuable to obtain: laboratory measurements and field measurements. Beginning the investigation in the laboratory may speed the process of obtaining data at a wide range of driving conditions to investigate the expected form of the models. However, a significant amount of the data (minimum of one month operation during cooling season) should also be obtained from a similar type of system operating in the field, since there may be unmodeled inputs affecting the system that are not present in the laboratory. The laboratory data is only intended to be used to better understand the form of the relationships between driving conditions and the operating states and as a test data set for the trained models. The training of the models must only be done with data that has been obtained from the system(s) operating the field. Enough field data should be collected in this step to ensure a rigorous testing of the modeling and training methods selected in Task 2. The use of existing data is also acceptable provided that the data was adequately documented at the time of collection and provided the data meets the requirements described above.

An experimental plan describing the data collection process shall be submitted to the PMS for approval prior to the commencement of this task (Task 3). If existing data is used, the experimental plan shall be replaced by appropriate documentation to assure the PMS that the data was collected during fault-free operation. Further, the documentation should include appropriate sensor information, driving conditions, sampling intervals, and other information deemed relevant by the PMS.

Using the data obtained in Task 3, the models and training techniques chosen for further study in Task 2 shall be compared and evaluated. The data should be divided into testing and training data in order to properly assess the capability of the modeling and training techniques to learn the normal operational behavior of the system. For example, one approach might be to let the methods train using the first week of operating data and then test for the rest the operating data.

Minimum methods of evaluation would be to compare the sum of the squares of the errors and the maximum errors produced the methods. In addition, some methods may be able to give an estimate of the uncertainty of their predictions in different operating regions based on the amount of training data in that operating region. The reliability of this uncertainty prediction should also be assessed. Methods should also be compared based on their abilities to make accurate predictions as a function of the training duration or amount of data used to train the model.

The contractor shall produce a comprehensive final report detailing all the work undertaken in the project. The report should contain recommendations regarding possible future research topics related to model identification and FDD for vapor compression equipment. All data sets collected and/or used during the course of this project along with all computer code should be thoroughly documented and delivered with the final report. All methods, tools, algorithms, computer code, and data used and developed as a part of this research shall be in the public domain.

1. Quarterly progress and financial reports.
2. Oral presentations of the Principal Investigators to the project monitoring subcommittee of TC4.11 at the annual and winter meetings.
3. Modeling and Training Techniques interim report describing mathematical fundamentals and summarizing strengths and weaknesses of the techniques considered.
4. A comprehensive final report that includes the literature review, the Modeling and Training Techniques interim report, detailed description of the demonstration equipment and instrumentation, and conclusions from the study.
5. Electronic copy of data collected during the study and documentation explaining the format of the data.
6. A documented paper copy and electronic copy of all computer code, a users manual guiding future researchers in the compilation and use of this code, and a detailed description of the role of any "canned" software tools (e.g. Numerical Recipes in C, MATLAB Neural Network Toolbox) used in the analysis.
7. A technical paper that summarizes the results of the project.

It is estimated that the project will require approximately 18 months to complete at a cost of about $90,000.

As the cost of hardware (e.g., sensors, micro-processors) goes down, interest in automated fault detection and diagnostics (FDD) for vapor compression systems grows. To date a number of methods have been developed and tested to perform FDD on vapor compression systems. For the most part, the methods that have been developed for on-line FDD involve the use of thermodynamic measurements to detect and diagnosis faults that degrade system cooling capacity and efficiency and impact the life of equipment. Methods based upon thermodynamic measurements have been documented by McKellar (1987), Stallard (1989), Yoshimua and Noboru (1989), Kumamaru et al. (1991), Wagner and Shoureshi (1992), Hiroshi et al. (1992), Grimmelius et al. (1995), Stylianou and Nikanpour (1996), and Rossi and Braun (1997). The faults considered include compressor valve leakage, heat exchanger fan failures, evaporator frosting, condenser fouling, evaporator air filter fouling, liquid line restriction, and refrigerant leakage. Typically, temperature and pressure measurements have been investigated because of their relatively low cost.

While a number of methods for FDD have been developed and documented, there does not seem to be as many publications that document the presence of faults and the corresponding need for FDD in vapor compression systems. It appears that the only publication to address causes of failure for air conditioning equipment was presented by Stouppe and Lau (1989). They summarized the cause of 15,716 failures that led to insurance claims in HVAC&R equipment over an eight-year period from 1980 - 1987. A few studies have addressed the effects of faults on overall air conditioning system performance. Bultman et al. (1993) showed a 7.6% decrease in system COP for a 40% reduction in air flow for an air conditioner. Krafthefer et al. (1987) estimated a 10-13% decrease in COP for typical evaporator filter fouling of a heat pump. Furthermore, they estimated operating cost savings of 10-25% through use of a high efficiency air filter upstream of the evaporator. Rossi and Braun (1996) compared the combined service and energy costs associated with optimal maintenance scheduling for cleaning of condensers and evaporator air filters for rooftop air conditioners. They estimated total cost savings between 5 and 15%. Breuker and Braun (1998) documented the frequency of occurrence and the total cost of some faults on rooftop air conditioners by surveying the database from a HVAC service company. They also quantified the effect of five faults on the capacity and efficiency of a rooftop air conditioning unit.

A number of techniques have been developed for FDD on vapor compression equipment but the demand for the application of this technology has not driven the product to significant levels of commercialization. In order to convince equipment manufacturers and building owners that there is a potential for saving money with the installation of FDD algorithms, the typically occurring levels of these faults and their impacts on performance need to be documented. Another added benefit of this study would be to highlight the need for routine maintenance practices for building operators and the role that artificial intelligence could play in helping to establish the proper scheduling of this maintenance (Rossi and Braun, 1996). The results of this research project could be used to justify the commercialization of automated FDD for vapor compression equipment.

The objective of this project is to study and document the degradation fault levels that occur in vapor compression systems operating in the field and their corresponding effect on the performance. The first phase of the project (included in this work statement) will determine the frequency and qualitative level of these faults. Since it is unreasonable to cost effectively visit and measure a statistically significant number of vapor compression systems in the field, the source of this data is expected to come from a combination of field surveys and interview/polling of service personnel. In a second phase (not included in this work statement), research will be conducted to determine quantitative levels and the effect of these degradation faults on the performance of the air conditioning system in terms of capacity, efficiency, and possible reliability problems.

In order to meet the objectives outlined above, a number of specific tasks need to be completed in two phases. This work statement outlines the first phase of the project. The work of the second phase will be bid after the completion of the first phase work. A suggested order for the first phase of the project is outlined below:

An important first step in this project is the survey of information that is already available on this topic. Publications that estimate the level and frequency of faults that typically occur in vapor compression equipment along with experimental studies on the effect of these faults on system performance should be investigated. Some examples of previous work are reviewed in the Background section of this work statement. Other possible sources of information may include equipment manufacturers, HVAC service companies, electric utilities, or previously published studies by academia or industry. Both industry estimates from qualified sources and specific qualitative and quantitative studies should be included in this work.

The two most desirable classes of vapor compression equipment to study in this project are unitary air conditioning/heat pump units and chillers. The contractor should specify in their proposal whether they will include either one or both of these systems their work. A number of FDD methods have been developed for different degradation faults in these vapor compression systems. By reviewing the body of FDD research work on vapor compression systems, some of the important faults to study should first be determined. Some previous studies indicate that important degradation faults to consider in these vapor compression systems are refrigerant leakage, condenser fouling, evaporator fouling, liquid line restriction, and reduction in compressor volumetric efficiency. All studies should include these five faults at a minimum. Additional faults to consider may be uncovered during Task 1 and the initial work for Task 2.

It is expected that fault data will be obtained by a combination of: 1) surveying the insights of field service personnel in several HVAC maintenance organizations and 2) collaborating with one or more HVAC maintenance companies to visit vapor compression systems in the field and report on the rough quantitative or qualitative level of faults that the systems are experiencing. The targeted faults to study established in Task 2 along with the detailed plan for obtaining the information should be presented to the PMS for approval before proceeding with the rest of the project.

The contractor will execute the plan approved at the end of Task 3 to obtain the fault data for the study.

The contractor will write a final report that contains a summary of the findings and any conclusions that can be drawn from the data. The report should comment on the effect of different environmental factors (application, installation procedures, region of country, last date of service) on the levels of faults that were found on the equipment. An appendix should include a well-organized copy of all the hard data from the study. Electronic copies of all the data and all reports (interim and final) should be given to the committee.

The second phase of the work which would start after completion of the first phase would include the following tasks: 1) Developing a plan to obtain quantitative fault data from a number of field installations, 2) Acquisition of field data, 3) Comparison of field data with the data obtained in the first phase of the project, 4) Estimating the effect of the faults (all faults even the once identified in the first phase) on performance the equipment and 5) Summary of results and final report.

1. Quarterly progress and financial reports shall be made to the Society through its Manager of Research.
2. Oral presentations of the Principal Investigators to the project monitoring subcommittee of TC4.11 at the annual and winter meetings.
3. A final report that summarizes the literature review, surveys, and conclusions from the study. The final report shall include an executive summary suitable for widespread distribution to the industry and the public. Unless otherwise specified, the final report shall be submitted as six bound copies; one unbound copy, single-sided and suitable for reproduction; and two electronic copies, one in ASCII and one in the word-processing format used to produce the report. The report should also include an appendix with hard data from the field surveys.
4. One or more technical papers shall be prepared in a form suitable for presentation at a Society meeting. The paper(s) shall conform to Section 5 of the Society's "Author's Manual for Technical and Symposium Papers.
5. A Technical Article suitable for publication in the ASHRAE Journal that summarizes the research results.

It is estimated that the project will require approximately 12 months to complete at a cost of about $60,000.

Mark Breuker (msbreuke@duke-energy.com)

Srinivas Katipamula (Srinivas.Katipamula@pnl.gov)

Large systems, including buildings, can be represented in a hierarchical structure where the entire system is divided into sub-systems, which are in turn divided into sub-sub-systems, etc., as illustrated in Figure 1.

Figure 1 – Hierarchical representation of a building HVAC system

A literature survey shall be conducted to identify existing methods for supervising and coordinating the interactions of multiple, hierarchically-related FDD or other software modules and associated issues.

In this task, the contractor shall select individual building systems, FDD methods that will be applied to them, and the faults that the methods will be capable of detecting and diagnosing. This task shall involve the following steps:

1. Identify at least four building sub-systems and at least two different FDD methods that will be used for examining FDD conflict resolution issues. These systems must be distinct from one another but sufficiently related that faults can occur that impact both systems simultaneously in such a way that FDD methods applied to both of them would produce related and in some cases contradictory results. These subsystems, might be at different levels in the building hierarchy. An example might be an air handler and the filter-coil section of it.

Illustrate in diagrams how these systems are hierarchically related and how they may potentially interact.

It is not the intention of this research project to develop new FDD methods. Contractors are encouraged to use existing FDD methods and software, if available.

2. Individual FDD methods will be applied to the selected HVAC subsystems (e.g., air handling units, chillers, and zone controllers) or subsections of them (e.g., filter coil section). Identify how the faults detected and diagnosed by the FDD methods for these systems interact with one another. Also, identify potential conflicts between the results obtained from these FDD methods, and how they would occur.
3. Identify what data (information) the FDD methods will require, what information they will provide, and how FDD results will overlap.
4. Develop four sets of input data each of which will cause at least 2 of the FDD methods to provide conflicting results. Apply the FDD methods to produce at least 4 sets of conflicting results with FDD applied to at least two HVAC subsystems for each data set. For example, one case might be an AHU FDD method complaining that the inlet chilled water is too warm while the FDD method for the chiller's controller calling for an increase in chilled water temperature.
5. The PMS will review and approve the selected HVAC subsystems, the FDD methods selected, and the interactions selected for further investigation.

The contractor shall prepare a comprehensive final report that describes all aspects and contributions developed in this project, including:

• the HVAC subsystems selected
• the FDD methods, the faults they identify, and how they are implemented
• detailed descriptions of the conflicts occurring between the FDD methods
• the data sets developed
• descriptions of any software or templates for using software (e.g., spreadsheet templates) used in the investigation
• copies of source code and executable versions for any software code developed as part of the project
• descriptions of all metrics used for evaluating conflict resolution methods
• descriptions and results of all tests and evaluations
• careful documentation of guidance for applying (i.e., using the conflict resolution methods in practice and as part of automated FDD systems.

The report should also identify and document problems and issues that need to be resolved before widespread deployment of independent automated diagnostic software.

2. A recommendation for the HVAC systems, FDD methods, and conflict resolution methods to be used in the project. This recommendation must be approved by the PMS, prior to further work.
3. Source code for any software programs, tools or templates developed in the project.
4. Monthly teleconferences to address project progress with the PMS. These teleconferences shall be initiated by the Contractor’s Principal Investigator and should include all members of the PMS. Minutes of the teleconferences shall be kept by the Contractor and distributed to all members of the PMS at the conclusion of each teleconference.
5. Written Progress and Financial Reports shall be made to ASHRAE through its Manager of Research at quarterly intervals – specifically on or before January 1, April 1, June 1, and October 1 of the contract period.
6. The Principal Investigator shall report in person to ASHRAE Technical Committee (TC) 4.11 at the Annual and Winter ASHRAE meetings and answer questions regarding the project as may arise.
7. A Final Report shall be prepared and submitted to the Manager of Research at the end of the contract period. This Report shall provide complete details of all research carried out on the project. Six (6) copies of the draft Final Report and demonstration product (disk or CD) shall be furnished for review by the Project Monitoring Subcommittee.
8. Following approval by the Project Monitoring Subcommittee and TC 1.5, final copies of the Final Report and demonstration product shall be furnished as follows:

• Six (6) copies of an Executive Summary of the project suitable for wide distribution to the industry and the public;
• Six (6) bound copies of the Final Report;
• One unbound copy of the Final Report, printed on one side only, suitable for reproduction;
• Four (4) 3.5" diskettes with copies of the Final Report -- two (2) with the Report in ASCII format and two (2) with the Report in the most current version of Microsoft Word or Corel WordPerfect;
• Six (6) copies of the final demonstration product on CD or 3.5" diskettes;
• A "master" copy of the CD or diskette for the demonstration product suitable for reproduction.

2. One or more Technical Papers shall be submitted in a form suitable for presentation at a Society meeting. The papers shall conform to ASHRAE’s "Submitting Manuscripts for ASHRAE Transactions" which may be obtained from the Special Publication s Section. (On the ASHRAE Home Page, these guidelines are titled "Meeting Paper Preparation" and can be found under "How to Participate."
3. All papers or articles submitted for inclusion in ASHRAE publications that result from this research project shall be submitted through the Manager of Research — not directly to the publication’s editor.
4. A Technical Article suitable for publication in the ASHRAE Journal may be requested by the TC or Journal Editor. Such an article would be considered a voluntary contribution to the profession and is not a project deliverable.