Energy and Buildings 36 (2004) 1075–1090

Improving building operation by tracking performance metrics

throughout the building lifecycle (BLC)
D.T.J. O’Sullivan a,∗ , M.M. Keane a , D. Kelliher a , R.J. Hitchcock b
a IRUSE, Department of Civil & Environmental Engineering, University College Cork, Cork, Ireland
b Lawrence Berkeley National Laboratory, Building Technologies Department, USA

Received 3 December 2003; received in revised form 19 February 2004; accepted 3 March 2004

Abstract

An industry foundation class based building product model of University College Cork’s “to be constructed” Environmental Research
Institute will be developed. This paper discusses combining such a building product model with a building management system and other
tools and technologies to create a framework for monitoring, analysing and controlling a building throughout its building lifecycle based
on a set of performance metrics. The framework will be known as the building energy monitoring, analysing and controlling (BEMAC)
framework. Current building performance assessment practices lack standardisation/continuity throughout the building lifecycle. An
environment such as the BEMAC framework described in this paper offers a means to achieving such standardisation/continuity by
documenting and communicating performance metrics data such that these data can provide value across the complete lifecycle of a
building project, from planning through design and construction into occupancy and operation.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Building performance; Performance metrics; Building product model; Building management system; Industry foundation classes; Building
lifecycle

1. Introduction Much work is ongoing in the area of developing build-

ing performance frameworks to encourage better assessment
A computer-based building product model (BPM) of Uni- practices. These building and facility performance frame-
versity College Cork’s (UCC) “to be constructed” Environ- works include ongoing efforts to prepare a comprehensive
mental Research Institute (ERI) [1] will be developed that, compendium of building performance models such as the In-
together with a building management system (BMS), and ternational Council for Building (CIB) Performance Based
other IT tools and technologies, creates an integrated envi- Building Program [2–4], to the development of a system for
ronment allowing for the storage, measurement and control assessing the environmental impact of a building over its
of certain performance-based sustainability criteria in rela- lifecycle such as the US Green Building Council’s LEED
tion to the buildings energy usage (Fig. 1). Green Building Rating System [5]. Other frameworks in-
This paper describes the BEMAC framework that provides clude: the ASTM/Whole Building Functionality and Ser-
a platform for integrating existing software tools at design, viceability [6,7]; the International Code Council (ICC) Per-
construction and operation stages of buildings so that mon- formance Code for Buildings and Facilities [8,9]; Build-
itoring, analysis and control of performance, with regard to ing Research Establishment’s Environmental Assessment
energy usage, at all stages of the building lifecycle (BLC) Method (BREEAM) [10]; and the US Department of Energy
can be achieved. The future of sustainable buildings lies in (DOE) High Performance Metrics Project [11–13].
such a performance-based approach at all stages of the BLC. Currently however most building performance assessment
Building performance assessment (BPA) is currently very is done at the design stage of a building through the use
fragmented. There is no continuity to the process of building of simulation tools, some assessment is carried out at the
performance assessment through the building lifecycle. construction and commissioning stage by means of com-
missioning tests, but thereafter there is little or no assess-
∗ Corresponding author. Tel.: +353-21-490-2913; ment carried out in the operation and maintenance phase of
fax: +353-21-427-6648. a buildings lifecycle. This despite the fact that current build-
E-mail address: dominicosullivan@[Link] (D.T.J. O’Sullivan). ing management systems facilitate trending and archiving of

0378-7788/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/[Link].2004.03.003
1076 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090

Fig. 1. ERI project overview.

all sorts of data providing a largely untapped resource that higher level of compliancy with such objectives and hence
can improve the design, operation and energy efficiency of a greater level of satisfaction with the built environment.
installed systems. Recent case studies [14,15] suggest that Sets of energy-related performance metrics have been
energy savings of between 15 and 40% could be made in defined by various efforts such as: the US DOE High-
commercial buildings by closer monitoring and supervision Performance Buildings Metrics Project—Performance Met-
of energy-usage and related data [16]. Fig. 2 sums up the rics related to Resource Consumption and Environmental
current state of building performance assessment. Loading of Energy Use [12,13]; the ANSI/ASHRAE Stan-
Performance objectives/criteria related to building en- dard 105-1984 (RA 99) entitled “Standard Methods of
ergy efficiency are common to all the previously mentioned Measuring and Expressing Building Energy Performance”
frameworks, however they often lack quantifiable metrics [18]; and as part of the Laboratories for the 21st Century
that can be used to specify and track the energy performance Program, an effort has been undertaken to “develop a stan-
of buildings. Work is also ongoing in this area, which is dard set of energy performance metrics that will become
defining performance metrics that are intended to explicitly commonly used in the design, commissioning and operation
represent the performance objectives for a building project of laboratories” [19,20].
using quantitative criteria, in a dynamic, structured format. If such performance metric sets are standardised and as-
To be useful across the building lifecycle, each metric must sociated with the earlier mentioned building performance
be capable of being either predicted or measured at various frameworks, the potential for performance tracking through-
stages of the project so that the achievement of the associ- out the building lifecycle, given the right Information Com-
ated objective can be evaluated [17]. Enabling better evalu- munication Technology (ICT) framework/environment, be-
ation of building project objectives will inevitably lead to a comes a reality. This paper proposes such a framework based

Fig. 2. Current state of building performance assessment.

 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090 1077

Fig. 3. BPM used to track metrics throughout the BLC.

around the concept of using a building product model to Fig. 4 illustrates a scenario for tracking performance met-
track performance metrics throughout the building lifecycle ric data across the lifecycle of a building that could be
(Fig. 3). The building product model offers the link or feed- achieved in the proposed framework. The scenario begins
back loop for information exchange throughout the BLC during pre-design planning where an initial set of perfor-
that is currently lacking in building performance assessment mance metrics are specified as documentation of the build-
practices (Fig. 2). ing program. During design, this initial set is elaborated and
The previous scenario offers a means of documenting and modified as the design evolves. The results of simulation
communicating performance metrics data such that these and other assessment methods are used to update the per-
data can provide value across the complete lifecycle of a formance metric data. These assessment results become the
building project, from planning through design and construc- benchmarks of expected performance for the final building
tion into occupancy and operation. The accumulated archive design. Carried forward to the commissioning phase, these
of data additionally captures a performance history of the design benchmarks identify both what to measure in the
project that can be analysed to evaluate the success of various constructed building, and the expected level of performance
decisions along the way, leading to an overall improvement with respect to each metric. Following commissioning the
in the delivery and maintenance of the built environment. updated metric data act as as-built benchmarks for use dur-

Fig. 4. Lifecycle performance metric tracking scenario.

1078 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090

ing ongoing operations, maintenance, and retrofit of the fa- holders (e.g. owners, designers, operators, occupants, etc.).
cility. Comparing these benchmarks to data collected by a While primary attention is generally given to space require-
building management system can then perform periodic as- ments and construction costs, a wide spectrum of objec-
sessment of ongoing performance. This continually updated tives may be at least informally considered at this stage, in-
archive of performance metric data thus serves to support cluding: energy-efficiency; environmental impact; lifecycle
numerous activities across the building lifecycle [17]. economics; occupant health, comfort and productivity; and
With an archive of building performance data, it is now building functionality, adaptability, durability and sustain-
possible to assess how well the simulation tools, initially ability. The process of elaborating on the objectives for a
used to assess design decisions, are emulating what is actu- given building project is often referred to as programming.
ally happening in the occupied building. Inevitably there will The intent of programming is to define the desired perfor-
be discrepancies due to assumptions made within the simu- mance for a facility so that design and operation decisions
lation tools, for example the assumed equipment schedules can be made to achieve this performance. The outcome of
may be inaccurate. By adjusting such assumptions the sim- programming is most commonly recorded in freeform text
ulation tool can now be calibrated to better model the real that becomes part of design and construction documenta-
situation. Such a calibrated model could now be employed tion. This documentation may be frequently referenced dur-
to test changes to the design or operation of a building to ing design, and occasionally referenced during construction,
improve its performance. Use of analysis/simulation tools but then most often is lost from that time forward. Perfor-
at later stages of the building lifecycle is an area of much mance metrics can be used to more clearly and quantitatively
research in recent times [16,21–25]. define the performance objectives for a building [17].
This paper will discuss the concept of performance met- One or more performance metrics may be identified for
rics and objectives. The paper also describes the standards, any given performance objective that building process par-
tools and technologies employed in developing an extensi- ticipants wish to specify and track. A guiding principle in se-
ble environment to implement the proposed BEMAC frame- lecting a performance metric is to identify a critical variable
work and the associated methodologies. It also examines the that measures, reflects, or significantly influences a partic-
architecture behind tying these various tool and technolo- ular performance objective. For example, a high-level per-
gies together. The paper then demonstrates a sample imple- formance objective may initially be identified using only a
mentation whereby the performance metric “energy use of qualitative statement such as the desire to “optimise energy
a fan” is tracked, and the results are visualised on a 2D plot. performance” in a building. In order to assure the achieve-
Finally the paper discusses how this work can be advanced ment of an objective statement like this, it is helpful to elab-
upon to ensure improvements in building performances. orate it using multiple metrics that influence its overall sat-
isfaction. This elaboration can be organised hierarchically.
The hierarchy in Fig. 5 shows one possible subset of per-
2. Performance metrics and objectives formance metrics that could be used to specify, track, and
maintain energy-efficiency in a building [17].
2.1. The methodology The performance objectives for a building, and their con-
stituent metrics, often change as a project moves through
A building project begins with a consideration of the var- time. Objectives become more fully elaborated and are mod-
ious performance objectives of interest to building stake- ified as conflicts are discovered and resolved. Similarly, ad-

Fig. 5. Performance objectives and metrics hierarchy.

 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090 1079

EnergyPlus
Simulating
Performance

Jan

Dec Dec

Performance Jan
Metric Jan
Dec
Ventilation
System
Energy
Use

Metric Benchmark
Simulated Metric Assessment
BPM1 Measured Metric Assessment

BPM2 BPM3 BPM4 BPM5

BPM0

14/02/2003 14/02/2003 14/02/2003 14/02/2003

Specification Design Construction & ..14.30 14.45 15.00 15.15

Commissioning Operation & Maintenance

The Building Lifecycle

Fig. 6. Tracking of a specific performance metric.

ditional performance metrics may be identified, their desired a specification mainly consisting of goals and objectives.
performance levels (benchmarks) may change, and the ac- BPM 1 is a more detailed design using CAD and simula-
tual performance of a building will vary over time. A build- tion tools to produce geometrical data and expected levels
ing product model for tracking performance metrics must of performance, as are BPMs 2 and 3. The design typi-
therefore be capable of archiving a history of these changes cally changes between these BPMs to better facilitate the
across the lifecycle of a building [17]. Developing a build- achievement of particular objectives. BPM 4 is built on from
ing product model that successfully achieves this will enable BPM 3 but will also contain all construction (as built) data
unprecedented tracking of performance metrics throughout such as material data and commissioning test results, which
the building lifecycle. will be used to update the performance metrics and objec-
tives. Finally the ultimate BPM 5 is built on from BPM 4
2.2. The goals but now also contains operational data from the building’s
BMS. Another point of note in the above is that simula-
As an example consider a building that is in its operational tions do not start until BPM 1 is produced because a some-
phase, one could elicit a specific performance metric from what detailed design is necessary to carry out a simula-
the BPM to evaluate that individual metrics performance tion. Furthermore simulation metrics for particular designs
over the entire BLC. Results could be illustrated on an x–y are carried out over hypothetical years. Benchmark metrics
plot where the y-axis represents the specific performance come from basic simulations or databases of expected per-
metric, say ventilation system energy use, and the x-axis formance for particular building types. Measured metrics
represents the metrics transition through the phases of the come from commissioning tests for the construction and
BLC. One might expect to get a plot along the lines of Fig. 6. commissioning stage and are retrieved by the BMS for the
To explain some features of the above plot; firstly it is operational stage of the building lifecycle. Finally, clearly
notable that a number of instances/versions of the build- there will only be measured data for the construction and
ing product model are produced, however all these BPMs commissioning and operation phases. Time steps for the
represent the same building at different stages of its lifecy- operation phase can be varied, 15 min intervals are shown
cle and they all reside in a single central database. Essen- above.
tially a new BPM instance needs to be produced for each Fig. 7 further clarifies the process of events and demon-
major design change, so in the above BPM 0 is merely strates how the central data repository captures the building
1080 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090

Fig. 7. The central data repository contains a number of BPM instances.

at all stages of the BLC. Furthermore it provides explicit First and foremost with regard to this BEMAC frame-
connections between each of the phases, this facilitates com- work the building product model takes a central pivotal role.
parisons between the different types of metrics (benchmark, In this the building is modelled using the industry founda-
simulated and measured) representing a particular metric tion classes (IFCs). The database is initially set-up using
value over the course of the BLC. the EDMsupervisorTM , which allows the creation of a new
Having established the need for and demonstrated the database using a graphical user interface (GUI). Then the
use of an integrated environment for obtaining, formatting, underlying schema (IFC2×), which will be used as a basis
storing, retrieving and controlling the data associated with for model development, is imported. Again this is achieved
the buildings energy usage, an Information Communication using EDMsupervisor’s GUI and importing an express file
Technology framework for its implementation must be de- that defines the entire IFC2× specification. The database
cided upon. For the resultant BEMAC framework to be use- now “knows” the rules and constraints that its model must
ful it should be extensible and applicable to other build- adhere to. Now the database needs to be populated in ac-
ings in the future, thus it is essential that it comply with cordance with the underlying schema and this introduces
the principals of standardisation and open systems. Among the first of the tools that needs to communicate with the
such standards employed were the standard for the exchange EDM building product model, namely ArchiCADTM devel-
of product model data (STEP) [26] and industry founda- oped by Graphisoft. ArchiCADTM is a 3D CAD tool [29]
tion classes (IFCs) [27] for the transfer of data between the and was chosen for this implementation ahead of other 3D
building product model and the various analysis and design CAD tools such as Bentley’s Micostation TriFormaTM [30],
tools that communicate with the BPM. To develop the BPM or AutoDesk’s Architectural DesktopTM [31], because it is
itself, the centrepiece around which the BEMAC framework well advanced with regard to its compliancy with the IFC
has been built, an industry standard tool known as Express schema. It was IFC2× compliant ahead of most/all other 3D
Data Manager 4.5TM (EDM) [28] was used. CAD tools.
The following section specifies how these various tools; When designing in ArchiCADTM , a model of the real
technologies and standards combine to give a generic solu- building is created. Instead of drawing lines, ellipses and
tion, applicable to any building. arcs, one raises walls, adds windows and doors, lays down
floors, builds stairs and constructs roofs. Most importantly
from this research point of view ArchiCADTM is able to
3. The framework architecture export this information, in accordance with the IFC specifi-
cation, in the form of a standard STEP physical file (P-21
The technologies central to the creation of an open, file). Thus given the design of the Environmental Research
generic and extensible environment have been listed. Fig. 8 Institute (ERI) developed in the ArchiCADTM tool, the
demonstrates how to join these elements to create a frame- EDMdatabaseTM can be populated by importing the P-21 file
work applicable to any building that adheres to these open generated by ArchiCADTM such that the EDMdatabaseTM
standards. now contains an IFC compliant virtual building (building
 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090 1081

CONTROLLING MONITORING ANALYTICAL /

TOOL TOOL DESIGN TOOLS

SENSORS /
SWITCHES

BMS
Front-End
Energy
Analysis &
C++ Simulation
Tools e.g.
applications to
EnergyPlus
communicate
TRNSYS
with the
Programming C++ various
application to program EDM packages eg.
elicit data to input DATABASE Energy
BMS required by data analysis tools Comfort
Controller Analysis
IFC (resides obtained Building etc. This data Tools e.g.
in controller) from Product Model can take the CFX
sensors form of
into EDM Express files,
DB. csv’s, P-21
files, XML
CAD packages
files and more.
e.g.
ArchiCAD
AutoCAD

Future Tools
e.g.
Prediction
Tools

Fig. 8. The BEMAC environment.

Fig. 9. Transition from 3D CAD model to P-21 file and into database.

product model) (Fig. 9). It is important to point out that this initial model is extended to accommodate the notion
this process is not quite as simple and straight forward as of performance metrics and objectives as proposed in this
suggested in Fig. 9 and the designer, using ArchiCADTM paper.
to develop the model, needs to appreciate certain aspects With the geometric building product model now in place,
of the compilation process from ArchiCADTM to P-21 file, the operation, control and monitoring of the ERI building
such that this transition occurs in an apparently seamless are the next considerations. Data elicited from the build-
manner. Initially this building product model will consist ing management system, must now be tied in with the IFC
mainly of geometrical and material data. In the next section, compliant BPM. In Fig. 8 the sensors/switches represent the
on implementation of the framework, it will be shown how various types of building automation controllers distributed
1082 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090

throughout the Environmental Research Institute. These are 4. Implementation/results

networked in accordance with a standard building controls
protocol such as BACnetTM [32], and controlled by the BMS Although as earlier described this is a framework to enable
controller, which sends messages, runs routines etc. A fa- monitoring of performance metrics over the entire building
cilities manager using the BMS front-end, with its associ- lifecycle, for the purpose of this research an existing building
ated graphical user interface, can in turn edit/program this is being used as a prototype and thus the operational phase
controller. The controller can therefore be set up such that of the BLC is the focus of attention. Hence with regard to
particular values, required by the central BPM, for instance Fig. 6, it is the BPM 5 instance of the BPMs that will be
various performance metrics, are elicited from the BMS at produced. The building used for this research is the Mardyke
certain intervals. sports arena in University College Cork (UCC). This is an
Using the programming implementation method of 18.5 million, 6511 m2 indoor sports complex located just
EDMinterfaceTM , these data values can now be passed into off the main campus on the banks of the river Lee in the
the BPM. This programming can be implemented in any of Mardyke area of Cork city. It is made up of a six-lane,
the following programming languages; C, C++, JAVA, or 25 m swimming pool, a 1400 m2 sports hall, a 425 m2 main
VB. C++ was chosen for a number of reasons. First and fitness suite, along with offices, meeting rooms, performance
foremost it is an object-oriented programming language. laboratories, dance studios, catering facilities and a second
This ties in well with the object-oriented IFC schema and newly created overflow fitness suite. This building is similar
object-oriented BPM, such that there is a seamless inte- in size to the proposed ERI building and more importantly
gration between the software engineering phases involved both buildings require quite strict climate control with regard
in the project, right through to implementation and main- to their HVAC systems. As a sample performance metric that
tenance. Secondly, C++ is well developed, tested and forms part of the ventilation system energy use metric shown
supported within the EDM environment. This programming in Fig. 6, the energy consumption of a supply fan in one of
method of implementing EDMinterfaceTM is the method the AHUs will be considered for this paper to demonstrate
by which most of the analytical/design tools communi- the implementation behind producing a plot along the lines
cate with the BPM at regular intervals during operation of Fig. 6.
of the building. By automating this process it reduces The process of producing this plot is quite complex and
the need for management of the BPM. Samples of these requires a good understanding of the BEMAC framework at
programs will be illustrated in the following section on a number of levels. For instance, one must understand: how
implementation. to control or program the building management system to
Finally with respect to the other tools that will commu- elicit the necessary data; how to extend/instantiate certain
nicate with the building product model these include en- areas of the IFC schema such that it can archive data and ac-
ergy analysis and simulation tools, e.g. EnergyPlusTM and commodate performance metrics; how to develop programs
TRNSYSTM [33,34], to carry out tasks such as monitor the to input data to and elicit data from the correct locations
building’s energy usage, comfort analysis tools, e.g. CFXTM in the database; and finally how to combine these programs
[35], to carry out a CFD analysis and any other tools that with plotting programs to illustrate the results. The follow-
need to communicate with the BPM. The easiest way for ing sections will elaborate on each of these tasks.
this communication to occur is if the particular software tool
talks the STEP compliant IFC language. This last statement 4.1. Eliciting data from the BMS
highlights the benefits of open standards once again. Because
the BPM has been developed in accordance with the IFC The building management system in the Mardyke sports
standard the onus is primarily on the architectural, engineer- arena was created by Cylon Controls [36] and installed by
ing, construction and facilities management (AEC/FM) in- ACE Controls Systems Ltd.1 It comes with a relatively easy
dustry software developers to make their tools IFC compati- to use graphical user interface. For this sample case where
ble, such that those tools can become a part of this encapsu- the performance metric of interest is the supply fan energy
lating framework. However in this implementation, because consumption, the reading from the BMS of interest is the
of the extensive functionality offered by Express Data Man- supply fan variable speed drive (VSD) signal, which is a per-
ager 4.5TM , even tools that are not IFC compliant can talk centage reading representing the fan speed. From this value
to the building product model by parsing the data repository one can determine the frequency of the fan, and thereafter
with an applicable C++ program, that either retrieves the by using the correct fan curve, the current used by the fan.
data required by that tool from the BPM or stores the data Given this information the energy consumption of the fan
obtained from that tool into the BPM. This is made possible is easily deduced. By simply ticking the value of interest as
due to the range of formats in which EDM can accept data. shown in Fig. 10, one can relatively easily set up the BMS
None the less, IFC compliancy is most definitely the way
forward for AEC/FM industry software developers enabling
direct communication without the need for any intermittent 1 ACE Controls Systems Ltd., Unit 7, Ballincollig Enterprise Park, Link

parsing programs. Road, Ballincollig, Co. Cork.

 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090 1083

Fig. 10. The BMS graphical user interface.

such that it archives the values required in the form of a is a simplified view of the objects that need to be created
comma separated variable (csv) file to a certain location at a to instantiate the performance metric and objective parts of
specified frequency. Later it will be shown how the Express the IFC schema. This illustrates the complexity in adding
Data Manager 4.5TM storage program then retrieves this csv and connecting performance metric and objective data to a
file and stores the data in its correct location in the database. single supply fan object.
In respect to the above, the C++ program newly creates
4.2. Instantiating the IFC schema all the elements other than the initial existing supply fan
object. The program consists of several pages of code, and
Because the IFC schema is so vast, AEC/FM software de- loops through all the objects in the building product model. If
velopers will typically instantiate the so-called “views” of the object will have performance objectives and metrics data
the schema. That is to say they will develop their software associated with it at a later stage, the program instantiates the
in accordance with the area of the IFCs that is most relevant necessary objects to create spaces for such data to be stored.
to their particular industry. Hence taking ArchiCADTM for Now new positions exist in the database for the Express Data
instance, that is initially used to populate the database via a Manager 4.5TM storage program to input the data from the
P-21 STEP file, while ArchiCADTM implements nearly all respective csv files.
the architectural features of IFC, it might not for instance With regard to Fig. 11 it shows there are three types of
implement/instantiate an IfcSensor since this is more in the metric data as was also illustrated in Fig. 6 earlier. Table 1
facility managers domain of interest. Likewise ArchiCADTM suggests where the data for each respective metric type might
does not implement performance metrics and objectives even come from for the sample performance metric, the fans en-
though they have been part of the IFC schema since Ver- ergy consumption.
sion 2.0. However as explained earlier the schema underly-
ing the building instantiation is the entire IFC2× schema,
thus it is possible to instantiate the performance metrics and Table 1
objectives part of IFC and tie them into the initial build- Source of data values for metric, “energy consumption of a fan”
ing product model as created by the P-21 file generated by Metric type Benchmark Simulated Measured
ArchiCADTM . This is achieved through a C++ program and metric assessment metric assessment metric
a good understanding of the IFC2× schema knowing what Source of data Manufacturers EnergyPlus BMS
to create and where to tie it in with the initial BPM. Fig. 11 catalogue simulation
1084 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090

Fig. 11. Instantiating performance objectives and metrics in IFC.

4.3. Extending the IFC schema made to the IFC data model in the IFC2×2 release. These en-
hancements will however be incorporated into later versions
The IFC2× release of the IFC data model used in this of the BEMAC framework by using IFC2×2 as the underly-
research is primarily concerned with the exchange of build- ing schema for the EDM building product model. However
ing geometry. IFC2× includes only rudimentary definitions for the purpose of this work there is no “ready-made” mech-
of HVAC equipment, systems and components. The limi- anism by which to archive performance history data. The
tations of the HVAC component definitions in the IFC2× archiving of this performance history data is vital in order
data model release made it impossible to exchange rich data to reap the benefits earlier described with regard to tracking
sets with HVAC content. This prompted the launching of performance metrics over a building’s entire lifecycle. With
the BS-8 (BS-8 is the designation for the IAI “Building this in mind it is important to realise the following with re-
Services project number 8”) project in 2001 to complete gard to the IFC2× schema; in the schema’s implementation
the IFC HVAC domain schemata and make such exchanges of performance metrics and objectives there exists one place
possible [37]. This project has since been completed and a for a measured metric value and one place for a simulated
major extension to the HVAC part of the International Al- metric value, as can be seen in Fig. 11. However in the pro-
liance for Interoperability’s (IAI) IFC data model has been posed framework, for the operational phase of a building
defined, reviewed and implemented, and is now part of the lifecycle, these values will be frequently updated. Because
latest IFC release, IFC2×2. This extension facilitates in- there is a need to archive these values it leaves two options:
teroperability between energy simulation programs, such as
1. Create a new instance of the BPM each time these values
EnergyPlus, and other software used in the design and oper-
are updated.
ation of HVAC systems. The work includes the development
2. Devise a method within the bounds of the IFC2× schema
of a generic model structure that provides for both product
by which when these values are updated the overwritten
types, as defined in manufacturers catalogues, and individ-
values are archived within the same BPM instance.
ual instances of those types, each of which has a unique
placement, connections and performance history. A time se- The first of these options is very wasteful in terms of
ries model has been defined to support both the performance memory usage since the same data (BPM) is being repeat-
histories of individual components and the various bound- edly stored with only very minor changes from the previous
ary conditions—weather data, occupancy schedules, etc. re- instance.
quired by simulation tools [38]. Hence the second of the options was investigated. The
Unfortunately with this work being based on the IFC2× method devised essentially archived the simulated and mea-
model, it is unable to take advantage of the enhancements sured metric values by putting them into IfcComplexProp-
 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090 1085

Fig. 12. Clip of program that stores and archives metric values.

Fig. 13. Schematic of how archiving is achieved.

erty objects, named by a timestamp and distinguished as knowing how to track down from a high level object such
simulated/measured. Thus as each value, for example a mea- as the root of an IfcBuildingElementProxy representing
sured metric value from the building management system a supply fan object, down to its associated properties
comes into the EDM database, as well as being placed in such as the performance metric representing its energy
its correct location in the database, it is also placed in a consumption.
suitably named IfcComplexProperty object. This IfcCom- Fig. 14 illustrates how nested such data can be.
plexProperty object is then associated with the relevant ob- Furthermore Fig. 15 is just one function in the program
ject, for instance the supply fan object, through an existing necessary to work down through the numerous objects
relationship object. Again this required a good understand- shown in Fig. 13. For the shown function one is at the point
ing of the underlying schema and was all carried out via a where they have got the correct IfcBuildingElementProxy
C++ program. A tiny snippet from this program is shown and are going from there to the IfcMetricValue.
in Fig. 12.
Essentially, in simplified terms, the results as depicted in 4.5. Plotting your results
Fig. 13, with the archiving IfcComplexProperty object being
tied to the object in question, in this case a supply fan.
Finally the programs developed in the previous section
With this method of archiving performance metrics in
need to work with a plotting program to produce graphs such
place, one can understand how in Fig. 6, a single instance
as that in Fig. 6. For this project the plotting program was
of the BPM can cover the entire operational phase of the
developed in Microsoft Visual C++ using Microsoft Foun-
building lifecycle, or at least until some major renova-
dation Classes (MFCs). The benefits of using MS Visual
tion/restoration takes place.
C++ for the plots was that it allowed all the programming
4.4. Developing programs to store/retrieve data to/from from, data retrieval from the EDM database, to data plot-
the BPM ting, to be carried out in a single environment. The resulting
program is quite long and quite complex and uses all the
Again with regard to developing these programs it is elements previously described. In the example case, that is,
all down to understanding the underlying schema and energy usage of a supply fan, the program produces the plot

Fig. 14. Schematic of how metric values are connected to relevant objects.
1086 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090

Fig. 15. Function for getting the metric value associated with a particular IfcBuildingElementProxy object.

Fig. 16. Plot of energy use for fan.

 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090 1087

as in Fig. 16 for a user specified period of the operational the possibility of using EnergyPlus at the more untraditional
phase of the building lifecycle. phase of operation and maintenance of the buildings lifecy-
cle. One can now test changes to the design or running of
a building or changes to its HVAC system to see what ad-
5. How such a framework can be used to improve justments will get the building running closer to its intended
building performance across the building lifecycle performance objectives (metric benchmarks). This could in
turn lead to a plot along the lines of Fig. 18.
The BEMAC framework can improve building perfor- Now not only are the measured and simulated metric
mance over the entire building lifecycle. From the design values in close agreement but more importantly the build-
stage the framework can be employed to focus attention on ing is operating very close to its metric benchmarks. This,
issues such as energy usage. Moving on to the construc- now very accurate model, offers a very detailed (at both
tion and commissioning stage the BEMAC framework can time step and system/equipment levels) measure against
provide expected performance benchmarks against which to which to compare day-to-day building performance as an
commission. This leads to the operation and maintenance evaluation of the success of design decisions. This closes
stage as was demonstrated in the previous section on im- the lifecycle loop by providing feedback from operation to
plementation. As earlier suggested, Fig. 16 shows there are design. Finally this model also offers a good basis from
discrepancies between what the simulation tool expects the which to experiment with the ideas of fault diagnosis and
performance to be and what the monitored values say it ac- predictive control, areas of much research in recent times
tually is. This would in fact be true for many of the per- [21–25,39,40].
formance metrics. These discrepancies could be due to any
one of a number of reasons: components not performing as
they should; incorrect assumptions made within the simula- 6. Future enhancements
tion tool (EnergyPlus in this case) etc. The next logical step
is to try and calibrate the EnergyPlus model to more closely The work described in this paper offers a basis for further
emulate the real situation. This could be achieved by alter- research and development of the ideas put forward. Firstly
ing various assumptions made when using EnergyPlus, such one could review the possibility of tying this work in with
as the buildings occupancy schedule. After such calibration one of the frameworks mentioned in Section 1, to give them
Fig. 6 may now look along the lines of Fig. 17. the quantitative basis and implementation procedure they
In Fig. 17 the calibrated EnergyPlus model is much more currently lack. Furthermore one could look at the possibility
closely modelling the real situation (i.e. what is actually of developing a standard template/procedure for document-
happening). With the calibrated model, one can now look at ing specifications. Pre-design planning should ideally lead

Calibrated
EnergyPlus EnergyPlus
Simulating Simulating
Performance Performance

Jan

Dec Dec

Performance Jan
Metric Jan
Dec
Equipment
Energy
Use

Metric Benchmark
Simulated Metric Assessment
BPM1 Measured Metric Assessment

BPM2 BPM3 BPM4 BPM5

BPM0

Specification Design Construction & Operation & Maintenance

Commissioning

The Building Lifecycle

Fig. 17. Tracking of a specific performance metric after model calibration.
1088 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090

Design Change
(as
Calibrated recommended
by calibrated
EnergyPlus EnergyPlus
model)
Simulating Simulating
Performance Performance

Jan

Dec Dec

Jan
Jan
Dec

Performance
Metric

Equipment
Energy
Use
Metric Benchmark
Simulated Metric Assessment
BPM1 Measured Metric Assessment

BPM2 BPM3 BPM4 BPM5 BPM6

BPM0

Specification Design Construction & Operation & Maintenance

Commissioning

The Building Lifecycle

Fig. 18. Changing design or operation as advised by calibrated model.

to a clear specification of the performance expectations for which to tie these areas in with the framework developed in
a new or to-be-renovated building. This process begins with this research.
identification of general qualitative performance objectives,
but should lead to quantitative performance metric bench-
marks that will guide subsequent design and operation de- 7. Conclusions
cisions.
Another extension of this work could be to use the cal- Current building performance assessment practice lacks
ibrated model as an energy-benchmarking tool like ARCH continuity throughout the building lifecycle. There is a con-
[41], Cal-Arch [42] and the Labs 21 BDT [43]. Such tools centration of assessment at earlier stages of the BLC with
can be used to determine a benchmark whole-building en- little or no assessment at the later stages of operation and
ergy use intensity (EUI) that would achieve the desired level maintenance. The BEMAC framework developed in this re-
of performance for a new building of a given type in a given search encourages a continuity in the assessment process
location, relative to the existing building stock (e.g. a 75th throughout the BLC by combining the data from the differ-
percentile ranking) [16]. Furthermore benchmarks for all key ent phases of the BLC in a single central data repository.
energy-related performance metrics could be determined us- Performance metrics can be used to more clearly and
ing the calibrated model. quantitatively define the performance objectives for a build-
Finally with regard to the Environmental Research In- ing. Documenting and communicating performance metric
stitute, the building for which this framework will be put data can provide value across the complete lifecycle of a
to the test, the aim is to monitor many more performance building project, from planning, through design and con-
metrics than have been monitored in the prototype build- struction, into occupancy and operation.
ing. This will effectively lead to a real time database of all A single database processing geometrical, material and
the buildings performance metrics, when the building is up real time data offers huge potential for analysis and more.
and running. As mentioned such a database captures a per- The accumulated archive of data in the proposed building
formance history of the project allowing for evaluation of product model captures a performance history of the project
the success of design decisions. Furthermore this database that can be analysed to evaluate the success of various de-
paves the way for other areas of research such as the idea of cisions along the way, leading to an overall improvement in
predictive control and fault diagnosis. Much work is on go- the delivery and maintenance of the built environment.
ing in these areas [21–25,39,40] and the proposed building The wealth of data that the building product model of-
product model would most certainly offer a good basis from fers in a fully monitored building provides the basis for: im-
 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090 1089

proving building performance; developing predictive control [13] US DOE, DRAFT—Resource Consumption and Environmental
routines and implementing fault diagnosis measures. Loading of Energy Use (breakout group meeting notes from Spring
2002 Workshop), US Department of Energy, Washington, DC, 2002.
Finally the BEMAC framework will in time lead to better [14] P. Herzog, L. LaVine, Identification and quantification of the impact
more accurate design decisions with regard to, for instance, of improper operation of midsize Minnesota office buildings on
the sizing of equipment for HVAC systems, by focusing energy use: a seven building case study, in: Proceedings of the
attention on the achievement of performance objectives from ACEEE 1992 Summer Study on Energy Efficiency in Buildings, vol.
an early stage of the design. III, ACEEE, 1992.
[15] D. Claridge, J. Haberl, M. Liu, J. Houcek, A. Athar, Can you achieve
150% predicted retrofit savings? is it time for re-commissioning, in:
Proceedings of the ACEEE 1994 Summer Study on Energy Efficiency
Acknowledgements in Buildings, vol. V, ACEEE, 1994.
[16] T. Salsbury, R. Diamond, Performance validation and energy analysis
The authors would like to acknowledge the financial sup- of HVAC systems using simulation, Energy and Buildings 32 (2000)
5–17.
port of the Higher Education Authority Programme for Re-
[17] R.J. Hitchcock, High-Performance Commercial Building Systems
search in Third Level Institutions administered by the HEA Program, Element 2-Project 2.1, Task 2.1.2, Standardized Build-
in Ireland. We would also like to thank Cylon Controls ing Performance Metrics, Final Report, Lawrence Berkeley National
Limited (Clonshaugh Industrial Estate, Clonshaugh, Dublin Laboratory, October 2002.
17, Ireland), ACE Controls Systems Ltd. (Unit 7, Ballincol- [18] ANSI/ASHRAE, Standard 105-1984 (RA99), Standard Methods of
lig Enterprise Park, Link Road, Ballincollig, Co. Cork) and Measuring and Expressing Building Energy Performance, ASHRAE,
Atlanta, GA (co sponsored by American National Standards Insti-
UCC’s Buildings and Estates Office (University College tute), Washington, DC, 1999.
Cork, Cork, Ireland) for their contributions during the course [19] Labs21, Labs for the 21st Century, 2002. [Link]
of the research. This work has also been supported by The [20] LBNL/Labs21, DRAFT Proposal 12/15/01Rev1.1—Laboratory En-
California Energy Commission (Public Interest Energy Re- ergy Performance Metrics, Lawrence Berkeley National Laboratory
search Program, Contract No. 400-99-012); The US Envi- A-Team, Berkeley, CA, 2001.
[21] J. Clarke, J. Cockroft, S. Conner, J.W. Hand, N.J. Kelly, R. Moore, T.
ronmental Protection Agency (Climate Protection Division);
O’Brien, P. Strachan, Simulation-assisted control in building energy
and The Assistant Secretary for Energy Efficiency and Re- management systems, Energy and Buildings 1449 (2002) 1–8.
newable Energy (Office of Building Technology, State and [22] G. Augenbore, C. Eastman, Needed Progress in Building Design
Community Programs of the US Department of Energy, Con- Product Models, White Paper, College of Architecture, College of
tract No. DE-AC03-76SF00098). Computing, Georgia Institute of Technology.
[23] R. Lahrech, P. Gruber, P. Riederer, P. Tessier, J.C. Visier, Develop-
ment of a testing method for control HVAC systems by emulation,
Energy and Buildings 1447 (2002) 1–9.
References
[24] J.L.M. Hensen, R. Lamberts, C.O.R. Negrao, A view of energy and
building performance simulation at the start of the third millennium,
[1] Environmental Research Institute, 2002. [Link] Energy and Buildings 34 (2002) 853–855.
[2] International Council for Building (CIB), Performance Based Build- [25] P. de Wilde, G. Augenbroe, M. van der Voorden, Energy system
ing Program. [Link] simulation in performance-based building design, in: Proceedings of
[3] G.C. Foliente, CIB, Performance Based Building Program, 2000. the Sixth International Conference on System Simulation in Build-
[Link] ings, University of Liege, Belgium, 16–18 December 2002. ISBN
[4] G.C. Foliente, R. Becker, CIB PBBCS Proactive Programme—Task 2-930322-55-1.
1, Compendium of Building Performance Models, CSIRO Building, [26] STEP Tools Inc., Introduction to STEP, 2002. [Link]
Construction and Engineering, Victoria, Australia, 2001. com/library/standard/introduction to [Link].
[5] US Green Building Council, LEED, Leadership in Energy [27] Industry Foundation Classes, 2002 [Link]
and Environmental Design Program, 2003. [Link] uk/iai/[Link].
LEED/LEED [Link]. [28] Express Data Manager 4.5TM , 2002. [Link]
[6] ASTM, ASTM Standards on Whole Building Functionality and Ser- newsletter/ew101/[Link].
viceability, 2nd ed., ASTM, International, West Conshohocken, PA,
[29] Graphisoft ArchiCADTM , 2002. [Link]
2000. [Link]
architecture and design/graphisoft archicad/.
[7] G. Davis, F. Szegeti, Introduction to the Standards on Whole Building
[30] Bentley’s Micostation TriForma, 2003. [Link]
Functionality and Serviceability, International Centre for Facilities,
[31] Autodesk’s Architectural Desktop, 2003. [Link]
Ottawa, Ont., Canada, 1999.
[32] BACnet—A Data Communication Protocol for Building Automation
[8] International Code Council. [Link]
and Control, 2002. [Link]
[9] ICC, Final Draft ICC Performance Code for Buildings and Facil-
[33] EnergyPlus, A New-Generation Building Energy Simulation Pro-
ities, International Code Council, Falls Church, VA, August 2000.
gram, 2002. [Link]
[Link]
[10] BREEAM, Building Research Establishment’s Environmental As- [34] TRNSYS, The Transient Energy System Simulation Tool, 2002.
sessment Method, 2003. [Link] [Link]
[11] US DOE, High-Performance Building Metrics Project, US Depart- [35] CFX, AEA Technology Engineering Software. [Link]
ment of Energy, Washington, DC, 2003. [Link] com/cfx/.
buildings/highperformance/performance metrics/. [36] Cylon Controls. [Link]
[12] US DOE, High-Performance Building Metrics Project Framework, [37] IAI BS-8 Project. [Link]
US Department of Energy, Washington, DC, 2002. [Link] [38] V. Bazjanac, J. Forester, P. Haves, D. Sucic, P. Xu, HVAC Component
gov/buildings/highperformance/metrics/pdfs/[Link]. Data Modelling Using Industry Foundation Classes, SSB, 2002.
1090 D.T.J. O’Sullivan et al. / Energy and Buildings 36 (2004) 1075–1090

[39] J.E. Pakanen, T. Sundquist, Automation-assisted fault detection of [41] ARCH, A Building Energy Reference Tool. [Link]
an air-handling unit implementing the method in a real building, [42] Cal-Arch, California Building Energy Reference Tool. [Link]
Energy and Buildings 35 (2003) 193–202. [Link]/cal-arch.
[40] T.I. Salsbury, R.C. Diamond, Fault detection in HVAC systems using [43] Labs21-LBNL A-Team, Benchmarking Database Tool, Labs for the
model-based feedforward control, Energy and Buildings 33 (2001) 21st Century, 2002. [Link]
403–415.