Model Based Control
Transcript of Model Based Control
-
7/30/2019 Model Based Control
1/8
SIMULATION-BASED APPROACHES FOR BUILDING CONTROL SYSTEM
DESIGN AND INTEGRATION
Stephen Treado, Payam Delgoshaei, and Andrew Windham
Department of Architectural Engineering
Pennsylvania State UniversityUniversity Park, PA 16802
ABSTRACT
This paper discusses how simulation-basedapproaches can improve the design of buildingcontrol systems, with the goal of providing guidance
to control system engineers for obtaining bettercontrol performance and more effective integration
of the space conditioning components, equipmentand systems. The discussion focuses on how to usesimulation methods to improve the design andoperation of the control system, including selectionof models and tools, along with weather conditions,load profiles, time steps and the number and/or
duration of simulations. Specific examples areimplemented using detailed control modelling withHVACSIM+, to demonstrate and compare themodelling input requirements, level of detail,application of results and potential impact on controlsystem design. The benefits of dynamic versus
quasi-static simulations are discussed. Limitations in
the methods, recommendations for appropriateapplications, and suggestions for improvements arealso described.
INTRODUCTION
The implementation of tightly integrated building
systems and subsystems has been increasinglyrecognized as an essential factor in providing energyefficient, healthy and productive interiorenvironments in buildings. This is particularly truefor building control systems, which must bear theburden of controlling a wide array of components,equipment and systems to achieve energy efficientoperation while also maintaining comfortable thermal
conditions, and responding to occupant requirements.Generally, building space conditioning systems aredesigned to be able to meet peak thermal loads, theoccurrences of which are relatively rare throughout atypical year. Most of the time, building systemsoperate at part-load conditions, which, if not properly
accounted for, can have a significant impact onperformance, frequently in a negative manner. Thisis because the part load efficiencies of the equipmentand components comprising the systems are lowerthan at full load. In addition, building controlsystems need to provide adequate response to
constantly varying environmental conditions and
occupant factors. As thermal loads vary, theaccuracy and stability of the control system also vary,leading to the possibility of loss of comfort control orincreased energy consumption for space
conditioning, or both.
Simulations can be useful in several respects forimproving the design and integration of building
control systems. The major responsibilities of thecontrol system designer are to size and select controlsystem components, determine control strategies, andspecify control sequences. Typically, these choicesare based on an analysis of a limited number of staticdesign conditions. This process is inevitably coupledto the equipment and systems being controlled, such
as heating and cooling coils, fans, pumps and thermalmachinery. The control system designer may havehad a hand in the design of the mechanical systemsas well, or may be working in conjunction with themechanical systems designer. Notably absent fromthe design process is a comprehensive evaluation ofthe part-load and dynamic performance of the control
system, the results of which could suggestimprovements and/or refinements in the design toachieve better overall performance. Theseimprovements could include changes in equipmentsizes and configurations, as well as control strategies.
This paper discusses how simulation-basedapproaches can be utilized to improve the design ofbuilding control systems, with the goal of providing
guidance to control system engineers for obtainingbetter control performance and more effectiveintegration of the space conditioning components,
equipment and systems. The discussion focuses onhow to use simulation methods to improve the designand operation of the control system, including
selection of models and tools, along with weatherconditions, load profiles, time steps and the number
and/or duration of simulations. Specific examples areimplemented using detailed control modeling withHVACSIM+, to demonstrate and compare themodeling input requirements, level of detail,application of results and potential impact on controlsystem design. The benefits of dynamic versus
quasi-static simulations are discussed. Limitations inthe methods, recommendations for appropriate
applications, and suggestions for improvements arealso described.
Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.
- 2544 -
-
7/30/2019 Model Based Control
2/8
LITERATURE SURVEY
The need for building control simulation is not newlyarticulated. Efforts have been underway since theearly 1970s to tackle the issue, but still tools arelacking that are efficient and effective for mostdesigners. Most of the tools that are available are
focused on the research community rather than thedesigners. Throughout the years, culminations ofknowledge have led to a variety of tools andapproaches. Haves, Norford, and DeSimone (1998)report on one such effort supported by the AmericanSociety of Heating, Refrigerating and Air
Conditioning Engineers (ASHRAE). The researchconducted and compiled was used to aid the
capabilities of HVACSIM+ (Park, 2008) andTRNSYS.
Trka & Hensen (2010) provide an overview ofHVAC system modeling and simulation. Specifically
regarding the controls, it is noted that there are fewoption for simulating more advanced controllerfunctions. Approaches may require the simulation of
supervisory control as with EnergyPlus down to localcontrollers in TRNSYS and ESP-r. As well,MATLAB and Dymola are noted for their ability totest controllers in simulations.
MATLAB and Modelica both exhibit capabilities ofsoundly modeling actual controllers and various toolsand libraries are being developed to facilitateutilization of their potential. Riederer (2005) coversthe developments of MATLAB/Simulink for building
and HVAC simulation up to 2005, including
SIMBAD, a MATLAB toolbox that facilitatesbuilding related simulations (Jreijiry, Inard, &Villenave, 2003). Likewise, the modeling language,Modelic, has garnered interest from the researchcommunity for its potential (Sahlin, Bring, &Eriksson, 2009; Wetter, 2009).
Ellis, Torcellini, and Crawley (2007), describe an
energy management system module for EnergyPlus.The purpose is to enhance the control system
simulation capability of EnergyPlus by facilitatingmore customizable supervisory control strategies.
Emerging efforts are moving towards co-operativesimulation environments with the goal of utilizing thecapabilities of a host of different simulation tools.Wetter and Haves (2008) describe efforts to developa middleware platform that would allow users to
mesh the capabilities of various simulation programsin order to achieve more rapid and detailed models.In doing so, controllers may be modeled in anenvironment such as MATLAB and interact withEnergyPlus to control the simulated HVACequipment. Likewise, co-simulation tools are beingdeveloped and evaluated for building performance
simulation (Trka, Hensen, & Wetter, 2009). Thereported prototype system utilizes EnergyPlus and
TRNSYS, making use of the advanced buildingmodels and large HVAC component libraryrespectively.
Component-based simulation environments, likeHVACSIM+, have also been utilized to evaluatebuilding systems performance. Systems like thisallow for more narrowly focused simulations thatfocus on building HVAC and other systems.Recently, Cui, Yang, Spitler, and Fang (2008) use
HVACSIM+ to simulate various performancecomparisons for a hybrid ground-coupled heat pumpwith domestic hot water.
SIMULATION OF BUILDING CONTROL
SYSTEMS
The use of simulation tools to help develop andevaluate building designs continues to increase inpopularity as more user-friendly tools with greatercapabilities are becoming available. Undoubtedly,
the wide use of building simulation tools has resultedin an improvement in the energy performance of newbuildings, and has enabled designers to investigate
more innovative design alternatives. One keyquestion that needs to be addressed, however, is howwell the simulations represent the actual performanceof the systems being modeled, and how important
this aspect is to effective design practices. Severalrecent investigations have shown that actual buildingenergy consumption can differ markedly fromsimulation results, probably for a variety of reasons.This is particularly true for system configurations andoperating strategies that differ from the more
typical designs (i.e. the more aggressive designconcepts). It is not always easy to determine thedetails of how different simulation tools handle the
modeling of system components, and the detailedperformance characteristics and interactions ofbuilding systems may not be incorporated in acomprehensive manner.
Simulation tools are composed of collections of
models, schedules and material properties wrapped ina user interface and incorporating equation solving
capabilities. All simulation tools will utilizeassumptions and approximations to a greater or lesserextent; these assumptions may or may not have asignificant impact on any conclusions drawn from a
particular building simulation. This is especially truefor control system simulations, since their operation
is frequently modeled in an idealized fashion. Forexample, the response time of the control system andassociated components may be ignored, as may bethe impact of sensor location and actuator operation.These real world characteristics can have anoticeable effect on installed performance versus
design intent.
The successful implementation of a building controlsystem requires contributions from and cooperation
among a number of disciplines. The process startswith the determination of the expectations of theowners, the needs of the occupants and the desired
building functions and services, such as spaceconditioning, electrical power, illumination, safety,security, communication and entertainment. This, in
Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.
- 2545 -
-
7/30/2019 Model Based Control
3/8
turn, leads to the initial design concepts for buildingenvelope, mechanical and electrical systems. Sincethe purpose of the control system is to ensure theproper operation of building systems and to maintainthe desired indoor environmental conditions(temperature, air quality, humidity, comfort), control
system design is dependent on and tightly coupled tothe design of the other building systems. As the
building design process progresses, system conceptsbecome more refined and more specific. Frequently,simulations are used to compare various designoptions and to evaluate building performancecharacteristics, such as energy usage, structuralparameters and construction costs. Various
simulation tools are available with varying degrees ofease of use, sophistication and application. Many ofthe issues motivating the use of building simulationsinvolve the impact of design trade-offs orinteractions, whereby the modification of one
building system (i.e. envelope) affects another (i.e.heating and cooling). The broad goal of achievingintegrated building design suggests a comprehensiveanalysis that incorporates the interactions between allof the relevant building systems. This approachwould require some consideration of the operation ofthe control system.
The job of the building control system designerencompasses a range of activities, including:
sizing and selecting actuators and controlcomponents
selecting and locating sensors specifying a sequence of operation specifying setpoints developing schematic diagrams developing points lists
This collection of information is incorporated intobid documents, plans and submittals, eventuallybeing instantiated as specific hardware andaccompanying software elements which are installed,programmed and commissioned. Each of these steps
presents an opportunity to utilize simulations toimprove the final product: a properly functioningbuilding control system. The nature and type of
simulation are dependent on the stage in the process,the ambition of the project and the available time andresources. Simulations do require a certain amount
of time and appropriate expertise, which may notalways be available. Larger, more ambitiousprojects will afford more opportunity for designrefinement and optimization.
One question regarding the simulation of buildingcontrol systems is related to the type and duration ofthe simulation. Building energy simulationsfrequently are conducted on an annual basis usinghourly increments and typical weather data files.
While this type of simulation provides usefulinformation regarding the effectiveness of generalcontrol and energy management strategies, it does
not directly address other important controlperformance issues such as dynamic response andstability. In order to assess those aspects of buildingcontrol systems it is necessary to conduct thesimulation using much shorter time steps and withgreater system detail. Fortunately, the duration of
such short time step, detailed simulations does nothave to be long; usually a few hours to a few days
will suffice to reveal any performance issues thatneed to be remedied. Also, the specific weather dataand other driving factors are not so important, sogeneric load profiles or arbitrary functions (impulse,step, ramp, square wave) can be used. Theseconsiderations allow the designer to focus more on
the details of the control system and its specificcomponents, including sensors, actuators, valves,dampers, coils, fans and pumps. Other importantfactors are the length of ducts and piping, and anyassociated transit times and time delays, along with
actuator hysteresis, leakage, valve and damperauthority, heat transfer and storage in fluid conduits,and temperature gradients.
In order to adequately address the control systemdesign and performance issues outlined above, asimulation tool focused on building control systemscan be utilized. It is not the intent of this paper toevaluate and compare the various software tools thatare available for this purpose, but rather to discusshow they could be used, and to present some
examples to demonstrate how the process wouldwork. The example simulations were conductedusing the HVACSIM+ computer program, which was
developed at the National Institute of Standards andTechnology specifically for modeling theperformance of building control systems. While thisprogram is very powerful and flexible, it is not
commonly used by building designers, but is more ofa research tool. However, it will suffice todemonstrate what could be done with acomprehensive building control system simulationtool given the proper combination of expertise, userinterface and support.
EXAMPLE SIMULATIONS
Control system simulation can require quite a bit
more detailed information than is typically used forbuilding energy simulations. It may also requireassumptions about set points and operatingschedules. In order to illustrate this, a constantvolume air handler for a single zone with a hydroniccooling coil was modeled using HVACSIM+, as
shown in Figure 1. The model consisted of 24components or units in the lexicon of HVACSIM+,as listed in Table 1. Each of the components has aspecific set of parameters to describe theircharacteristics relevant to the simulation, includingsuch things as physical dimensions, thermodynamicproperties and mass flow values, some of which are
also listed in the table. For these simulations, typicalvalues were used that may not represent the best, but
Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.
- 2546 -
-
7/30/2019 Model Based Control
4/8
could serve as starting points for refining a design.Some key points to note are the significant thermalcapacitance of the supply air duct and conditionedzone, the time constants of the sensor and controller,and the valve actuator hysteresis. Thesecharacteristics will all combine to make it difficult to
provide for stable control, as errors will tend toaccumulate and control signals overcompensate.
Zone loads and outdoor air temperature were
explicitly specified, and other variables or set points
were selected or modified to suit each simulationusing a boundary value file. The model includes atotal of 66 variables (21 pressure, 21 temperature, 6flow, 10 control, 6 power and 4 others).HVAVSIM+ solves for the values of all dependentvariables for each time step based on physical models
of components and the fundamental principlesunderlying heat and mass transfer. Many aspects of
Figure 1. Schematic of single zone air handler
Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.
- 2547 -
-
7/30/2019 Model Based Control
5/8
Figure 2. Zone air temperature and cooling load for a steady heat gain of 12 kW (a), 8 kW (b), and 4 kW (c)
the simulation, such as initial values, time steps andoutput data can be specified by the user. Eachsimulation needs to have a starting condition.
Sometimes, these types of simulations are startedwith all variables set to zero, but for building control
system modeling, this approach may not work as thenumerical solver may not converge to the correctsolution. Starting with initial values in the generalballpark of final values will usually be beneficial in
encouraging a convergent solution, but numerical
challenges will sometimes present themselvesregardless.
For a transient simulation, it is useful to start bysimulating an equilibrium condition before movinginto the transient boundary condition, therebyallowing the control system to attain a goodoperating state before having to respond to the testcondition. Then, the test condition can be applied,
either as a onetime event (step change, ramp orpulse) or as a steady periodic profile (sinusoidal,cyclic) change in one or more boundary variables(setpoints, outdoor air temperature, heat gain, etc.).Variables of interest, such as space temperatures,thermal loads and control signals, can be output fromthe simulation and plotted for analysis.
For the first example, a constant zone load wassimulated. This is useful for verifying the sizing ofthe components identifying any modeling errors andchecking the execution of the simulation. For thismodel, the chilled water valve on the cooling coil is
controlled by the room air temperature, whichmodulates the flow of chilled water through thecooling coil to try to maintain the zone airtemperature at its setpoint. Figure 2a shows zone airtemperature and cooling load for a steady zone heat
gain of 12 kW simulated over approximately oneday. The initial cyclic behavior is a consequence of
the models converging to an equilibrium conditionfrom the initial condition. The zone temperature thenstabilizes as the room cooling load drifts slowlyhigher to compensate for heat storage and loss fromthe ducts and zone. The cooling load of 50 kW (coilload) includes both sensible and latent loads from the
zone heat gain and outdoor makeup air, as well asheat gains from the fans and ducts. This well isabove the simple loads due to zone heat gain andventilation air. Figure 3 shows the control signalfrom the controller to the control valve, and the
chilled water supply temperature, the latter of which
Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.
- 2548 -
-
7/30/2019 Model Based Control
6/8
is being held constant. The control signal and valveseem to be operating properly.
Figure 3. Chilled water supply temperature and
control signal to valve for a steady zone heat gain of
12 kW.
Similar simulation was conducted for steady zoneheat gains of 8 and 4 kW, as shown in Figures 2b and2c. For the 8 kW case, the simulation takes a bitlonger to reach equilibrium, but then both zone airtemperature and cooling load stabilize as expected.However, for the 4 kW case, oscillating behavior is
observed for the entire simulation time. It is notimmediately apparent whether this is due to anumerical problem or a system problem. In otherwords, is the control system itself oscillating due tothe tuning of the PI controller and other factors, or isthe numerical solver of the simulation tool failing toconverge to the correct solution. It may be that a
combination of the two factors are actuallyresponsible, as when the physical system becomesunstable or approaches instability, the numericalsolution becomes more difficult. In this case, it is
Table 1 Components of Example Control System Simulation
UnitNumber
Component and Selected Parameters
1 Inlet pipe to coil V=0.00251 m3C=2.277 kJ/C
2 Bypass pipe V=0.00023 m3 C=0.2116 kJ/C
3 Outlet pipe from coil V=0.002278 m3
C=2.07 kJ/C
4 Three-way valve =10 s hysteresis = 0.10
5 Return pipe V=0.0355 m3 C=32.2 kJ/C
6 Air temperature sensor= 20 s
7 PID controller KP=6.0 KI=0.08 =2 s
8 Cooling coil C=34.669 m3
9 Control signal inverter
10 Return fan Pressure/flow coefficients
11 Duct
12 Flow split
13 Exhaust air damper Leakage 0.012
14 Recirculation air damper Leakage 0.012
15 Mixing box
16 Outdoor air damper Leakage 0.012
17 Outdoor air duct
18 Damper (filter flow resistance)
19 Damper (heating coil flow resistance)
20 Damper (cooling coil flow resistance)
21 Supply fan Pressure/flow coefficients
22 Supply air duct C=220.0 kJ/C V=7.75 m#
23 Conditioned zone V=1440 m3 CEXT=40000 kJ/C CINT=4000 kJ/C
Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.
- 2549 -
-
7/30/2019 Model Based Control
7/8
24 Return air duct V=7.75 m3 m C=220.0 kJ/C
Key V-volume C-thermal capacitance -time constant
likely that the physical system itself is at leastpartially responsible, since equilibrium was achievedat the higher load conditions. This would suggestinvestigating the gains for the PI controller, and theeffects of the various time constants.
It is not the intent of this paper to try to optimize thisparticular example system, but rather to explore theuse of simulations to design control systems. Thus,
we will continue with the example by implementingvariable zone heat gain and outdoor air temperature
to look at the impact on zone air temperature andcooling load.
Figure 4. Variable zone heat gain profile
Figure 4 shows a variable heat gain profile andFigure 5 shows the resulting zone air temperature andcooling load, which appear to be acceptable (i.e.,zone air temperature tracks the setpoint and coolingloads are smooth.
Figure 5. Zone air temperature and cooling load for
the variable zone heat gain profile
There are many other variables that can be of interest
including sensible heat gain (Figure 6) and supply airtemperature (Figure 7).
Figure 6. Sensible cooling for the variable zoneheat gain profile
Figure 7. Supply air temperature for the variable
zone heat gain profile
CONCLUSION
Detailed simulations of building control systems canbe useful for designing, refining and verifying theperformance of the systems under realistic operating
conditions, provided that sufficient information isavailable to accurately model the components.
Although not a simple task, evaluating the effects oftime constants, thermal storage, hysteresis and heatloss can identify performance issues that otherwisemight be overlooked. However, simulation tools
with these capabilities are generally lacking,particularly for building control system designers.The current emphasis on simulation tooldevelopment may help to overcome this barrier andenable more realistic performance evaluations ofbuilding control systems, thereby promoting the
design of better building control systems and moreenergy efficient and comfortable buildings.
REFERENCES
Cui, P., Yang, H., Spitler, J., & Fang, Z. (2008).
Simulation of hybrid ground-coupled heat pump
Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.
- 2550 -
-
7/30/2019 Model Based Control
8/8
with domestic hot water heating systems using
HVACSIM+. Energy and Buildings, 40(9),
1731-1736.
Ellis, P. G., Torcellini, P. A., & Crawley, D. B.
(2007). Simulation of energy management
systems in EnergyPlus. Building Simulation
2007 (pp. 1346-1353).
Haves, P., Norford, L., DeSimone, M. (1998). A
Standard Simulation Test Bed for the Evaluation
of Control Algorithms and Strategies. ASHRAE
Transaction, 104(1), 1-14.
Jreijiry, D., Husaunndee, A., Inard, C., & Villenave,
J. G. (2003). Control of ventilation in buildings
using SIMBAD building and hvac toolbox.
Eighth international IBPSA conference (pp. 591-
598).
Park, C. (2008). HVACSIM+ User s Guide Update.
Building (pp. 1-108). Retrieved fromhttp://www.fire.nist.gov/bfrlpubs/build08/PDF/b
08030.pdf.
Riederer, P. (2005). Matlab/Simulink for building
and HVAC simulation - State of the art. Ninth
International IBPSA Conference (pp. 1019-
1026).
Sahlin, P., Bring, A., & Eriksson, L. (2009). Real
controllers in the context of full-building, whole-
year simulation. Eleventh International IBPSA
Conference (pp. 72-79).
Trcka, M., Hensen, J., & Wetter, M. (2009). Co-
simulation of innovative integrated HVACsystems in buildings. Journal of Building
Performance Simulation, 2(3), 209-230.
Trka, M., & Hensen, J. L. M. (2010). Overview of
HVAC system simulation. Automation in
Construction, 19(2), 93-99.
Wetter, M., & Haves, P. (2008). A modular building
controls virtual test bed for the integration of
heterogeneous systems. Third National
Conference of IBPSA-USA (pp. 69-76).
Wetter, M. (2009). Modelica-based modelling and
simulation to support research and development
in building energy and control systems. Journal
of Building Performance Simulation, 2(2), 143-
161.
Proceedings of Building Simulation 2011:
12th Conference of International Building Performance Simulation Association, Sydney, 14-16 November.
- 2551 -