R.Yang,L.Wang/SustainableCitiesandSociety6(2013)16–21

17

Eachzoneagentisresponsibleforenergymanagementinitsspeci czone.Zoneagentcommunicateswithoccupantsanddeter-minestheamountofpowerthatshouldbedispatchedtoeachlocalcontroller-agentthroughasuitablecontrolalgorithm.Inaddition,thezoneagentcollectstheenvironmentalparametersfromthelocalcontroller-agentstoanalyzethecomfortconditionofthiszoneandreportsittothecentralagent.

Ineachzone,threetypesoflocalcontroller-agentsincludinglocaltemperaturecontroller-agent,localilluminationcontroller-agentandlocalairqualitycontroller-agentareutilizedtocontrolthethermalcomfort,thevisualcomfortandtheIAQcom-fort,respectively.Theyaredistributedlocallyineachzoneformonitoringandcontrollingthebuildingenvironmentthroughcor-respondingsensorsandactuators,andreturningtheenvironmentalparameterstothezoneagents.Toimplementcontrolinthephysi-calenvironment,heaters,airconditioners,electricallights,blinds,andventilatorsareutilizedasactuatorstocontrolthezoneenvi-ronment.

3.Mathematicalmodelandalgorithmforagents

3.1.Centralagent

Fig.1.Multi-agentcontrolarchitecture.

2.Agent-basedsystemformulti-zonebuildingcontrol

2.1.Agentsandthemulti-agentsystem

Agentsinamulti-agentsystemaredesignedtoaccomplishcom-plextasksinacollaborativefashion.Inamulti-agentsystem,toattainthecapabilityofseamlesscooperationbetweenagents,theyrequireabilitiessuchascollaborationbetweenindividuals,coordi-nationofactions,andresolutionofcon ictsbythemselves(Ferber,1999;Qiao,Liu,&Guy,2006).Herecollaborationdealswiththedistributionoftasksandresourcesamongagents.Coordinationisconcernedaboutorganizingtheactionsofdifferentagentsintimeandspace.Inaddition,resolutionofcon ictsusuallyrequiresnego-tiationtechniquestoenhancethesystem’sperformance.

Thecentralagentaimsatmaximizingthebuilding’soverallcom-fortwithacertainamountofenergysuppliedbythedistributedrenewableenergysources.Throughahuman–computerinterface,thecentralagentreceivesthemanager’sinstructionabouteachzone’spriority,whichismathematicallyrepresentedasaweight-ingcoef cient.Thecentralagentwillmakedecisionsregardingpowerdistributiontoeachzone.Thecontrolobjectiveistomaxi-mizetheoverallcomfortofthebuildingandthecontrolvariablesareafunctionofpowerdistributedtoeachzoneP(i).Aparticleswarmoptimizerisintegratedintothecentralagenttooptimizethecontrolsystem.The tnessfunctionis:

OverallComfort=

n i=1

w(i)comfort(i)

(1)

2.2.Architectureofthedesignedmulti-agentsystem

subjectto:

P(i)≤Pdemand(i)i=1,...,n

(2)(3)

Thecontrolsystemforamulti-zonebuildingmodelmainlyfocusesonmaximizingtheoccupants’comfortwhileminimizingtheenergyconsumption.Itresolvesthecon ictbetweentheenergyconsumptionandtheoccupants’comfort,aswellasthecon ictbetweendifferentzones’demands.Thesystemobtainsoccupant’spreferenceandmakesdecisionstodeterminetheenergydispatchschemeforallthezones.Fig.1showsthedesignedmulti-agentarchitectureforcontrollingamulti-zonebuilding.Theproposedmulti-agentsystemcomprisesthreetypesofagents:thecentralagent,zoneagentsandlocal-controlleragents.

Thecentralagenthastwomajorfunctions:communicationanddecisionmaking.Thecentralagentcommunicateswiththebuild-ingmanagertocollectrelevantinformationandhelpsthemanagerinmonitoringthebuildingoperationsthroughahuman–computerinterface.Thecentralagentalsoneedstocommunicatewiththezoneagentsandtheenergyresourcestoobtainthestatusofeachzoneandpowersupplyforinformeddecision-making.Thecentralagentmakescorrespondingdecisionsbasedoninformationcol-lectedandsendsthedecisionsbacktothezoneagentsandthebuildingmanager.

n i=1

P(i)≤Pgiven

whereOverallComfortistheoverallcomfortofthebuilding;nisthenumberofzonesinthebuilding;w(i)istheweightingcoef -cientforzonei,whichisprede nedbythebuildingmanagerandthesummationofalltheweightingcoef cientsequalsto1;com-fort(i)isthecomfortvalueinzonei,whichisprovidedbyeachzoneagent;P(i)isthepowerdistributedtozonei;Pdemand(i)isthepowerdemandofzonei.Pgivenistheavailablepowersupplyfromdistributedenergysources.

3.2.Zoneagentandlocalcontroller-agents

Thecomfortvalueofazoneiscalculatedbyitszoneagentviaanalyzingtheenvironmentalparametersandtheoccupants’prefer-ences.Thecomfortvalueofzoneiisrepresentedasfollows(Wang,Yang,&Wang,2011):

comforti=

ti[1 |(Ti Tseti)/Tseti|]+ li[1 |(Li Lseti)/Lseti|]+ ai[1 |(Ai Aseti)/Ai|]

ifAi>Aseti

ti[1 |(Ti Tseti)/Tseti|]+ li[1 |(Li Lset)/Lseti|]+ ai

ifAi≤Aseti

(4)