NIFS 7 (2001),l,10-21

Multi-objective Optimisation in Air-Conditioning Systems: Comfort/Discomfort Definition by IF Sets

Plamen Angelov
Building ServicesEngineering ResearchGroup Departmentof Civil and Building Engineering, UK Loughborough University, Loughborough LEl l 3TU, Leicestershire, fax: +44 (1509) 223981 e-mail: P.P.Angelov@Lboro.ac.UK tel: +44 (1509) 223774

The problem of multi-objective optimisation of air-conditioning (AC) systemsis treated in the Abstract: paper in the framework of intuitionistic fvzzy (lF) set theory. The nature of the problem is multi-objective one with requirements for minimal costs (generally, life cycle costs; more specifically, energy costs) and maximal occupants' comfort (minimal discomforl). Moreover, its definition by conventional means is bounded to a number of restrictions and assumptions,which are often far from the real-life situations. Affempts have been made to formulate and solve this problem by means of the fuzzy optimisation [4]. The present paper makes further step by exploring the innovative concept of IF sets [6] into definition of the trickiest issue: comfort and discomforr definition. The new approach allows to formulate more precisely the problem which compromises energy saving and thermal comfort satisfactionunder given constraints.The resulting IF optimisation problem analytically Il]-t21. An example illustratesthe viability could be solved numerically or, under some assumptions, of the proposed approach.A solution which significantly (with 3S%) improves comfort is found which is more energeticallyexpensive with only 0.6Yo.This illustratesthe possibility to use the approachfor trade-off analysis in multi-objective optimisation of AC systems. IF sets, IF optimisation, multi-objective optimisation, air-conditioning systems, Keywords: comfort/discomfort

1. Introduction In the last few decadesthere are an increasinginterestin the optimum design and control of AC systems I4l, t9l-t191. Besides of economical efficiency, it has ecological aspects lowering the "green house" effect, fuel consumption and, hence, reducing emission of pollutants while keeping occupants' comfort U 01, [22]. The aim is to design and control AC system such that to minimise the life cycle costs, including capital, operational costs (mainly energy consumption),maintenanceand commissioningcostsetc. while minimising the feeling of discomfort of occupantsand keeping as maximum as possible productive environmental conditions. Solution of this problem is usually based on random search or gradient-based applicationof GA to this problem has been techniques[9]-[0]. In last few years successful reported[5], [9]. The specific of the AC systems,by differ from most other technical systems,is the fact that human and its comfort are in the centre of the problem and, as a result, control and optimisation objectives as well as some constraints are highly subjective, ill defined and So far as the laws of mathematics refer to reality, they are not certain. And so far Albert Ehstein they are certain, they do not refer to reality. '

different factors like clothing, metabolism, air movements,occupants'activity, solar radiation etc. have been seriously studied [15]-[20]. Human comfort is a complex non-linear, timevarying function of temperature, humidity, airflow rate, clothing level, activity, lighting, shading, noise etc. It is important to mention that the comfort requirements are specffic for each individual or group of individuals and are not constant.The fact that natural temperature could be different in different locations and could vary with cultural and other differences, that occupantscould have different feeling of comfort has beenreported 122l-1231. Current practice is to design and operate AC systemsunder assumptionsof occupants' comfort requirements based on statistical design criteria contained within standards like ISO7730 l2l), known as PPD and PMVr. In reality, the thermal sensationsand needs of occupants vary from simple design standards [23]. Occupants'preferences are subiective, individual, their presencein the occupied zone andlevel of activity is time varying tl5]-[19]. Another important issue in the AC systems control is their energy efficiency. The latest datafrom the International Energy Agency (http:i/www.iea.org) shows that significant part of the primary energy is used in buildings and AC systemaccountfor the major part of it. This problem is increasingly important nowadays, when ecological and economical concurrent requirements (the ozone hole, gas emissions and surging petrol prices, cost effectiveness as a condition for higher competitivenessetc.) are pressing and have to be balanced with the comforl needs.Additional potential for higher energy efficiency exist in finding and keeping more economical set-points or regimes, based on more adequate representationof humans'feelings of comfort and discomfort, of keeping a really productive environment[15]. It could lead to a twofold improvement: ./ more customised,specific user-orientedAC system; ,/ higher energy efficiency. A general formulation of the optimisation problem of an AC system in the framework of IF setstheory is proposedin the presentpaper.IF setshave been introducedrecently [6] as an extension and generalisationof the fuzzy set concept. They consider not only the degree of membershipto a given set, but also the degreeof non-membershipsuch that the sum of both values is less than one [7]. Due to the symmetrical treatment of objectives and constraints this multi-objective problem is transformed to a non-fuzzy (crisp, conventional) one [l-2], that in some casescould be, solved analytically I25l.The building thermal model 124l-1261 plant models are not consideredin this presentationin order to simplify the problem. and An example of a multi-objective IF optimisation problem illustrates the viability of the proposedapproach.A solution, which significantly improves comfort and reducesdiscomfort feeling, is found. This illustratesthe possibility to use IF optimisation for trade-off analysisin multi-objectiveoptimisationof AC systems. 2. Some definitions over IF sets IF set- basic definition [6]: Definition I The IF set A is a mappingR + [0;l]x[0;l], which is definedby a pair <pra(x);va(x)>

t6l
where pra(x)denotesthe degreeof membershipof x to the set A; va(x) is the degreeof non-membership(rejection) of x to A; pn(x) + vo(x) < 1.

' from Predicted Percentage Dissatisfiedpeople and PredictedMean Vote respectively of

1l

It is interestingto note, that A becomea conventionalfuzzy set,when pu(x)+vu(x) :1. 2.1 Crispification: de-fuzzification over IF sets[3J Crisptfication operatorshave been introducedby the author in [3] as a mapping [0;1]x[0;l] + R in analogy to de-fuzzificationoperationover conventionalfuzzy sets.The for result of this operation is a crisp value which is representative the given IF set as a whole. final control action (fuzzy or IF set This operation is necessaryin controllers to derive the point can not be given to the servo in a control system)which will be realised.It is necessary also in decision making and expert systemsfor elicitation of information. A brief summary of definitions of four crispification operatorsand their main propertiesare presentedbellow. Definition 2a l3l The centre-of-gravity (CorC)crispification operationover IF setsis defined as follows
0 *lF -t=l coG

-v(x,))x, Z1t@,) -v(x,)) ZUt@,) j=l

N

N : card(x); for p(x)> v(x)

(1)

This operator's definition is based on the difference between the degree of membershipand non-membership(rejection),which have to be positive. Definition 2b l3l The mean-of-maxima(MoM) crispification operatoris defined in analogy to the MoM de-fuzzification operatorover (p(xi) -v(x')).
I.

:t, *,,,,_ru, 2c Definition l3l

>'-t,
;

xr:{* | p(*')-v(*r):

N
Tg"

(pr("i)-v(xJ)}

Q)

Mean-of Equilibria (MoE) is introduced as a new operator in [3], which differ from de-fuzzification of conventional fuzzy sets). It averagesall points with maximal p(x1) (the points) and thesewith minimal v(xi) (the lessnon-acceptable): most acceptable

, MoE i=l * ?. .. _ 2M

Z*|' >.:
+ i=t 2K

(3)

p(*)}; x*:{xl tt6*) - Mmax,':'

xo:{*l({:*min,':'

,(*)}

Definition2d It for is of Finally, an analogue BADD operator introduced crispification. is over the p(x')-v(x') : difference
x9-_BADD : IF o,,-,.

-v(x,))" x, ZUt@,)
l=l

-v(x,))" ZUt@,)

for p(x) > v(x)

(s)
Basic featuresof theseoperatorsare similar with theseof their analogies: The CoG crispification operator gives all possible solutions in which the degree of (p(xi)>v(*i)). However, it averages acceptanceis higher than the degree of non-acceptance good and poor solutions (although it gives them different weights). The MoM and MoE crispification operationsgive information about the best solution(s) (MoE gives it in the sense However they ignore of higher degree of acceptanceand lower degree of non-acceptance).

t2

information about the other possible solutions. The BADD crispification operator has the analogousproperty as BADD de-fuzzification operator: - it implies CoG cispificationoperatorfor cr:l (it implies CoG, if v:0 also); - it approaches MoM crispification operator for o + c and approaches MoM crispification when v:0 also; - it implies middle average(MA) when cr:O: IF_BADD(a: I ; v> 0) : IF_CoG IF_BADD(a: I ; v:0) : IF_CoG IF_BADD(a + a;v>0) : IF MoM IF_BADD(q -+ *;v:0) : IF_MoM IF_BADD(a: 0) : IF_MA a/v

0 I

0 IF_CoG IF_MoM
IF-MA

Positive

IF_CoG IF MoM

Table I Crisptfication definitions t3] in combination with the conjunction and disjunction operationsover IF sets [7] allow to developnew engineeringapplicationsof IF setssuch as IF optimisation, IF decision support system,IF controllers,IF expert systemsetc. 2.2. IF optimisationf ll-[2] The concept of IF optimisationproblem is introducedand investigatedin tll-t2l by the author. Ideas about IF sets and optimisation and decision making have been also considered in [7]-[8]. The original definition of tll-t2l and the approachproposedthere for solving the IF optimisation problem will be briefly presentedhere, as it will be used in AC system optimisation problem later. The benefit of the IF optimisation problems is twofold: they give richest apparatus for formulation of optimisation problems and, on the other hand, the solution of IF optimisation problem can satisfy the objective(s) with higher degreethan the analogousfuzzy optimisation problem and the crisp one. Quite often, it is difficult to describethe constraintsof an optimisation problem by crisp relation's (equalities andlor non-equalities)[35]. Practically, a small violation of a given constraint is admissible and it can lead to a more efficient solution of the real problem [35]. Objective formulation represents,in fact, a subjective estimation of a possible effect of a given value of the objective function. Fuzzy optimisation formulations f29,32,34f are more flexible and allow finding solutions,which are more adequate the real problem. One of the to poorly studied problems in this field is definition of degreesof membership and rejection (non-membership)[30,36]. The principles of fuzzy optimisation problems are critically studied in pal-126]. However, mainly the transformationsand the solution procedures1231, Applying IF setsconceptit is possibleto reformulate 1251, [31], [33] have been investigated. optimisation problem using degreesof rejection of constraintsand values of the objective, which are non-admissible. In general, an optimisation problem includes objective(s) and constraints. In fuzzy optimisation problems (fuzzy mathematical programming [35], fuzzy optimal control [28], linear programming with fuzzy parameters[32] etc.) the objective(s) and/or constraints or parametersand relations are represented fuzzy sets.Thesefuzzy sets explain the degreeof by satisfactionof the respectivecondition and are expressed their membershipfunctions [36]. by Let us consideran optimisationproblem: (6) i: I ,...,p f,(x)---> min

l3

subjectto where x f;(x) g:(x) p q

i:I,...,q Sj6)<0 denotesunknowns denotesobjective functions denotesconstraints(non-equalities) denotesthe number of objectives denotesthe number of constraints

The solution of this crisp optimisation problem satisfiesall constraintsexactly. In the analogousfvzzy optimisation problem (7) the degreeof satisfactionof objective(s) AND of is constraints maximised(8). i:I,...,p Q) f,(x) --+ frin to subject where ftin gi6) 7 0 denotesfuzzy minimisation i:I,...,q

It is transformed via Bellman-Zadeh's approach [27] to the following crisp (nonfuzzy) optimi sation problem: To maximise the degree of membership (acceptance)of the objective(s)AND constraintsto the respectivefuzzy sets: (8) i:I,...,p+q xeR' maxpi6) subjectto 0Strti@)<I

of degreeof acceptance x to the respectivefuzzy sets where p1(x)denotes In the case when the degreeof rejection (non-membership)is defined simultaneously with the degree of acceptance (membership) and when both these degrees are not complementary, IF sets can be used as a more general and reach tool for describing this uncertainty [6]. It is possible to representdeeply existing nuances in problem formulation defining objective(s) and constraints(or part of them) by IF sets,i.e. by pairs of membership (pi(x)) and rejection (vi(x)) functions. An IF optimisation problem is formulated as follows

AND constraints of the degreeof acceptance lF objective(s) To maximise AND of of the ANDto minimise degree rejection lF objective(s) constraints.
max, x{lt,(x)} min,r{vi(x)} to subject v,(x)>,_0 It{x) >,_ v,(x) F,6) + v,(x)<,_ I where Fi(x) vt(x) crisp. as IF optimisation problem such as fuzzy optimisation problems can be represented a two-stage process which includes aggregation of constraints and objective(s) and defuzzifrcation (maximisation of aggregated function) I24, 261. Usually applied BellmanZadeh's approach [27] for fuzzy optimisation problem solving realises min-aggregator. Conjunctionof IF setsis definedas [6]: xeRn i : I,...,P+q i:I,...,p+q i: I,...,P+ql -- I i ,...,p+q -- I,...,p+ql i (9)

[3]:

denotesdegreeof membershipof x to the i-th IF set denotesdegreeof rejection of x from the i-th IF set

It is possible that a part of constraintsand objective(s) are IF and other are fuzzy or

t4
I

where G C

G ,tC : { <x, ltc6) n pc(x), vc6) ctvr(x)>, xeRn} denotesan IF objective (gain) denotesan IF constraint

(10)

This operatorcan be easily generalised and applied to IF optimisation problem: D: {<x, lto(x), vo6)>, xeRn}
p+q

(11)

lto:)lt
i:l

v D :f.1, , y,
where D denotesthe IF set of the decision
p*.s

Min-aggregatorcould be used for conjunction and max-operatorfor disjunction:

ItD :

\i! lt,
q

x e R'

(12)

ltn S lti

vD:^,?f''

xeR'

vo2v, It can be transformedto the following systemof equations: e 1,_ lti(x) i:I,...,p+q B>,_ v,(x) i:I,...,p+ql

(13)

a > , _p p>,_0 a ),-0 a+ p<,_l

where a denotesthe minimal acceptable degreeof objective(s)and constraints B denotesthe maximal degreeof rejection of objective(s)and constraints Finally IF optimisation problem can be transformed tll-t2l to the following crisp (non-fuzzy) optimisation problem which can be easily solved numerically by simplex, gradient-based genetic algorithms [5] or, in somecases[25] analyically: or (cr-p) -> max (14) subject to 0 (,_ pri(x) i: 1,...,p+q i:l,...,p+q B >,_ vi(x) cr >r- B

B>,_0 cr+B(,_1
In general,the solution of IF optimisation problem is different than the solution of the analogousfuzzy problem and the degreesof satisfactionof a given objective or constraint in an IF optimisation problem can be higher or lower. It depends on the formulation of the respectivefunctions of acceptance and rejection. This property of the IF optimisation problem is analogousto the property of fuzzy optimisation problems when in some cases(it depends on the membership function) the solution can satisfy the objective function better, but the price for this is the worse satisfaction of some constraints.A practical example of multiobjective optimisation of AC systemsis consideredin the next section.

l5
I

I
3. optimisation of AC systemsas an IF optimisation problem 3.1 AC optimisation- an multi-objectiveproblem In general, the problem of AC system optimisation is formulated as a multi-objective optimisation, which aims to minimise costs (life-cycle costs,maintenance, commissioning, capital and running - mainly energy costs) and to maximise the comfort (minimise discomfort) subject to number of constraints t4l. Building and thermal plant models are consideredas constraintsas well as different physical and technical limitations over different componentsof the AC system(fans, coils, heaters,air-handling units etc.): Jt: Costs (AC systemvariqbles: T, RH, m, ...) -+ min (1s) J2: Comfort (AC systemvariables: T, RH, m, ...) -> max J3: Discomfort (AC systemvariables: T, RH, m, ...) + min Subjectto where Constraints (AC systemvariables: T, RH, m, ...) : 0 RH denotesrelative humi dity, o/o m denotesmassflow rate,kg/s

Practically, the optimisation problem is solved as an uni-objective one where comfort requirements considered a constraint[4]-[5], t9l-tl0l. are as : Costs (I, RH, m, ...) + min J (t 6) I PPD (f, RH, m, ...) | < I0% Constraints (f, RH, m, ...):0 This simplification makes the problem a rough approximation of the real intention of the designer.It is so, if conventionaltechniqueis applied, becauseit simplifies the solution of the problem. Fuzzy sets and IF sets, specifically, offer a tool for coming to grip with this problem. Discomfort, dissotisfaction as well as comfort, which is normally a more restrictive category, could be formulated with an appropriate, subjectively specified membership function (Fig.2). The problem of AC systemoptimisation becamefuzzy multi-objective one: Costs(f, RH, m, ...) -+ min mrn -)max
i pcosts I Vdiscomfort i pcomfort I /lconstraints

Subjectto

(r7)

Discomfort (f, RH,m, ...) + Comfort (I, RH, m, ...) Subjectto Constraints (f, RH, m, ...) z 0

where - denotesfuzzy It should be noted that constraintsand costs could be fuzzy, IF or crisp. We consider fuzzy description for them without loss of generality for the sake of simplification of the expressions.Building model (which could be of the fuzzy rule-basetype) as well as models of componentsof the thermal systemare not considereddue to simplicity. 3.2 Partial Load Ratio definition by IF sets Formulations of constraintswith IF sets allow a higher flexibility to describe better the really existing uncertainty. For example, fans and coils are used with a part of their design load range: the loads lower than given thresholdsare not so effective. ParameterPartial Load Ratio (PLR) representsthe load used as a ratio to the designedload. Practically, loads lower than given threshold are not practically used (dashedline on the Fig.l). Very often, plBcritical is 20oh of the design load (Lo"'tn). Fuzzy constraintsover the fan's load (bold line p; on Fig.l ) give additional flexibility in description. According to this fuzzy constraint, PLR in a range of [0.15; 0.25] are acceptable, but with degree less than 1. Introduction of a non-

I6

membership (rejection) function, however, (bold line v1 on Fig.l) enhanceseven further this flexibility: combinationof prt andvsdefines an IF set over PLR (Fig.1). Its meaningis that:

the rate of rejection (non-acceptance) decreasingsteeplyuntil 20% of the I.desisn' not fully acceptable(acceptablewith somereservations)as pr1<I. The last zone makes the main distinction with the conventional fuzzy sets and as it is seen it enriches the options for description of uncertainties,which exists in practice and the real life*.

Fig.l Partial load ratio constraint 3.2 Comfort/discom-fort defintionby IF sets Definition of comfort/discomfortrequirements(normally treated by statistical categories of PPD and PMV) could be described more realistically using IF sets. A typical case of divergencein comfort and discomfort (which is not simply a complementaryto the comfort) is given on Fig.2. Similar representationscould be made for different physical variables describing human comfort and discomfort (T, relative humidity, lighting, shading, noise, air quality, solar radiation, air movement etc.). The meaning of this graphic is: conditions are definitely uncomfortable;

This representation is more natural, because people tent to be more pretentious, demanding and easier say 'it is uncomfortable', even if, certainphysi cal variable under considerationlay inside the pre-scribedby ISO standardband. One possible reasonis that some other parameter of the internal climate is out of the range of the comfort zone. Another is the individual pessimism,conservatismand demands.

Fig.2 IF descriptionof comforI/discomfort

t7

(l - conventional case;2 - comforl description; 3 - discomfort description) From the Fig. 2 one could also seethat for this particular example of thermal comfort, slightly higher temperatures, which are still in the standardcomfort band (25-26 oC) are with higher degree of rejection v1 (dissatisfoction, discomfort) than slightly lower temperatures (22-23 oC). That means that the non-membership (discomfort) function could be nonsymmetrical. The practical meaning is that this person(s) tend to prefer slightly lower temperaturethan normal to slightly higher. Such degreeof flexibility and freedom is difficult, if not possible,to have even using descriptionwith conventionalfvzzy sets.In practice, when statistical measuresare used instead of fuzzy and IF sets, such nuancesare simply ignored. Our point is that using more complex and realistic description possesspotential for much more customisedand effective systemsto be build in engineeringpractice increasingmachine intelligencequotient. 4. Solving Multi-objective IF optimisation problem of AC system control Solution of the optimisationproblem (17) could be found numerically Il-2].For some special casesit could be found analyically l25l.We considera simple example of an AC systemoptimisation in order to illustrate the proposedapproach. Supposethat cumulative costs (capital and running) dependson PPD as follows (Fig.3) [4]:

I Costs:9.9+ IPPDI

(18)

and suppose for simplicity that constraints are satisfied for all variables considered, the following optimal solution is achieved: Costso: I0 kt (l e) PPDO: TI O% Practically, the dependence Costs on PPD is more complex and includes building and of plant models often it is dynamic problem [4]-[5], t9l-ll0l. Variables are usually control parametersof the plant (supply air temperature,mass flow rate etc.). However, the nature of the problem is the same. If apply conventional optimisation problem (15) comfort characteristics will be in the required range,but on one of its marginal values (+ or - 10% dissatisfied). Supposethat comfort and discomfort are describedin a similar way as shown on Fig.2 (with the only difference that instead of T and marginal values of 22 and 26 oC the dependence in respectto PPD with tl\% dissatisfaction boundaryconditions). is as
Lrcosts/comfort 1

PPD)2;I PPD l<r0% Ir - ro.l

1 0 . . . . . . . . . . . . . .I.P.P.D I > 1 0 % . . ;
Costsfunction

(20)

C+l

u; o o

11 1 0 .8 1 0 .6 10.4 10.2 10 9.8 9.6 9.4

9 T q o N $ ( o 3 PPD,o/o

Fig.3 Costsas a function of PPD

18

Supposethat costs are given with the following membershipfunction [4]:

I . . . . . . . . . . . . . . . . . . .C.o.s t < 1 0k f . . ; s > 0.ll(9.9 Costs); Costs 10kf

: Costs 9.9+ I

(2r)

I PPDI
Costsfunction
1.2 I

#oe
= 04 0 .2 0
r

8,0'o

Fig.4 Costsfunction (conventionalfuzzy set) Then the following IF multi-objective optimisation problem arise [4]: -> Fresutt max
Vdiscomfort min + subject to presutt : FcostsAND pcomfort AND pconstraints

(22)

Its solution could be found analytically for some cases[25] from: (PPD) : min(trtromfortgPD,vdir"o.for) trt"os, Q3) P P D " + I 0 I P P D I- 1 0 0 : 0 (24) Note that objective function is a conventional fuzzy set, comfort/discomfort are IF sets and other constraints could be.eitherfuzzy,IF or crisp sets. It is supposedfor simplicity that constraintsare satisfied for all values of consideredvariables. The fuzzy optimal solution is: : Costsoq' f, l0 060 (-0.6%) (2s) : !4.2% (+3Bo%comparingto the standard10% dissatisfaction) PPDopt It is seenthat significant improvement with more than ll3 of the comfort requirementscould be achieved with marginally higher (with only 0.6% or f60) costs. For the more complex (and realistic) IF optimisation problems the solution could be found numerically using gradient-based or geneticalgorithms [5]. 5. Conclusions Optimisation of air-conditioning systems is treated in the paper in the framework of IF optimisation. This new approach allows to formulate more precisely the problem, which compromises energy saving and thermal comfort/discomfort satisfaction under given constraints. This fuzzy optimisation problem is solved analytically under some assumptions. IF optimisation of AC systemsallow us to formulate problem more flexibly, user-friendly and to analysemore precisely and deeply the whole range of possible solutions.It helps us to avoid necessity of scaling and normalisation applied in other approaches(when penalty functions are used) as well as infeasible solutions (infeasible solutions has degreeof rejection I and are simply not consideredin the final result). The problem is formulated in more natural IF multi-objective fashion and is solved efficiently. This is possible analytically in some casesand either numerically in general. A practical engineering example illustrates the viability of the proposed approach. More customised solution (significantly improving comfort - with more than 38%) is found for the marginally higher energy expenseof f60 or 0.6% more. In general, it is possible to find less

t9

energetically expensivesolutions, which in the sametime are more customised.This, however, dependson the specific preferencesof occupantsand could not be guaranteedfor all practical cases. 6. Acknowledgenent

The authorwould like to thank EPSRCfor the financial supporttrough the grant GNM27299. 7. References

in c Fwzy Sets and Systems,86, [] AngelovP., Optimization an Intuitionisti Fvzzy Environment, 299-306.1997 Noteson IntuitionisticFuzzySets,1,27-33, 1995 [2] AngelovP., IntuitionisticFuzzyOptimization, Angelov P., CrispiJication: Defuzzifrcation over IntuitionisticFuzzy Sets,BUSEFAL,64,5l[3] 55,1995 [4] Angelov P., A Fuzzy Approachto Multi-Objective Optimization of Building Thermal Systems, Proc. ofthe 8'oIFSA WorldCongress, Taiwan,Taipei,17-19 August 1999,I,429-434 [5] Angelov P., Evolving Fuzzy Rule-basedModels, Journal of CIIE, special issue on Industrial Application of Soft Computing,17, 2459-468,2000,ISSN l0I7 -0669 K., FuzzySets, FuzzySets Systems,20,87-96, and 1986 [6] Atanassov Intuitionistic AtanassovK., Ideasfor Intuitionistic Fuzzy Equations,Inequalitiesand Optimisation,Notes on [7] IntuitionisticFuzzySets,l,l7-24, 1995 Bustince H., Handling Multicriteria Fuzzy Decision Making ProblemsBasedon Intuitionistic [8] FuzzySets,Noteson IntuitionisticFuzzySets (l) (1995)42-47 | Wrieht J.A., HVAC Optimisation Studies: Sizing by GA, Building SystemsEngineering l9l Reseqrch Technologt,17 (1996)7-14 and (1) 0]Ren M.J., Optimal Predictive Control of Thermal Storagein Hollow Core Ventilated Slab [ Systems, h.D.Thesrs, P Loughborough University(1997) Kajl S., P.Malinowski,E.Czogala, M.Blazinski, Flzzy logic and NN approach thermal to [11] description of buildings, Proc. of the 3th European Congress on Fuzzy and Intelligent Techniques EUFIT'9S, Aachen, Germanyv.l (1995) 299-303 SkrjancI. et.al.,Flzzy Modeling of Thermal Behavior in Building, Proc. of the 6th European [2] Congress IntelligentTechniques on EUFIT'98,Aachen, Germany, (1998)878-882 2 LovedayD.L., G.S.Virk, Artificial Intelligence Buildings, for lpplied Energt,4l, 1992,201[13] 221 [ 4] Culp C.et al.,The Impact of AI Technology within the FIVAC IndustryISHRAE JournaI,l2,l 990,I2-22 M. Fuzzy GeneticMulati-agent [15]HagrasH., V. Callaghan, Colley, G. Clarke,A Hierarchical Architecture for Intelligent Buildings Sensingand Control, Conferenceon Advancesin Soft Computing, Leicester, June29-302000 SharplesS., V. Callaghan, Clarke, A Multiagent Architecturefor Intelligent Building G. [6] Sensing Control,Sensor and Review,19(2) 135-140, 1999 [17]CallaghanV., G.Clarke, M.Colley, H.Hagras,A Soft-ComputingDAI Architecture for Intelligent Buildings, Journal of Studies in Fuzzinessand Soft ComputingAgents, Spinger Verlag,July 2000 Proc. 9th [8]Coen M.H., Building Brains for Rooms:DesigningDistributedSoftware Agents, InnovativeApplications ofAI Conference, AAAI Press,1997 [l9]Mozer M., The Neural Network House:An Environmentthat Adapts to Its Inhabitants,Proc. AAAI SpringSymposium IntelligentEnvironments, on 1998, AAAI Press, 10-l 14, I

20

to [20] Garg V., N.K. Bansal,Smart OccupancySensors ReduceEnergy Consumption,Energt and -87 Buildings,32, 2000,8I [21] Tayloa Francis,Moderate ThermalEnvironments Determinationof the PMV and PPD thermalcomfort, EH ISO 7730,(1995),EO BS of indicesandspecification theconditiowfor [22] Roulet C.-A., Indoor Environment Quality in Buildings and its Impact on Outdoor Environment,Energt and Buildings, 33, 2001 I 83- I 9 1 , ThermalComfort:useof Controlsin M.A. Humphreys, RajaI.A., J.F.Nicol, K.J. McCartney, [23] 33, NaturallyVentilated Buildings,EnergyandBuildings, 2001,235-244 [24] P. Angelov, A Generalized Approach to Fuzzy Optimization, International Journal of Intelligent Systems (4) (1994)261-268. 9 [25] P. Angelov, An Anall'tical Method for Solving a Type of Fuzzy Optimization Problems, 3 Controland Cybernetics (1995),to appear [26] P. Angelov, Approximate ReasoningBased Optimization, YugoslavJournal on Operations Research (l) (1994)ll-17. 4 Science17 [27] R. Bellman andL. Zadeh,Decision Making in a Fuzzy Environment,Managetnent (r970)141-160. 47 and [28] D. Filev andP. Angelov,FuzzyOptimalControl, FuzzySets Systems (1992)151-156. J. Kacpnyk, A Generalization of Fuzzy Multistage Decision Making and Control via 129] LinguisticQuantifiers, rnationalJournalofControl3S (1983)1249-1270. lnte 30 An LuhandjulaM., Fvzzy Optimization: Appraisal,Fuzzy Setsand Systerns (1988)257[30] 282. in: An [31] M.R. Maleki and M.Mashinchi, Algorithm for Solvinga Fvzzy Linear Programming, (Seoul, Korea,4-9July 1993)| 641-643. Proc. of V IFSA WorldCongress [32] H. Rommelfanger,Inequality Relations in Fv4r Constraintsand its use in Linear Fuzzy ArtiJicial Intelligence Optimization, in: Verdegay J.L., DelgadoM. Eds: TheInterfoceBetween (VerlagTUV, Rheinland, Koln, 1989)195and Operational Research FuzzyEnvironment in 211. [33] M. Sakawaand H. Yano, An Interactive Fuzzy Satisficing Method for Multiobjective Non Fuzzy Sets and Systems (1989) 22130 Linear ProgrammingProblemswith Fuzzy Parameters, 238. [34] H. Tanakaand K. Asai, Fuzzy Linear ProgrammingProblemswith Fuzzy Numbers,Fuzzy Sets & Systems (1984)l-10. 13 g, H.-J. Zimmermann,Fuzzy MathematicalProgrammin Computersand OperationsResearch [35] 10 (1983) 29r-298 1985) H.-J.Zimmermann, FuzzySetsTheoryandits Application(Kluwer,Dordrecht, [36]

21