A Fuzzy Control System Based on the Human Sensation of Thermal Comfort

8/2/2019 A Fuzzy Control System Based on the Human Sensation of Thermal Comfort

1/6


2/6

and to indicate how the environmental parameters shouldbe combined in order to create optimal thermal comfort.The fuzzy rule base is formulated on the basis of learningFanger's thermal comfort model which is considered as themost important and common used one [2].

This paper is organized as follows. First the problemlimitation of HVAC conventional control strategies isexposed. Then, the design of the thermal comfort fuzzysystem is described and applied to the control of the indoorclimate of a single zone building. Finally, the superiorityan9 the effectiveness of the proposed fuzzy system isverified through computer simulation using MAXAB@and TRNSYSO lgorithms.2. Thermal comfort fuzzy system

Recently, it has been pointed out that controllers thatdirectly regulate human's thermal comfort have advantagesover the conventional thermostatic controller [ l , 3, 4, 91.The main advantages are increased comfort and energysavings. In addition, thermal comfort regulation provides acomfort verification process. Although mathematicalmodels are available to predict the human sensation ofthermal comfort [2, 51, only the air temperature and therelative humidity are controlled in the majority of theconventional residential W A C systems. The thermalcomfort level and the other variables are difficult toquantify and therefore not used in classic controltedhniques. Presently, thermal comfort is ensured by theoccupants who have to adjust the air temperature setpointdepending on their perception of the indoor climate. Thispractice is found to be inadequate to satisfy occupantsdesire to feel thermally comfortable. Occupants sitting nearsunny windows or underneath air conditioning ducts orunder hot and humid conditions will find the HVAC controlstrategy based only on air temperature is not adequate.

Thermal sensation indicator

k dFigure.1 PMV and thermal sensation'Over the past decades, numerous studies of thermalcomfort have been achieved. The widely accepted

mathematical representation of thermal comfort is thepredicted mean vote (PMV) index[2]. This index is a realnumber and comfort conditions are achieved if the PMV

belongs to the [-OS, .51 range 12, 81. Fig. 1 shows thesubjective scale used to describe an occupant's feeling ofwarmth or coolness. However, since the human sensationof thermal comfort is a subjective evaluation that changesaccording to personal preferences, the development of aHVAC control system on the basis of the PMV model hadproven to be impossible [1,4,9]. In fact, all classicaltechniques, including adaptive optimal controllers,requiring a crisp determination of the comfort conditions,are not suitable for handling this problem.

HVAC systemRH

MADu model mTCL"rfThe occupan t perception

Figure 2. TCL-based control of HVAC systemEven if the vagueness and the subjectivity of thermal

comfort are the main obstacles in its implementation inclassical HVAC controllers, fuzzy logic is well suited toevaluate the thermal comfort sensation as a fuzzy concept.The comfort range can be therefore evaluated as a fuzzyrange rather than a crisply defined comfort zone. Presently,the fuzziness is not eliminated with the conventionalHVAC control techniques, it is simply ignored byconsidering them as air temperature control problems. Inthese conditions, the HVAC control system goal is tomaintain a desired air temperature in a given indoor space.However, in everyday life, what is desired is not constantair temperature but constant comfort conditions. The fuzzymodelling of thermal comfort could be of importance in thedesign of such a control system that regulates thermalcomfort levels (TCL) rather than temperature levels. Thecontrol strategy based on comfort criteria will regulate thethermal comfort-influencing factors to provide thermalcomfort in the indoor space. The TCL-based fuzzycontroller establishes the desired setpoint values of theenvironmental variables to be supplied to the HVACsystem and distributed in the building to create acomfortable indoor climate.

\ of the indoor climate

2.1. System architectureThe TCL-based fuzzy system starts with the evaluation

of the indoor thermal comfort level depending on the stateof the six parameters: air temperature (Tu), relative

488


3/6

humidity (RH), air velocity (Vair),mean radiant temperature(Tmrt), the activity level of occupants (MADu) and theirclothing insulation (Id). hen, if the estimated thermalcomfort level is out of the comfort range, the controlalgorithm will provide the air temperature and the air velocitysetpoints that should be supplied to the HVAC system in orderto create indoor thermal comfort. Fig.2 shows the blockdiagram of the HVAC control system based on the TCL.

LO W Medium

TaVairTmrt I I Environmental mode

High

RH

1-1

3s -3.s.sE8

Personal-dependanmodel I

Light Very High High MediumMedium High Medium LowHeavy High Low VeryLowVery heavy Medium Lo w VeryLow

-

MA-

TCLTa-setVair-set-t

Figure 3. TCL-based fuzzy system.Once the TCL is calculated, it is compared to the users

actual thermal sensation in order to improve the fuzzyapproximation of the specific users thermal comfort level(UTCL) on-line.So that, with time the fuzzy model of thermalcomfort sensation exactly matches the specific occupantsactual thermal sensation. The on-line adaptation of thethermal comfort model is justified by the fact that eachoccupant possesses different attributes that will affect his orher thermal comfort due to biological variance. Thisadjustment is realized by changing the fuzzy state of theactivity level of the occupant to take into account his or herspecific metabolic rate.

The proposed architecture of the TCL-based fuzzy controlsystem has two main advantages. It is equipped with an on-line comfort verification process and the possibility of theoccupants-participation in the formation and the definition ofthe comfort range according to their personal-preferences.These characteristics could be of importance in thedevelopment of modern HVAC systems by using thermalcomfort sensors to quantify the users degree of thermalcomfoddiscomfort.2.2. Fuzzy logic system rules

The fuzzy thermal comfort system is composed of threemain subsystems which are interconnected as shown in Fig.3.The personal-dependant model is used to approximate the airtemperature range [Tal, Tu21 around which the users shouldbe in thermal comfort according to the state of the activitylevel and the clothing insulation. This subsystem uses thetriangular membership functions given in Fig.4 to describe theinput and output variables. The fuzzy rule base shown in table1represents the set of fuzzy rules that are activated to evaluatethe optimal temperature range.

The 12 fuzzy rules are expressed such as:0 IF the clothing insulation is Light AND the activity level

is Low THEN he air temperature range should be VeryHigh.

0 IF the clothing insulation is Heavy AND the activity levelis High THEN the air temperature range should be VeryLow.

etc.

1

0.5

Air temperature0 range (QC)2 6 105 11 19s 23 24 27.5Figure.4 Membership functions used in thepersonal-dependant uzzy subsystem.Once the air temperature range is evaluated, it is supplied

to the environmental model to determine the air temperatureand the air velocity setpoints that will create indoor thermalcomfort. The next subsections describe how to derive thesetwo parameters for any combination of the four environmentalvariables.

The air temperature setpoint that will provide indoorthermal comfort is estimated according to the state of the airvelocity, the mean radiant temperature and the relativehumidity. This is realized in two steps. First, the air velocity isused to evaluate the air temperature setpoint for RH=50%.Then, the air temperature setpoint is adjusted to compensateany deviation of the relative humidity from 50%.To this end,

489


4/6

S Th Tm r t and the operative temperature (TuO), which is theoptimal temperature that will create thermal comfort whenRH=50% and Tmrt=Ta, are estimated by using themembership functions and fuzzy terms (Vl, ..,V7), (Tl, ..T7)and (ATl,...Al7) as shown in Fig.5. The fuzzy rule baserepresents the effect of the air velocity and the mean radianttemperature on the necessary air temperature that shouldcreate optimal thermal comfort. For a given air velocity, if themean radiant temperature in a room is altered, e.g. due tochanged outdoor conditions, or to crowding, or because lightsare turned on, a different air temperature setpoint is requiredto maintain the indoor thermal comfort. This statement istransformed into a fuzzy reasoning composed of the followingseven fuzzy rules:0 IF Vuir is V1 THEN Tu0 is T1 and (GThTmrt) s AT10 IF Vuir is V2 THEN Tu0 is T2 and (GThTmrt) s AT2

IF Vair is V3 THEN Tu0 is T3 and (GThTmrt) s AT3IF Vair is V4 THEN Tu0 is T4 and (SThTmrt ) s AT4

0 IF Vair is V5 THEN Tu0 is T5 and (GTfiTmrtj is AT50 IF Vuir is V6 THEN Tu0 is T6 and (GTfiTmrtj is AT60 IF Vair is V7 THEN Tu0 isl7 nd (STfiTmrt) s AT7

d e irst rule can be interpreted as if the air velocity is verylow, the operative temperature is close to Tu1 and an increasein the mean radiant temperature by 1C must be compensatedfor by a decrease of the temperature by 1C. However, therequired air temperature increases and the GThTmrt falls withrising velocity. The defuzzification process is done using thecentre of area method and the air temperature setpoint istherefore calculated as:

Ts e t = Tu0 + ( T m r t - T a O) . ST/GTmrt (1)The air velocity setpoint required to maintain thermal

comfort conditions is evaluated by using the mean airtemperature ( (Ta+Tmrt)R) as the input of the fuzzysubsystem. If the mean air temperature is in the temperaturerange [ Ta l , TU^], then the air velocity may vary between 0.1 -1.5 d s . The same membership functions of Fig.5 are used todescribe the mean air temperature and the velocity setpoint.The fuzzy rule base used to evaluate the air velocity setpointaccorfling to the mean air temperature state is deduced on thebasis of analysing the effect of each of them on the humansensation of thermal comfort. In all seven fuzzy rules areselected and expressed as:0 IF the mean air temperature is TI THEN Vuir is V10 IF the mean air temperature is T2 THEN Vuir is V 20 IF the mean air temperature is T3 THEN Vair is V30 IF the mean air temperature is T4 THEN Vuir is V 40 IF the mean air temperature is T5 THEN Vair is V 50 IF the mean air temperature is T6 THEN Vuir is V60 IF the mean air temperature is T7 THEN Vair is V7

Once the fuzzy rules are evaluated, the air velocity setpointis calculated by using the centre of area method in thedefuzzification step.

Ta2lGT/GTmrt0 1 -08 -0.65 -0.55 -0.5 -035 -03

Figure 5. Membership functions used to evaluatethe optimal air temperature setpoint.The effect of the relative humidity on the air temperature

and the air velocity setpoints is relatively moderate since achange from absolutely dry air (RH=O%) to saturated air(RH=lOO%) can be compensated for by a velocity increase AV=0.1 m/s or a temperature decrease of 1.5 "C. Thesestatements are added to the environmental model to adjust theoutput variables when the relative humidity deviates from50%.3. Simulation results

The TCL-based fuzzy system has been successfully testedfor the control of HVAC system. Simulation results for thecontrol of a single room climate are investigated. Fornumerical simulations, TRNSYS' algorithm which is acommon used tool in the study of the interactions between thethermal environment and buildings and MATLAB@algorithmare used to verify the effectiveness of the proposed thermalcomfort level fuzzy system.

Fig.6 shows the outdoor temperature profile (a Januaryday) and heat gains that the building was subject to. Theseincluded gains due to climatic factors such as solar gains andto lighting and machines. The profiles of the activity level ofthe occupants and their clothing insulation used in thesimulation is given in Fig.7. Fi g3 shows a full day'ssimulation results of the TCL-based fuzzy system whenapplied for heating mode. It shows the air temperaturesetpoint in the top graph, the temperature tracking in thecentre and the thermal comfort level in the bottom graph.These simulations show that the TCL-based fuzzy system is

490


5/6

able to adjust the necessary air temperature setpoint tomaintain the indoor thermal comfort as soon as thepersonal-dependant variables change.

In order to verify the superiority and the effectiveness ofthe proposed thermal comfort fuzzy system, twocommonly used conventional techniques are simulated forthe same indoor and outdoor conditions: night setback andconstant setpoint thermostat systems. For night setback, thethermostat setpoint was simulated at 70 F (21.1 C) from 6a.m. to 10 p.m. and at 60F (15.6 C) from 10p.m. to 6 a.m(Fig 9). On the other side, the thermostat setpoint wassimulated at 70F (21.1 C) for constant setpoint system(Fig.IO). The air temperature tracking and the resultedthermal comfort level of the occupants versus the hour ofthe day are given in figures 9 and 10.

For comparison purposes, the performance of the threeHVAC control systems are studied. Table 2 gives thenumber of hours-per-day in which the occupants arecomfortable (the thermal comfort level is in the [-OS,.51comfort range), the energy consumption and the percentageenergy savings for each of the above-studied systems. Theheating energy consumption is calculated by the integral ofthe simulated system energy demand when operating in theheating mode. Savings in energy consumption wheredetermined by subtracting the totals obtained with the TCLfuzzy system and the night setback system from the totalsobtained with the constant setpoint thermostat.

Number of comfort hours Energy[per-dayl consumptionConstant setpointthermostat

Night setbackthermostatTCL- basedfuzzy system

5 hours a day

10 hours a day24 hours a day

8.25 kWh

7.10 kWh6.6 kWh

Savings

14%20 %

~~ ____ ~~

Table 2. Performances of the three HVACcontrol systems for one day simulations.This study shows that the TCL-based fuzzy control

system provides better thermal comfort of the users withthe possibility of energy savings. However, the nightsetback technique provides energy savings at the expenseof the occupant's thermal comfort since the thermalsensation is out of the comfort range 10 hours per a day.The TCL fuzzy system is able to maintain the thermalcomfort level in the comfort range during all the day evenif the consumed energy is lower than that required by theclassical techniques. Finally, it is important to note that thethermal comfort level is out of the comfort zone when theconstant thermostat setpoint is used.

4. ConclusionsA new HVAC control strategy that regulates indoor

thermal comfort levels is presented. Fuzzy logic is appliedto the evaluation of the air velocity and the air temperaturesetpoints that should be supplied to the HVAC system inorder to create indoor thermal comfort. The design of thethermal comfort-based fuzzy system is realized byextracting knowledge from Fanger's thermal comfortmodel. The architecture of the proposed control systemallows easier evaluation of the indoor climate by usinglinguistic description of the thermal comfort sensationwhich make it simpler to understand and to process thanhaving to solve iteratively a complex mathematical model.The simulation results show that the control based onthermal comfort levels provides better comfort at a lowercost than that provided by thermostatic control techniques.AcknowledgmentsThis research has been supported by FCAR and Hydro-QuCbec. We thank Franqois Michaud for his contributionsto the development of the fuzzy thermal comfort model.References[l] C. H. Culp, Rhodes, M. L. Krafthefer, B.C. and M. Listvan"Silicon infrared sensors for thermal comfort and control",ASHRAE Journal, p. 38-42, April 1993.[2] P. 0. Fanger, Thermal Comfort Analysis and Applications inEnvironmental Engineering, McGraw Hill, 1972.[3] C. Federspiel, "User-adaptable and minimum-power thermalcomfort control", Ph.D. thesis, MIT, Department of

Mechanical Engineering, June 1992.[4] S . Funakoshi and K. Matsuo, "PMV-Based train airconditioning control system", ASHRAE transactions, part 1,[5] A.P. Gagge, A. P. Fobelets and L.G. Berglund, "A standardpredictive index of human response to the thermal

environment", ASHRAE Transactions, vol. 92, no. 2B, pp.[6] M. Hamdi, G. Lachiver and F. Michaud, "A new predictivethermal comfort sensation index of human response" Paper to

appear in Energy and buildings joumall998.[7] M. Hamdi, F. Michaud and G. Lachiver, "Modkle flou duconfort thermique" Rapport inteme, D6partement de gCnie

Clectrique et de genie informatique, Facult6 des sciencesappliquCes. Universitk de Sherbrooke, April 1996.[8] IS 0 7730, "Moderate thermal environments determination ofthe PMV and PPD indices and specification of the conditionsfor thermal comfort, 1987.[9] J. W. MacArthur, "Humidity and predicted-mean-vote-basedcomfort control", ASHRAE Trans., no. 1, pp. 5-17, 1986.[lo] M. Sherman, "A simplified model of thermal comfort"Energy and Buildings joumal, vol. 8, pp. 37-50, 1985.

pp. 423-430, 1995.

709-731, 1986.

49 1


6/6

Documents

A Fuzzy Control System Based on the Human Sensation of Thermal Comfort