DEVELOPMENT OF OPTIMAL OPERATION OF THERMAL STORAGE TANK AND THE VALIDATION BY SIMULATION TOOL

Transcription

1 DEVELOPMENT OF OPTIMAL OPERATION OF THERMAL STORAGE TANK AND THE VALIDATION BY SIMULATION TOOL Yoshihito Gotou* and Harunori Yoshida Department of Global Environmental Enineerin, Kyoto University Sakyo-ku, Kyoto , Japan ABSTRACT Optimization of thermal storae tank is one of the major areas for enery conservation in HVAC systems. In the present paper, an alorithm is developed for optimum of thermal storae tank, by minimizin non-linear cost function. Results are validated under stationary and random load conditions. It is concluded that the present alorithm is quite robust and provides an optimal scheme of. INTRODUCTION To meet the national requirement of 6% reduction in CO2 emission made at the COP3 Kyoto in 1997, thermal storae used for air-conditionin systems was desinated one of the major CO 2 reduction means in buildin sectors by the Japanese overnment. Thermal storae tank are widely used in Japan and elsewhere to store enery at off-peak hours, thereby flattenin the electricity demand and increasin the chiller COP when outside wet bulb temperature is lower than day time. However, the statistics show buildins with such systems consume more enery as compared to these without it. One of the main reasons of this problem is that air-conditionin system operators try to store imum thermal enery at the lowest temperature level to avoid shortae on the followin day. Therefore, computerized optimal usin predicted future thermal load profile of a followin day miht be a powerful tool because optimal of the system is very complicated and beyond human experience[3],[4],[8]. The three most important factors influencin electric power consumption by an HVAC system with thermal storae tank are, 1) The operatin time of the chiller: Ordinarily the operatin time of the chiller is not controlled and the thermal storae tank is filled with heat without any limit. Therefore, water temperature in the tank remains low and consequently the chiller COP oes down and hiher heat transfer loss from the thermal storae tank * presently workin for Buildin System s Desin Section, Nikken Sekkei Ltd, Japan takes place. 2) The operatin schedule of chiller: When the chiller operates at midniht, the water temperature in the thermal storae tank remains low after completin the heat storae till its use. So the heat transfer loss from the thermal storae tank becomes larer than that when the chiller operatin schedule shifts to early mornin. In this shifted, the chiller can be operated at hih COP takin advantae of the mornin outside low wet bulb temperature. 3) The chilled water temperature: Enery consumption in the chiller can be saved by settin the evaporator outlet water temperature as hih as possible because of the improvement of the chiller COP. When the followin day thermal load is not so lare, too low chilled water is needless. On the other hand, if the chilled water temperature is above a fixed temperature, it cannot handle the load satisfactorily. In the present paper, authors propose an optimal scheme which was developed as computerized software on MATLAB/SIMULINK environment and results are obtained usin al data of a real buildin. Optimization is achieved by determinin optimal chiller outlet water temperature takin account the performance of storae tanks, chiller, coolin tower, pumps and AHU coil. The optimum is determined by minimizin nonlinear cost function toether with simulation of the models. Four cases of are considered (Table 1). OPTIMAL OPERATION ALGORITHM The optimal alorithms for HVAC systems with a thermal storae tank are made up of four blocks (Fi. 1), prediction of coolin load, determination of required heat storae, system simulation, and optimal system control. The first block predicts coolin load for the followin day. The second block determines heat storae requirement on the basis of coolin load prediction. The third block comprises quasi-stationary model in which components are expressed as numerical expressions. By usin this model power consumption of the followin day at the 1

2 Table1.Types of thermal storae system Operatin ways time control Operatin schedule Set point temp. Diff. in inlet and outlet cold water temp. N- P1- P2- O- at full capacity until power consumption falls to 2% control usin coolin load prediction control usin coolin load prediction control usin coolin load prediction Midniht Midniht Early mornin Early mornin 7 o C 5 o C 7 o C 5 o C 7 o C 5 o C not fixed 5 o C systems are simulated. The last block determines how to control the systems optimally. Variable here considered is set point temperature of the chiller outlet water and constraint is mass flow rate of the pump, which supplies water to the coolin coil, and operatin time of the chiller. The flow chart of the alorithms is shown in Fi. 2. Prediction of coolin load: The followin day coolin load prediction alorithm has been reported in our previous work[1],[2]. Determination of required heat storae: Thermal load of an HVAC system comprises coolin load ( q p ), heat losses from thermal storae tank and pipes ( q l ), and heat eneration at the pumps ( q h ). However, available thermal storae heat depends on the coolin coil performance, and cannot be determined by one temperature level. When the water temperature level in the thermal storae tank is too hih, it cannot handle heat load well and room temperature becomes hiher than the set point temperature. Hence, the larer is heat load, the lower is this limit required. Storin just the followin day HVAC heat load is not enouh when the storae starts from hih water temperature level in the thermal storae tank due to over-consumption of the heat storae the day before. next day coolin load prediction required heat storae ƒloop for ptimum calculation chiller outlet water set point temperature al time and chiller schedule dicision next day coolin load is predicted by usin the alorithms which requires the observed data such as past actual coolin load, room temperature, and weather forecast. heat loss from the thermal storae tank and pipes, and heat enerated by the pumps. heat requirement to restore the temperature potential in cold side tank at beinin of the. chiller set point temperature optimization by nonlinear prorammin to minimize electrical power consumption. al flas of the heat source equipment transferred to the block of simulatin systems the block of predictin coolin load the block of determinin required heat storae the block of simulatin systems the block of controllin optimally system Fi. 1 Block diaram of optimum control system consumption of electrical power by the system optimum set point temperature of the chiller outlet water Fi. 2 : optimal system alorithms and composition of the blocks in thermal storae system 2

3 Table 2 : Desin outline of the thermal storae system 1) thermal resource thermal storae systems water storae tank number of rows 1 total volume 76m 3 thermal transmittance 3.37kJ/m 2 h o C temperature outside tank 2 o C main resources hermetic turbo refrierator chilled water set point temp. 7 o C refrieratin capacity 58kW electric power input 139kW coolin water mass flow rate 124m 3 /h chilled water mass flow rate 99.8m 3 /h coolin tower electric power input 5.5kW 2) air conditionin systems CAV systems temperature and humidity set point 26Ž, 5 mass flow rate of supply air 5.25x1 4 m 3 /h intake of outside air fixed at 25% 1.75x1 4 m 3 /h In order to avoid this problem, this block keeps water temperature level in the thermal storae tank at startin point of time to the level (θ d ) on which storae heat can be taken out. Therefore, required heat storae ( q r ) is considered as the sum of the HVAC thermal load and the heat which need to restore the temperature level in the cold side tank. n r p l h p k = 1 q = q + { q + q } + c m( θ θ ) where, c p : specific heat of water [kj/k o C] m : mass flow rate of the chilled water [k] θ k : water temp. at the k th tank from the cold side [ o C] n : rows of the cold side tank θ d : temp. at which storae heat can be taken out [ o C] SYSTEM SIMULATION Composition of the thermal storae systems that are examined in this study is shown in Fi. 3, and the outline of this systems are shown in Table 2. This systems are desined to operate chiller in the niht and to store the whole heat neccesary for the followin day air conditionin. Therefore, chiller is not operated in the daytime. This block is made up of several components on MATLAB SIMULINK(Fi. 4). In this study, each of the components except a thermal storae tank is made without considerin its dynamic characteristics. These components are considered as momentary stationary units and are expressed by introducin modifications in standard system simulation models [6]. Optimal system control: In this block, the set point temperature of the outlet chilled water and k d its operatin schedule are decided. Dependin upon the combination of constraints and purposes, many ways of can be planed. However, as mentioned earlier, four basic cases of are examined in this study. N- is normal. P1- and P2- are the usin predicted load. Only O is an optimum. θ R &m w W m1 Room W S θ cl, Coolin Tower θ ci, θ el, θ ei, Thermal Storae Tank Fresh Air 2 Way Control Valve Air Handlin Unit Fi. 3 : Block diaram of HVAC system with a thermal storae tank Mixin Valve θ S θ 1 θ 2 θ n θ F θ AC W AC θ ra x ra θ oa x oa θ aout, θ ain, x ain, &m a θ ain, &m a W S W R θ aout, 3

4 [F] R_penalty [G] [A] chiller Ptot R_pulse 1 Operation Schedule [B] [cpump(r_ct)] cpump2_pw cpump2_pw Pump(R_CT) Memory1 Memory2 Tel Wr1 Tei Tcl Tset Tci Qtot Cop penalty1 chiller_penalty chiller [B] cpump1 twin chiller_schedule [C] cpump(r_ct) C.tower pw_total_sum twout CT [D] cpump2 R_pulse Qloss_save_cpump2 cpump(r_st) [E] vpump Ctower_pw [G] vpump penalty penalty [H] Qsum [C] Ctower_pw pw q [I] Tst Coolin Tower1 (CT) [J] Tst Qtot cpump1_pw Qtot Power_sum_save [D] cpump1_pw [cpump(r_st)] twin Memory3 Pump(R_ST) twout Wr Memory4 (chiller) R_pulse Ts [A] Qloss_save_cpump1 Ts Wac Ms Wm1 _chiller_pw_ [J] Wr Tr Tei chiller_pw Tset Qtot1 a_qload_sa Tac Mixin Valve.1 step tank_temp Qload ya_cop_sav step Thermal Storae Tank 1 Out1 _total_pw_ total_pw Scope ts [F] tf Tunk temperature s_ya_tst_old [I] Tst2 tei tel tci tcl t_ct_in t_ct_out water_temp [H] q1 Cop s_ya_coil_schedule Coil Operation Schedule1 Tain Tain Xain Xain Ma Ma Q Qtot Out Data N-: This corresponds to normal. operates at imum capacity untill power consumption falls down to 2% of rated performance. P1-: This may be desinated midniht. It s commences with the beinnin of the niht time tariff, and stops when the heat storae meets the demand. P2-: This may be desinated early mornin. The operatin mode of the chiller is same as in previous case. However, timin is set such that its ends at the beinnin of air conditionin. O-: This is similar to the early mornin. However, set point temperature of outlet chilled water is optimized in order to increase chiller s COP. The problem is solved by nonlinear prorammin. Operatin schedule of the chiller is kept same as in early mornin. The followin constraints were applied, * Mass flow rate at the pump, which supplies water to the coolin coil, is smaller than its desined rate. * Operatin time of the is shorter than the imum time lenth ( t ) to which niht time tariff is applied. (In this examination, t is setted for 1 hour). Mw1 Air Handlin Unit1 Memory Mac Fi. 4 : Block diaram of optimum control system F HG r a m & m & p = 1 + m& I KJ 2 b b bt b m& m& = 1 m& < m& F HG r b t t time = 1+ t I KJ 2 t = 1 t < t Where, r p : coefficient of pump power supply &m: mass flow rate [k/h] & tac Memory6 vpump_pw vpump_pw Memory5 Wac m : desined water mass flow supplied to AHU coils [k/h] r t : coefficient of chiller power supply t : operatin time of the chiller [h] t : imum operatin time of the chiller [h] a,b : weiht of the constraints W vpump Tst save a_tstdata_s tstdata Wac [E] (vpump) twin twout Mw C_pulse Qloss_save_vpump Cost function includin the penalties is as follows, J = E + rp Ep + rt Eref Where, E : consumption of electric power at each of the components except chiller and pump E p : consumption of electric power of at the pump E ref : rated power consumption at the chiller These constraints are replaced by penalties when solved by nonlinear prorammin to make the problem nonrestricted. 4 RESULT AND DISCUSSION In the present study, for the purpose of validatin optimum alorithm, four basic cases of are compared on the basis of power consumption under similar stationary and random load

5 [ o C] Fi. 5 : variation of dry bulb temperature with time [ o C] load day Medium load day Liht load day load day Medium load day Liht load day Fi. 6 : variation of wet bulb temperature with time conditions. In an actual system, predicted load has the probabilistic error in a certain rane. However, in this study, it is assumed that the predicted load has no probabilistic error. The study considerin this probability may be considered in future developments. [kw] 14 Power consumption 12 Chilled water supply pump x1 5 [kcal/h] Medium Liht Fi. 7 : variation of coolin load with time under stationary load condition x1 5 [kcal/h] 3.5 Medium Liht Fi. 9 : Variation of coolin load with time under random load condition Periodically Stationary Load Condition: Three types of load profile for examinin several cases of under periodically stationary load condition are prepared on the basis of load level (Fi. 7). Weather data that correspond to three types of load, heavy, midium, and liht load, are shown in Fi. 5 & 6. Time [yen] 14 Cost of electricity* 12 Chilled water supply pump Liht Medium Liht Medium Fi. 8 : Comparison of power consumption and cost of electricity in periodically stationary load condition [kw] 14 Power consumption 12 Chilled water supply pump 1 [yen] Cost of electricity* Chilled water supply pump Liht Medium Liht Medium Fi. 1 : Comparison of power consumption and cost of electricity in random load condition From the left, Maximum,Midniht,Early mornin,early mornin-optimum caution*optimization about the cost of electricity is not examined

6 step for the simulation is kept 9 seconds. (At O-, time step is 36 seconds to save computational time.) At O-, optimum set point temperature of the chiller outlet water remains at 1.9 o C, 13.7 o C and 16.9 o C under heavy, medium, and liht load conditions respectively. Enery consumption reduces by 8.6%, 17.% and 26.8% in heavy, medium, and liht load conditions respectively as compared to that in N-. However, this turns out to be the most costly mode of (Fi. 8). It is because power consumption at the pump in the daytime is quite lare, as chilled water temperature is hiher. Random load condition: To examine several cases of under random load condition three load level were also prepared. 1 days examples are shown in Fi. 9. (These are made by processin the load data measured at a real buildin [5].) Time step for the simulation is fixed at 9 seconds as in the previous case. Under these conditions, enery consumption reduces almost in the same manner as observed under periodically stationary load condition. However, in heavy load condition, cost disadvantae at O- is not found (Fi. 1). It also demonstrates that it is possible to operate optimally with stability even under random load condition. In this case, water temperature in the tank is stirred like actual. CONCLUSIONS It may be concluded that the proposed alorithm controls the set point temperature, operatin time, and operatin schedule of the chiller optimally with stability. Alorithm presented can be applied to real systems. Therefore, amon the four cases of s, the O- attains the larest enery conservation about 18%. However, investiations into the followin issues at optimum conditions are further solicited, 1) probabilistic reliability of the coolin load prediction, 2) addition of the restrictions in order to deal with the several problems that occur in adoptin to the real systems, 3) purpose of the optimum. Nevertheless the basic concept of the present work can be applied with small modifications. For example, CO 2 reduction can be achieved by replacin cost function. Prediction for Optimal Operation of Thermal Storae Systems, Summaris of Technical Papers of Annual Meetin, The Society of Heatin, Air-Conditionin and Sanitary Enineers of Japan, pp , ) Yoshida.H and Goto.Y,,,Validation of Thermal Load Predictin Alorithm for Thermal Storae Systems usin Real Buildin Data, Transactions of the Society of Heatin, Air-Conditionin and Sanitary Enineers of Japan, No.73, pp.11-11, ) Yoshida.H and Inooka.T,,, Rational Operation of a Thermal Storae Tank with Load Prediction Scheme by ARX Model Approach, IBPSA, 2, pp.79-86, ) Inooka.T and Suzuki.T et al,,,study on Control Systems for Enery Conservation of Thermal Storae Air-conditionin Systems (Part3), Summaris of Technical Papers of Annual Meetin, The Society of Heatin, Air-Conditionin and Sanitary Enineers of Japan, pp , ) Nakahara.N, Optimal Operation of Thermal Storae System and Thermal Load Prediction, The Society of Heatin, Air-Conditionin and Sanitary Enineers of Japan, ) Manuals of TRNSYS, HVACSIM+, HASP/ACSS 7) Kawashima.M and Ito.N,,,Study on Smart HVAC System with Load Prediction Adjustment of Heatin Start-up Time, Summaris of Technical Papers of Annual Meetin, Architectural Institute of Technoloy, pp , ) Hokoi.S and Matumoto.M,,,Stochastic Optimal Control of Thermal Storae System, Summaris of Technical Papers of Annual Meetin, Architectural Institute of Technoloy, pp , ) Mori.S et al, Thermal Storae System,The Society of Heatin, Air-Conditionin and Sanitary Enineers of Japan, 1982 ACKNOWLEGEMENTS Authers are rateful to Tokyo Electoric Company, Japan for supportin the present study and Toyo Thermal Enineerin Ltd, Japan for providin data. REFERENCES 1) Yoshida.H and Goto.Y,,,Air-Conditionin Load 6