Optimal Operation of a HVAC System with a Thermal Storage Water Tank Harunori Yoshida, Hiromasa Yamaguti

Control Optimization 2

Seventh International IBPSA Conference

Rio de Janeiro, Brazil
August 13-15, 2001

OPTIMAL OPERATION OF A HVAC SYSTEM WITH A

TTHERMAL STORAGE WATER TANK

Harunori Yoshida and HiromasaYamaguti *

Department of Global Environmental Engineering, Kyoto
University
Sakyo-ku, Kyoto - 6068501, Japan

operation of the system. One cause of the problem is

ABSTRACT immaturity in the control strategies developed without
adequate understanding of system’s sophistication. To
A methodology for the optimal operation of a HVAC cope with this problem development of new optimal
system with a thermal storage water tank has been operation strategy is needed, where enhancing the
proposed by one of the authors1). In the present paper recent development of computer control technology and
improvement of the methodology is explained together the fast price dropping of computer products will be
with simulation results obtained by using real building key issues.
data. The main results are as follows. 1) This method is The authors have previously proposed a basic
sufficiently robust to be applied to a real HVAC system methodology for the optimal operation of a thermal
having time varying air-conditioning (AC) load. 2) storage water tank at the Building Simulation ‘99
Predicted AC load must include some error and it was conference utilizing simulation technology1). In the
found that storage operation ignoring the error often present paper the following improvements added to the
experiences shortage of thermal energy and as a result previous method and the investigation results related
unsatisfied air-conditioning creating room air to the new development are explained.
temperature deviation. To avoid the problem increasing
in the load by taking the error level into account should 1) Instability and local minimum occurrence in the
be considered. 3) On peak load days daytime storage process of optimization was found in the previous
operation is normally necessary due to the limitation in methodology, therefore, computational algorithm was
tank capacity. The optimal operation method improved.
considering this constraint was developed and it was 2) The optimal operation was searched assuming that
found that the performance is fine. AC load prediction is perfect. If predicted load is
smaller than the resultant load, satisfactory air-
conditioning cannot be achieved, therefore, it is
INTRODUCTION recognized that increasing the predicted load by some
percentage is safer decision. In this study magnitude
Japanese government has controlled its electric of the percentage defined as a safety margin is
industry in order to provide stable supply and to attain investigated and recommended.
nationwide price uniformity. However the current 3) In the previous study all the thermal energy required
worldwide movement known as restructuring of to manage AC load is stored during the nighttime.
electricity market forced the government to change the But in the actual system storage tank capacity is
policy drastically and the market has been partly opened designed not to store whole energy for a peak day,
in 2000. The issues in restructuring differ from country that is, on heavy load days chillers must be operated
to country based on its constraints in energy resources even during the daytime and electricity used for this
and conditions of power plants, however, minimizing a additional operation must be taken into account in
large demand gap between the daytime and the nighttime optimization. In this research this is investigated and
is the Japanese top priority issue. Therefore nighttime quasi-optimal operation strategy is proposed.
low electricity price has been adopted before the current 4) The aim of the optimization was to minimize electric
restructuring movement and the thermal storage systems power consumption, however, reduction of electricity
for HVAC systems have been widely applied since 1960s cost is the most important aim for building users and
mainly to large commercial buildings. In spite of the the amount of CO2 emission for global warming issue.
long experience in thermal storage system operation, The effect by the different objectives is analyzed
however, operators still find difficulty in optimal and compared.
5) The AC loads used for investigation were typical
* presently working for Kansai Electric Power Company, periodic diurnal and artificially produced random
Japan loads. In the present investigation AC load taken

- 1249 -
 from a real building and the load predicted are used block of thermal load
prediction
additionally to demonstrate the feasibility of our
To predict the following day's load utiliz-
method. block of thermal ing past loads, room temperatures and
load prediction weather forecast information.

As shown in the previous report the total HVAC block to determine

system including major components, such as a storage required thermal energy
storage
tank, air-handling units, cooling towers, and water heat dissipation due to thermal
loss around a storage tank and
pumps is modeled to simulate the performance on water supply determining required thermal
MATLAB/ SIMULINK environment (Figure 1). By this energy to recover the thermal
thermal potential potential of the lower tempera-
implementation wide variety of HVAC systems can be calculation at the ture tank at the beginning time
starting time of of storage operation
easily modeled by the aid of user-friendly graphical storage operation
interface. The variables to be optimized are the chilled
determining required thermal energy
water temperature and the duration of chiller operation.

IMPROVEMENT OF OPTIMIZATION
initial calculation of operation hour tope
ALGORITHM
block of optimal
operation
Several improvements were added to the previous
optimizing algorithm in order to stabilize calculation, to find optimal chilled water tem-
optimal calculation of θ set perature by applying nonlinear least
avoid local minimum occurrence and reduce (child water temperature of square method to minimize system
a chiller) objective function
computational time by applying more rationale
constraints and more precise model for water circulation.
pw ( n+ 1) > pw (n)
1) Improvement in determining required thermal storage
Yes
The required thermal storage for a following day is
No
defined as follows.
elongation of operation hour determining optimal value of
of refrigerator tope and θ set
Qd = Q p + Qc + Q l − Qs (1)
Where,
Q p : predicted building load block of system
simulation
Qc : generated heat by circulation pumps for chilled
determining electricity con-
water sumption of the total system actual load
Ql : heat loss through storage tank walls and temperature profile within measurement
Qs : stored heat in a tank the tank

Although Qd is mainly subject to Q p the impact of

Qc and Ql on Qd cannot be ignored. In the previous optimal chilled water temperature
algorithm Qc and Ql were estimated as a given fraction
Figure 1 Improved optimal argolithm
of the predicted heat load Q p but this caused instability
of simulation when time varying real load was fed as the 3) Rationalization of constraints in optimization
input for simulation. In the new algorithm those values The constraints in optimization are;
are estimated based on simplified simulation and this
0 ≤ m
≤ m
d , 0 ≤ Tch ≤ 10 and Q ch ≤ Qd
improvement can reduce error level down to about 10%
rather than the previous magnitude of 70%. where
m
: chilled water flow rate
2) Avoiding local minimum in optimization m
d : design value of m

In the previous algorithm the optimal chilled water Tch : chiller operation time
temperature θ was determined by searching the Q ch : total thermal energy production by chiller
temperature itself and the duration of chiller operation Qd : total chiller capacity based on design value
Tch simultaneously, however, this approach sometimes
could not find the optimal values due to local minimum The second constraint is automatically satisfied by
occurrence in optimization process. To avoid the problem the new optimization search method mentioned above.
it was found that the two variables should be searched To incorporate the first and the third constraint in
independently. In the new algorithm θ and Tch are optimization a penalty function method is used.
searched starting from the initial value 5 oC and 10 hours
respectively. 4) Improvement of water circulation model
The functions of bypass circuit of chilled water
circulation system and on-off control of pump units are

- 1250 -
 Table 1 Reduction rate for three objective functions from the reference operation
reduction rate [%]
load type control method
electricity
electricity cost CO2 emission
consumption
optimal electricity consumption 41.6 34.0 41.7
light load optimal electricity cost 41.7 34.2 41.6
optimal CO2 emission 41.1 33.6 41.7
optimal electricity consumption 24.9 16.1 25.4
medium load optimal electricity cost 24.5 18.2 24.9
optimal CO2 emission 24.8 15.5 25.5
optimal electricity consumption 11.7 3.1 11.7
heavy load optimal electricity cost 10.7 6.9 11.4
optimal CO2 emission 11.6 4.8 11.8

[kWh]
5000
incorporated to make the simulation model realistic. The 4000
actual load

characteristic curves of pumps are also improved to predicted load

3000
make the performance more precise.
2000

TOTALLY NIGHTTIME OPERATION 1000

0
For a basic investigation the storage tank capacity is Jun. 3 Jun. 24 Jul. 15 Aug. 5 Aug. 26 Sep. 16 Oct. 7 Oct. 28
date
assumed large enough to store whole thermal energy Figure 2 Air-conditioning load taken from a real
required for air-conditioning on the following days. building and the predicted load
Although this assumption is not realistic in common
building systems the investigation is useful to
understand the basic performance and characteristics functions is very small. The reasons of this result are as
of the present operation method. follows.
1) To reduce electricity cost daytime electricity usage
1) Comparison of Objective Functions with Periodic should be minimized because daytime price is higher.
Load Decreasing in chilled water temperature is the most
To investigate the effect brought by the deferent effective way to achieve this because water circulation
objective functions which correspond to minimizing rate during daytime air-conditioning can be reduced and
electricity consumption, electricity cost and CO2 as a result electricity used for pumping is reduced.
emission simulations were carried out using three However decreasing the temperature requires more
standard and periodic AC loads; heavy, medium and electricity due to COP drop in chiller operation. Due to
light load. For the simulations the following conditions these contradictory effects the temperature level defers
are used. The current electricity prices in the Tokyo from that of the operation for electricity consumption
region are 6 JPY/kWh for the nighttime (22:00 - 8:00), 20 optimization.
JPY/kWh for the daytime (8:00 - 14:00, 16:00 - 22:00) and 2) Hourly CO2 emission rates in unit electricity usage
22 JPY/kWh for peak hour (14:00 - 16:00). The larger the do not differ so significantly that much difference
percentage of generation by nuclear power plants is, between two objective functions; electricity
the less the CO 2 emission rate by unit electricity consumption and CO2 emission, does not occur.
consumption is because nuclear power is believed to If we evaluate the total effectiveness of incorporating
emit less CO 2. In the present research hourly CO 2 the optimal operation based on the medium load
emission rates reported by Tokyo Electric Company are condition, about 25% in electricity consumption and
used. For the comparison of the results the reference CO2 emission and 16% in electricity cost can be reduced.
operation is defined; namely non-optimal operation that
corresponds to chilled water temperature of 7 0C (design 2) Simulation using Real Air-Conditioning Loads
value) and full storage without load prediction function. To verify the feasibility and the stability of the present
The reduction rates from the reference vale are shown method simulation using real air-conditioning loads
in Table-1. measured at a real building was performed. The record
It can be seen that the reduction of all the three optimal of the integrated loads in daily basis is shown by the
operations is substantial. This says incorporating solid line of Figure 2. The period of air-conditioning is
optimal operation is very beneficial. However the from June 3 to October 31 and the daily air-conditioning
deference between the reduction rates of three objective period is from 9:00 to 18:00. The dotted line in Figure 2 is

- 1251 -
 Table 2 Results of perfect prediction case for total period
in average power consumption [MWh]
o
operation hour [h] set temperature [ C] total refregirator secondary pump others
maximum storage operation 5.4 7.0 96.8 73.7 5.0 18.1
optimal operation 3.5 15.2 67.3 49.1 6.4 11.8

Table 3 Monthly compumption and reduction rates in

electricily
maximum storage
optimal operation reduction rate
operation
[MWh] [%]
[MWh]
Jun 17.3 11.3 34.5
Jul 26.4 21.3 19.4
Aug 25.1 17.8 28.9
Sep 16.4 10.5 35.6
Oct 11.6 6.3 45.6
total 96.8 67.3 30.4

Table 4 Results of operation based on real load with prediction error

in average power consumption [MWh]
o
operation hour [h] set temperature [ C] total refregirator secondary pump others
maximum storage operation 5.1 7.0 79.8 60.2 4.4 15.2
optimal operation 3.1 15.2 50.9 36.3 5.3 9.3

the predicted load by the method of the previous report a day is defined as an insufficient storage day. In the
4)
. The prediction error is 7.32% (a0) of the maximum present study the number of such days was 17. Table-4
load. The a 0 is defined as EEP (Expected Error shows the results and it can be seen that the reduction
Percentage) value. rate (36%) excluding the insufficient storage days is
similar to that of the perfect prediction case (32%). But
2-1) Simulation with Perfect Prediction the existence of 17 days of insufficient air-conditioning
Load prediction is not perfect in a real situation; would be a serious problem to building users.
however, basic investigation was made assuming Furthermore it was found that inadequate temperature
perfection. The results for the whole period of simulation disturbance in the tank is produced and this causes
are shown in Table-2 compared with the reference unstable air-conditioning operation.
operation defined at the previous investigation. It is
found that substantial conservation 33.1MWh (=96.8 - 3) Simulation Considering Prediction Error
63.7MWh) can be attained, that corresponds to 31% Two points are considered in the investigation of
reduction. This is mainly due to chiller COP increase prediction error; the effect of prediction error EEP level
with high chilled water temperature of 15.2 0C in average. and the effect of a margin of safety to reduce occurrence
Table-3 shows monthly results. It can be found that the of faulty air-conditioning days.
reduction rates are larger in intermittent season. This is
why in heavy load period the temperature must be low 3-1) EEP level of prediction error
to manage large AC load and as a result COP decreases. The EEP level of prediction error of the present
example is 7.32% as shown previously, however,
2-2) Simulation with Prediction Error according to the benchmark trials operated by SHASE
Taking the real situation into account simulation with (The Society of Heating, Air-Conditioning and Sanitary
load prediction error was carried out. Namely thermal Engineering of Japan) the average EEP level of the best
storage operation is performed based on the predicted methods was around 5%. Therefore in the present
load and consumption of the thermal energy or air- investigation predicted loads with three levels of EEP,
conditioning operation is achieved based on real loads. namely 5% (average), 10% (2 times) and 20% (4 times),
As mentioned previously the EEP value of the present were generated by the following method and the
example is 7.32%. Therefore due to the existence of the simulation using them was performed.
prediction error the stored thermal energy is not
necessarily sufficient to manage air-conditioning for a ed ( n ) = Q r , d ( n ) − Q p , d ( n ) (2)
whole day on certain days. The shortage of thermal
energy results in the increase of stored chilled water (aL a0 ) ⋅ ed (n) − Qr , d ( n)
k ( n) = (3)
temperature, and consequently room air temperature Q p , d ( n)
becomes higher than the set point for some hours. Such

- 1252 -
 q p , L ( n , j ) = k ( n ) ⋅ q p ( n, j ) (4) maximum load condition. This implies maximizing the
duration of chiller operation and consequently
minimizing chiller capacity. In other words the tank
where,
capacity is designed so as to store a half of the maximum
ed (n) : prediction error in day total
Qr ,d : day total load, although this does not necessarily ensure
actual load in day total
Qp,d : optimal size. When daytime storage operation takes place
predicted load in day total
q p, L : the conditions concerning optimal operation become
predicted hourly load
more sophisticated. On a day when total air-conditioning
n : indication of day
load is less than the tank capacity nighttime operation
j : indication of hour
meets the requirement and the operation strategy can
aL : EEP level ( a1 : 5%, a2 : 10%, a3 : 20%)
be the same as explained for the totally nighttime
operation. However when total load exceeds the tank
As shown in Table-5 the optimal chilled water
capacity optimal operation becomes more complex
temperature decreases and the number of days of
because other variables such as operation schedule of
insufficient storage or unsatisfactory air-conditioning
chillers in addition to the chilled water temperature must
increases substantially with EEP, although electricity
be considered. In the present study four conditions are
consumption does not increase significantly. When EEP
analyzed with different levels as listed in Table-7. They
level is 10% 20 days out of 122 days or 16% of cooling
are chiller operation schedule, chilled water temperature
season days will have unsatisfactory air-conditioning.
set point, reference tank temperature, and a margin of
This figure will not be acceptable by buildings users.
safety for storage. Eleven scenarios (S1 -- S7, S11 -- S14
shown in Table-8) based on different combinations of
3-2) Operation with a margin of safety
the four conditions and levels were investigated as
Increasing the stored thermal energy by a margin of
potential optimal strategies. For example each scenario
safety for the predicted load can decrease the number
is abbreviated such as S1( σ ), if the scenario is S1 and
of unsatisfactory air-conditioning days. To investigate
the safety margin is σ . Simulation was carried out by
the effect of margin level simulation was performed with
taking electric power consumption as the objective
five levels; 0%, σ (=7.32%), 2 σ , 4 σ and 8 σ . As shown
function and daytime chilled water temperature was
in Table-6 no unsatisfactory air-conditioning day occurs
optimized. To evaluate the results of the scenarios the
by applying the margin level greater than 4 σ ,
performance of non-optimal operation is taken as the
nevertheless the electricity consumption increases very
reference, that is, the chilled water temperature in the
little (3%).
nighttime is set at 7 oC and the temperature of the last
tank is maintained between 8 and 9 oC by on-off control
PARTLLY DAYTIME STORAGE of chillers. The reduction rates shown in Table-9 mean
OPERATION the reduction from the reference value. As seen they
range from 11 to 18 % with 15 % in average.
According to a design guide a standard storage tank
should be sized so that chillers are operated for 10 hours 1) Daytime chiller operation schedule
both in the nighttime and the daytime on the day of To evaluate the schedule of daytime chiller operation

Table 5 Comparison of optimal operation with different levels of EEP

set temperature power consumption insufficient storage
EEP value o
[ C] [MWh] increase rate [%] storage day needless day
0% 15.9 42.1 0.0 0 0
5% 15.4 43.1 2.2 11 5
10% 14.7 43.2 2.4 20 12
20% 13.2 43.7 3.6 27 27

Table 6 Results of operation with a margin of safety

power consumption [MWh] insufficient storage
reduction effect [%]
total refregirator secondary pump others storage day needless day
0 56.0 40.1 5.7 10.2 11 5 35.0
s 56.5 40.8 5.5 10.3 9 5 34.4
2s 57.3 41.6 5.3 10.4 5 6 33.6
4s 58.6 43.0 5.2 10.5 0 6 32.0
8s 61.2 45.6 4.9 10.7 0 5 29.0
full storage 86.2 65.3 4.6 16.3 0 0 0.0

- 1253 -
the performance of early time operation (OE) and late mostly in the daytime. The differences between S1, S4
time operation (OL) were compared. The OE means that and S5 are not significant, but comparing three aspects
chillers are operated without break after nighttime of power consumption, cost and CO2 emission the
operation, namely, operated in the morning. On the S4(2 σ ) would be decided as an optimal scenario and
contrary OL is reverse or operation in the afternoon. reductions of 12.9% in power consumption, 6.8% in cost
Heat loss is reduced by OL operation but the possibility and 12.4% in CO2 emission can be attained.
of the occurrence of thermal energy shortage is higher.
Although the difference between the two strategies is 3) Reference temperature change with month
negligible as shown in S1(0) for OE and S2(0) for OL of As investigated previously by changing the tank
Table-9. reference temperatures with month the system efficiency
will increase. To investigate this effect simulation S6
2) Chilled water temperature in the nighttime and S7 were performed. By comparing with the best
In the nighttime chillers are operated at maximum given result S4(2 σ ) it can be found that additional
capacity in partially daytime storage operation. To reductions of 3.5% in power consumption, 2.5% in cost
investigate the effect of the chiller temperature set point and 3.2% in CO2 emission is possible.
in the nighttime operation, simulations were performed
at three levels of set point; namely 7, 9 and 11 oC. They 4) Difference in objective function
are abbreviated as S1, S4 and S5 respectively and the The objective function of the above studies is the
effect of three levels of safety margin for load prediction electric power consumption. To investigate the effect
is analyzed together. Fundamentally the set point of applying different objective functions simulations
should be the design value of 7 oC to store maximum through S11 to S14 were performed. It can be seen that
thermal energy during night, however, it can be set higher the effect is very little, therefore taking the electric power
in real operation. In this case the stored energy will be consumption as the objective function is acceptable.
consumed excessively in the daytime and the risk of
unsatisfactory air-conditioning or insufficient storage CONCLUSIONS
increases. In addition to the three scenarios the scenario
S3 in which the set point was optimized as a uniform An optimal operation method for a HVAC system
value for day and night was performed. The results are equipped with a water storage tank was proposed by
shown in Table-9. All the scenarios of S3 show good the author’s in the previous work1). In the present report
reduction in power consumption, however, electricity improvement of the methodology is explained and the
cost increases and the number of insufficient storage effect and the feasibility of the method is investigated
days is large. This is because the chiller operation is and evaluated. The evaluation is made in terms of

Table 7 Conditions for partial daytime operation

investigated condition a b c d
earlier time later time
1. operation schedule
operation (OE) operation (OL)
night time: 7oCday night time: 9oCday night time: 11oC
optimizing
2. set temperature time: optimizing time: optimizing day time: optimizing
for whole day (TW)
for day time (T7) for day time (T9) for day time (T11)
same for all setting for each
3. reference temperature
months (SA) month (SE)
4. margin for storage safety 0 σ 2σ
Table 8 Combination of conditions for each scenario
operation set reference margin for
operation senario objective function
schedule temperature temperature storage safety
S1 OE T7 SA 0, s , 2s power consumption
S2 OL T7 SA 0 power consumption
S3 OE TW SA 0, s , 1s power consumption
S4 OE T9 SA 0, s , 0s power consumption
S5 OE T11 SA 0, s , 1s power consumption
S6 OE T7 SE 0, s , 2s power consumption
S7 OE T9 SE 0, s , 2s power consumption
S11 OE T7 SE 2s power cost
S12 OE T7 SE 2s CO2 emission
S13 OE TW SA 2s power cost
S14 OE TW SA 2s CO2 emission

- 1254 -
electricity consumption, electricity cost and CO2 following scenario is the best; nighttime storage
emission based on the present electricity tariff and operation at 9 oC followed by daytime operation in the
diurnal variations in CO2 emission for unit electricity morning, monthly varied reference temperatures for
usage taking nuclear power generation into account. storage and 10% of a margin of safety for load prediction.
Firstly totally nighttime operation is investigated. This Approximate reduction by this near-optimal operation
is the case that the tank capacity is designed large is 15% in power consumption, 9% in cost and 14% in
enough to store whole energy to satisfy air-conditioning CO2 emission. The effect of different objective function
of a peak day. If air-conditioning load is predicted is also analyzed but not so significant effect is detected.
perfectly the reduction of electric power consumption
from non-optimized operation can be about up to 30%. ACKNOWLEDGEMENT
In reality perfect prediction is not possible and if the
error level is 5% the system encounters 11 days of The authors acknowledge the favor of Tonetsu Co.
unsatisfied air-conditioning or insufficient storage. In in providing measured data from a real building.
general this figure may not be accepted by building
users. Therefore the cases in which predicted load is REFERENCES
increased with a margin of safety are investigated and it
is found that no insufficient storage day occurs with 1) Yoshida.H. and Goto.Y., Development of Optimal
20% of a margin of safety. Nevertheless reduction effect Operation of Thermal Storage Tank and the Validation
remains at about 30%. by Simulation Tool”, Building Simulation ’99 Kyoto,
Secondary more common operational condition is Japan, BS99-088, 1999
investigated, that is, partially daytime storage operation. 2) Yoshida, H. and Goto, Y., Air-Conditioning Load
According to a design guide the tank capacity is sized Prediction for Optimal Operation of Thermal Storage
so that chillers are operated 10 hours both in the Systems”, Summaries of Technical Papers of Annual
nighttime and the daytime on a peak load day. Meeting, The Society of Heating, Air-Conditioning
Consequently the tank is in storing mode only for and Sanitary Engineers of Japan, pp.177-180, 1998
nighttime on the days of light load. On these days the 3) Yoshida, H. and Goto, Y., Validation of Thermal Load
optimal strategy can be the same as for the totally Predicting Algorithm for Thermal Storage Systems
nighttime operation. On the contrary storing operation using Real Building Data”, Transactions of the
must be done during daytime also on the days of heavy Society of Heating, Air-Conditioning and Sanitary
load, then conditions becomes complex. Eleven Engineers of Japan, No.73, pp.101-110, 1999
scenarios are made as feasible combinations of 4) Yoshida, H. and Inooka, T., Rational Operation of a
conditions and they are evaluated. In conclusion the Thermal Storage Tank with Load Prediction Scheme

Table 9 Comparison of different scenarios for partially daytime operation

power consumption electricity cost CO2 emission
operation margin for insufficient
senario storage safety reduction reduction reduction storage day
[MWh] [JPY*1000] [kgC]
rate [%] rate [%] rate [%]
0 79.9 12.7 740 0.3 6450 11.1 0
S1 s 80.3 12.2 699 5.8 6411 11.6 0
2s 81.1 11.4 689 7.1 6450 11.1 0
S2 0 80.1 12.4 744 -0.2 6470 10.8 1
0 76.6 16.3 870 -17.2 6451 11.1 1
S3 s 76.7 16.2 849 -14.4 6428 11.4 12
2s 77.1 15.7 855 -15.2 6468 10.8 13
0 78.4 14.3 731 1.5 6340 12.6 0
S4 s 78.7 13.9 691 6.9 6297 13.2 0
2s 79.6 12.9 692 6.8 6358 12.4 0
0 77.3 15.5 764 -2.9 6308 13.0 0
S5 s 77.7 15.1 747 -0.6 6310 13.0 1
2s 78.3 14.4 771 -3.9 6397 11.8 3
0 77.2 15.6 733 1.2 6265 13.6 5
S6 s 76.5 16.4 677 8.8 6130 15.5 7
2s 78.0 14.8 674 9.2 6224 14.2 1
0 75.7 17.3 722 2.7 6148 15.3 5
S7 s 75.1 17.9 671 9.7 6025 17.0 7
2s 76.4 16.4 673 9.3 6121 15.6 1
S11 2s 78.0 14.8 672 9.4 6221 14.3 1
S12 2s 78.1 14.7 674 9.3 6221 14.3 1
S13 2s 79.9 12.7 697 6.1 6377 12.1 0
S14 2s 79.4 13.2 718 3.3 6369 12.2 0

- 1255 -
 by ARX Model Approach”, IBPSA, 2, pp.79-86, 1997
5) Inooka, T. and Suzuki, T. et al, Study on Control
Systems for Energy Conservation of Thermal Storage
Air-conditioning Systems” (Part3), Summaries of
Technical Papers of Annual Meeting, The Society of
Heating, Air-Conditioning and Sanitary Engineers of
Japan, pp.241-224, 1996
6) Mori, S. et al, Thermal Storage System, The Society
of Heating, Air-Conditioning and Sanitary Engineers
of Japan, 1982
7) Ibamoto, T et al, Thermal Storage System-basis and
application, The Society of Heating, Air-Conditioning
and Sanitary Engineers of Japan, 1995

- 1256 -