Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
A R T I C LE I N FO A B S T R A C T
Keywords: This paper presents a comprehensive methodology helpful to engineers and plant managers for designing and
Approach temperature reset schedule achieving optimized operating conditions for cooling systems. The most widely accepted and utilized theories,
Cooling tower such as the Merkel and Poppe models, has brought deviation of modeling performance or computational burden.
Condenser water leaving temperature Thus, the effectiveness cooling tower model by Braun is adopted to model the cooling tower performance with
Building cooling system
field measurements. A typical electric chiller model based on the condenser entering temperature is applied from
the EnergyPlus Engineering Reference. These two models are integrated to determine the optimal operation for
the cooling system. Correlations between the power consumption of chillers and cooling tower fans and ap-
proach temperatures are analyzed to determine the “near-optimal” operation. Furthermore, an optimized ap-
proach temperature reset schedule is evaluated in a case study. About 15.6% of the actual power consumption
can be saved for the 68 sampled days with the proposed optimal control.
1. Introduction models [16–20] even though it does not achieve the complete state of
saturation. Zheng et al. found that the outlet air conditions were un-
The heating and cooling of buildings use 14.6% of U.S. energy use saturated at inlet air temperatures of 290 K, 300 K and 310 K for am-
[1,2]. Central plants providing heating and cooling for building groups bient relative humidity varying from 8% to 97% and were all super-
is a common practice and there has been substantial research devoted saturated at an inlet air temperature of 280 K for the specific cooling
to reducing plant energy use [3–11]. Take Texas A&M University as an tower they investigated [21]. Kloppers and Kröger stated that it does
example. According to the Texas A&M University Energy Management not matter how much water vapor is present in the supersaturated air
Report [12], the primary energy consumption was 211 kBtu per gross for a specific air enthalpy because the lines of constant air enthalpy in
square foot, totaling 4.98 trillion Btu during the fiscal year 2014. The the supersaturated region are very close to vertical, and the results
total annual cost was more than $75 million. Even a small percentage calculated from supersaturated governing equations are fairly close to
decrease in the plant energy requirements could produce significant unsaturated ones [22]. However, the assumption of saturation has been
dollar savings when the total cost is this large. adopted in models to avoid computational burdens.
Cooling towers are widely used in large heating and cooling plants
to reduce the temperature of circulated condenser water to improve 1.2. Cooling tower methods and models
overall plant efficiency [13,14]. Cooling towers work on the principle
of evaporative cooling and hence the condenser water temperature is Several classical methods have been used to model the performance
limited by the wet bulb temperature of the cooling air [15]. Usually, of a cooling tower. The first practical theory and equation set was de-
engineers are primarily interested in the water inlet and outlet tem- veloped by Merkel in 1925 [23]. This theory is widely used for the
peratures of cooling towers and seek to reach the desired approach thermal evaluation of cooling towers. It relies on three critical as-
temperature. sumptions to simplify the calculations involved: (1) the water loss by
evaporation is negligible; (2) the air exiting the cooling tower is satu-
1.1. Outlet air conditions in cooling tower rated; and (3) the Lewis number relating heat and mass transfer is unity.
Eqs. (1) and (2) for the Merkel theory are obtained from mass and
Outlet air from a cooling tower is treated as saturated air in most energy balances of control volumes shown in Figs. 1 and 2.
⁎
Corresponding author.
E-mail address: jjliao@tamu.edu (J. Liao).
https://doi.org/10.1016/j.enconman.2019.111950
Received 21 June 2019; Received in revised form 14 August 2019; Accepted 16 August 2019
Available online 28 August 2019
0196-8904/ © 2019 Elsevier Ltd. All rights reserved.
J. Liao, et al. Energy Conversion and Management 199 (2019) 111950
dw
dTw
c pw (ws, w − w) mw
= /[h s, w − hma + (Le − 1)
ma
{h s, w − ha − (ws, w − w) hv } − (ws, w − w) c pw Tw] (3)
dha
dTw
c pw mw
= [1+(ws, w − w) c pw
ma
Tw /[h s, w − ha + (Le − 1){h s, w − ha − (ws, w − w) hv }
− (ws, w − w) c pw Tw] (4)
Fig. 1. Schematic diagram of a counterflow cooling tower. h 2 ws, w + 0.622 ws, w + 0.622
Le = = 0.865 3 ⎛ − 1⎞/ ln ⎛ ⎞
h d c pa ⎝ w + 0.622 ⎠ ⎝ w + 0.622 ⎠ (5)
where ws,w is the saturation humidity ratio of air at the local bulk water
temperature, Le is the Lewis number, cpa is the specific air of the air,
and hv is the enthalpy of the vapor.
A modified non-dimensional model developed by Halasz [27]
transformed the complex system of differential equations to a pure non-
dimensional form (θ, ∊) show in Eqs. (6) and (7). In order to obtain the
simplest possible non-dimensional model, the Lewis number is assumed
to equal to 1 while the outlet air condition is not assumed saturated.
When the temperature difference between water inlet and outlet is
greater than 18°F (10 °C), this model yields more accurate results than
Merkel method. However, it is not validated when the water inlet and
outlet temperature are close. In addition, since an approximate formula
Fig. 2. Control volume of the counterflow fill. for the enthalpy of unsaturated air is used in deriving this model, errors
are inevitable.
2
J. Liao, et al. Energy Conversion and Management 199 (2019) 111950
dwa NTU
=− (wa − ws, w)
dV VT (11)
dha LeNTU 1
=− [(ha − h s, w) + h g, w ⎛ − 1⎞ (wa − ws, w)
dV VT ⎝ Le ⎠ (12)
Fig. 3. Air Heat Transfer Effectiveness Comparisons versus NTU.
where hg, w is the enthalpy of water vapor at the local water tempera-
ture, wao is the outlet air humidity ratio, and Tref is the reference tem-
2. Methodology
perature for zero enthalpy of liquid water.
Whillier introduced a fundamental concept called the tower capa-
2.1. Model comparisons
city factor, R, as a basis for correlation of test data and for predicting
performance and presented an optimal ratio of water flowrate to air
Comparisons between cooling tower model results are obtained
flowrate for a certain set of outside conditions to minimize the average
with (1) detailed analysis with Lewis number of unity, (2) The Merkel
water temperature across the cooling tower [37].
model, (3) The Braun Model, and (4) The Halaz model.
Thermodynamic cooling tower models have been developed to
Fig. 3 shows the cooling tower air heat transfer effectiveness (εa )
predict the tower performance. Naik and Muthukumar developed a
versus NTU when the mass ratios of water to air flowrate are 0.5, 1, and
thermodynamic analytical model and analyzed the heat and mass
2 respectively. The operating condition is fixed at 70°F (21.1 °C) dry-
transfer in a cooling tower to calculate water evaporation loss in a
bulb, 60°F (15.6 °C) wet-bulb, and 90°F (32.2 °C) water inlet tempera-
cooling tower [38]. Ghazani et al. used the first and second laws of
tures. Overall, the Merkel, the Braun, and the Halasz models all agree
thermodynamics to calculate the air properties at the inlet conditions of
well with the detailed analysis under all conditions. The Merkel model
the cooling tower which was validated against test results from a lab-
slightly underpredicts the heat transfer effectiveness and the water
scale cooling tower [39]. Lu and Cai developed a model based on the
evaporation rate but the error decreases for increasing ambient air
Merkel and Braun models as well as the laws of mass and energy bal-
temperatures. Errors associated with the Braun model are primarily a
ance, which allowed for the use of fewer input variables and simpler
result of the assumption of a linear saturation enthalpy relationship.
calculations due to the lack of iterative computation required by
The Halasz model also assumed a linear air saturation line, so similar
Braun’s model [40].
errors exist for the Halasz model. And the overestimation of the Halasz
model is more distinct as the ratio of water to air mass flow decreases.
1.3. Chiller and cooling system models Fig. 4 shows the cooling tower water temperature effectiveness (εw )
under the same conditions as
For the chiller models, thermodynamic models of reciprocating Fig. 3. The water temperature effectiveness is defined as the ratio of
chillers [41], centrifugal chillers, and a “universal” thermodynamic the temperature difference between the inlet and outlet water to the
model for chillers [42] are available. maximum possible temperature difference if the leaving water is at the
There have been several attempts to optimize the operating meth- entering air wet-bulb temperature. The Merkel model overstates the
odology of a cooling system. Van utilized the condenser water flowrate water temperature effectiveness slightly due to the neglect of the water
and cooling tower fan speed as the control variables [43] while loss. The εw values predicted by the Halasz model are higher than those
Schwedler and Bradley used average water temperature to determine predicted by the detailed analysis when the ratio of water to air is 0.5
the capacity of the tower [44]. Most of these attempts only optimized and 1 but the gap decreases as the water-air ratio goes to 2.0. The Braun
the operation of one component such as the chiller or the cooling tower model gives results consistently closer to the detailed thermodynamic
instead of the whole system. analysis results for εw than those of Merkel and Halasz.
Studies have also addressed the optimization of the entire cooling Therefore, the Braun model performs better than the Merkel and
system as a whole with partial success. For instance, an optimum water Halasz models and offers a good compromise between simplicity,
flowrate was used to achieve the lowest possible average temperature
of water when dissipating a specific amount of heat into a specific
stream of air. Braun developed an algorithm, based on an open-loop
control equation, for the near-optimal control of the cooling system
[45]. He pointed out that the minimum total power occurs at a point
where the rate of increase in the fan power with airflow is equal to the
rate of decrease in the chiller power. Zhang et al. then demonstrated
that the optimal cooling tower approach temperature reset schedule can
be approximated with two straight lines, which can produce significant
energy savings compared with the scenario with a constant cooling
tower condenser water leaving temperature [14]. However, currently,
no single model in the published literature performs the operation op-
timization of all components in the cooling system [46]. Cortinovis
et al. summarized specific situations to increase either the water flow-
rate or the air flowrate for optimal operation from their case studies
[47]. Fig. 4. Water Temperature Effectiveness Comparisons versus NTU.
3
J. Liao, et al. Energy Conversion and Management 199 (2019) 111950
Table 1
Design operating conditions of the cooling towers.
Inputs Water Dry-Bulb Wet-Bulb Water Inlet Water Outlet
Flowrate Temp Temp
4
J. Liao, et al. Energy Conversion and Management 199 (2019) 111950
3 ṁ a + 1092.6 2
⎢
(
67.369 )
⎡− 0.0000014 ∗ ṁ a + 1092.6 + 0.00073 ∗ ( 67.369 ) +⎤
⎥
⎢ ⎥
P
0.0021 ∗ ( 67.369 )
ṁ a + 1092.6
+ 0.57
εfan = fan = ⎣ ⎦
Q [ṁ w (Twi − Two )]
= f2 (m w , Tapp, Twi, Tai, TWB, NTU) (24)
If = m w , Twi, Tai, TWB and NTU are fixed at a given set of values and
inserted in Eqs. (18)–(24), the functions that relate the approach tem-
perature (Tapp) to the fan power (kW) and power input per capacity
(kW/ton) can be obtained, as shown in Fig. 9. As more electricity is
consumed by the fan, a lower approach temperature can be achieved.
parameters given in Eqs. (20) and (21). The Braun model gives a good prediction for water outlet condi-
tions. Fig. 10 plots the predicted and actual outlet water temperatures
Tapp = f (mw , ma , Twi, Tai, TWB, NTU) (20)
versus dry-bulb temperature, as well as the corresponding wet-bulb
temperatures and design condensing water supply temperatures. The
ma = f (mw , Tapp, Twi, Tai, TWB, NTU) (21)
minimum outlet water temperature is set to 65°F (18.3 °C) because of
Insert Eqs. (16) and (21) to Eq. (17), and then the relationship be- the by-pass operation. Although the actual outlet water temperatures
tween fan power and VFD speed becomes: are generally slightly higher than the design values, the cooling tower
exit temperatures are almost all less than 3.5°F (1.9 °C) above the design
ṁ + 1092.6 ⎞3 ṁ + 1092.6 ⎞2 points. When the dry-bulb temperature is lower than 60°F (15.6 °C), the
Pfan = −0.0000014 ∗ ⎛ a + 0.00073 ∗ ⎛ a
⎝ 67.369 ⎠ ⎝ 67.369 ⎠ outlet water temperatures are relatively constant at 65°F (18.3 °C) be-
ṁ a + 1092.6 cause of the by-pass setting for the cooling tower.
+ 0.0021 ∗ ⎛ ⎞ + 0.57
⎝ 67.369 ⎠ The optimal cooling tower approach setpoint reset schedule can be
approximated with a 4-parameter change-point model [14]. This kind
= f1(mw , Tapp, Twi, Tai, TWB, NTU) (22)
of schedule can help achieve significant energy savings compared with
the use of a constant cooling tower condenser water leaving tempera-
Cooling Capacity Q = Cpw ∗ mw ∗ (Twi − Two) (23) ture (CWLT). Meanwhile, chiller part load ratio (PLR), chilled water
Fig. 8. Cooling range and water flowrate through the year of 2015.
5
J. Liao, et al. Energy Conversion and Management 199 (2019) 111950
Fig. 11. Cooling tower approach temperatures versus ambient wet-bulb tem-
peratures.
The input power curves of the chillers can be generated based on the
typical electric chiller model in the EnergyPlus Engineering Reference
[49] as shown in Equation (27). Since the chiller input power varies as a
function of PLR and the chilled water leaving temperature, a more
accurate model with the different chilled water leaving temperature
(Tcw,l) intervals applied is introduced, and an example result of PLR
higher than 80% and the Tcw,l=(41, 42)°F or (5, 5.6)°C is given in Eq.
(28). Fig. 12 shows the distribution of the scattered points of Chiller kW
and kW/ton versus approach temperatures, respectively. Note that most
of the points are clustered in the vicinity of an approach temperature of
about 6°F (3.3 °C) as designed.
Fig. 10. Comparison of predicted, design, actual water outlet, and wet-bulb
temperatures.
leaving temperature, and the climate have minor effects on the coeffi-
cients of the optimal CWLT reset schedule [14]. The approach setpoint
reset schedule with two straight lines is applicable for the cooling tower
as plotted in Fig. 11. Thus, when the chiller PLR is 80%, and the chilled
water leaving temperature is about 42.0°F (5.6 °C), the approach tem-
perature reset schedule can be expressed as Eqs. (25) and (26) instead
of operating the cooling tower with a constant approach temperature
(Tapp) at 6°F (3.3 °C).
Tapp = −0.2 Twb + 21.88 if Twb > 54°F (12.2 °C) (26) Fig. 12. Chiller power consumption versus approach temperature at
Tcw,l = (41, 42)°F or (5, 5.6)°C.
6
J. Liao, et al. Energy Conversion and Management 199 (2019) 111950
Q̇ ref
PChiller = *(ChillerCapFTemp)*(ChillerEIRFTemp) dTotal Power ( ) = 0.0000356 ∗ 3 ∗ (T
kW
ton 2
COPref app ) + 0.00058175 ∗ 2
dTapp
*(ChillerEIRFPLR) (27)
∗ (Tapp) − 0.04909153 (32)
PChiller (kW) = −0.00007357 ∗ (Tcond,e ) 4 + 0.02218499 ∗ (Tcond,e )3
Let Eq. (32) equal to 0, solve for Tapp , substitute its value into Eq.
− 2.48569803 ∗ (Tcond,e )2 + 125.93271371 ∗ Tcond,e (31) and then the lowest total power consumption and the corre-
− 2419.84 (28) sponding optimal approach temperature is obtained. The lowest value
of kW/ton for this case is 0.701 at Tapp = 16.7°F (9.3 °C). If the opti-
According to the water pump power consumption data throughout
mized range is defined as 5% higher than the lowest power consump-
the year 2015, pump power consumption is relatively constant with an
tion, the corresponding optimized Tapp is within the range of (12.7,
average consumption of 7.19 kW. Therefore, the water pump power can
20.5)°F or (7.1, 11.2)°C. Similarly, when the wet-bulb temperature is
be neglected when optimizing the total power consumption of the
75°F (23.9 °C), the lowest value of kW/ton is 0.764 at Tapp = 13.1°F
cooling system.
(7.2 °C) or within the range of (9.9, 16.1)°F or (5.5, 8.9)°C; when the
wet-bulb temperature is 80°F (26.7 °C), the lowest value of kW/ton is
3.5. Cooling system modeling and analysis
0.925 at Tapp = 7.3°F (4.1 °C) or within the range of (5.1, 9.2)°F or
(2.8, 5.1 °C) .
The optimum operation of the cooling system should consider both
As the wet-bulb temperature increases from 70°F (21.1 °C) to 80°F
the chiller power and cooling tower fan power. By adding the perfor-
(26.7 °C), the optimized approach temperature decreases from 16.7°F
mance curves of the chiller and the cooling tower fan, the dependence
(9.3 °C) to 7.3°F (4.1 °C) while the optimized total power over the
of the total energy consumption on CWLT and the approach tempera-
tonnage of the cooling tower increases from 0.701 kW/ton to
ture can be obtained. The lowest total energy consumption reveals the
0.925 kW/ton.
optimal operation point.
Since the wet-bulb temperature has a major impact on the perfor-
mance curves, the total power consumption curves are plotted at three 3.6. Optimization of the total power consumption
different wet-bulb temperatures levels [70°F (21.1 °C), 75°F (23.9 °C),
80°F (26.7 °C)]. Take the case of Tcw,l interval of (41, 42)°F or (5, The optimized approach temperature values are substituted into the
5.6)°C and the wet-bulb temperature of 70°F (21.1 °C) as an example. total power (kW) equation so that the minimum total power can be also
The power input per capacity (kW/ton), as a function of the approach achieved. Specifically, when the wet-bulb temperature is 70°F (21.1 °C),
temperature, is obtained by adding Eqs. (29) and (30). the total power is 31.6 kW; when wet-bulb temperature is 75°F
Pfan (kW/ton) = −0.0000356 ∗ (Tapp)3 − 0. 000904 ∗ (Tapp)2 (23.9 °C), the total power is 38.4 kW; when wet-bulb temperature is
80°F (26.7 °C), the total power is 73.7 kW. Repeat the above calculation
+ 0.0057 ∗ (Tapp) + 0.0181 (29) steps for other ambient conditions throughout the year 2015, and then
the optimized total power (kW) for the whole year can be obtained. The
Pchiller (kW/ton) = 0.00297992 ∗ (Tapp)2 − −0.09992197 ∗ (Tapp) sampled 68 days, used for NTU calculation and chiller modeling, were
+ 1.43860069 (30) also utilized to show the saving potential for the cooling system.
Fig. 14 plots the comparison between the actual and optimized total
If the approach temperature varies from 1°F (0.6 °C) to 30°F
power, and the power savings for these sampled 68 days. It shows that
(16.7 °C), a curve showing the corresponding power (kW) and the
the optimization methodology can help save more energy at the high-
power input per capacity (kW/ton) at different approach temperatures
temperature range, while the total power consumption after optimiza-
can be plotted. Similarly, the curves at eight other cases of different
tion is close to the actual power consumption at the low-temperature
wet-bulbs and chilled water leaving temperatures values are obtained.
range. The total power consumption and power savings for these
Fig. 13 shows that the kW/ton curves at different chilled water leaving
68 days are listed in Table 2.
temperatures are very close and a higher wet-bulb temperature even
Meanwhile, when this optimal approach temperature calculation
reduces the impact of the Tcw,l on these curves.
process is repeated for all ambient air wet bulb temperature values, the
When the wet-bulb temperature is 70°F (21.1 °C), the total power in
optimized approach temperature is plotted versus wet-bulb tempera-
kW/ton is listed in Eq. (31).
ture in Fig. 15, which is quite different from the actual reset schedule in
Total Power (kW/ton) = 0.0000356 ∗ (Tapp)3 + 0.00058175 ∗ (Tapp)2 Fig. 11.
Fig. 13. Comparison of kW/ton at different Tcw,l and Twb. Fig. 14. Optimized chiller plant power comparison.
7
J. Liao, et al. Energy Conversion and Management 199 (2019) 111950
8
J. Liao, et al. Energy Conversion and Management 199 (2019) 111950
reciprocating chillers and their relation to thermodynamic modeling and chiller Air-Conditioning Engineers);(United States) 96(CONF-9006117–) (1990).
design. Int J Heat Mass Transf 1996;39(11):2195–204. [46] Graves RD. Thermodynamic modeling and optimization of a screw compressor
[42] Gordon J, Ng KC, Chua HT. Centrifugal chillers: thermodynamic modelling and a chiller and cooling tower system. Texas A&M University; 2004.
diagnostic case study. Int J Refrig 1995;18(4):253–7. [47] Cortinovis GF, Paiva JL, Song TW, Pinto JM. A systemic approach for optimal
[43] Dijk J, POLL C, Poldervaart P. Investment in cooling tower control pays big divi- cooling tower operation. Energy Convers Manage 2009;50(9):2200–9.
dends. Process Eng 1985;66(10):57–60. [48] Xie X. Modeling and Optimization of the Cooling Tower under Unsaturated Outlet
[44] Schwedler M, Bradley B. Departments-data digest “Uncover the hidden assets in Air Conditions, 2016.
your condenser water system”. HPAC Engineering 2001;73(11). 68-68. [49] U. DoE, Energyplus engineering reference, The reference to energyplus calculations
[45] Braun J, Diderrich G. Near-optimal control of cooling towers for chilled-water (2010).
systems, ASHRAE Transactions (American Society of Heating, Refrigerating and