New procedure for estimating seasonal energy efficiency ratio of chillers.

ABSTRACT

This paper proposes a new index of the seasonal energy efficiency for chillers. The new index, the CSE (chiller seasonal efficiency) index, can be used by building designers for specifying the required chiller performance to manufacturers. This index has an advantage in that it is adaptable to multiple-chiller systems by setting six rating points to consider the difference in the COP due to the entering condenser water temperature as well as the part load. In addition, the use of the weighted harmonic mean equation improves the accuracy in predicting the seasonal COP. In this paper, the rating condition for the CSE index is determined from the cooling load profile of office buildings in Japan. Using the CSE index, the advantage of a variable-speed compressor chiller is presented. In the final part of the paper, the effect of the difference in chiller sequence control is discussed.

INTRODUCTION

The seasonal energy efficiency ratio is an important index for predicting the energy efficiency of a chiller under actual operating conditions. It is also a useful tool for building designers when specifying the required chiller performance to manufacturers. In the USA, the Air-Conditioning and Refrigeration Institute (ARI 2003) has a standard for the performance rating of water-chilling packages using the vapor compression cycle (ARI Standard 550/590). In this standard, the seasonal energy efficiency of a chiller is expressed as the integrated part-load value (IPLV). IPLV is calculated from four measured coefficients of performance (COPs) at loads of 100%, 75%, 50%, and 25%. Further, in Japan, the Society of Heating, Air-Conditioning and Sanitary Engineers of Japan (SHASE) had proposed the use of the IPLV, but it has not been used commonly; rather, the COP at maximum capacity is usually used as the energy efficiency index of chillers. Since the efficiency of a chiller depends on the part-load ratio and the entering condenser water temperature (ECWT) and chillers in buildings are operated at part load during most of the operating hours, evaluation based on the seasonal energy efficiency is important for designing an energy-efficient HVAC system.

As an index for the energy efficiency of chillers, the IPLV has some limitations:

* The IPLV is designed for the part-load performance of single-chiller systems and does not consider multiple-chiller systems. In the present ARI standard, the weight of COP at 100% load is only 0.01, but it is supposed to be much larger in multiple-chiller systems and heat storage systems.

* Since the IPLV is designed for single-chiller systems, the ECWT is supposed to increase with the part-load ratio. In other words, the cooling load is in proportion with the wet-bulb temperature. However, this relationship is not always valid in multiple-chiller systems.

In addition, to apply the IPLV to Japanese buildings, the rating conditions should be modified in accordance with the Japanese weather and building characteristics.

In the first part of this paper, the annual distribution of the chiller operating conditions (part-load ratio and ECWT) of each chiller in single- and multiple-chiller systems is calculated from the cooling load data of Japanese office buildings. Using these data, the rating conditions of the chiller for estimating the seasonal energy efficiency of various multiple-chiller systems are proposed. In the second part of this paper, the shortcomings of the present IPLV formula are pointed out and a new equation is proposed. In the final part, the accuracy and applicability of the index are discussed.

PROCEDURE TO DEFINE THE RATING CONDITIONS OF THE CHILLER PART LOAD

In ARI's procedure to derive the IPLV formula, the following assumptions were used (ARI 2003):

* Modified ASHRAE temperature bin method for energy calculation was used.

* Weather data were a weighted average of 29 cities across the USA.

* Building types were a weighted average of all types.

* Operational hours were a weighted average of various operations.

* A weighted average of buildings with and without some form of economizer was used.

* The bulk of the load profile used in the last derivation of the equation was again used, which assumed that 38% of the building's load was the average internal load.

* One point is predetermined to be the design point of 100% load and 29.4[degrees]C (85[degrees]F). Other points were determined by distributional analysis of MJ-hours (ton-hours).

On the other hand, we defined the part-load and ECWT conditions for the rating from the actual cooling load profile of office buildings and local weather data. From hourly cooling load data, the chiller operating condition distribution, which consists of the part-load ratio and ECWT, is calculated by simulation of the cooling plant of a single-chiller system and some representative multiple-chiller systems.

Cooling Load Data Used in This Research

The cooling load profiles of 18 office buildings are used in this survey. These data are provided from the district heating and cooling plant in Tokyo and were recorded from April 2002 to March 2003 (Matsushima et al. 2005). The weather data during this period are used to calculate the wet-bulb air temperature. Figure 1 presents the duration curve of these cooling load data. In most of the buildings, the cooling load occurred throughout the year. The existence of cooling load during midnight hours is a special characteristic of these buildings, which are supplied with heat from the district heating and cooling plant. Except for this, since most of the large cities in Japan have the same climatic condition as Tokyo, these data are supposed to represent the cooling load profile of most of the office buildings in Japan.

[FIGURE 1 OMITTED]

Chillers Evaluated in This Study

To obtain the load of each chiller from the total building cooling load, a chiller sequence control simulation model was developed. The chillers simulated in this model are defined as follows:

* Three types of cooling plants are simulated: a single-chiller plant, a two-chiller plant, and a three-chiller plant.

* The total cooling capacity of the plant is assumed to be equal to the annual maximum cooling load of the building. In addition, every chiller in the multiple-chiller system is assumed to have the same capacity.

* Each chiller in the multiple-chiller plant is assigned a particular role. In the case of the three-chiller plant (as shown in Figure 2), the base chiller is started first and operated whenever a cooling load exists. When the cooling load exceeds the capacity of the base chiller, the middle chiller is operated. When the cooling load exceeds the total capacity of these two chillers, the peak chiller is operated.

* The rotating chiller switches its role on a daily basis. For example, in the three-chiller system, the rotating chiller alternately operates as the base, middle, and peak chiller. In this case, every chiller in the plant is assumed to have the same characteristics.

[FIGURE 2 OMITTED]

* As a result, eight kinds of chillers are evaluated: The chiller in the single-chiller system (1); the base (2-b), peak (2-p), and rotating (2-r) chillers in the two-chiller system; and the base (3-b), middle (3-m), peak (3-p), and rotating (3-r) chillers in the three-chiller system.

Chiller Sequence Control Simulation

The flow of the chiller sequence control simulation is as follows:

* Based on the cooling load measured every hour, the number of operational chillers is decided. The cooling load is equally divided into the operational chillers. The chiller control sequence is indicated by the solid line (ideal case) in Figure 2.

* The ECWT is calculated from the wet-bulb temperature and the amount of exhaust heat from the condenser of the chiller using the cooling tower model. The exhaust heat from the chiller is the sum of the cooling load and the energy consumption of the chiller (compressor) and is a function of the chiller's COP. On the other hand, the chiller's COP is a function of the cooling load and ECWT. Therefore, the ECWT and COP of the chiller must be calculated simultaneously using the cooling tower model and chiller model, respectively. The relationship among the part load, ECWT, and COP of the turbo chiller and inverter-driven variable-speed compressor turbo chiller (inverter turbo chiller) is shown in Figure 3. The relationship among the outdoor wet-bulb temperature, temperature of water entering the cooling tower, and ECWT is shown in Figure 4.

[FIGURE 3 OMITTED]

* To ensure stable operation of the chiller, most cooling plants in Japan control the speed of the cooling tower fan to maintain the ECWT above a certain value. This is the common practice in Japan, and one of the reasons is that most of the cooling plants have at least one absorption chiller that must be operated at high ECWT. Therefore, we set 22[degrees]C (71.6[degrees]F) as the minimum ECWT. This value should be reconsidered when this index is applied in other countries.

* When the part load of the chiller is less than 20%, intermittent operation at a part-load ratio of 20% is assumed.

* Repeating these calculations throughout the year, the part-load ratio, ECWT, and COP of the chiller per hour are calculated for all the chillers.

* In this simulation, each chiller's seasonal COP (SCOP), which refers to the ratio of annual cooling load to the annual energy consumption of the chiller, is also calculated for validating the proposed index.

Rating Points of the Chiller

Figure 5 shows the result of the chiller sequence control simulation for the base, middle, and peak chillers in the three-chiller system for one of the buildings. The characteristics of the turbo chiller, as shown in Figure 3, are used for this simulation. Each plot in these figures shows the operating part-load ratio and ECWT per hour (operating point).

In the ARI's procedure to derive the IPLV formula, these operating points are classified into four groups only by the part-load ratio. After that, the representative point (rating condition) and the duration of operation (weight) are calculated for each group. However, from Figure 5, for the base and middle load, it is clear that many operating points exist in areas where the ECWT is low and the part-load ratio is high. These points never appear in the case of the single-chiller system. Further, this area is supposed to be important when a cold-water storage tank is used. Therefore, we propose that each of the two higher part-load groups be divided into two groups by the ECWT to accurately express the difference in the COP of the chiller due to the ECWT.

[FIGURE 4 OMITTED]

[FIGURE 5 OMITTED]

Based on the distribution of the operating points for each chiller and building, we divided the group based on part loads of 25%, 50%, and 70% and on an ECWT of 26[degrees]C (78.8[degrees]F) to maintain a significant weight for each group of at least one type of chiller. Therefore, we defined the following six groups based on the operating conditions:

[FIGURE 6 OMITTED]

* Group 1: part-load ratio is 70%-100%, ECWT is higher than 26[degrees]C (78.8[degrees]F)

* Group 2: part-load ratio is 70%-100%, ECWT is lower than 26[degrees]C (78.8[degrees]F)

* Group 3: part-load ratio is 50%-70%, ECWT is higher than 26[degrees]C (78.8[degrees]F)

* Group 4: part-load ratio is 50%-70%, ECWT is lower than 26[degrees]C (78.8[degrees]F)

* Group 5: part-load ratio is 25%-50%

* Group 6: part-load ratio is 20%-25%

For each building and type of chiller, the part-load ratio and ECWT of the operating points in each group are averaged. The result of the averaged operating points for the cooling load profiles of 18 buildings and eight chiller types are shown in Figure 6. With the exception of a few points, the distribution of the averaged points is small. Therefore, the rating condition for each group is obtained by re-averaging these averaged operating points. The result is shown in Tables 1a and 1b. In addition, the weight of each group is calculated according to the ratio of the sum of the cooling loads in each group to the annual cooling load. Table 2 shows the weight for each group and chiller type. It can be seen that each group has a weight of more than 10% for at least one type of chiller.

By rounding off the "total average" values in Table 1, we obtain the rating condition for groups 1 to 6 as shown in Table 3. Even if the characteristics of the inverter turbo chiller, as shown in Figure 3, are used, the total average values do not change. For each rating point, the COP of the chillers calculated from Figure 3 is also shown in Table 3.

DEFINITION OF THE CSE INDEX AND ITS APPLICATION

CSE Formula

In the IPLV formula, the weighted arithmetic mean equation is used for calculating the IPLV value from four measured COPs at loads of 100%, 75%, 50%, and 25%, as shown in Equation 1.

IPLV = 0.01A + 0.42B + 0.45C + 0.12D (1)

where

A = COP at 100% load and an ECWT of 29.4[degrees]C (85[degrees]F)

B = COP at 75% load and an ECWT of 23.9[degrees]C (75[degrees]F)

C = COP at 50% load and an ECWT of 18.3[degrees]C (65[degrees]F)

D = COP at 25% load and an ECWT of 12.8[degrees]C (65[degrees]F)

However, by definition,

SCOP = [annual total cooling load]/[annual total energy consumption]

= [[n.summation over (i = 1)] [Qc.sub.i]]/[[n.summation over (i = 1)][E.sub.i]] = [[n.summation over (i = 1)][Qc.sub.i]]/[[n.summation over (i = 1)][[Qc.sub.i]/[CO[P.sub.i]]]]=[[n.summation over (i = 1)][Qc.sub.i]]/[[6.summation over (j = 1)] ([[r.sub.j].summation over (i = 1)] [[Qc.sub.i]/[CO[P.sub.i]]])], (2)

where

SCOP = seasonal COP;

[Qc.sub.i] = cooling load at time i, kWh;

[E.sub.i] = energy consumption at time i, kWh;

CO[P.sub.i] = COP of the chiller at time i;

n = total duration of the chiller operation, h;

[r.sub.j] = total duration of the chiller operation in group j, h;

j = group number (j = 1 to 6).

If the difference in COP in each group is ignored and substituted by the measured value at the rating point of each group shown in Table 1, Equation 2 becomes

SCOP = [[n.summation over (i=1)][Qc.sub.i]]/[[6.summation over (j=1)]([[r.sub.j].summation over (i=1)][Qc.sub.i]]/[bar.CO[P.sub.j]]), (3)

=1/[[6.summtion over (j = 1)]([[[r.sub.j].summation over (i = 1)][Qc.sub.i]/[n.summation over (i = 1)][Qc.sub.i]]/[bar.CO[P.sub.j]])] = 1/[[6.summation over (j=1)]([W.sub.j]/[bar.CO[P.sub.j]])]

where

[bar.CO[P.sub.j]] = COP of the chiller measured at the rating point of group j and

[W.sub.j] = weight of group j.

Equation 3 shows that the SCOP should be estimated by the weighted harmonic mean of COPs at each rating point. To emphasize the difference between the IPLV formula and our formula, we named our formula CSE (chiller seasonal efficiency). CSE is calculated by the following equation:

CSE = 1/[[[W.sub.1]/[bar.CO[P.sub.1]]] + [[W.sub.2]/[bar.CO[P.sub.2]]] + [[W.sub.3]/[bar.CO[P.sub.3]]] + [[W.sub.4]/[bar.CO[P.sub.4]]] + [[W.sub.5]/[bar.CO[P.sub.5]]] + [[W.sub.6]/[bar.CO[P.sub.6]]]] (4)

CSE Index Under Standard Condition

Figure 7 shows the CSE index calculated using Equation 4 with the COPs of the chiller shown in Table 3 and the weights shown in Table 2 for the eight types of chillers. In addition to the conventional turbo chiller, as shown in Figure 3a, the inverter turbo chiller (variable-speed compressor), as shown in Figure 3b, is also evaluated. The COP at 100% load and the IPLV index calculated using Equation 1 are also shown in Figure 7. Since the minimum ECWT is set at 22[degrees]C (71.6[degrees]F) in this study, the COP at part loads of 25% and 50% is calculated at this temperature. Therefore, the term NPLV (non-standard part-load value) is used instead of IPLV in Figure 7. The conventional turbo chiller shows a very low CSE value in the case of the chiller types mainly operating at lower part loads. For example, CSE values of the chiller in the single-chiller plant and the base chiller in the two-chiller system are 3.68 and 4.31, respectively. The maximum difference of CSE value between the chiller types is 40%. Since the NPLV does not reflect this difference, the CSE index is advantageous.

The CSE values for the inverter turbo chiller (variable-speed compressor turbo chiller) are almost constant for all the chiller types. This means that the use of the inverter turbo chiller is effective for mainly operating at low part loads, but the advantage is small for chillers always operating at high part loads, such as peak chillers. Evaluation by the CSE index can suggest these recommendations to the HVAC designer. The NPLV value is much larger than these CSE values.

Validation of CSE index

To validate the superiority of the CSE based on the weighted harmonic mean of COPs in comparison to the IPLV index, three types of SCOPs are compared for a specified building (building A shown in Figure 1). The first one is the CSE index calculated from the weights calculated for building A by the chiller sequence control simulation model using Equation 4. The second one is the IPL[V.sub.six], which is the weighted arithmetic mean given by Equation 5 using the same weights and COPs as the CSE equation.

[FIGURE 7 OMITTED]

IPL[V.sub.six] = [W.sub.1] x CO[P.sub.1] + [W.sub.2] x CO[P.sub.2] + [W.sub.3] x CO[P.sub.3] + [W.sub.4] x CO[P.sub.4] + [W.sub.5] x CO[P.sub.5] + [W.sub.6] x CO[P.sub.6] (5)

The third one is the SCOP that is calculated as the ratio of annual total cooling load to annual total energy consumption (the third term of Equation 2) from the hourly cooling load and electricity consumption simulated by the chiller sequence control simulation model. This result is shown in Figure 8. The IPL[V.sub.six] index becomes larger than the simulated SCOP by approximately 10% in the case of chiller types mainly operating at low part loads. On the other hand, the CSE index shows a very good agreement with the simulated SCOP.

This result shows that the weighted arithmetic mean equation is overestimated when at least one COP is considerably lower than the other COPs. Therefore, it can be concluded that the CSE, which uses the harmonic mean equation, is more accurate for estimating the SCOP of chillers.

Application of CSE to Realistic Chiller Sequence Control

For the chiller sequence control simulation used to calculate the weights shown in Table 2, the ideal situation is assumed, as indicated by the bold line in Figure 2. However, the part load of the chiller, which is actually operated in the cooling plant, is lower than the ideal situation. The reasons are as follows (Shimoda et al. 2005):

* Usually, the total capacity of the plant is designed to be larger than the peak load by the "safety factor."

* In order to respond to a sudden increase in the cooling load and to avoid short-cycling of the chiller operation and since the temperature difference between the supply and return water is smaller than the designed value, chillers are operated such that they have a "margin of load factor."

[FIGURE 8 OMITTED]

Setting the safety factor as 20% of the peak load and the margin of load factor as 20% of the chiller capacity, the chiller sequence control is simulated for the conventional turbo chiller, as indicated by the dashed line in Figure 2, which corresponds to the actual case.

Figure 9 compares the three estimation results for the eight chiller types and building A as follows:

* CSE index calculated from the standard weights shown in Table 2.

* CSE index calculated from the weights obtained by the actual chiller sequence control model for building A.

* Simulated SCOP in the case of actual chiller sequence control.

By applying the actual chiller sequence control, the SCOP decreases by about 5% for each chiller type. The CSE index calculated from the weights obtained by the actual chiller sequence control program for building A shows good agreement with the simulated SCOP. This fact suggests that if we can predict only the weights for each group, it would be possible to estimate the CSE index correctly without changing the rating points.

CONCLUSION

In this paper, a new index for estimating the seasonal energy efficiency of chillers is proposed. This index has some advantages as compared to the conventional IPLV index, as follows:

1. This index is adaptable to multiple-chiller systems by setting six rating points for considering the difference in the COP due to the ECWT.

[FIGURE 9 OMITTED]

2. The use of the weighted harmonic mean equation improves the accuracy in predicting the SCOP as compared with the weighted arithmetic mean equation adopted by the IPLV equation.

3. This index can be flexibly applied to various patterns of chiller operation by setting the weights predicted by the chiller sequence control simulation. For this purpose, a detailed simulation of the COP and heat balance of the chiller and cooling tower, respectively, is not required. Only the part load and ECWT distribution of each chiller is necessary. Therefore, this index is useful to select the optimum chiller depending on the chiller type in the cooling plant and the cooling load characteristics of the building.

To develop this index as the standard, more discussion is needed about the "standard" procedure to define the rating conditions, such as the representativity of the cooling load data, weather data, chiller control sequence, definition of the minimum ECWT, and so on.

ACKNOWLEDGMENTS 

This work was conducted under the SHASE (The Society of Heating, Air-Conditioning and Sanitary Engineers of Japan) research project, "Performance Requirement Method of HVAC Equipment." The authors wish to thank the members of the project for their valuable advice.

REFERENCES

ARI. 2003. ARI Standard 550/590-2003, Performance Rating of Water-Chilling Packages Using the Vapor Compression Cycle. Air-Conditioning and Refrigeration Institute, Arlington, VA.

Matsushima, T., T. Akiba, T. Nobe, T. Sato, Y. Tanaka, K. Fujii. 2005. Study on building cooling and heating load database, Part 1: Prototype of building heat and chilled load data sheet. Technical Papers of Annual Meeting, The Society of Heating, Air-Conditioning and Sanitary Engineers of Japan, pp.1705-08 (in Japanese).

Shimoda, Y., T. Nagota, N. Isayama, and M. Mizuno. 2005. Verification of energy efficiency of district heating and cooling system using realistic parameters. Proceedings of the Ninth International IBPSA Conference, pp. 1123-30.

Yoshiyuki Shimoda, DEng

Nattapon Choonchuachan

Minoru Mizuno, DEng

Yoshiyuki Shimoda is an associate professor and Minoru Mizuno is a professor in the Division of Sustainable Energy and Environment, Osaka University, Osaka, Japan. Nattapon Choonchuachan is an engineer with Fuji Seiki Co., Ltd., and a former graduate student at Osaka University.

Table 1a. Averaged Operating Point for Each Group and Chiller Type (Load
[%], ECWT [[degrees]C])

                          Chiller Type
         1           2-b         2-p         2-r         3-b

Group 1  81.2, 30.7  84.0, 29.8  80.6, 30.4  83.0, 30.1  83.5, 29.7
Group 2  79.3, 22.7  78.7, 24.4  75.6, 24.3  79.0, 24.5  80.5, 23.6
Group 3  60.1, 28.4  60.5, 28.4  60.9, 28.3  60.6, 28.4  61.4, 28.4
Group 4  53.9, 24.6  57.7, 23.3  55.7, 24.4  57.3, 23.5  57.9, 22.9
Group 5  34.7, 23.6  34.7, 22.7  --,   --*   34.6, 22.7  35.6, 22.7
Group 6  20.3, 22.3  20.4, 22.3  --,   --    20.4, 22.3  21.2, 22.2

                          Chiller Type
         3-m         3-p         3-r         Total Average

Group 1  82.7, 29.8  80.7, 30.4  82.7, 30.0  82.3, 30.1
Group 2  76.5, 24.7  76.1, 24.2  80.2, 23.7  78.2, 24.0
Group 3  62.1, 28.4  68.3, 29.0  62.1, 28.5  62.0, 28.5
Group 4  58.4, 23.6  68.5, 24.9  58.1, 23.0  58.4, 23.8
Group 5  --,   --    --,   --    35.7, 22.7  35.1, 22.9
Group 6  --,   --    --,   --    21.2, 22.2  20.7, 22.3

* "--" indicates that no operating point exists for this group and
chiller type.

Table 1b. Averaged Operating Point for Each Group and Chiller Type (Load
[%], ECWT [[degrees]F])

                          Chiller Type
         1           2-b         2-p         2-r         3-b

Group 1  81.2, 87.3  84.0, 85.6  80.6, 86.7  83.0, 86.2  83.5, 85.5
Group 2  79.3, 72.9  78.7, 75.9  75.6, 75.7  79.0, 76.1  80.5, 74.5
Group 3  60.1, 83.1  60.5, 83.1  60.9, 82.9  60.6, 83.1  61.4, 83.1
Group 4  53.9, 76.3  57.7, 73.9  55.7, 75.9  57.3, 74.3  57.9, 73.2
Group 5  34.7, 74.5  34.7, 72.9  --,   --*   34.6, 72.9  35.6, 72.9
Group 6  20.3, 72.1  20.4, 72.1  --,   --    20.4, 72.1  21.2, 72.0

                          Chiller Type
         3-m         3-p         3-r         Total Average

Group 1  82.7, 85.6  80.7, 86.7  82.7, 86.0  82.3, 86.2
Group 2  76.5, 76.5  76.1, 75.6  80.2, 74.7  78.2, 75.2
Group 3  62.1, 83.1  68.3, 84.2  62.1, 83.3  62.0, 83.3
Group 4  58.4, 74.5  68.5, 76.8  58.1, 73.4  58.4, 74.8
Group 5  --,   --    --,   --    35.7, 72.9  35.1, 73.2
Group 6  --,   --    --,   --    21.2, 72.0  20.7, 72.1

* "--" indicates that no operating point exists for this group and
chiller type.

Table 2. Calculated Weights for Each Group and Chiller Type

                             Chiller Type
         1     2-b   2-p   2-r   3-b   3-m   3-p   3-r

Group 1  0.19  0.24  0.55  0.29  0.24  0.58  0.87  0.35
Group 2  0.00  0.06  0.01  0.06  0.12  0.06  0.01  0.11
Group 3  0.11  0.09  0.31  0.13  0.06  0.14  0.11  0.09
Group 4  0.04  0.13  0.13  0.13  0.19  0.21  0.00  0.18
Group 5  0.26  0.23  0.00  0.20  0.19  0.00  0.00  0.14
Group 6  0.40  0.24  0.00  0.20  0.19  0.00  0.00  0.13

Table 3. Proposed Rating Condition and COP of the Chiller

                          ECWT [[degrees]C]  COP (Turbo  COP (Inverter
Group      Part load [%]  ([[degrees]F])     Chiller)    Turbo Chiller)

Group 1     80.0          30.0 (86.0)        6.09        6.39
Group 2     80.0          24.0 (75.2)        6.96        8.45
Group 3     60.0          28.5 (83.3)        5.69        6.29
Group 4     60.0          24.0 (75.2)        6.39        8.50
Group 5     35.0          23.0 (73.4)        4.54        7.06
Group 6     20.0          22.5 (72.5)        2.54        5.29
Full load  100.0          32.0 (89.6)        5.92        5.65