Effect of load-sharing operation strategy on the aggregate performance of existed multiple-chiller systems

Effect of load-sharing operation strategy on the aggregate performance of existed multiple-chiller systems

Applied Energy 135 (2014) 329–338 Contents lists available at ScienceDirect Applied Energy journal homepage: www.elsevier.com/locate/apenergy Effec...

769KB Sizes 1 Downloads 36 Views

Applied Energy 135 (2014) 329–338

Contents lists available at ScienceDirect

Applied Energy journal homepage: www.elsevier.com/locate/apenergy

Effect of load-sharing operation strategy on the aggregate performance of existed multiple-chiller systems Hosny Z. Abou-Ziyan, Ali F. Alajmi ⇑ College of Technological Studies, Mechanical Engineering Department, P.O. Box 3831, Mishref, Kuwait

h i g h l i g h t s  The tested operation strategies obtain aggregate performance from 1.06 to 1.41.  The load-sharing operation strategies have been affected by operation parameters.  The best load-sharing operation strategy is better than conventional by 22–33%.  COP of individual chiller ranged from lower than 1 to about 5.  Testing different operation strategies involves monitoring the COP of chillers.

a r t i c l e

i n f o

Article history: Received 24 July 2013 Received in revised form 9 May 2014 Accepted 28 June 2014

Keywords: Chiller operation strategies Chiller performance Load-sharing operation strategy Part and full load operation

a b s t r a c t This paper examines the effect of load-sharing operation strategies on the aggregate performance for existing multiple-chiller system under different partial loads and environmental conditions. The various tested load-sharing operation strategies obtain aggregate performance from 1.64 to 2.18 during the day hours and from 1.06 to 1.41 for the full day indicating significant effect of the operation strategies on the aggregate performance. The conventional (same part load ratio) strategies attain aggregate performance that is lower than the best strategy by 22–33%. At very low system partial load, the performance of the multiple-chiller system falls to less than quarter its value at large load whereas the performance of individual chiller drops to about one sixth of its large load value. The load sharing strategy is influenced by many parameters such as the condition of the chillers and compressors, the piping arrangement, and the heat loss from the chilled water piping where these parameters may overwhelm the individual chiller performance. Accordingly, the load-sharing operation strategy may vary from case to another and should be periodically examined to verify proper system operation, rectify the existing chiller performance and identify chiller faults. Therefore, the need for maintenance can be predicted and the standby chiller may be eliminated. Ó 2014 Published by Elsevier Ltd.

1. Introduction In hot climate countries, the use of air-conditioning equipment during summer is essential in every building which is reflected on large power consumption. According to official report, heating, ventilating, and air-conditioning (HVAC) sector consumes about 70% of the national generated power [1]. Thus, efficient design and operation of HVAC systems may result in substantial energy saving both nationally and worldwide particularly in hot climate countries. For large buildings, multiple chiller systems are practiced to operate in parallel to meet large cooling requirements. The aggre⇑ Corresponding author. Tel.: +965 22314200; fax: +965 24811753. E-mail address: [email protected] (A.F. Alajmi). http://dx.doi.org/10.1016/j.apenergy.2014.06.065 0306-2619/Ó 2014 Published by Elsevier Ltd.

gate performance of chiller systems depends on dynamic factors such as heat rejection medium, ambient conditions and compressor efficiency in addition to another important factor which is the load carried by each operating chiller (load sharing). Bekker and Carew [2] stated that there is little understanding of the factors that influence chillers performance, due to many interrelated variables. Many studies were conducted to improve the chiller system performance. Sun et al. [3] proposed an optimal strategy for operating chillers in steps in response to the changing building cooling loads. The Gordon and Ng’s thermodynamic model correlates the chiller COP with the temperature monitored at the evaporator and condenser sides [4]. Chan and Yu [5] analyzed how the chiller component interact with each other and discussed the use of

330

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

Nomenclature cw COP mcw msw RH Tci Tco Tsi Tso

specific heat of chilled water (J/kg K) coefficient of performance (–) mass flow rate of the chiller chilled water (kg/s) mass flow rate of the system chilled water (kg/s) relative humidity (–) inlet chilled water temperatures to the chiller (°C) outlet chilled water temperature from the chiller (°C) inlet chilled water temperatures to the system (°C) outlet chilled water temperatures from the system (°C)

floating condensing temperature control to improve chiller performance. Various approaches are reported in literature to handle the complexity of the many interrelated factors that affect the chillers performance. Some studies coupled empirical models and genetic algorithms with other advanced techniques to optimize physical and operating characteristics of chillers [6–9]. Other studies introduced approach based on statistical analysis to investigate the effectiveness of these parameters [10–15]. Among the parameters that highly influence the multiple-chiller aggregate performance is the load-sharing operation strategy. It addresses how the building cooling load should be allocated to individual chillers operating in order to optimize their aggregate performance. Usually, the multiple-chiller systems are integrated with sequencing control to react to the building load demand [16]. It determines how many and which chillers are to be put into operation according to the instantaneous building cooling load. Under the arrangement of chiller sequencing, all operating chillers are running at the same part load ratio, and no additional chillers start to operate until each of the operating chillers is running at full load [17,18]. In a conventional pumping system, different numbers of chillers and constant-speed pumps operate in pairs. In such configuration the total flow of chilled water should satisfy various building cooling loads. Since the flow of chilled water passing through individual chillers is fixed, the load which each of the chillers carries is related directly to the temperature rise of the chilled water. In order to achieve the same chilled water temperature across all the chillers, they have to operate at the same part load ratio. Such an even load sharing strategy has long been used in multiple-chiller systems for equal size or different size chillers. Instead of operating chillers at the same part load ratio, it is preferable to develop load sharing strategies that accounting for their part load performance characteristics to maximize the COP of the entire system. Yet there is limited research work relating to load sharing strategies for multiple-chiller system [19]. Chang et al. [17] employed a genetic algorithm to achieve the optimal chiller loading problem for two centrifugal chiller plants serving a semiconductor plant and a hotel. The genetic algorithm was utilized to search for their optimum loading points for any given system load. The authors pointed out that the optimum loading points mean that the individual chillers should operate at different part load ratios without explaining the rational for uneven load allocation to the chillers. Austin [20] emphasized that optimum loading points for the part load operation of chillers should be based on their performance rating at any given constant condenser water temperature, rather than on the part load performance curve at ARI rating conditions [21]. Yet, the author did not justify the use of individual chillers at different part load ratios to meet the system load. Kaya [22] evaluated the optimum chiller load allocation considering

Abbreviations AHU air handling unit ARI air conditioning and refrigeration institute CAV constant air volume FCU fan-coil-unit HVAC heating, ventilation and air conditioning VAV variable air volume

cost-versus-load characteristics as an objective function. A maximum saving of 24% in the cost per ton of refrigeration is achieved using the optimum load allocation compared to the equal percentage load allocation. Therefore, the aim of this paper is to experimentally investigate the effect of load sharing strategies for existing multiple-chiller system on their aggregate performance. Different load-sharing operation strategies are suggested, tested and compared based on their aggregate performance. The work is conducted on existing multiple-reciprocating chillers system under different partial loads and environmental conditions. In order to study the load effect on chiller and system performance under constant chiller capacity, a fixed number of compressors remain working during the experiments. The proposed approach is easy to implement as it requires only power consumption measurement and needs no additional cost. In addition, elimination of standby chiller may be possible by the use of performance monitoring to predict chiller malfunction prior to any failure or breakdown. In order to achieve the above goals, the building and chillers were equipped with the necessary instrumentation to evaluate the performance of the chillers under the tested conditions.

2. Characteristics of the building and HVAC system In order to understand the building systems, both architectural and mechanical drawings, equipment specifications and control sequences of chiller system in the building are scrutinized. The following description is based on a review of existing documents and site survey. 2.1. Building description The building under consideration is the Mechanical Engineering Department, College of Technological Studies, Kuwait. The building is a two-story institutional facility with total floor area of 7020 m2. The long side of the building is oriented toward east–west direction with four main entrance doors in the east side. The building is used by about 350 students, and 50 staff with irregular occupancy pattern between 8 am and 5 pm, five days per week during the academic semester. Less number of people uses the building during summer semester whereas no occupancy occurs during summer vacation. The building wall construction can be considered as heavy mass and good insulation with overall heat transfer coefficient of 0.562 W/m2 K. The roof of this building was made from light mass construction that is well insulated (0.187 W/m2 K). The windows and entrance doors are aluminum framed constructed from 6mm double-tinted glazing with overall heat transfer coefficient of 3.42 W/m2 K. The lighting in the building space is predominantly 4-foot fluorescents T12 lamps with electronic ballasts and manual

331

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

switches whereas the building equipment are mainly for computer, laboratory and workshop facilities. 2.2. HVAC system characteristics The installed HVAC system in the building consists of four aircooled reciprocating (semi-Hermetic) chillers, three chillers should be in duty and one is on standby. The chillers use environmental friendly refrigerant R-407c with rated capacity of 429 kW. Each chiller comprises four refrigeration circuits that may work in parallel, a shell-and-tube evaporator of the flooded type and air-cooled condenser with constant-speed fans. Each refrigeration circuit includes one reciprocating compressor and one electronic expansion valve. Since the circuit may operate at full or partial capacity via unloaded option in one of the four compressors, the cooling capacity can be controlled in 12 steps. The chilled water mass flow rate through the chiller is designed at 17.0 kg/s with supply temperature of 7.5 °C and temperature rise of 5.5 °C at full load. The air distribution system consists of fourteen air-handling units (AHUs), constant air volume (CAV), which serves the staffs’ offices, classes, and laboratories; thirty fan-coil-units (FCUs) that serves the workshops whereas the only portion of the building that is designed to utilize a variable air volume (VAV) is the administrative section. 3. Experimental program The experimental program of this work starts by testing different operation strategies based on their integrated system performance in order to determine the most efficient partial load sharing operation strategy of the chillers. All the tested operation strategies are conducted on constant cooling capacity that was achieved by operating 6 compressors (less than 2 chillers). The constant cooling capacity may be achieved by operating the same number of compressors from different chillers assuming that they are working at the same condensing and evaporating temperatures and supply similar chilled water temperatures. It should be noted that the tests are conducted during a period when the six running compressors were fairly adequate to carry out the building load and ensure that the desired indoor temperature is maintained. The preparation and installation of instrumentation were made during weekends to avoid any troubles to the building service during experimentation. In order to investigate the effect of different possible operation strategies, these six compressors are selected from 2 or 3 of the designated chillers. Among the suggested arrangements are; two compressors from each chiller or three compressors from two chillers (same part load ratio) and four compressors from one chiller (full load) and two compressors from another chiller (half load). Each operation strategy is designated according to the number of compressors that are used from chillers 1, 2 and 3, i.e. when two compressors are operated from each chiller, this will be given a token 2-2-2 strategy, other tested strategies include two-one-three (2-1-3), three-three-zero (3-3-0), or four-two-zero (4-2-0). Different combinations of these strategies were tested to evaluate their effect on energy consumption and aggregate system performance and to find out the best load-sharing operation strategy for the chillers. Each possible strategy has been tested repeatedly by alternating the running chillers, i.e. in the case of 3-3-0 means that chillers 1 and 2 are running with 3 compressors each while chiller number 3 stays off. For the same strategy, chiller 1 or chiller 2 may become off while the other two chillers are running with three compressors each (0-3-3 and 3-0-3). This has been done for all possible strategies while the hourly energy consumption of chillers,

pumps, and air-handling units (AHUs) is recorded by the main building meter. Also, the corresponding hourly inlet and outlet chilled water temperatures are recorded. Each strategy was applied for 2 days where the system performance during the busy hours (8 am–3 pm), day time (7 am–7 pm), night hours (7 pm– 7 am) and the average daily performance were computed using the following equation:

aggregate system performance ¼

system cooling capacity system energy consumption ð1Þ

The system energy consumption accounts for the whole system including the AHUs and pumps. The system cooling capacity is the cooling produced by all working chillers that is obtained using Eq. (2). Under steady state where the conditions inside the building kept unchanged and the operation of the chillers system is not altered, the difference between the building load and the system cooling capacity is negligible.

_ sw cw ðT si  T so Þ building load  system cooling capacity ¼ m

ð2Þ

where msw the mass flow rate of the chilled water which is kept constant at 17 kg/s; cw is the specific heat of chilled water and Tsi and Tso are the inlet and outlet chilled water temperatures across the system which range from about 6 to 11 °C. The experimental error for aggregate system performance that is evaluated according to Moffat [24] shows maximum uncertainty in chilled water temperatures, system consumption and aggregate system performance of 1.67%, 2.70%, and 5.27%, respectively. In order to determine the appropriate load-sharing operation strategy for multiple chiller system, the control is disabled and manual operation is allowed to examine the different sequences of operation. Then, the new achieved load-sharing operation strategy could be implemented in the control system. In addition to the examination of the different load-sharing operation strategies, the effect of partial load operation on the performance of individual chiller is also studied. This was carried out to provide more explanation for the multiple-chillers performance results, particularly at part loads. The characteristics of the chiller are measured for 2 consecutive days during day and night hours for variable operating conditions such as the ambient temperature and the chilled water inlet temperatures. One compressor of the chiller deliberately runs all the time to monitor the performance at different loads while the chiller capacity remains constant. In order to evaluate the characteristics of the chillers, a number of sensors have been installed in each compressor to measure the refrigerant pressures and temperatures at suction and discharge points of a compressor and at the condenser exit for a chiller. Also, the compressor power, the chilled water flow rate, inlet and outlet chilled water temperatures along with ambient temperature and relative humidity are measured and recorded using another set of instruments. These readings were recorded every 5 min and averaged over a period of 30 min. The data are collected for steady state operation that was assumed when the variations of readings within the time interval is less than 0.5 °C. The chiller cooling capacity and chiller coefficients of performance (COP) were computed according to the following equations:

_ cw cw ðT ci  T co Þ chiller cooling capacity ¼ m

ð3Þ

chiller cooling capacity compressors electric power

ð4Þ

COP ¼

where mcw the mass flow rate of the chilled water; cw is the specific heat of child water and Tci and Tco are the inlet and outlet chilled water temperatures across the chiller (range 12–20 °C, see Fig. 7).

332

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

The cooling capacity of the chiller, as defined by Eq. (3), was calculated from the chilled water side to ensure higher accuracy than that calculated from refrigerant side. The necessary measuring instruments for the experimental operating parameters such as temperature, pressure, flow rate, and running current are calibrated and installed. For the chilled water temperature measurements the already existed holes/plugs in the pipelines are used. An ultrasonic meter is used to measure the chilled water flow rate and an energy meter is installed to measure the chiller consumption. All the measuring points are acquired using a data acquisition system and its compatible software. Calibration of the temperature sensors along with the use of the data acquisition system have made temperature readings accurate to ±0.1 °C for chilled water and refrigerant temperatures, ±0.04 bar for suction and discharge refrigerant pressures and ±3.5% for relative humidity. The dry bulb temperature and relative humidity are recorded with their own Mini Data Logger (HOBO U-12-012) [23]. The experimental error for the performance of individual chiller is also evaluated according to Moffat [24]. The maximum uncertainty in refrigerant (10 to 55 °C) and chilled water temperatures, chiller consumption and COP are 1.0%, 0.83%, 0.0025% (4 digits meter) and 2.66%, respectively.

sors) in the building. The different strategies discussed in Section 3 (2-2-2, 1-2-3, 3-0-3 and 4-2-0) were tested and evaluated. 4.1.1. Hourly characteristics Typical variations of the hourly ambient temperature, relative humidity and global incident and reflected solar radiation, while conducting the strategy 4-2-0 are shown in Fig. 1. The ambient temperature varies from about 35 °C at 7 am to over 45 °C between 1 and 3 pm then decreases to around 41 °C at 8 pm. The effect of relative humidity during that day is limited as it varies from about 5% between 2 and 3 pm to about 13% in the early morning. Both incident and reflected solar radiation follow the same trend showing maximum values of about 3300 for incident and 1150 W/m2 for reflected solar radiation between 11 am and 1 pm. The effect of global solar radiation turns out to be negligible after 6 pm as the values for both types of radiation become small. Fig. 2 shows the changes in the hourly building cooling load, system consumption and aggregate performance corresponding to the environmental data shown in Fig. 1 for the same strategy 4-2-0. At 7 am the temperature inside the building is re-set at 24 °C instead of 28 °C during the night. This creates relatively large building load in the early hours of the morning that decreases up to 10 am. Thereafter, the building cooling load increases between 11 am and 3 pm, and then decreases. The thermal capacity of the building causes some time lag between the building cooling load and the environmental conditions. Thus, the maximum cooling load occurs around 3 pm while the maximum ambient temperature takes place between 1 and 3 pm and the maximum radiation is around noon time. The pattern of the cooling load is due to the combined effects of ambient temperature, global solar radiation, setback temperature during the night and the thermal inertia of the building. The heat stored in the building during the night because of the temperature setback to 28 °C causes the system consumption to increase between 7 am and 10 am while the building cooling load is decreasing (Fig. 2). Then the consumption continues to increase because of the increase in the building cooling load between 10 am and 1 pm. Afterwards, the consumption slightly decreases until 7 pm. The consumption shows opposite trend to that of the performance between 7 am and 4 pm, then the effective decrease in the building cooling load overcomes that of COP causing the consumption to decrease. While the COP starts at its highest value at 7 am because of resetting the temperature inside the building, the decrease in the building cooling load and the increase in the consumption cause the aggregate system performance to decrease between 7 am and noon time. Then the COP increases up to 4 pm then starts to decrease because of the effective decrease in the building load.

4. Results and discussion The results presented in this section include the search for the load-sharing operation strategy that accomplishes the best aggregate performance of the multiple-chiller system under its current internal condition. The results are also explained in light of the characteristics of individual chiller under different operating conditions when constant chiller capacity is used. 4.1. Load sharing operation strategies As reported by many researchers, the multiple-chiller system is running at partial load most of the time either due to variation in weather conditions or building use. In such circumstances, the search for the appropriate partial load sharing operation strategy for the multiple-chiller system under different building usage and conditions is essential in order to achieve the best performance of the integrated system. These appropriate operation strategies may lead to considerable energy saving of the existed multiple-chillers when operated at partial load. All the load sharing operation strategies are tested by running six compressors (half the available compressors) that may be selected by many different combinations (strategies) of compressors from 2 or 3 of the working chillers (each chiller has 4 compres-

3500

50 45

3000

40 2500

35

Incident

2000

Reflected

1500

Temp

30 25 20

%RH

15

1000

10 500

5 0

0

6

7

8

9

10

11

12

13

14

15

16

17

18

Fig. 1. Environmental data during conducting strategy 4-2-0.

19

20

333

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

800

3

700 600 2

500 400 300

1

200 100

Consumpon

Load

COP

0

0 6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

Fig. 2. Hourly aggregate system characteristics for strategy 4-2-0.

Table 1 Average ambient temperature (°C) for the tested operation strategies. Operation strategy

Daily

8 am–3 pm

7 am–7 pm

7 pm–7 am

2-2-2 2-1-3 2-3-1 1-2-3 1-3-2 3-1-2 3-2-1 3-0-3 3-3-0 0-3-3 4-2-0 2-4-0 0-2-4

39.8 40.5 37.9 38.1 39.3 38.7 38.0 39.7 39.2 39.0 39.0 37.8 39.7

42.3 43.1 41.3 42.7 41.7 41.8 43.1 43.1 41.4 41.5 42.0 41.4 42.9

42.5 43.1 41.1 42.7 41.8 41.6 42.4 43.0 41.4 41.4 42.0 41.1 42.9

37.0 37.9 34.7 33.6 36.8 35.7 33.6 36.3 37.0 36.5 35.9 34.5 36.5

Fig. 2 shows that the COP tendencies follow those of the building load more strongly than it reflects the opposite trends of the HVAC consumption. It may be stated that the building load adopt the performance while the consumption is influenced by both parameters. Since the chiller capacity is fixed for the same strategy, the performance improved as the building cooling load increases. It should be noted that similar trends to those for 4-2-0 strategy are observed for the hourly characteristics of the other tested strategies. 4.1.2. Daily characteristics Table 1 lists the average ambient temperature for the tested strategies during various tested intervals of the day. The highest average ambient temperature takes place between 8 am and 3 pm or between 7 am and 7 pm periods for some days whereas the lowest temperature is always between 7 pm and 7 am. The difference between the average day and night temperatures ranges from 4.4 °C (3-3-0) to 9.2 °C (1-2-3). It should be noted that the effect of relative humidity is limited because the variations during the day range from 2.7% to 6.8%. Table 2 shows that the average measured inlet and outlet chilled water temperatures, for the tested strategies during the day time are lower than the design values except for outlet temperature of strategy 3-3-0 which produces the lowest temperature difference. Thus, the measured inlet and outlet chilled water temperatures along with the chilled water temperature difference indicate possible partial load operation. Fig. 3 shows the coefficient of performance (COP) of the integrated system for the different tested operation strategies during the heavy occupied period (from 8 am to 3 pm) and during the full day. The daily coefficient of performance varies from about 1.06 to 1.41 with variation of about 25% between the different strategies.

Table 2 Average chilled water inlet temperature (°C) for the tested operation strategies. Operation strategy

2-2-2 2-1-3 2-3-1 1-2-3 1-3-2 3-1-2 3-2-1 3-0-3 3-3-0 0-3-3 4-2-0 2-4-0 2-4-0 0-2-4

Chilled water temp. Inlet

Outlet

10.4 10.3 10.6 10.7 11.2 10.9 10.6 10.4 11.3 10.9 11.0 10.6 10.8 9.8

6.6 5.88 6.35 6.75 7.00 7.06 6.55 6.15 8.80 6.94 7.00 6.98 7.13 6.21

Similarly, the COP during the occupied period varies from about 1.64 to 2.18 with about 36% variation. Both results confirm the dependency of the system performance on the load-sharing operation strategy. Also, the results are in agreement with Yu and Chan as they obtained aggregate performance between 0.08 and 5.22 [15]. Fig. 3 also shows that for each strategy, the COP during the occupied period is larger than that for the full day. This is because the rates of external and internal heat gain to the building during the occupied period are larger than that during the full day. Clearly, the average environmental temperatures during the occupied period are larger than those of the full day (Table 1). Thus, the variations in the building cooling load due to outside or inside conditions contribute to the variations in the aggregate system performance. The dependence of the system performance on the load-sharing operation strategy may be attributed to some possible causes such as the chiller percentage loading (part/full load), evaporator and condenser performance, status of the different chillers and compressors and finally the heat gained to the piping system of the chillers because the piping arrangement has different lengths from each chiller to the child water collection point. The chilled water out of chiller 1 or chiller 2 has a piping length of about 11 m or 25 m, respectively, whereas that of chiller 3 is about 95 m. Thus, the chilled water out of chiller 3 may gain more heat from surroundings than the water out of chiller 1 or 2 before it reaches the collection point where it is distributed to AHUs. The six strategies that use 3 chillers; each of different percentage loading (75%, 50% and 25%), produce various system performance during both the occupied period and the full day (Fig. 3).

334

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

2.5

Daily

8am-3pm

2 1.5 1 0.5

0--2--4

2--4--0

4--2--0

0--3--3

3--3--0

3--0--3

3--2--1

3--1--2

1--3--2

1--2--3

2--3--1

2--1--3

2--2--2

0

Fig. 3. Aggregate system performance for the tested operation strategies.

Strategy 3-1-2 has the best performance among this group followed by strategies 1-2-3 and 2-1-3 whereas strategies 1-3-2 and 2-3-1 produce the lowest performance. Since, the percentage loading is the same for these cases, the different results of system performance indicates that the percentage loading is not the only governor parameter for the aggregate system performance. Thus, the difference in COP is due to combined effects of the chillers and compressors condition and the different heat loss from the piping of each chiller to the collection point as stated earlier. However, the best strategy 3-1-2 has the highest inlet and outlet chilled water temperatures among the six strategies in this group (Table 2). Comparing the strategies that use full load of one chiller and half load of another shows better daily performance when chiller 3 was off and lower COP during the occupied period when chiller 3 was working on full capacity. This is mainly because more heat is gained by the chilled water of chiller 3 due to the long piping of this chiller, particularly during the day hours where ambient temperature is relatively high. The best COP for all cases in this group is achieved by 2-4-0 strategy which has the highest outlet chilled water temperature among the strategies in this group (Table 2). The strategies that use 75% of the capacity of 2 chillers have different performance that depends on which chiller is switched off. Again the best performance among this group is achieved when chiller 3 is not working, followed by the strategy that uses chillers 2 and 3. Again, the best strategy 3-3-0 has the highest inlet and outlet chilled water temperatures among the strategies in this group (Table 2). The comparison between all strategies that use 2 chillers (3-3 or 4-2) indicates better daily performance for strategy 2-4-0 but better performance during the occupied period for the strategy 3-3-0. Both of these strategies do not include chiller 3 in the operation. On the other side, comparing the strategies that use 3 chillers shows that the strategy that uses half the capacity of the three chillers (2-2-2) produces the lowest performance during the occupied period and is among the lowest aggregate performance for the full day. The tested strategies that use the same part load ratios are 2-22, 3-0-3, 3-3-0 and 0-3-3. Although, strategy 3-3-0 achieves reasonable aggregate performance, two strategies (3-0-3, 2-2-2) in this group attain the lowest performance among all the tested strategies. Thus, the tested strategies that use the same part load ratio never achieve the best aggregate performance but may attain the lowest. The best strategy provides daily aggregate performance that is higher than strategies 2-2-2 and 3-0-3 by 22% and 33%, respectively. Considering the aggregate performance of all the tested strategies shows comparable system performance during the occupied

period of the three strategies 2-4-0, 3-1-2 and 3-3-0 with better daily COP of the first two strategies 2-4-0 and 3-1-2. It is unexpected to find that one of the strategies that has the lowest percentage loading (75%, 50% and 25%) produces the best daily performance. This indicates that the effects of the piping arrangement, heat losses and perhaps the chillers and compressors condition may overwhelm the effect of the percentage loading of individual chiller. In conclusion, while the COP of a chiller increases with the percentage loading particularly for reciprocating type, the above discussion indicates that for all load-sharing operation strategies the aggregate performance is influenced by many other parameters in addition to the percentage loading such as the internal condition of the machines, the piping arrangement and the heat loss from the chilled water. The combined effects of these parameters influence the aggregate performance and overcome the effect of percentage loading in many cases. Therefore, it is recommended to experiment different operation strategies to ensure that the system operates at its best performance particularly for partial loading operation where many operation strategies are possible to apply. This becomes more important for existing systems where the condition of chillers or part of the chillers turns out to be more influencing. The above discussion compares the operation strategies of the chillers during the occupied period and during the full day. In order to further differentiate between the strategies under consideration, the aggregate performance during day time (7 am–7 pm) and during night time (7 pm–7 am) is compared in Fig. 4. Clearly, the performance during the day time is more than 3 times higher than that at night. This is due to the reduction in the internal and external building cooling load during the night where no occupancy in the building and the ambient temperature becomes relatively lower than that during the day beside the temperature setback to 28 °C. Thus, the effect of partial loading becomes more obvious during the night where the individual chiller COP is dramatically decreased for partial loading as the six compressors are still running. This is mainly because the chiller performance depends on the chilled water temperature difference. Thus, low chilled water temperature difference due to running many chillers at low partial load (low DT syndrome) should be avoided in order to obtain efficient chiller operation. The data in Fig. 4 shows that, the strategies 3-1-2 and 2-4-0 still produce the best system performance at day and night whereas the strategy 3-3-0 that was among the best during the occupied period as shown in Fig. 3 becomes lower than these 2 strategies. Fig. 5 shows the ratios of cooling load, energy consumption and coefficient of performance during night time to that during day time. While the ratio of cooling load varies from 0.15 to 0.20, the ratio of energy consumption varies from 0.72 to 1.0 indicating

335

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

2.5

7am-7pm

7pm-7am

2 1.5 1

0.5

0--2--4

2--4--0

4--2--0

0--3--3

3--3--0

3--0--3

3--2--1

3--1--2

1--3--2

1--2--3

2--3--1

2--1--3

2--2--2

0

Fig. 4. Aggregate system performance for the tested operation strategies during day and night.

1.2

Cooling Load

Consumpon

COP

1 0.8 0.6 0.4 0.2

0--2--4

2--4--0

4--2--0

0--3--3

3--3--0

3--0--3

3--2--1

3--1--2

1--3--2

1--2--3

2--3--1

2--1--3

2--2--2

0

Fig. 5. Comparison of the aggregate system characteristics during day and night.

relatively lower COP that varies from 0.15 to 0.26 of its value during the day time. This again confirms the large deterioration of the aggregate performance at night because of the low partial load operation. This will be explained in terms of the characteristic performance of individual chiller in Section 4.2. However, it should be mentioned that the high ratios of night to day consumption (>0.8) is related to large solar radiation in such days which stored more energy in the building walls during the day. Based on the above discussion, it can be concluded that for the load sharing operation strategies the chillers operate at high performance when working at large building cooling load during the occupied period of the day. On the contrary, the chillers operate at low performance when working at partial load outside the working hours (when the building is unoccupied and the ambient temperature is relatively lower than that during occupied period). 4.2. Chiller characteristics Understanding the previous results for the integrated multiplechiller system requires depicting the characteristics of the existing individual chillers under full and partial loading. In order to assure full loading during day hours, the tests performed while only one refrigerant circuits of the chiller was working. The chiller is monitored for 2 days under different ambient temperatures and load conditions. Fig. 6 shows typical variations of ambient temperature and relative humidity for the reported testing period of time (24 h) starting from midnight. Ambient temperature varies from about 35 °C at midnight to 44°°C around noon time. The highest temperature during the day occurs from 1 to 3 pm whereas the lowest occurs

around dawn time. On the other side, the relative humidity is lowest (about 15%) around noon time from 12 to 3 pm and increases before and after this period of time. Highest relative humidity occurs 1–2 h after mid night and may reach about 35%. It can be seen that the trend of relative humidity is opposite to that of the ambient temperature because as the air temperature increases, its ability to carry moisture increases. Fig. 7 shows the variations of some chiller characteristics such as COP, inlet chilled water temperature and compressor power with the ambient temperature and relative humidity. Clearly, as the ambient temperature increases, the COP, the inlet chilled water temperature and compressor power increase. This is attributed to the increase in the building load and consequently, the required cooling capacity of the chillers as the ambient temperature increases. It should be noted that the scatter in the relative humidity may be attributed to the fact that different values of relative humidity occurred for the same ambient temperature within the testing period of the day (see Fig. 6), i.e. the same ambient temperature but with different relative humidity during time. This scatter causes some corresponding scattering in the other parameters shown in Fig. 7, particularly the COP. Using partial capacity of the chiller for large loads causes the inlet chilled water temperature to reach higher values indicating insufficient cooling capacity of the working chiller, but with high performance as shown in Figs. 7 and 8. On the other side, for partial loads the inlet chilled water temperature was lower than 16 °C. While the chilled water temperature difference should be between 5.5 and 7.5 °C, the actual measured values are lower than these limits during partial load operation causing the cooling capacity and consequently the COP to decrease. This is linked to low DT

336

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

Ambient temp. & Relative humidity

50 45 40 35 30 25 20 15 10 Temp. (°C)

5 0

0

2

4

6

8

10

12

14

16

18

20

RH (%)

22

24

Local Solar Time Fig. 6. Variations of ambient temperature and relative humidity with local solar time for the individual chiller characteristics.

% RH

presents the average data for 30 min, the chiller operates long time at COP below 1.0. The low COP of the chillers may be explained with respect to the chillers system control which is designed so that chillers are running in combination of their partial capacities via the controlled outlet chilled water temperature to maintain it at the design value (7.5 °C). The operation of the chillers under partial capacities causes the difference between inlet and outlet chilled water temperatures to be small; what is called low-DT syndrome. Fig. 8 shows the same data of COP in Fig. 7 but plotted against local solar time instead of ambient temperature. Clearly, the small values of COP (lower than 1) occur during night time (unoccupied period and relatively low ambient temperatures) whereas the large values of COP take place during the working hours of the day time (occupied period and high ambient temperatures). Thus, the COP increases as the internal and external loads increase and the chiller operation during periods of low internal and external loads is considerably inefficient. Again this explains the sharp reduction in the aggregate system performance for all tested strategies during the night compared to that during the day (Figs. 4 and 5).

4.3. Further discussion

Fig. 7. Individual chiller characteristics versus ambient temperature.

syndrome particularly when the chilled water flow rate is kept constant for various loads. In general, COP varies from lower than 1.0 for relatively small ambient temperature to about 5 for large ambient temperature (Fig. 7). This is in agreement with values reported in literature which range from about 0.2 to 9 [15]. Thus, the chiller COP increases with the cooling capacity of the chillers which increases with the rise in the inlet chilled water temperature and the chilled water temperature difference (Fig. 7) providing that the chilled water flow rate remains unchanged. The increase in the cooling capacity is a consequence of the increase in the building cooling load. This explains the similar trend between the building load and the aggregate performance in Fig. 2. It should be noted that few points of large COP are existed whereas the majority of the COP data points are below 1.0. Considering that each point

4.3.1. Performance monitoring and standby chiller elimination It is well known that the performance monitoring of equipment is one of the effective techniques to determine the operational problems in equipment. The performance of chillers provides a good insight on their internal conditions. Thus, testing different operation strategies provides the performance monitoring of a chiller or part of a chiller to find out any problems or faults in the machine prior to their development into major breakdown or failure. Thus, the predicted required maintenance is planned and scheduled to avoid urgent repairs when the machine breakdown takes place. This in turn may lead to avoiding the need for standby chiller because the availability of the working chillers is ensured. For example, comparing the performance of strategies 0-3-3 and 3-0-3 (Fig. 3) indicates that the condition of chiller 2 is better than that for chiller 1. This is deduced because while chiller 3 is common in both strategies, the performance of 0-3-3 strategy is better than the other. The condition of chiller 1 may further confirmed by comparing the performance of strategies 4-2-0 and 2-4-0 because full loading of chiller 1 produces lower performance than full loading of chiller 2 while chiller 3 is not working in both cases. This is more established by the fact that the heat loss from chilled water of chiller 2 is more than that of chiller 1. Another way of using performance monitoring is to examine the

337

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

Coefficient of performance

6

Unoccupied Period

Occupied Period

Unoccupied Period

5 4 3 2 1 0 0

2

4

6

8

10

12

14

16

18

20

22

24

Local solar time Fig. 8. Individual chiller performance versus local solar time.

performance of 2 refrigeration circuits out of the four in a chiller. This is carried out by operating 2 compressors and switching the others off, then reverse the situation and compare the performance in both cases. This indicates the condition of each two compressors and repeating the experiment again using one of the two compressors in the less efficient group will define the faulty or the less efficient compressor. 4.3.2. Even and uneven size chillers Although one of the multiple-chillers concepts is to allow one chiller or more to work at full capacity while the system load is partial, small number of even capacity chillers makes that concept more difficult to accomplish. For example, the current multiplechiller system that composed of 3 chillers can work efficiently on partial loads of 1/3 and 2/3 of the full load while one or two chillers are working at their full capacity. However, uneven sizes of chillers may provide better application of the concept ‘‘full chiller capacity for partial building load’’. Multiple-chiller-systems may use four chillers, two of 30% and the other two of 20% of the load. Thus, partial load of 20% up to 80% with a step of 10% can be covered while one or more chillers are working at full capacity. Thus, close track of the various partial building loads and more efficient operation can be achieved using this arrangement. In this case, the standby chiller may be chosen for 30% of the load to provide alternative for the maximum needed chiller. 5. Conclusion Based on the reported results of the load-sharing operation strategies for multiple-chiller systems and the individual chiller characteristics, the following conclusions may be drawn:  The various tested operation strategies obtain aggregate performance from 1.64 to 2.18 during the day hours and from 1.06 to 1.41 for the full day indicating significant effect of the operation strategies on the aggregate performance of the system.  The tested strategies that use the conventional same part load ratio never achieve the best aggregate performance but may attain the lowest. The best load-sharing operation strategy achieves daily aggregate performance that is better than the same part load ratio strategies by 22–33%.  The load-sharing operation strategies that operate under the same percentage loading of chillers attain different aggregate performance of the system. Thus, operation and installation parameters such as the piping arrangement, performance of evaporators and condensers, heat loss from chilled water piping and condition of chillers and compressors have significant effects on the system aggregate performance.

 The system aggregate performance produced by the load sharing operation strategies under different percentage loading of the chillers indicate that the effect of the installation and operation parameters may overwhelm the effect of individual chiller performance. In other words strategies with low chiller loading (i.e. low chiller performance) produce performance that is similar or higher than some strategies with high chiller loading.  Comparing the system characteristics under constant cooling capacity during day and night shows that as the building load decreases to about 20% of its value, the average consumption decreases only to about 80% of the value at the higher load causing the aggregate performance to drop to about 20% of its value.  The building load adopts both chiller and aggregate performance of the system and their trends follow that of the building load. Low performance of individual chillers and multiple-chiller systems are due to partial load operation. Values of individual chiller COP that range from lower than 1 to about 5 are experienced under variable building load and constant chiller capacity.  Testing different operation strategies involves monitoring the performance of different chillers and compressors that may be useful to diagnose the less performing ones. Applying performance monitoring as one of the predictive maintenance techniques leads to prevention of chiller breakdown and urgent repairs. Thus, standby chiller can be eliminated as the availability of the working chillers is guaranteed.  Better performance of a chiller can be achieved if the partial load of the building can be matched with a full capacity of a chiller or more. This requires close track of the building load either by increasing the number of even-size units or by using four uneven size chillers, two of 30% and two of 20% of the load.

Acknowledgment This work was supported by the Public Authority of Applied Education and Training, Kuwait, project number TS-07-06. References [1] Ministry of energy. Report on prediction of energy consumption, Kuwait; 2005. [2] Bekker B, Carew P. Reducing energy consumption by designing for chiller efficiency. In: 10th REHVA World Conference; 2010. [3] Sun Y, Wang S, Huang G. Model-based optimal start control strategy for multichiller plants in commercial buildings. Build Serv Eng Res Technol 2010;32(2):113–29. [4] Gordon JM, Ng KC. Cool thermodynamics. Cambridge: International Science Publishing; 2000. [5] Chan KT, Yu FW. Thermodynamic-behaviour model for air-cooled screw chillers with a variable set-point condensing temperature. Appl Energy 2006;83:265–79.

338

H.Z. Abou-Ziyan, A.F. Alajmi / Applied Energy 135 (2014) 329–338

[6] Fong KF, Yuen SY, Chow CK, Leung SW. Energy management and design of centralized air-conditioning system through the non-revisiting strategy for heuristic optimization methods. Appl Energy 2010;87:3494–506. [7] Rezk ARM, Al-Dadah RK. Physical and operating conditions effects on silica gel/ water adsorption chiller performance. Appl Energy 2012;89:142–9. [8] Ma Z, Wang S. Supervisory and optimal control of central chiller plants using simplified adaptive models and genetic algorithm. Appl Energy 2011;88:198–211. [9] Sun Y, Wang Sh, Huang G. Chiller sequencing control with enhanced robustness for energy efficient operation. Energy Build 2009;41:1246–55. [10] Chung W. Review of building energy-use performance benchmarking methodologies. Appl Energy 2011;88:1470–9. [11] Wong LT, Mui KW, Law LY. An energy consumption benchmarking system for residential buildings in Hong Kong. Build Serv Eng Res Technol 2009;30(2):135–42. [12] Lam JC, Wan KKW, Cheung KL. An analysis of climatic influences of chiller plant electricity consumption. Appl Energy 2009;86:933–40. [13] Lee WS. Benchmarking the energy efficiency of government buildings with data envelopment analysis. Energy Build 2008;40:891–5. [14] Xu X, Fu X, Wang S. Enhanced chiller sensor fault detection, diagnosis and estimation using wavelet analysis and principal component analysis methods. Appl Therm Eng 2008;28:226–37.

[15] Yu F, Chan K. Improved energy management of chiller system by multivariate and data envelopment analyses. Appl Energy 2012;92:168–74. [16] Yongjun S, Shengwei W, Gongsheng H. Chiller sequencing control with enhanced robustness for energy efficient operation. Energy Build 2009;41:1246–55. [17] Chang YC, Lin JK, Chuang MH. Optimal chiller loading by genetic algorithm for reducing energy consumption. Energy Build 2005;37:147–55. [18] Gidwani BN. Optimization of chilled water system. Energy Eng 1987;84(5):30–50. [19] Chan K, Yu F. Optimum load sharing strategy for multiple-chiller systems serving air-conditioned building. Build Environ 2007;42:1581–93. [20] Austin SB. Optimum chiller loading. Ashrae J 1991;33(7):40–3. [21] Air-conditioning & Refrigeration Institute. Standard 550/590:performance rating of water chilling packages using the vapour compression cycle. Arlington (VA): Author; 2003. [22] Kaya A. Improving efficiency in existing chillers with optimization technology. ASHRAE J 1991;33(10):30–8. [23] EDMI. EDMI Ltd. Automatic meter reading systems manufacturing company, http://www.edmi-meters.com/index; 2008. [24] Moffat R. Describing the uncertainties in experimental results. Exp Therm Fluid Sci 1988;1:3–17.