Feature Story Unexpected Savings: Variable-speed chillers cut energy use in multiple applications Roy S. Hubbard, Senior Marketing Manager, Johnson Controls Inc. S ome chiller plant owners and designers believe there are certain applications in which variablespeed drive (VSD) chillers do not save energy. One such application is a multichiller plant, where the chillers have traditionally operated at high average loads. It is assumed that a multichiller plant cannot benefit from VSDs, which are thought to improve efficiency only at lower chiller loads. Another application is a plant located where the design outdoor temperatures are high. Because VSD chillers are known to save energy at lower outdoor temperatures, it is assumed that they are not suitable for higher-temperature applications. Contrary to these assumptions, VSD chillers are capable of saving energy in both multichiller plants and high-ambient locations. The key to understanding how VSDs save energy in both cases is to examine the impact of chiller load versus lift (aka “head”) on chiller efficiency, the impact of the sequencing logic used and whether or not optimization algorithms are employed. This article explains how lift affects VSD chillers, how VSD chillers should be sequenced differently from constant-speed drive (CSD) chillers and why that means VSD technology can offer substantial energy savings in a wider range of chiller plants than expected. Understanding Chiller Load and Lift The assumption that VSD technology cannot save energy in multichiller plants or in high-ambient applications is usually based on a misunderstanding of how load and lift affect VSD chiller efficiency. Table 1. Load-Based Sequencing of Chillers in a Five-Chiller Plant With a Water-Side Economizer. Facility Load (%) No. Chiller Systems Operating Chillers’ Load Ranges (%) 100 5 100-80 80 4 100-75 60 3 100-67 40 2 100-60 25 0 — Source: Johnson Controls Inc. A chiller in a single-chiller plant sees a full range of cooling loads, from its design load (100 percent) down to its minimum load (typically 10 percent). However, chillers in a multichiller plant typically see higher average loads. This is a result of the sequencing logic typically used. As the cooling load decreases in a multichiller plant, the typical sequencing strategy turns off a chiller © 2012 International District Energy Association. ALL RIGHTS RESERVED. and its auxiliaries (pumps and tower) as soon as the remainder of the chillers can carry the facility load. This control strategy will be called “load-based sequencing” and is illustrated in table 1. Note that, as in the table 1 example, when a plant uses a water-side economizer, the chillers will cycle off when the outdoor ambient temperature reaches about 40 degrees F, which means the last two chillers will cycle off at around 25 percent building load and about 60 percent of their capacity. Also note that the chiller systems would not be sequenced on and off at exactly these percent load points – that would cause hunting. Instead, the systems would be turned off at a slightly lower percent load and turned on at a slightly higher percent load. (This applies to all subsequent examples illustrated in the tables and figures below.) As the load-based sequencing control cycles off a chiller, the loads on the remaining chillers increase to compensate. Consequently, in a multichiller plant, the individual chillers see higher average loads than would a chiller in a single-chiller plant. The higher average loads seen by chillers in multichiller plants concern some chiller plant owners and designers; they believe that VSD chillers will District Energy / Third Quarter 2012 13 not save energy in this application. However, they are neglecting to account for changes in lift. “Lift” is defined as the pressure differential against which the compressor operates. In a water chiller, the lower pressure is driven by the refrigerant evaporating temperature, which is determined primarily by the leaving chilledwater temperature. The higher pressure is determined by the refrigerant condensing temperature, which is primarily a function of the weather conditions. Specifically, it is driven by the dry-bulb temperature of the air entering an aircooled chiller or the wet-bulb temperature of the air entering the cooling tower, which in turn determines the entering condenser water temperature (ECWT) for a water-cooled chiller. Here is a key point: The cooling towers all experience the same weather, and therefore they supply water at the same instantaneous ECWT to the chillers in the plant, regardless of the load. To analyze the impact on energy performance caused by changes in load and lift, figure 1 will be used. It shows energy performance curves that are “industry average,” reflecting averages of published curves for multiple brands of water-cooled VSD centrifugal chillers. How do load changes impact energy consumption of a VSD chiller? In figure 2, the lift is held constant at 85 F ECWT, and the load is allowed to vary. The result is that the kilowatts per ton of cooling vary relatively little – only about 15 percent. On the other hand, how do lift changes impact energy consumption? In figure 3, the load is held constant at 100 percent, and the lift is allowed to vary. The variation in chiller efficiency is around 45 percent. Compared to the 15 percent impact of variable load, that is a 3-1 ratio! This demonstrates that lift has a major impact on chiller efficiency, while load has a much lower impact. Therefore, VSDs can save energy in multichiller plants, even though the average chiller loads are higher. One key to saving energy with VSD chillers is the lift variation. One key to saving energy with VSD chillers is the lift variation. Performance in Multichiller Plants The influence of load and lift on VSD and CSD chillers was examined by this author in an analysis of a hypothetical large U.S. district cooling plant. The plant was assumed to contain five 2,250-ton (7.9 MW) water-cooled centrifugal chillers, each with dedicated auxiliaries: chilled-water pump, condenser water pump and cooling tower. A typical load line was assumed, as was 24/7 operation. A water-side economizer was also assumed. The choice of weather data used in this analysis required a compromise. Rather than drawing on data from any one particular location, the concept of “average U.S. weather data” was employed to make the analysis as widely applicable as possible. This is the same weather data used by the AirConditioning, Heating, and Refrigeration Institute (AHRI) to calculate the IPLV/ NPLV (integrated part-load value/nonstandard part-load valve) ratings for chillers in AHRI Standard 550/590-98. The data was created by looking at weather data in the 29 cities across the United States that correlated with 80 percent of all the centrifugal chiller sales over a 25-year period. The data for the 29 cities were then averaged. The concept of average U.S. weather data has been approved by the major chiller manufacturers as an acceptable compromise for use in analyzing chiller energy consumption. Only chiller energy consumption was examined in the study as well as for this article. Inclusion of the auxiliaries (cooling towers and pumps) would have required a large number of variables, such as how they are driven (CSD or VSD), the piping scheme (headered or dedicated pumping, primary-secondary or variable primary, etc.) and the operating method used (manual, automated or Figure 1. Performance of an “Industry Average” Water-Cooled Figure 2. Efficiency Changes With Variable Load on a VSD Centrifugal VSD Centrifugal Chiller. Chiller. .80 Kilowatts per ton of cooling .60 .50 .40 .30 n n n n .20 .10 85 F ECWT 75 F ECWT 65 F ECWT 55 F ECWT | |||||| || .00 20% 30%40%50%60%70%80% 90%100% Load Source: Johnson Controls Inc. District Energy / Third Quarter 2012 .70 .60 Kilowatts per ton of cooling .70 14 .80 Design performance 100% Load 85 F ECWT 0.600kW/ton 15% .50 .40 .30 .20 .10 n n n n 85 F ECWT 75 F ECWT 65 F ECWT 55 F ECWT | |||||| || .00 20% 30%40%50%60%70%80% 90%100% Load Source: Johnson Controls Inc. © 2012 International District Energy Association. ALL RIGHTS RESERVED. optimized). Since the intent of this article is to examine the impact of adding VSDs to the chillers, the impact of the auxiliaries has not been calculated. Drawing on data from the analysis, the following scenarios demonstrate the performance of CSD and VSD chillers using different sequencing methods: Scenario 1: CSD Chillers With Traditional Load-Based Sequencing In scenario 1, the chillers were all CSD and used the performance curves in figure 4, which are averages of published curves for multiple brands of water-cooled CSD centrifugal chillers. The load-based sequencing strategy was also used. The tool/method used to do the energy analysis was the ASHRAE modified temperature bin method (Simplified Energy Analysis Using the Modified Bin Method, ASHRAE, David E. Knebel, principal investigator, 1983). Table 2 shows how the five chillers unload when controlled by the traditional load-based sequencing control strategy, heretofore considered the most efficient method of sequencing a plant. The red rectangles show the chillers cycling off, while the green rectangles show the loads on the remaining on-line chillers increasing to pick up the remaining building load. The water-side economizer enabled the chillers to be completely cycled off when the outdoor air was below 40 F. Table 3 shows how the efficiency values from figure 4, the loads from table 2 and the weather data were used to calculate the annual energy consumption for CSD chiller No. 1. It is important to note that the 95.0-99.0 F temperature bin does not represent design conditions only. It is an average of the few hours spent at design conditions plus a number of hours spent at off-design conditions. That is why the load and efficiency values are lower than at design. The annual energy consumption for CSD chiller No. 1 was 5,972,580 kWh. The same calculation process was used for CSD chillers Nos. 2-5 (not shown because of limited space). The total annual energy consumption for all five chillers was 18.35 million kWh. Scenario 2: VSD Chillers With Traditional Load-Based Sequencing In scenario 2, all the chillers were equipped with VSDs, and the energy performance curves in figure 1 were used. Built into figure 1 is a 3 percent loss for the VSDs. The load-based sequencing strategy was used again. The total annual energy consumption for all five VSD chillers was 15.4 million kWh, a savings of 16 percent. Scenario 3: VSD Chillers With EnergyBased Sequencing In scenario 3, the chillers were again equipped with VSDs, but the Figure 3. Efficiency Changes With Variable Lift on a VSD Centrifugal Chiller. sequencing strategy was modified. Because VSDs offer the most energy savings at lower loads and lower ECWTs, it makes sense to keep them on line for as long as possible while the cooling load of the facility decreases. Therefore, a chiller system was not turned off until the energy consumption of operating all the on-line chillers exceeded the energy consumption of operating all the on-line chillers minus one. This control strategy – which is really an optimization algorithm – will be called “energy-based sequencing™.” Table 4 shows how the five chillers unload when optimized by the energybased sequencing strategy. Once again, the red rectangles show the chillers cycling off, while the green rectangles show the loads on the remaining on-line chillers increasing to pick up the remaining building load. The sequencing points are significantly different from the points used with the load-based sequencing strategy. For instance, the most cost-efficient point at which the first chiller system should be turned off was when the facility load had fallen to about 33 percent. The sequencing points are not linear; they are determined by the energy consumption and not just the load. Note also that the chiller loads, indicated by the green rectangles, are lower than when load-based sequencing strategy is used. Determining the sequencing Figure 4. Performance of an “Industry Average” Water-Cooled CSD Centrifugal Chiller. .80 .80 .70 .70 .60 45% .50 .40 .30 .20 .10 n n n n 85 F ECWT 75 F ECWT 65 F ECWT 55 F ECWT | |||||| || .00 20% 30%40%50%60%70%80% 90%100% Load Source: Johnson Controls Inc. © 2012 International District Energy Association. ALL RIGHTS RESERVED. Kilowatts per ton of cooling Kilowatts per ton of cooling .60 Design performance 100% Load 85 F ECWT 0.582 kW/ton .50 .40 .30 .20 .10 n n n n 85 F ECWT 75 F ECWT 65 F ECWT 55 F ECWT | |||||| || .00 20% 30%40%50%60%70%80% 90%100% Load Source: Johnson Controls Inc. District Energy / Third Quarter 2012 15 Table 2. Chiller Loading When Using Load-Based Sequencing. Temperature Building Bin (°F) Load (Tons) Building Load (%) Chiller Chiller Capacities (Tons) Capacity No. 1 No. 2 No. 3 No. 4 No. 5 (%) 95.0-99.9 10,810 96.1 96.1 2,1622,1622,1622,1622,162 90.0-94.9 9,935 88.3 88.3 1,9871,9871,9871,9871,987 85.0-89.9 9,060 80.5 80.5 1,8121,8121,8121,8121,812 80.0-84.9 8,180 72.7 90.9 2,0452,0452,0452,045 0 75.0-79.9 7,304 64.9 81.2 1,8261,8261,8261,826 0 70.0-74.9 6,426 57.1 95.2 2,1422,1422,142 00 65.0-69.9 5,550 49.3 82.2 1,8501,8501,850 0 0 60.0-64.9 4,674 41.5 69.2 1,5581,5581,558 0 0 55.0-59.9 3,800 33.7 84.4 1,9001,900 0 00 50.0-54.93,272 29.1 72.61,636 1,636 0 00 45.0-49.93,066 27.2 68.11,533 1,533 0 00 40.0-44.92,858 25.4 63.61,429 1,429 0 00 Table 3. Annual Energy Consumption for CSD Chiller No. 1 With Traditional Load-Based Sequencing. Temp. Average Tower Entering Load Efficiency Draw Hours Energy Bin (°F) Wet-Bulb Approach Condenser (Tons) (kW/Ton) (kW) (kWh) Usage Temp. (°F) (ºF) Water Temp. (°F) points when using energy-based sequencing requires a central plant optimization (CPO) system, which can analyze the multitude of variables that impact the decision beyond just load. The total annual energy consumption for all five VSD chillers, controlled by energy-based sequencing, was 12.87 million kWh. That is a savings of 30 percent. By controlling the VSD chillers with energy-based sequencing using a CPO system, the savings was increased an additional 14 percent. This analysis shows that VSDs can significantly reduce chiller energy consumption in multichiller plants. Even if the average chiller load is high, as when using the load-based sequencing, the chiller lift decreases, and lift has a bigger impact on chiller efficiency than does load. 95.0-99.9 72.0 7.9 79.9 2,1620.5601,212 37 44,844 Energy-based sequencing is 90.0-94.9 71.0 7.6 78.6 1,9870.5441,081120129,671 85.0-89.9 7.4 76.4 1,812 another key to maximizing 80.0-84.9 68.0 8.4 76.4 2,0450.5331,089515561,075 75.0-79.9 8.0 74.0 1,826 70.0-74.9 63.0 9.8 72.8 2,1420.5141,101928 1,021,684 65.0-69.9 59.0 9.6 68.6 1,850 0.484 895 895 801,228 60.0-64.9 55.0 9.2 64.2 1,558 0.458 713 856 610,421 55.0-59.9 50.0 12.0 62.0 1,898 0.447 848 776 658,062 50.0-54.9 45.0 11.9 56.9 1,635 0.411 672 679 456,483 45.0-49.9 41.0 14.0 55.0 1,533 0.397 608 586 356,535 40.0-44.9 37.0 18.0 55.0 1,430 0.397 568 549 311,903 69.0 66.0 0.526 0.514 953 938 303 780 288,897 731,777 Subtotal 5,972,580 Table 4. Chiller Loading When Using Energy-Based Sequencing. Temperature Building Bin (°F) Load (Tons) Building Load (%) Chiller Chiller Capacities (Tons) Capacity No. 1 No. 2 No. 3 No. 4 No. 5 (%) 95.0-99.9 10,81096.196.1 2,162 2,162 2,162 2,162 2,162 90.0-94.9 9,93588.388.3 1,987 1,987 1,987 1,987 1,987 85.0-89.9 9,06080.580.5 1,812 1,812 1,812 1,812 1,812 80.0-84.9 8,18072.772.7 1,636 1,636 1,636 1,636 1,636 75.0-79.9 7,30564.964.9 1,461 1,461 1,461 1,461 1,461 70.0-74.9 6,42557.157.1 1,286 1,285 1,285 1,285 1,285 65.0-69.9 5,55049.349.3 1,110 1,110 1,110 1,110 1,110 60.0-64.9 4,67541.541.5 935 935 935 935 935 55.0-59.9 3,80033.833.8 760 760 760 760 760 50.0-54.9 3,272 29.1 36.4 45.0-49.9 3,066 27.3 45.4 818 818 818818 0 1,0221,022 1,022 00 40.0-44.9 2,85925.442.4 953 953 9530 0 Tables 2, 3 and 4 source: Johnson Controls Inc. 16 District Energy / Third FourthQuarter Quarter2012 2011 savings in VSD chiller plants. Resequencing VSD chillers based on energy means they will run in their sweet spot – or sweet span – as much as possible, which often means lower loads. Energy-based sequencing is another key to maximizing savings in VSD chiller plants. This strategy leverages the savings capabilities of VSDs at lower loads and, most importantly, lower ECWTs. The auxiliaries can marginally add to or subtract from these savings, depending on the piping configuration and control. Although this analysis was done on a large chiller plant, the percentages of savings are scalable to any size plant; and, because industry average efficiency curves were used, the analysis is valid for any brand of VSD chiller. The cost-effectiveness of adding VSDs to the chillers will depend on a number of factors, including the capital and installation costs of the VSDs, local electricity cost and cooling load profile. These factors are highly individual and outside the scope of this article. © 2012 International District Energy Association. ALL RIGHTS RESERVED. Performance in HighAmbient Locations Are VSD chillers a good investment in locations that experience high outdoor temperatures, such as the Middle East? The answer is, “Often.” Once again, it depends on the weather. The design dry-bulb and wet-bulb temperatures are not the critical factors. Most important is whether there is a significant variation in the temperatures throughout the cooling season. Dhahran, Saudi Arabia, has design temperatures of 110 F dry bulb and 86 F wet bulb, making it one of the warmest and most humid cities in the world. However, the outdoor temperatures can also fall to 35 F dry bulb and 32 F wet bulb. This substantial change in outdoor temperatures indicates VSDs are likely to provide significant energy savings. The same 11,250-ton (39.5 MW) plant analyzed above was relocated to Dhahran, with the efficiency of the chillers adjusted for the higher lift. The total annual energy consumption for all five CSD chillers, controlled with load-based sequencing, was 32.08 million kWh. When the chillers were equipped with VSDs, and operated with CPOdriven energy-based sequencing, the total annual energy consumption was 23.31 million kWh – an energy savings of 27 percent. So, even in high-ambient locations, putting VSDs on the chillers and using an energy-based sequencing strategy reduced plant energy consumption significantly. The key is that the location has temperatures that vary throughout the day and the cooling season; the design temperatures are rarely applicable. Conclusion Water-cooled VSD chillers are capable of saving energy in both multichiller plants and high-ambient locations. If the leaving chilled-water temperature and condenser water flow are constant, the key to their performance is the variation in entering condenser water temperature, which lowers the chiller lift. Additional energy savings can be obtained by sequencing the chillers according to an energy-based sequencing strategy, designed to get the maximum performance from the VSDs. Energy-based sequencing requires a central plant optimization system, which can analyze the multitude of variables that impact the decision beyond just load. As a result, chiller plant owners and designers will find VSD chillers yield substantial energy savings in a surprisingly wide range of chiller plants. Roy S. Hubbard is senior marketing manager with Johnson Controls Inc., a global provider of HVAC equipment, controls and service solutions. He has more than 30 years of experience in sales, marketing and applications, with particular focus on chilled-water systems. He is also the architect for YORKcalc, Johnson Controls’ operating cost software for chiller plants. Hubbard holds a bachelor’s degree in engineering from the U.S. Military Academy at West Point. He can be reached at roy.s.hubbard@jci.com. The Water Treatment Choice of Over 100 Colleges & Universities Chem-Aqua, Inc. • 1-800-527-9921 • www.chemaqua.com Irving, TX • Monmouth, NJ • Jacksonville, FL • Indianapolis, IN • Macon, GA • Las Vegas, NV • Sunnyvale, CA • Brampton, ON IDEA_DistrictEnergy_Q3_2012.indd 1 © 2012 International District Energy Association. ALL RIGHTS RESERVED. 7/5/2012 10:42:06 AM District Energy / Third Quarter 2012 17