Chiller plant optimization

Transcription

1 Chiller plant optimization CLEANROOM Terrence Morris & Steve Blaine PE, CH2M HILL, Oregon, USA ABSTRACT Outside of the process tools themselves, the chilled water plant is typically the single largest consumer of electrical energy in a semiconductor facility [1]. This load includes not just chillers but also cooling tower fans, primary pumps, secondary pumps and condenser pumps. In order to meet the cooling requirements for any particular heat load, many different combinations of this equipment can be run. However, electricity consumption varies considerably depending on the combination of equipment used and the operating levels of the individual components. Selecting the optimal mix of equipment and operating levels presents a substantial challenge for an automatic control system and plant operators. Typically, no method is available to predict the effect of interactions and variations in load demand and outside air. This makes it challenging, if not impossible, to find an equipment mix that achieves optimal energy use. In response to this challenge, we set out to create a model/tool that would allow operators to automatically determine the optimal equipment mix to satisfy cooling requirements and minimize energy use. This paper describes how this model was created and how it works. Introduction A recent ASHRAE article by Mark Hydeman and Guo Zhou describes the use of a chiller plant optimization model based on parametric modeling [2]. Thomas Hartman has published several articles describing his patented LOOP chiller plant design [3] and [4]. Ben Erpelding describes implementation of a system based on this work [5]. Our model performs a similar optimizing function but is accomplished using only a standard Microsoft Excel spreadsheet. The model can calculate the predicted energy consumption for any valid combination of equipment serving a range of cooling loads at any outside air condition. The type and quantity of equipment modeled is based on an actual semiconductor facility chilled water plant. The model s loaddetermining inputs include outside air (wet bulb) temperature, distribution system flow and distribution system return water temperature. Based on vendor data for chillers, cooling towers and pumps, the model calculates outputs for each component of the system, including system bypass and condenser water flow, entering and leaving chiller condenser and cooling tower temperatures, entering evaporator temperature and leaving evaporator temperature (assumed 44ºF (6.7ºC)). Based on these flows and temperatures, the individual and total system energy usage is computed. The model also includes another powerful feature: a spreadsheet function that optimizes the selection of equipment. The results from this exercise show the optimal choice of equipment at any outside air condition and system load. Further investigations will seek to automate this optimization so that it can be used in a control algorithm. Figure 1a. Evaporator water system schematic. Figure 1b. Condenser water system schematic. W W W.FA B T E C H.ORG 31

2 TABLE 1: SAMPLE CHILLER PART LOAD PERFORMANCE Pct CAP Pct Inp Pwr EEFT ELFT CEFT CLFT Sys Perf Load (TR) Power (KW) (ºF) (ºF) (ºF) (ºF) (KW/TR) TABLE 2: SAMPLE CHILLER/COOLING TOWER EQUILIBRIUM TABLE (EEFT: EVAPORATOR ENTERING FLUID TEMPERATURE; CEFT: CONDENSER ENTERING FLUID TEMPERATURE; CLFT: CONDENSER LEAVING FLUID TEMPERATURE) EEFT: 52.4ºF 100% flow 100% fan CEFT CLFT 85ºF (WB) 80ºF (WB) 75ºF (WB) 70ºF (WB) 65ºF (WB) 60ºF (WB) x x x x x x x x x x x x x x x x x x x x x x x x x Creating the model We selected a typical chiller plant with decoupled primary and secondary flow loops (see Figures 1a and 1b). The system includes the following equipment and variable speed drives: 7 chillers 2000 tons (7000kw) 7 primary pumps 4000gpm (250L/s), 100hp (75kW) 10 cooling towers 4000gpm (250L/s), 100hp (75kW) fans with variable speed drives 7 secondary chilled water pumps with variable speed drives 8671gpm (547L/ s), 300hp (223kW) 7 condenser water pumps with variable speed drives 8671gpm (547L/s), 300hp (223kW) Calculating condenser water temperatures The first step in creating the model was to calculate the entering and leaving condenser water temperatures at equilibrium. The performance of both the chiller condenser and the cooling tower performance were taken into account for this purpose. A constant evaporator discharge temperature of 44ºF (6.7ºC) was assumed, which is consistent with plant operating practice. Chiller vendor data showing part load performance per ARI 550/590 and ASHRAE 90.1 was used as a starting point. This data was available for CEFT from 85ºF (29ºC) to 53ºF (12ºC). A sample of the chiller data is shown in Table 1. Given a certain entering evaporator fluid temperature (EEFT) and an entering condenser fluid temperature (CEFT), the data shows the resultant leaving condenser fluid temperature (CLFT). Note that chiller performance data per ARI 550/590 allows a considerable credit to be taken for manufacturing tolerance, especially at full load. Most manufacturers do not need this much manufacturing tolerance, so the predicted performance may show the chiller to be more efficient than it really is. The result is that low load efficiency in particular may be inaccurate. Future investigations will begin with a request for zero tolerance data. Cooling tower selection software was used to predict the temperature of the water leaving the cooling towers (CEFT) at different flow rates and cooling tower fan speeds. We created 162 Excel tables similar to Table 2 covering a range of parameters, including entering evaporator temperature (i.e. return water from the fab) from approximately 46ºF to 56ºF; cooling tower flow from 80% to 110% of rated design; cooling tower fan speeds from 50% to 100 %; and outside air (wet bulb) from 55 to 85ºF. Chiller performance data per ARI 550/590 allows a considerable credit to be taken for manufacturing tolerance. These tables were used to pinpoint the moment where the temperature change was the same for both condenser and cooling tower, for each combination of variables. It was assumed that an energy balance existed during steady state conditions (when the temperature rise in the chiller condensers is the same as the temperature drop in the cooling towers). Each table essentially acts as a temperature balance that can be used to find equilibrium temperature at steady state conditions, as outlined in the Example below. Repeating ter ms signify that rough equilibrium temperature has been reached; a precise value will be determined later through the use of trend line extraction from a curve fit of the data. The values from the condenser side of the table come directly from the chiller manufacturer s data sheets. The EXAMPLE: FIND THE EQUILIBRIUM TEMPERATURE FOR 60ºF OSA (WB). THE STARTING POINT DOES NOT MATTER, AS LONG AS IT IS REASONABLE. FOR THIS EXAMPLE, LET US START AT CEFT = 80 CEFT: 80 CLFT: CTEFT: CTLFT: CEFT: 75 CLFT: CTEFT: CTLFT: CEFT: 70 CLFT: CTEFT: CTLFT: CEFT: 70 CLFT: CTEFT: CTLFT: SEMICONDUCTOR FABTECH 38 TH EDITION

3 Figure 2a. Four pumps at 96% speed = 910kW. Figure 2b. Five pumps at 89% speed = 827kW. TABLE 3: MODEL INPUTS AND OUTPUTS INPUTS OUTPUTS LOAD SUPPLY TEMP (T1) SCW PUMP SPEED LEAVING EVAPORATOR TEMP (T4) CW PUMP SPEED OUTSIDE AIR TEMP (WET BULB) (T5) COOLING TOWER FLOW LOAD RETURN TEMP (T2) BYPASS FLOW (F2) SYSTEM FLOW (F1) ENTERING EVAPORATOR TEMP (T3) ENTERING CONDENSER TEMP (T6) LEAVING CONDENSER TEMP (T7) OPTIMIZER CONTROLLED INPUTS TONS-REFRIGERATION QUANTITY OF CHILLERS CHW PUMP POWER QUANTITY OF SCW PUMPS CHILLER POWER QUANTITY OF COOLING TOWERS SCW PUMP POWER COOLING TOWER FAN SPEED CW PUMP POWER QUANTITY OF CW PUMPS COOLING TOWER FAN POWER CONDENSER FLOW (F3) TOTAL POWER % FROM OPTIMAL SYSTEM KW/TON values on the cooling tower side come from the cooling tower manufacturer s selection software program. This solves for the cooling tower leaving flow temperature (CTLFT) based on cooling tower flow rate, outside air (OSA) wet bulb temperature, fan speed and cooling tower enter ing flow temperature (CTEFT). From this data, a series of equations was created that related condenser water temperatures to the previous variables. By graphing the data and extrapolating an equation for the trend line, these relationships were converted into a set of simultaneous equations that solve the entering and leaving condenser water temperatures, assuming full flow conditions through the condenser. One feature of this model is the ability to vary condenser water flow. Therefore, the actual condenser temperature can be higher than the values computed with the energy balance above. The correct value is calculated by proportionately adjusting the previous result to account for the varying condenser flow. Calculating pump operating points The next substantial calculation involved computing operating points for the secondary chilled water (SCW) and condenser water (CW) pumps. Based on assumed pipe diameters and lengths, system curves were entered for the secondary and condenser water systems. Using vendor data, pump curves were calculated for each possible quantity of operating pumps. The required flow rate is noted and, using curve fit techniques, the operating speed of the pump(s) is calculated. The calculation takes into account pump, motor and VFD efficiency using a method explained in a previous ASHRAE Journal article [7]. These curves dynamically update to display the intersection of the system and pump curves. Note that the condenser water system curve reflects the 10 feet of static head (30kPa) and 11.6 feet of pressure head (34.6kPa) (see Figures 2a and 2b). All variables required to calculate the power consumption have been determined at this stage. Chiller power is calculated by trending the data provided by the manufacturer as a function of entering evaporator and condenser temperatures. The model The model has two purposes: determining the power consumption of a specific equipment configuration and determining W W W.FA B T E C H.ORG 33

4 the most energy efficient equipment configuration based on the load and outside air (OSA) wet bulb temperature. In the first of the two functions, the user enters the cooling load flow (i.e. SCW), return water temperature, OSA and the quantity and speed of equipment running. The model then computes the power consumed by each set of equipment and the total kw. The model also displays the overall system kw/ton. In the second case, the user also specifies the current load and OSA temperature; the model then finds valid equipment configurations that will meet the load. An optimizing function then chooses the lowest energy cost option. The optimizer processes more than 2,000 combinations of equipment configurations until it finds the best result. Optimizing Microsoft Excel software includes a suite of what-if analysis tools based on the Generalized Reduced Gradient (GRG2) nonlinear optimization code developed by Leon Lasdon, University of Texas at Austin, and Allan Waren, Cleveland State University better known as the Solver function [6]. With these tools, the optimal value for a formula in one cell - called the target cell - on a worksheet can be found. The tools work by adjusting the values in the specified changing cells known as the adjustable cells - to produce the specified result from the target cell formula. Constraints can be applied to restrict the values used by the tools in the model, while the constraints can refer to other cells that affect the target cell formula. TABLE 4: MODEL CONSTRAINTS MIN BYPASS FLOW = 0 GPM MAX CONDENSER WATER FLOW = 42,000 GPM MIN ENTERING CONDENSER TEMP = SUPPLY TEMP - (TEMP RISE) + 17ºF MAX ENTERING CONDENSER WATER TEMP = 110ºF MAX INDIVIDUAL CONDENSER FLOW = 6851 GPM MIN INDIVIDUAL CONDENSER FLOW = 2195 GPM AVAILABLE CHILLERS = 7 AVAILABLE SCW PUMPS = 7 MAX SCW PUMP SPEED = 100% AVAILABLE CW PUMPS = 7 MAX CW PUMP SPEED = 100% AVAILABLE COOLING TOWERS = 10 MAX COOLING TOWER FLOW = 110% MIN COOLING TOWER FLOW = 80% MAX COOLING TOWER FAN SPEED = 100% The optimizer processes more than 2,000 combinations of equipment configurations until it finds the best result. In this model, the what-if tools return the selection of the most energy efficient equipment set-up that meets the load and user-entered constraints. To use this feature, the user simply inputs the load conditions and OSA, and presses the optimize button. The constraints are programmed as shown in Table 4. Constraints The normal programmed operating sequence for the chilled water plan in such a semiconductor factory is as follows: Chiller discharge temperature is set constant at 44ºF (6.7ºC) Figure 3. Normal operating conditions (prior to optimization). Two-way valves in the distribution create a variable flow demand based on system load Chillers are sequenced based on the bypass flow (F2) and/or chiller % load readings Primary chilled water pumps are sequenced on with the associated chiller Secondary chilled water pumps are sequenced and their speed is varied to maintain system pressure across the supply and return headers Condenser water pumps are sequenced and their speed is varied to maintain condenser water system pressure across the condenser supply and return headers Cooling towers are sequenced to match the required condenser capacity of the chillers (note that there are 10 towers and 7 chillers) Cooling tower fans are operated at a speed necessary to attain the condenser water temperature setpoint. The setpoint is established at the OSA wetbulb plus a 5ºF (2.8ºC) approach down to a minimum of 55ºF (13ºC). 34 SEMICONDUCTOR FABTECH 38 TH EDITION

5 Figure 4. Optimized operating conditions. Consider the example in Figure 3. The system is running with normal operating logic as described above with a 66ºF (19ºC) outside air wet bulb temperature, secondary chilled water flow of 21750gpm (1372L/s), and a 54ºF (12ºC) return water temperature. With these operating conditions, the model predicts an energy consumption of 6076kW. The condenser flow at these conditions is 16000gpm (1001L/s). Now the model is asked to recommend more energy efficient operating conditions (Figure 4). Although still producing the same predicted 9063 tons of refrigeration, the model proposes an operating method that is approximately 3.7% more efficient. The number of secondary pumps is increased from 5 to 6. The number of condenser pumps is changed from 6 to 5, reducing the condenser flow to 13170gpm (831L/s) and the number of cooling tower fans is reduced from 5 to 4. The chillers are now using 123kW more in this scenario, but the overall system is predicted to consume 224kW less energy. The most important Figure 5. System conditions and load influence potential savings when using optimizer. TABLE 5: NORMAL VS. OPTIMIZED OPERATING CONDITIONS Normally Programmed System Equipment kw Optimized System Equipment kw 6 - CHW Pumps CHW Pumps Chillers Chillers 4432 Optimizer 5 - SCW 78% Speed SCW 75% Speed Condenser 62% Speed Condenser 56% Speed Cooling Tower 100% Speed Cooling Tower 84% Speed 213 Total 6076 Total 5852 W W W.FA B T E C H.ORG 35

6 differences between the normal and optimized operating conditions for this example are illustrated in Table 5. By focusing on minimizing the power consumed by the system, the optimizer is able to find the most efficient mix of equipment to run that satisfies the system load and the programmed constraints. Results By utilizing the optimizer, predicted savings of close to 4% can be reached, depending on the system conditions and load (Figure 5). In most cases, this is achieved by decreasing the cooling tower and condenser water pump operating points. In response, the electrical consumption of the chillers will increase because of a larger compressor lift. In many cases, the decrease in electrical energy consumed by the condenser system is much greater than the increase in electrical consumption by the chillers. As a further exercise, tables were created for varying system flow, return supply temperature and outside air wet bulb temperature. By providing optimal equipment configurations based on various load and outside air conditions, these tables provide operators with a quick and easy way to operate the plant as energy-efficiently as possible. Using this guide, an operator could simply look up the current load and outside air conditions and see what the model proposes as the optimal mix of equipment and speeds. An automatic version of this feature is the subject of future investigation. Possibilities for future development While the system modelled in this case was simple in nature, future versions could potentially increase in complexity by, for example, adding heat recovery components. Another addition would be the ability to enter different sizes of equipment for each area. To make the model more applicable in an actual plant, real data could be substituted in real time to replace existing data and equipment efficiencies. REFERENCES [1] Naughton, P. 2006, Greening our Cleanrooms Tools to Help Us Improve Cleanroom Energy Performance, Future Fab International, 21. [2] Hydeman, M. & Zhou, G. 2007, Optimizing Chilled Water Plant Control, ASHRAE-Journal. [3] Hartman, T. [online], Optimizing All- Variable Speed Systems with Demand Based Control, available at [4] Hartman, T. [online], All-Variable Speed Centrifugal Chiller Plants: Can We Make Our Plants More Efficient? available at automatedbuildings.com. [5] Erpelding, B. 2006, Ultraefficient All-Variable Speed Chiller Water Plants, HPAC Engineering. [6] Flystra, D., Lasdon, L., Watson, J. & Waren, A. 1998, Design and Use of the Microsoft Excel Solver, Institute of Operations Research and the Management Sciences. [7] Bernier, M. & Bourret, B. 1999, Pumping Energy and Variable Frequency Drives, ASHRAE-Journal, December ABOUT THE AUTHORS Terrence Morris is an engineering intern with CH2M HILL. He i s c u r re n t l y e n ro l l e d a s a mechanical engineering student at Oregon State University. Mr. Morris engineering interests include energy conserving technology and modeling. Steve Blaine is an instrumentation and controls specialist with CH2M HILL. He has more than 25 years experience designing control systems for processes a n d f a c i l i t i e s, p a r t i c u l a r l y semiconductor factories. Mr. Blaine holds a B.S. degree in electrical engineering from the University of Pennsylvania and an M.S. degree in engineering management from Portland State University. He is a registered professional engineer in the states of Oregon, New Mexico, Arizona, Florida, Texas and Utah, and is also a licensed electrician in Oregon. ENQUIRIES Website: 36 SEMICONDUCTOR FABTECH 38 TH EDITION