Waterside Economizer Retrofit for Data Center.
The facility is a windowless data center whose internal cooling load is nearly a constant 1,000 tons (3520 kW) year-round, with the skin load being a very minor component. The data center operates 24/7/365 and cannot be without cooling for all but the briefest of outages (on the order of a few minutes).
Figure 1 (Page 62) or diagrams very similar to it have been published several times, (1,3,4) and it has become somewhat of a de facto standard for piping integrated waterside economizers. Thus, when we began implementation of the project in this case study, we attempted to replicate the piping arrangement accordingly. But since this is a retrofit condition, with a chiller plant designed by a different engineering firm, there were some immediate differences noted.
* Whereas Figure 1 has condenser water and chilled water pumps on the discharge of each chiller, our chiller plant features pumps on the inlets to the chillers.
* Whereas Figure 1 depicts condenser water pumps and chilled water pumps with an intervening header between the pumps and chillers, our chiller plant featured each pump dedicated to a specific chiller with normally closed manual cross-connections between pumps for flexibility of operation.
* Finally, Figure 1 for the integrated economizer depicts the economizer heat exchanger piped on the chilled water side in series with the chillers (with a bypass) such that chilled water can flow first through the heat exchanger and then through a chiller; but the condenser water remains piped in parallel such that any one unit volume of condenser water flows through a heat exchanger or a chiller but not both. We elected to pipe both the chilled water and condenser water such that the heat exchanger can be flowed in series, for reasons which will be explained here.
Figures 2 and 3 (Page 63) depict the retrofit chiller plant for chilled water piping arrangement and condenser water piping arrangement, respectively. Piping that existed prior to this retrofit project is shown with lighter, thinner lines, while new piping is shown with heavier, darker lines. HX-2 is the new 1,000 ton (3520 kW) plate-and-frame waterside economizer heat exchanger, with its capacity chosen to provide the winter cooling load, essentially equaling the capacity of any two of the 500 ton (1760 kW) chillers.
Nomenclature. AS Air Separator CH Water-Cooled Centrifugal Chiller CWP Condenser Water Pump CWR Condenser Water Return CWS Condenser Water Supply PCHWP Primary Chilled Water Pump PCHWR Primary Chilled Water Return PCHWS Primary Chilled Water Supply HX Plate-and-Frame Water-to-Water Heat Exchanger SCHWR Secondary Chilled Water Return SCHWS Secondary Chilled Water Supply TCV Temperature Control Valve [??] Connection Point of New Work to Existing
Beginning with Figure 2, we note that each existing primary chilled water pump is piped to serve a chiller, with any two chillers and their respective two chilled water pumps expected to operate at a given time. This actually made construction of the project easier than would have been the case with the Figure 1 arrangement, since one pump/chiller leg at a time could be isolated for cutting and patching of piping while the other two remained in service. We were also fortunate that the pumps and chillers were located physically far enough apart, with enough intervening pipe between them, to allow for the new tie-ins shown.
Again referring to Figure 2, the new piping arrangement accomplishes the major goals of the project:
* Flow from any two primary chilled water pumps can be diverted to the heat exchanger and then bypass the chillers altogether (full economizer); or
* Flow from any two primary chilled water pumps can be diverted to the heat exchanger and then also directed through chillers before being sent to the load (partial economizer); or
* The heat exchanger can be isolated with flow routed only through chillers (traditional summer operation).
Table 1 illustrates valve positions to accomplish each of those three scenarios.
One key consideration in this arrangement is the fact that the partial economizer mode adds pressure drop above and beyond the original system design because flow must route through both a heat exchanger and a chiller in series. Therefore, the heat exchanger was selected with as little pressure drop as practical. We were lucky that the existing chilled water pumps were sized somewhat conservatively and had enough capacity to nearly overcome the added pressure drop. A small reduction in flow rate due to the added pressure loss was offset by the fact that some building loads are reduced during weather that allows for partial economizer (outdoor air and wall/roof conductance, for example) so the serviced load is not truly 1,000 tons (3520 kW) in this mode anyway.
Turning to Figure 3, we again see that each existing condenser water pump is piped to serve a chiller, with any two chillers and their respective two condenser water pumps expected to operate at a given time. And again, flow from any two condenser water pumps can be diverted to the heat exchanger and then bypass the chillers altogether (full economizer). Or, flow from any two condenser water pumps can be diverted to the heat exchanger and then also directed through chillers before being sent to the load (partial economizer). Or, the heat exchanger can be isolated with flow routed only through chillers (traditional summer operation). Again, refer to Table 1.
The key discussion point in this arrangement is the fact that, unlike Figure 1, we elected to implement a series arrangement for the condenser water piping such that any one unit volume of condenser water can flow through a heat exchanger and a chiller sequentially. It would not be necessary to do this, as it is possible to operate partial economizer with condenser water being fed to the chiller and heat exchanger in parallel. So why did we choose the series arrangement?
Both Taylor (3) and Stein (4) have recommended head-pressure control at the chillers as a means of permitting chiller operation using cold condenser water during partial economizer. The condenser water must be cold enough to chill water in the heat exchanger via temperature gradient only, but chillers have difficulty staying online with cold condenser water. So the head-pressure control strategy allows a reduction in flow of cold condenser water through an operating chiller such that the chiller leaving condenser water is warm enough to maintain the chiller online. But head pressure control and a reduction in condenser water flow through an operating chiller made our customer very nervous, uncomfortable, and more than a little skeptical. So by first allowing the condenser water temperature to rise inside the heat exchanger during partial economizer, this problem of feeding cold condenser water to a working chiller is minimized.
Since this specific waterside economizer serves data center loads, the chilled water setpoint can be higher than might be used for office building service. That fact plus the warming of condenser water in the heat exchanger during partial economizer results in condenser water warm enough to feed the chillers without head pressure control. To illustrate, on partial economizer we may be able to have condenser water enter the heat exchanger at 50[degrees]F (10[degrees]C) and leave at 58[degrees]F (14[degrees]C), while precooling return chilled water from 60[degrees]F (16[degrees]C) to 53[degrees]F (12[degrees]C) before sending it on to a chiller. When the condenser water is so cold to allow full economizer, head pressure control concerns are non-applicable since all chillers shut off, and the heat exchanger provides full cooling.
But we did leave ourselves a backup plan. If partial economizer flow through the heat exchanger results in condenser water still too cold for service to the chillers, motorized isolation valves 16, 26, and 36 (intended to be open or closed) were specified with modulating actuators so they could, if necessary, function as the head pressure control valves depicted in Figure 1.
Another consideration is that we did not want to increase the overall system flow rate. We were concerned that parallel operation during partial economizer could potentially increase total system flow rate because flow, rather than pressure drop, would be additive among two chillers and the heat exchanger. As stated in the chilled water discussion, the heat exchanger was selected with as little pressure drop as practical. The existing condenser water pumps were sized somewhat conservatively and had enough capacity to nearly overcome the added pressure drop. We were more comfortable increasing overall system pressure drop instead of overall system flow, since an increase in flow may have unforeseen consequences at the cooling towers themselves.
Not the Only Solution
The piping solution presented here adds valves and a heat exchanger to the central plant, but no new pumps. Two of the three chilled water pumps and two of the three condenser water pumps operate 24/7/365, and a simple monthly schedule rotates duty among the three pumps. This solution does add quite a large number of valves, but the operating concept is simple. The temperature control system only has to ask itself two questions: (a) Which two of the three pairs of chiller/condenser pumps are operating? and (b) which of the three economizer modes is applicable (winter, summer, or hybrid)? Knowing the answer to those questions, the temperature control system decides which valves to open and close based on a table (of which Table 1 is an sample). The valves are spring-return and all of them fail in the position that allows mechanical cooling via the summer mode.
However, other solutions are available, using no new valves, but instead the addition of new pumps to route chilled water and condenser water through the heat exchanger. Floor space for additional pumps was scarce. These pumps would have been rather large, because they would need to handle the chilled water and condenser water flow equivalent of two chiller's worth. Electrical capacity for an additional pump was questionable as well. So the solution that adds valves but no new pumps was found to be preferable in this case, but may not be in other cases.
Phasing of Construction
Now that we've explored the desired piping arrangement, we turn to a discussion of the sequential phasing of construction that was required to implement the desired piping arrangement. This chiller plant serves a data center critical to the national defense, and must operate on a 24/7/365 basis without interruption. Because of its critical nature, there is a spare (often called "N+1") of each major component: chiller, each type of pump, and cooling tower. Therefore, the customer permitted one and only one chiller (plus its associated primary chilled water pump and condenser water pump) to be taken out of service at any one time, and construction phasing became a key point of this case study. The contractor was instructed to phase the construction activity as follows.
When taking CH-1 out of service, confirm that CH-2 and CH-3 (plus their associated primary chilled water pumps and condenser water pumps) are available for duty. Perform Steps 1 through 11 (see "Piping Implementation Steps" sidebar) as quickly as possible in one continuous 24/7 work operation, not stopping for nights or weekends, using multiple shifts of installers as necessary until Steps 1 through 11 are completed. Confirm that all parts and materials necessary are on site and preassembled to the greatest extent possible before beginning Step 1.
With this phasing of construction, only two new-to-existing connections would require a shutdown of operation: the one near TCV-41 in Figure 2, and the one near TCV-42 in Figure 3. After discussing this with a reputable mechanical contractor and learning that this delay would be far longer than the data center could tolerate for an outage, these two connections were specified to be performed by "hot tap" or pressure tap. (5)
Cold Weather Sequences of Operation
Finally, we turn to the challenges of cold weather operation of cooling towers. While these cooling towers were previously operated in the winter, the condenser water temperatures were much warmer. We were now asking those cooling towers to produce water cold enough for heat exchanger duty 40[degrees]F to 45[degrees]F (4[degrees]C to 7[degrees]C), so the risk of freezing is much greater. A good primer on cold weather cooling tower operation is presented by Lindahl. (6) The greatest risk of freezing occurs when one attempts to minimize flow through the cooling tower media on a cold day, so any condenser water tower bypass should be all-or-nothing when using waterside economizer in cold weather.
The sequence begins by attempting to bypass all condenser water flow straight to the cooling tower basin. Depending on the wind, just how cold the outdoor temperature is, and similar factors, this might be enough to cool the condenser water a few degrees, and that might be good enough depending on the load. Next, if the condenser water temperature leaving the tower's basin is too warm, shut the bypass completely and put the full flow over the cooling tower fill, but without turning on the tower's fan. Again, that might be enough to meet setpoint, simply from natural convection and additional contact between water and cold ambient air. If that is still not enough, then start the cooling tower fan on very low speed, just above stall, and increase fan speed only to the extent necessary.
I recommend the use of variable frequency drives (VFDs) on any cooling towers intended for use in cold-climate waterside economizing. Full-speed fan operation increases the risk of freezing the condenser water in cold weather.
At first glance, it is tempting (on paper) to modulate the condenser water tower bypass valve and mix some flow over the tower media with some flow bypassed, but that presents the greatest risk for freezing because of the small amount of flow over the media. So during this mode, it is recommended to be all-in or not at all. Some may initially feel this appears abrupt and unstable, but that is generally not the case, at least not on larger systems. If the cooling tower were located immediately adjacent to the economizer heat exchanger, such that there is comparatively little volume of water in the system, this scheme may be abrupt and unstable. But the majority of condenser water systems have a larger volume of water due to large diameter pipe and/or significant lengths of pipe between tower and heat exchanger, so the thermal mass of water is great enough to ride out the bumps.
Sump heaters are often specified for open-cell cooling towers where winter cooling tower operation is anticipated, and in an irony sometimes they must operate during economizing--which partially defeats the purpose of an economizer. However, the condenser water won't typically freeze in the sump during use due to the rate of circulation, and the sump heaters typically operate only during off periods (nights/weekends) for non-24/7/365 buildings.
Some engineers specify a secondary or auxiliary indoor condenser water sump. These indoor sumps are great for non-24/7/365 buildings because one can store all the condenser water indoors on nights/ weekends when the systems are off, and reduce or eliminate use of the sump heaters. But indoor sumps do nothing to help you during waterside economizer operation. Bypassing water to an indoor sump results in no cooling of the condenser water, so you must run water over the tower fill to achieve any cooling at all. Bypassing water to an outdoor sump might be enough to cool the condenser water a few degrees as stated earlier, depending on conditions. Since many waterside economizers are applied on 24/7/365 applications such as data centers and four-pipe fan-coil buildings (e.g., hotels, dormitories, multifamily housing, and hospital patient rooms), the indoor sump may not buy you much. Finally, if the indoor sump is very large (i.e., holds a large water volume) then it can result in significant time delays due to the thermal mass of water when changing over back-and-forth from warm-weather non-economizer mode to mild-weather partial economizer mode--a lesson learned on this project.
When retrofitting waterside economizers onto existing cooling towers, one must make sure the water level in the sump is set high enough so the sump isn't drawn down empty when the bypass valve is initially closed and all flow is directed over the fill, since there is a time lag as the water needs time to trickle through the fill and refill the sump. On the other hand, one must make sure the water level in the sump is not set too high, lest the sump overflow each time the system is shut off, as again the time lag for the residual water in the fill continues to fill the sump for a period of time. Finding the ideal cooling tower sump level can be a trial-and-error process in the field that needs to be built into the expectations of the project.
The strategy of head-pressure control at the chillers as a means of permitting chiller operation using cold condenser water during partial economizer might appear to reduce net condenser water flow rates in violation of my earlier caution not to reduce condenser water flow over the cooling tower fill during freezing weather. But when the temperature outside is cold enough that one must be concerned about freezing, the plant should typically be operating on full economizer without any flow through the chillers. (Nall, (7) however, has challenged whether full economizer is always as available as we typically think it is.)
On an integrated economizer when one needs partial chiller operation, it will most often be the case that the air temperature outside is above freezing, and then it is okay to reduce flow over the fill. A good strategy is to first reset the chilled water temperature setpoint up to the extent possible while still meeting cooling loads. This increases the number of full waterside economizer hours and helps ensure that the integrated operation with partial chiller operation all occurs above freezing.
One final consideration: earlier in the article, I emphasized how critical this data center is and how important it was for all pumps, chillers, and cooling towers to have an "N+1" or redundant component. So why did we not include a second spare "N+1" waterside economizer heat exchanger in the design? What if the heat exchanger fails during the winter? Well, we felt this was exceptionally unlikely since the heat exchanger has no moving parts, but moreover we actually did include an "N+1": the chillers provide the necessary redundancy. While it will not be energy friendly, the chillers can operate in the winter just as they did before this retrofit project, should the heat exchanger ever fail. The energy penalty for doing so will be much smaller-given the very few hours this should actually occur --compared to the cost of providing a redundant heat exchanger.
Retrofit case studies present interesting challenges as existing conditions often dictate alternative thinking outside of what the textbook or ideal engineering guidance might say. In this one, we studied why one might pipe condenser water in a waterside economizer such that the heat exchanger is in series with, rather than parallel to, the chillers. We examined a very specific construction phasing protocol to allow continuous chiller plant operation while the retrofit piping was constructed, installed, and connected. And we described sequences for operating cooling towers in cold weather to guard against freezing of water in the tower fill.
Piping Implementation Steps
1. Disable CH-1, PCHWP-1, and CWP-1.
2. Close the manual isolation valve at the discharge of PCHWP-1. Close the manual isolation valve at the chilled water inlet of CH-1. Confirm that all normally-closed cross-connection valves are indeed closed.
3. Perform the piping connections and tie-ins of TCV-11, 12, and 13.
4. Close new TCV-11 and 13 and open TCV-12 (wiring need not be complete).
5. Provide temporary caps on the downstream side of TCV-11 and 13.
6. Close the manual isolation valve at the discharge of CWP-1. Close the manual isolation valve at the condenser water inlet of CH-1. Confirm that all normally closed cross-connection valves are indeed closed.
7. Perform the connections and tie-ins of TCV-14, 15, and 16.
8. Close new TCV-14 and 16 and open TCV-15 (wiring need not be complete).
9. Provide temporary caps on the downstream side of TCV-14 and 16.
10. Then reopen the manual isolation valves at the discharge of PCHWP-1 and CWP-1. Reopen the manual isolation valves at the chilled and condenser water inlets of CH-1.
11. Restore CH-1 to service.
12. After completing the above work at CH-1, repeat all of the above for similar work at CH-2 (substituting the appropriate equipment tag numbers).
13. After completing the above work at CH-2, repeat all of the above for similar work at CH-3 (substituting the appropriate equipment tag numbers).
14. Perform remaining piping work to/from HX-2 at any time. When system is complete, remove temporary caps and make final connections.
(1.) Duda, S. 2013. "Lessons from energy audits." ASHRAE Journal (11).
(2.) ANSI/ASHRAE Standard 90.1-2016, Energy Standard for Buildings Except Low-Rise Residential Buildings.
(3.) Taylor, S. 2014. "How to design and control waterside economizers." ASHRAE Journal (6).
(4.) Stein, J. 2009. "Waterside economizing in data centers: design and control considerations." ASHRAE Transactions 115 (2).
(5.) American Petroleum Institute. "Procedures for welding or hot tapping on equipment in service." API Recommended Practice 2201, Third and Fourth Editions, October 1985 and September 1995.
(6.) Lindahl, P. 2014. "Cold weather operation of cooling towers." ASHRAE Journal (3).
(7.) Nall, D. 2014. "Waterside economizers and 90.1." ASHRAE Journal (8).
Stephen W. Duda, P.E., is senior mechanical engineer at Ross & Baruzzini, Inc. in St. Louis.
Caption: FIGURE 1 Integrated waterside economizer.
Caption: FIGURE 2 Chilled water flow.
Caption: FIGURE 3 Condenser water flow.
TABLE 1 Economizer control status. ECONOMIZER MODE CONTROL ITEM WINTER PARTIAL SUMMER TCV-11 OPEN OPEN CLOSED TCV-12 CLOSED CLOSED OPEN TCV-13 CLOSED OPEN CLOSED TCV-14 OPEN OPEN CLOSED TCV-15 CLOSED CLOSED OPEN TCV-16 CLOSED OPEN CLOSED TCV-21 OPEN OPEN CLOSED TCV-22 CLOSED CLOSED OPEN TCV-23 CLOSED OPEN CLOSED TCV-24 OPEN OPEN CLOSED TCV-25 CLOSED CLOSED OPEN TCV-26 CLOSED OPEN CLOSED TCV-31 CLOSED CLOSED CLOSED TCV-32 CLOSED CLOSED CLOSED TCV-33 CLOSED CLOSED CLOSED TCV-34 CLOSED CLOSED CLOSED TCV-35 CLOSED CLOSED CLOSED TCV-36 CLOSED CLOSED CLOSED TCV-41 OPEN CLOSED CLOSED TCV-42 OPEN CLOSED CLOSED CH-1 OFF ON ON CH-2 OFF ON ON CH-3 OFF OFF OFF CWP-1 ON ON ON CWP-2 ON ON ON CWP-3 OFF OFF OFF PCHWP-1 ON ON ON PCHWP-2 ON ON ON PCHWP-3 OFF OFF OFF Note: The above assumes that Chiller #3 is the spare or "N+1" chiller. In reality, any two of the three chillers may operate.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||COLUMN: ENGINEER'S NOTEBOOK|
|Author:||Duda, Stephen W.|
|Article Type:||Case study|
|Date:||May 1, 2017|
|Previous Article:||Creating a Contract for Success.|
|Next Article:||Damage Function Stress Relief.|