TES Tanks Critical for Cooling Data Centers: Reliability under Failure Conditions
TES tanks critical for cooling data centers behave as a passive reserve. Stored cold covers the plant through a chiller trip or a utility failure while generators and UPS units come up, because an atmospheric water tank needs no external power to hold pressure and only needs power to circulate. Size the reserve from the load that must survive, the outlet temperature the load tolerates and the hours of autonomy, after a 70 to 90 percent usability factor. The failure modes that end that reserve are stratification collapse, freezing at nozzles and supports, wet insulation and a stuck diffuser, not the shell itself.
Reliability design starts with the questions nobody wants asked during an outage: how long must the plant stay cold with the chillers down, and what takes the load until the generators are at frequency? For a conventional plant the answer is mechanical: flywheel, UPS, emergency chilled water or a second utility feed. Adding thermal storage changes the arithmetic, because the cooling reserve already sits in the building as water in a tank.
That change is useful and not free. Storage introduces its own failure modes: a tank that stratifies into a lukewarm mixture, nozzles that freeze at the shell, cladding that hides a wet wall, a diffuser that sticks in the wrong position.
This article covers the failure conditions a mission critical cooling tank must survive, how autonomous run time is calculated, which failure modes shorten it, and what the specification should require.
What the Tank Is Required to Do
Carry the plant through the transfer window. A utility failure is not an outage of seconds; generator start, ATS sequencing, load shedding and the return of the second feed take minutes, and the critical load must not see a temperature excursion in that window. Stored cold covers the gap with no moving parts involved.
Support the generator start sequence. Emergency power arrives in stages. Chilled water pumps, cooling tower fans and control power come up as the generators stabilise, and during the ramp the tank is the only source feeding the critical branch.
Absorb the chiller trip. A single chiller going off on low refrigerant pressure or a VSD fault is routine, and storage sized only for generator transfer is smaller than the vessel needed for a two-chiller trip on a hot afternoon.
Hold temperature, not just volume. A tank gives energy only while the outlet stays inside the design band; once the mixed outlet passes the top limit the reserve is spent with cold still at the bottom.
For TES tanks critical for cooling data centers these duties set four sizing inputs: survival load, autonomy window, usability factor and the acceptable outlet band. Survival load is the cooling the critical halls need, less than total IT load once halls are shed. Autonomy is minutes for transfer or hours if storage also rides a chiller plant outage. Usability is stored kilowatt-hours times the 70 to 90 percent factor, never nameplate capacity. The outlet band is the temperature the load tolerates, which defines when the reserve counts as gone.
Calculating Autonomous Run Time
Start from the load, not the tank. Multiply survival load in kilowatts by autonomy in hours for deliverable kilowatt-hours, divide by the usability factor and by medium density to reach volume. At a 6 K delta water holds about 7 kWh per cubic metre before the factor, so a 2 MW survival load over three hours needs roughly 850 m3 of usable storage, or about 1,100 m3 of tank.
Then confirm the discharge rate. Capacity is rarely the binding constraint at these sizes. Nozzle diameter, diffuser area and pump duty set how fast cold leaves the vessel; a tank that discharges too slow reaches the end of its autonomy with cold still inside.
Check the return condition. Full charging needs the chiller plant to bring the return down to the supply band within the available hours. If charging only reaches partial at dawn, the plant should not be promised the same hours of discharge.
Credit the tank honestly in the reliability model. Designers normally give storage a defined autonomous contribution inside the concurrent maintainability narrative, alongside the UPS and generator path, rather than an unlimited runtime.
Condition | Storage role | Planned duration | Design input |
Utility drop, generator transfer | Primary cold source | Minutes | Transfer time plus margin |
Generator start and ramp | Primary cold source | Minutes into ramp | Pump power availability |
Single chiller trip | Absorb plant transient | Minutes | Trip coincidence study |
Chiller plant outage | Backup cooling | Hours | Usability factor and outlet band |
Pump or control failure | Not covered | Not credited | Isolate and diagnose |
Failure Modes That Shorten the Reserve
Stratification collapse is the quiet one. For TES tanks critical for cooling data centers the inlet diffuser must be correctly sized and positioned; when it is undersized or stuck open, warm return mixes into the cold layer and the outlet warms early. Design inlet velocity at 0.5 to 1 m/s, size the diffuser for peak flow, and keep the tank free of internal obstructions that encourage mixing.
Freezing at supports and nozzles ends a tank quickly. A cold bridge at a saddle or an uninsulated nozzle drops the local surface below zero, the surrounding water freezes and the expansion cracks the connection or pipe. Thermal breaks under saddles, insulated nozzle sleeves and a continuous vapour seal are the standard treatment.
Wet insulation hides corrosion. Condensation entering a cladding seam wets mineral wool, the wool compacts, standby loss climbs and a coating holiday corrodes behind a surface that still looks fine. Vapour sealing and a vented cladding keep the envelope dry.
Power-dependent accessories fail before the shell does. Level transmitters, motorised outlet valves, diffuser actuators and mixers all need power, so every powered accessory sits on the emergency bus with a defined position on loss of power. A drain that freezes or a trace heater that lost its supply takes the vessel out of service at the worst moment.
Technical Specification
Parameter | Typical value or option |
Duty | Generator transfer backup, chiller trip absorption, short autonomy |
Survival load | Design kilowatts of cooling for critical halls only |
Autonomy | Minutes for transfer, hours if credited with plant outage |
Usability factor | 70-90 percent of theoretical stored energy |
Supply / return band | 5-7 degC supply, 12-14 degC return |
Tank type | Near-atmospheric, or pressurised if the design specifies it |
Inlet velocity | 0.5-1 m/s at the diffuser |
Cold bridge control | Saddle thermal break, insulated nozzle sleeves, vapour seal |
Powered accessories | Level transmitter, motorised valve, actuator on emergency bus |
Isolation | Supply and return isolation valves plus tank bypass |
Acceptance | Delivered kilowatt-hours and outlet temperature on a real transfer |
TES tanks critical for cooling data centers are specified against this table before the vendor is asked to quote, so that autonomy, usability and the acceptance test are all defined in the same document.
Project Case
Field | Value |
Project | China Sichuan fire water storage (P-C) |
Product | 01 Glass-Fused-to-Steel tank |
Capacity | 8,930 m3 total |
Dimensions | 2 tanks, diameter 19.87 m, height 14.4 m |
Completion | November 2023 |
While our delivered reference projects in the water and wastewater sector include this fire water scope, the data-center TES scope is engineered to the same standards. Two bolted tanks at 19.87 m diameter built to hold water reliably without a powered pressurising system is the same reliability argument a chilled water reserve makes: a large near-atmospheric vessel whose content is held by the medium and the insulation rather than by an active device.
Center Enamel Engineering Capability
Center Enamel (Shijiazhuang Zhengzhong Technology Co., Ltd) has designed and fabricated bolted storage tanks since 2008. As the first glass-fused-to-steel (GFS) tank manufacturer in China, the company holds close to 200 enamel-related patents, produces roughly 300,000 enamel-coated steel plates a year, has completed more than 30,000 installed projects and supplies its tanks to over 100 countries. The new 150,000 m² production base was added to raise output capacity, and single tanks are supplied up to 60,000 m³. Manufacturing runs under ISO 9001 and ISO 45001, with product certification including NSF/ANSI 61, WRAS, FDA, LFGB, CE (EN 1090), ISO 28765, FM, BSCI and EUROCODE, and design referenced to AWWA D103-09, AWWA C550 and NFPA where the application requires it.
Passive reserve behaviour
· An atmospheric tank has no pumps, gas space control or pressurised boundary in its own right; when the plant loses power the stored cold is still there, which is exactly the reliability property a mission-critical hall is buying.
· Enamelled or welded carbon steel shells with standard insulation keep that property through the whole storage volume, and a bolted shell can be inspected compartment by compartment while the rest stays in service.
· Fire-water and emergency water tanks in our record, such as the Sichuan pair holding 8,930 m³, exist for the same reason: the reserve must stay available while other subsystems are down.
Frequently Asked Questions
Q1: Can a storage tank really keep a data center cold with no mains power?
A1: Yes, for a defined window. An atmospheric tank holds cold without any active device; only circulation needs power, and that runs on the emergency bus. Credit it with the autonomy the sizing study gives, not with unlimited runtime.
Q2: How do I size autonomy for a generator transfer?
A2: Multiply the survival cooling load by transfer time in hours plus margin, divide by the usable factor and the medium density. For a few minutes of transfer the vessel is small; for hours it becomes a tank farm.
Q3: Is a pressurised or atmospheric tank better for standby duty?
A3: Atmospheric is the simpler reserve because it holds content without a pressurising system. A pressurised vessel adds a dependency on that system during the exact condition where you need the tank.
Q4: What is the most common reason a tank underdelivers in an outage?
A4: Mixing. When the inlet or outlet arrangement does not hold temperature layers apart, the mixed outlet warms early and the reserve is spent while cold water remains at the bottom.
Q5: Do the tank instruments need to be on emergency power?
A5: The ones that matter for failover should be. Level indication, the outlet valve actuator and any diffuser actuator must actuate on the emergency path, with a defined safe position on loss of power.
Q6: How is tank performance verified before handover?
A6: Run a transfer on the real plant: isolate a chiller or the mains feed, move the load to the tank, and record delivered kilowatt-hours, outlet temperature and time to limit. Nameplate capacity alone is not evidence.
Q7: Does the insulation envelope fail during long idle periods?
A7: It degrades slowly through condensation. A vented cladding, a vapour seal and a dew-point calculation for thickness keep the wool dry between cycles.
A tank earns its place in a reliability study only if the failure analysis covers it. Store enough cold for the surviving load over the transfer window, apply the usability factor honestly, size the discharge path so cold can leave, and remove the failure modes that shorten the reserve, which are mixing, freezing at cold bridges and wet insulation. TES tanks critical for cooling data centers then behave as intended: a passive reserve that buys the minutes the plant needs to start its own machines.
Talk to an Engineer
Send the survival load in kilowatts, the transfer window in minutes or hours, the acceptable outlet temperature and the emergency power arrangement, and we will return the autonomy calculation, the tank volume with the usability factor applied, and the nozzle, diffuser and cold-bridge specification. For an existing plant, request a failover review of the current storage sizing and control sequence before the next commissioning window.