TES Tanks: A Quick Reference for Thermal Storage in Data Center Cooling
Created on Today
TES Tanks: A Quick Reference for Thermal Storage in Data Center Cooling
TES tanks shift cooling work rather than create it. In data center cooling, thermal energy storage holds chilled water, ice or phase change material made during off-peak hours so the chiller plant can shrink or idle during the peak. The three numbers to read are usable capacity in kilowatt-hours, usable factor as a percentage of theory, and autonomy hours at design load. A water tank stores roughly 1.16 kWh per cubic metre per kelvin of temperature difference and is usually designed at 70 to 90 percent usable. TES does not reduce total site electricity; it moves and reshapes it.
Procurement teams arriving at thermal storage for the first time usually meet the same vocabulary in the same week: TES, SOC, delta T, COP, PUE, N+1. Each of those terms carries a specific meaning in a chilled water plant, and a mismatch in definitions is where a storage project goes wrong. One party quotes nameplate tank volume, another quotes deliverable kilowatt-hours, and the contract ends up with a tank that meets its drawing and misses its duty.
This reference sheet covers the practical definitions used on data center storage projects: what the four TES routes are, which metrics matter in a specification, which beliefs turn out not to be true, and what each abbreviation means. Use it before the tank order, so the TES tanks on the drawings and the TES tanks on the site schedule are the same vessel.
Use it as a shared baseline between the design engineer, the contractor and the tank supplier.
The Four TES Routes at a Glance
Sensible chilled water is the reference. Water is cooled, held in an insulated tank and returned to the plant. Capacity follows `Q = V x rho x cp x dT`, and the design temperature band is usually 5-7 degC to 12-14 degC, giving a delta T of roughly 6 K.
Ice raises density by latent heat. Ice formed on coils or by a harvester inside a water tank stores about 334 kJ per kilogram of water turned to ice, which is far more per unit mass than sensible cooling. The trade is thermal shorting between cycles and an indirect loop.
PCM holds a phase change temperature. Encapsulated salts or paraffins melt and freeze at a set point, which lets the storage sit at a different temperature than the chilled water loop. A heat exchanger is always in the path.
Thermocline is water with a temperature front. Two water temperatures sit in one tank with a moving interface between them, charged by pumping hot return to the top and cold supply to the bottom. It uses the same medium as the chilled water route and simplifies the nozzle layout.
Route | Medium | Temperature band | Indirect loop | Typical use |
Chilled water | Water | 5-7 to 12-14 degC | Sometimes | Peak shave, buffer, emergency cooling |
Ice | Ice in water | Near 0 degC | Always | Space-constrained retrofits |
PCM | Encapsulated salt or paraffin | Set by phase point | Always | Narrow temperature band duties |
Thermocline | Water, two tiers | Split by front | Sometimes | Long storage, single vessel |
The Metrics That Belong in a Specification
Usable capacity in kilowatt-hours. Not tank volume, not nameplate. Size from load in kilowatts multiplied by the storage window in hours, divided by the usable factor, then divided by the volumetric density of the chosen medium. A 5 MW hall over an eight-hour window needs 40 MWh of deliverable cold before losses.
Usable factor and why it is below 100 percent. Stratification collapse, mixing at the inlet, thermal shorting and ambient heat leak all reduce the drawable fraction. The water route is normally designed at 70 to 90 percent of theory; ice and PCM land lower once their transfer limits are counted.
Autonomy hours and the failure case. Storage bought for the grid peak is different from storage bought for a generator start. A tank credited with, say, four hours of autonomy at partial load is a very different vessel from one credited with a full tier of runtime.
1. Stored energy: megawatt-hours of deliverable cooling at design return temperature.
2. Storage window: hours per day the plant expects to run on storage.
3. Charge and discharge rate: kilowatts in and out, which sets nozzle size, diffuser area and pump duty.
4. Standby loss: percentage of charge lost per day, replaced on the next charge.
5. Usable factor: percentage of theoretical capacity actually withdrawable.
Common Misconceptions
"TES reduces my electricity bill by storing free cold." It does not create cold. The chiller still produces every kilowatt-hour, some of it more efficiently off-peak when ambient is lower. The saving comes from tariff structure and demand charges, not from free energy.
"A bigger tank is always better." Storage that is never fully charged is dead weight. Charge time and chiller headroom must match tank size, or the vessel fills to 60 percent every morning and delivers 60 percent.
"Ice always beats water for footprint." Only if the indirect loop, the machine room height and the defrost duty are counted. The density gain is real; the delivered capacity after losses needs checking.
"The tank needs power to stay cold." A well-insulated vessel holds temperature for days. What needs power is circulation, not storage, which is why an atmospheric water tank helps during a generator start.
Abbreviations on a Cooling Schedule
TES thermal energy storage; in a data center the cold made off-peak and held in a tank.
SOC state of charge as a percentage of usable capacity, the same idea as a battery.
dT or delta T the difference between supply and return in kelvin or degC; the driver of water capacity.
COP coefficient of performance, cooling output over electricity input for a chiller.
PUE power usage effectiveness, total site energy over IT energy.
CHW / CW chilled water supply and the condenser water loop.
N+1 / 2N redundancy arrangement of the chiller plant.
Technical Specification
Item | Value or option |
Duty | Peak shave, chiller downsizing, or emergency cooling autonomy |
Medium | Chilled water, ice, PCM, or thermocline water |
Supply / return | 5-7 degC supply, 12-14 degC return as a starting band |
Delta T | 4-6 K conservative, 5-7 K typical for water |
Density (water) | About 1.163 kWh per m3 per K, around 7 kWh/m3 at 6 K |
Usable factor | 70-90 percent for water, lower for ice and PCM |
Tank condition | Near atmospheric, or low gauge pressure if designed as a pressure boundary |
Insulation | Mineral wool sized to keep the skin above dew point |
Usable confirmation | Measured capacity at the plant boundary over a full cycle |
Project Case
Field | Value |
Project | Namibia drinking water storage (P-A) |
Product | 01 Glass-Fused-to-Steel tank |
Capacity | 44,900 m3 total |
Dimensions | 4 tanks, diameter 41.26 m, height 8.4 m |
Completion | September 2022 |
While our delivered reference projects in the water and wastewater sector include this potable scope, the data-center TES scope is engineered to the same standards. Four bolted tanks at 41.26 m diameter are the reference for how a large near-atmospheric water vessel is panelised, shipped and assembled on site, which is the same construction basis a TES tank uses for its shell and coating.
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.
Terminology mapped to hardware
· TES is an umbrella term, and for a data center the useful split is by storage medium and by pressure class, not by marketing naming; the shell requirements follow that split.
· Definitions worth fixing early in a project: usable capacity (after mixed and dead zones), standing loss, charge/discharge rate, and pressure class (atmospheric versus pressurised boundary).
· Our atmospheric bolted shells for bulk water storage, delivered up to 60,000 m³ per tank, are the physical counterpart behind most of those definitions in practice.
Frequently Asked Questions
Q1: Do TES tanks reduce PUE?
A1: Not directly. Lower condensing temperature at night and longer chiller run hours at the design point can improve machine efficiency, and the main benefit is on the energy cost side. Calculate the change in machine COP against the change in tank losses before claiming a PUE gain.
Q2: What is a reasonable delta T for a chilled water tank?
A2: Five to seven kelvin is typical for a 6 to 12 degC band; four to six is the conservative choice when stratification loss is expected to be high. A wider band lowers the required volume for the same stored kilowatt-hours.
Q3: How is state of charge shown on a data center control screen?
A3: As a percentage of usable capacity, from the measured supply and return temperatures, the stored volume and the usability factor. Set the upper and lower limits so the plant never promises more cold than the tank holds.
Q4: Does a TES tank need to be a pressure vessel?
A4: No, not usually. A near-atmospheric water tank sits outside the pressure boundary, so the applicable code and design are those specified in the project specification and contract. A pressurised design moves the vessel, its openings and its inspection regime into a different scope.
Q5: Can one tank serve as both buffer and storage?
A5: Yes. Short-term buffer for chiller trips and pump transients and longer-term storage can share one vessel, provided the nozzle and diffuser layout supports it and the operator understands the combined state of charge.
Q6: Why is usable capacity lower than the theoretical figure?
A6: Stratification, mixing at the inlet, ambient heat leak and the outlet temperature band reduce what can be drawn. Designers apply 70 to 90 percent for water and measure the result at the plant boundary.
Q7: What should the acceptance test measure?
A7: Deliverable kilowatt-hours at the plant boundary over a full charge and discharge cycle, plus charge time, standby loss and the outlet temperature reached at design discharge rate. A TES tank contract that only quotes cubic metres leaves the acceptance test to argument later.
Read the storage project as a set of definitions before it becomes a set of drawings. Four routes exist, five metrics matter, and the abbreviations must mean the same thing to the client, the designer and the supplier. TES tanks do not make a data center use less electricity; they move that electricity into cheaper hours, shrink the chiller plant and give the plant a chance to ride through an outage. Get the definitions right and the vessel sizing follows.
Talk to an Engineer
Send the design IT load, the target storage window, the chiller supply and return temperatures and the available footprint. We will return the calculation behind this reference: stored kilowatt-hours, required cubic metres, the usable factor applied, and the outline nozzle, diffuser and insulation specification. If you are comparing routes, request the same calculation for ice and PCM so the comparison is made on deliverable capacity.
WhatsApp