TES Tanks: A Quick Reference For Thermal Storage In Data Centre Cooling
Created on Today
TES Tanks: A Quick Reference For Thermal Storage In Data Centre Cooling
A TES tank stores cooling as either sensible heat in stratified chilled water, latent heat in ice, or latent heat in a phase-change material, and it lets the chiller plant shave peak demand or ride through an event; the right architecture is chosen from the available temperature lift, the space and the duty the storage must perform.
Introduction
Thermal energy storage (TES) in a data centre cooling plant is a tank that stores cooling when it is cheap or abundant and gives it back when the load or the tariff demands it. It is the single largest lever most cooling plants have for cutting peak electrical demand and for riding through a grid event without generator fuel, and it works with the chillers you already own rather than replacing them.
This is a compact reference, not a design manual. It tells you what a TES tank is, the three architectures in common use and the numbers that define each, how storage reshapes the chiller plant, what it costs and what it returns, the cases where it is not worth doing, and the questions you should take to a specialist before you commit capital. The physical basis is the ASHRAE Handbook; the performance testing of storage tanks follows the uniform testing procedures published by IAPMO.
Read this first when you are orienting, then take the questions at the end to the engineer who will size the actual tank.
What a TES tank is in one paragraph
A TES tank is a store of cooling capacity, not a generator of it. It holds a volume of fluid at a temperature that is useful to the plant, charged by the chillers when there is spare capacity or cheap power and discharged to the cooling coil or the distribution when the load or the tariff calls for it. The tank itself does no cooling; it decouples the time the cooling is made from the time it is used.
For a data centre the value is twofold. Operationally it carries the plant through a grid disturbance or a chiller trip for the minutes or hours the thermal mass can cover, and economically it lets the chillers run at night or during off-peak tariff to fill the tank, then trim or avoid the peak electrical demand during the day. The tank is the buffer between the chillers and the load in time, rather than only in flow as a buffer tank does.
The three architectures
Three architectures dominate. Stratified chilled water storage is the simplest: a tall tank of water charged cold at the bottom and warm at the top, separated by a thermocline, so the plant draws cold water from the bottom and returns warm to the top. It runs across a modest temperature lift of a few degrees and uses ordinary water as the medium.
Ice storage uses the latent heat of fusion: a much larger energy density per volume than water, built as ice-on-coil where the ice forms around tubes the brine flows through, or as a dynamic harvester that makes and stores ice separately. Phase-change material storage sits between the two, using a salt or eutectic that melts and freezes at a chosen temperature, giving latent storage at a set point without the very low brine temperatures ice demands.
The numbers that define each architecture
The numbers that define each architecture are the energy density and the temperature lift it needs. Stratified chilled water stores roughly 1.16 kWh per cubic metre per degree Celsius of delta-T (rho x cp x delta-T), so a 5 degC (9 degF) swing gives around 5.8 kWh/m3, and the tank is large but the chillers stay at normal chilled-water temperatures.
Ice stores about 93 kWh/m3 of latent heat at the freezing point, so the tank is far smaller for the same energy, but the chillers must make brine near 0 degC (32 degF) or below, which costs lift and efficiency. Phase-change material stores at its melt temperature, typically tens of kWh/m3, between the two extremes. The selection is a trade of tank volume, chiller lift and the temperature the distribution can accept. The choice is rarely about which stores the most energy per cubic metre, because the plant can usually make space for a tank; it is about the chiller lift the chosen architecture demands and the distribution temperature the existing coils will actually accept.
How it changes the chiller plant
Storage changes the chiller plant by letting it run against the tank instead of against the instantaneous load. In charge mode the chillers fill the tank, sized to the average load over the charge window rather than the peak, so a smaller or slower chiller plant can serve a larger peak load than it could without storage.
In discharge mode the tank supplies part or all of the cooling, so the chillers can be throttled or stopped during the peak or the event. The plant control gains a degree of freedom: when to make cooling versus when to use it, decided by tariff, grid signal or resilience plan rather than by the live coil demand. That freedom is the asset; the tank is the hardware that holds it.
What it costs and what it saves
What it costs is the tank, the extra pump and distribution, and the controls to charge and discharge it; what it saves is peak electrical demand, the associated demand charge, and in some markets the ability to avoid a generator or reduce its fuel. The return is a function of the tariff structure and how often the peak is actually shaved, not a fixed percentage.
The honest statement is that the saving depends on the duty: a plant that peaks hard against a steep demand charge recovers the tank faster than one with a flat profile, and a campus that needs resilience through grid events values the tank for the ride-through it provides regardless of tariff. Quantify both the demand reduction and the resilience value before the capital is committed, because one or the other usually carries the case.
When it does not make sense
Thermal storage is not worth doing when the load profile is already flat, the tariff carries no demand charge worth shaving, or the site has no space and no tolerance for the chiller lift the chosen architecture needs. A plant that never peaks against a charge has nothing for the tank to smooth, and the capital sits idle.
It also fails when the distribution cannot use the storage temperature: stratified chilled water needs a delta-T the coils accept, ice needs brine plant the owner will not operate, and phase-change needs a set point the system actually crosses. If the only way to use the stored cooling is to rebuild the distribution, the storage is the wrong first investment. Match the architecture to the plant you have before sizing the tank.
Questions to take to a specialist
Take a short, specific list to the specialist so the conversation is about numbers, not categories. The questions are about the duty, the temperature, the space and the money, and each one constrains the architecture that survives.
· What duty must the tank perform: demand shave, ride-through, or both, and for how long?
· What delta-T can the distribution actually use, and what chiller lift is acceptable to make the storage?
· What volume and footprint are available on the site, and what is the slab or floor load rating?
· What is the tariff structure, and what is the demand charge or resilience value the tank would recover?
· Which architecture fits the temperature and space, and what is the installed cost versus the recovered value over the plant life?
Key terms and standards
A few terms and standards frame every TES discussion. Thermocline is the graded boundary between warm and cold water in a stratified tank; full charging keeps it sharp. Latent heat of fusion is the energy ice or a phase-change material stores at constant temperature as it freezes or melts, distinct from sensible heat stored by a temperature change.
The engineering references are the ASHRAE Handbook - HVAC Systems and Equipment for the plant physics and the IAPMO uniform testing procedures for thermal storage tanks for how performance is measured. A tank is specified by its energy capacity in kWh of cooling, its operating delta-T, its charge and discharge rates, and the temperature the distribution accepts; keep those four on every drawing and the architecture choice stays honest.
TES Architecture Comparison: Medium, Energy Density, Chiller Lift
WhatsApp