Chilled Water Storage Tank Data Center: How Charging and Discharging Work
Created on Today
Chilled Water Storage Tank Data Center: How Charging and Discharging Work
A chilled water storage tank data center charges by pumping 5-6 °C water into the shell through a low-velocity diffuser, letting the cold mass sink and a thermocline form, then discharges from the bottom nozzle as the level falls and the thermocline descends. Charge flow is set by chiller capacity, discharge flow by the IT load, and both by the same energy equation. Real usable capacity lands between 70% and 90% of theoretical after mixing and standing losses. Switching between the two phases, not the tank volume, is what decides whether supply temperature stays inside the band.
Operators rarely lose storage capacity on the day the tank was commissioned. They lose it gradually, as the thermocline widens and the switching logic fires at the wrong point, until the same shell delivers two peak hours instead of four. That decay is a charging problem, not a vessel problem.
The two phases are simple in isolation; the difficulty is the transition, and the fact that a tank entering discharge partly mixed behaves differently from one entering at full charge. What follows describes both phases, the arithmetic behind the flow rates, and the logic that decides when to cross between them.
Charging: Filling the Shell Without Destroying the Layers
Charging is a slow, low-velocity operation. The chillers make cold water, the plant pump moves it into the shell, and stratification does the rest.
1. Charge flow follows chiller capacity. Q = m × cp × ΔT sets it; a chiller producing 3 MW into water at cp 4.186 kJ/kg·K with a 5 K rise needs about 143 kg/s, roughly 515 m³/h.
2. Enter through the diffuser. The inlet must discharge at 0.5-1 m/s so the cold jet does not stir the whole shell. A jet that plunges through the warm layer mixes everything and costs the next cycle its usable volume.
3. Keep the thermocline in the middle. Cold water is denser, so it settles; warm return floats. The band between the two, the thermocline, is where the mixing happens, and a thin thermocline means a tank that gives more usable cold.
4. Stop on level, not on time. The charge ends at the high setpoint, leaving a top reserve so the thermocline never touches the inlet.
A chiller that cannot reach 4-6 °C on a hot day will not fill a tank; the charge simply stalls at a warmer level and the plant behaves as if the tank were smaller.
Discharging: Drawing Cold From the Bottom
Discharge reverses the direction of flow, and the useful cold leaves from the bottom of the shell while the level drops.
· Discharge flow follows the load. For the same 3 MW at a 6 K usable ΔT the loop needs about 600 m³/h. If the pump cannot hold that flow from a cold shell, supply temperature rises during the worst hour of the peak.
· The top reserve matters more. Only the cold below the thermocline is usable, so a tank discharging to an empty shell spends its last hour on mixed water.
· Watch the outlet temperature, not the level. Two sensors, one near the bottom and one at the top of the cold zone, tell the operator when usable cold is nearly gone.
Switching Control and the Blending Margin
The handover between the two phases is the part most plants get wrong, and it is decided by setpoints rather than by hardware.
· Enter discharge on supply temperature. A 1-2 K deadband around the supply setpoint is the trigger, with hysteresis so the plant does not oscillate between phases every few minutes.
· Open before the chiller unloads. Moving the discharge valve and raising the chiller setpoint in the same scan is what causes the dip; sequencing them, and holding 2 K of blending margin in the common pipe, is what avoids it.
· Keep the isolation valves interlocked. A chiller running while the discharge path is open pushes warm water back into the shell and undoes a night of charging.
Under a failure condition the same logic applies without the chillers: the tank discharges against the load while the generators start, and usable depth decides how long supply stays in band.
Mixing Losses, Standing Losses and the Usable Fraction
Three mechanisms take capacity from the theoretical figure, and each has a different remedy.
Loss mechanism | Typical effect | What reduces it |
Charge and discharge mixing | 10-25% of theoretical | Low inlet velocity, correct diffuser, adequate top and bottom reserve |
Standing heat gain through the shell | Small per cycle, larger over long holds | Thicker insulation, continuous cladding, thermal bridge detail at supports |
Thermocline growth during hold | Grows with hold time | Shorter holds, still water, no flow through the idle path |
The usable fraction is a design output. A chilled water storage tank data center with a 0.5 m/s inlet and a proper diffuser sits near the top of the 70-90% band; a tall shell with a hard-charging chiller sits near the bottom. Insulation is what operators notice first, because a shell that sweats or freezes at the supports has already lost its standing loss argument.
Technical Specification
Parameter | Typical value or choice | Why it is specified |
Charging supply temperature | 4-6 °C from the chillers | Chiller approach and IPLV |
Discharging supply temperature | 6-8 °C | Keeps IT inlet inside the band |
Usable ΔT across a cycle | 5-7 K | Sets energy per cubic metre |
Charge flow | From chiller capacity and ΔT | Sizes the inlet and diffuser |
Inlet velocity through the diffuser | 0.5-1 m/s | Protects the thermocline |
Top and bottom reserve | A few percent of shell volume each | Keeps the thermocline off the nozzles |
Usable fraction claimed | 70-90% of theoretical | Governs gross volume |
Standing loss target | State in kW, not thickness alone | Verifies the insulation package |
Project Case
Our delivered reference work in the water and wastewater sector includes large-diameter bolted shells built for long service lives, such as the installation below. The data-center TES scope is engineered to the same standards.
Project Case Summary
Project | Product | Capacity | Dimensions | Completion |
Namibia - drinking water storage (P-A) | 01 GFS bolted glass-fused-to-steel tank | 44,900 m³ total, 4 tanks | 4 × φ41,260 mm × 8,400 mm | September 2022 |
One shell at that installation holds roughly 11,000 m³, the order of magnitude a data center plant starts with for two to four peak hours. What carries across to a chilled water storage tank data center is the shell behaviour: enamel-coated walls with a roughness below Ra 0.8 µm, a lining inspected panel by panel, and a 30-year life under insulation.
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.
Charge and discharge envelope
· Charging and discharging swap the tank between inflow and outflow duty while the shell, insulation and enamel surface stay unchanged, so operation is a valving and control sequence rather than a process change.
· Atmospheric bolted shells are typically specified with enamelled or aluminium tops, keeping the tank free of the electrical equipment that a pressurised vessel would require for gas space control.
· Large single-shell atmospheric tanks in our delivery record, such as the Sichuan fire-water pair of φ19.87 m × 14.4 m (8,930 m³, 2023-11), are run for years in continuous duty with the same envelope.
Frequently Asked Questions
Q1: Why does the tank need a top reserve?
A1: Because the thermocline rises as the shell fills. If it reaches the inlet, newly charged cold water travels straight to the outlet and the plant draws mixed water. A few percent kept above the operating level keeps the layers separated.
Q2: What flow rate should the discharge pump be sized for?
A2: The peak load divided by the usable ΔT, taken from the same energy equation used for charging, plus margin for the terminal units calling at once. Sizing from the average load is how plants find supply drifting during the last hour of the peak.
Q3: Why did the tank deliver less cold than it did last year?
A3: Either the inlet velocity went up, which widens the thermocline, or the charge stopped short because the chillers could not reach 4-6 °C. Check the charge curve and the diffuser first.
Q4: How do I know the usable fraction before commissioning?
A4: Model it, then verify it. Ask each vendor for the usable depth, the reserve volumes and the resulting percentage, and confirm the inlet velocity and diffuser type.
Q5: Can the tank be discharged without the pumps running?
A5: Only with enough static head and an open path to the load, which is uncommon in a distributed plant. Most designs keep the pumps running so flow, filtration and temperature control stay predictable during the discharge window.
Charging and discharging are one system with two flow rates. Charge flow comes from chiller capacity, discharge flow from the IT load, and the usable cold between them from the depth of the thermocline and the discipline of the switching logic. Specify the diffuser, the reserves and the deadband, and a chilled water storage tank data center will hold its calculated usable capacity for the life of the shell.
Talk to an Engineer
Send us the chiller capacity, the IT load profile and the supply and return temperatures. We will work through the charge and discharge flow rates with you, show the usable depth for the proposed geometry, and return a specification sheet covering inlet velocity, reserves, level instrumentation and the insulation package. Where the plant has run a season, bring the charge curve and we will look at where the usable fraction is being lost.
WhatsApp