Data Centre Buffer Expansion And TES Vessels: Retrofit And Capacity Upgrade

6_1.jpg

Data Centre Buffer Expansion And TES Vessels: Retrofit And Capacity Upgrade

A live retrofit is governed by the changeover, not the tank: survey the existing loop, build a temporary bypass so cooling never stops, route the new pipework within the live plant constraints, and commission against a risk register that names the failure that matters most, which is loss of cooling.

Introduction

Adding storage to a data centre cooling plant is a different problem once the plant is already running. A new build starts from a blank slab and a shutdown that nobody notices; a retrofit has to thread new volume, new pipework and a new control path through a loop that is carrying the heat of live computing load every minute it is open. The constraint is not the tank, it is the changeover.

 

This article works through the retrofit the way a plant engineer plans it: why it is harder than new build, what the existing loop has to tell you before you cut a pipe, how a temporary bypass keeps cooling alive during the cut-over, where the pipework and the structure get difficult, when a phased approach beats a single event, and how you commission a change that the building never stopped using. The objective is capacity you can prove without an outage you cannot afford.

 

Throughout, the reference is the plant as it is, not the plant as drawn. The survey, the risk register and the sequence below are what turn an intention into a controlled change.

Why retrofitting is harder than new build

A retrofit is harder than new build for one reason: the system you are modifying is also the system that is protecting the load. In new construction the chiller plant is commissioned before the IT arrives, so a slip in the schedule costs money but not uptime. In a retrofit every cut, weld and fill happens while the hall is drawing its full design heat, so the margin for error is the margin the running plant had spare.

The second difference is information. A new plant has a coordinated model; a running plant has the model it was built to, the changes made since, and the reality of what is actually in the ceiling and under the floor. Retrofit engineering starts by closing that gap, because a bypass designed against the drawing and built against the field will not fit. The tank itself is the easy part; the tie-in is the project.

Surveying the existing loop

Before any pipe is cut, survey the existing loop for the things the model will not tell you: the true routing of the supply and return mains, the location and capacity of the existing pumps, the valve lineup that lets you isolate a section, the available connection points with their pressures, and the spare floor or roof load path to the proposed tank location.

You also need the operating envelope: the minimum and maximum flow the live plant can tolerate while a leg is isolated, the control sequence the building management system runs today, and the alarm and trip points that will fire if you disturb pressure or temperature. The survey output is a constraint map, not a drawing request, and it is what makes the temporary bypass and the changeover window realistic rather than optimistic.

Temporary bypass and the changeover window

The temporary bypass is the device that lets you open the live main without losing cooling. It is a short, fully valved alternative path around the section you are cutting, built and pressure-tested in advance, so that when you close the isolating valves the flow simply diverts and the chilled water keeps reaching the coils.

The changeover window is the time you allow for the cut, the tie-in and the reversal, and it is set by how long the plant can run on the bypass plus its residual buffer, not by how long the welder estimates. You stage the work so the window is shorter than that limit, with a rollback path if the new connection does not seal. The bypass is removed only after the new leg has proven itself holding pressure and flow.

Pipework routing in a live plant

Routing new pipework in a live plant means working inside space that is already occupied by cable trays, the existing mains, fire protection and the access paths maintenance already uses. The new run has to clear all of them, keep its fall and venting correct, and stay reachable for the isolation valves the next job will need.

The difficult cases are the penetrations: through a firewall, through a slab, or into a live plant room where the only route crosses active equipment. Each penetration is a small project with its own fire-stop, support and coordination requirement, and each has to be agreed before the changeover window, because you cannot discover a firewall in the middle of a bypassed main. Building services piping follows ASME B31.9, and the live plant adds the constraint that none of it can be closed for longer than the window allows.

Space, access and the structural constraint

Space is the obvious constraint, but the structural one is the one that stops projects. A buffer or thermal storage tank is a heavy, water-filled object, and the floor or roof it sits on has a load rating that was set when the building was designed, often for a different use. The proposed location has to be checked against that rating with the full operating weight, not the empty shell.

Access is the other hidden limit. A tank that cannot be brought through the door, up the stair or onto the roof in the pieces it ships in will not arrive, and craning onto a live roof carries its own exclusion zone over operating equipment. The survey has to confirm both the resting place and the path to it, including turning radii and lift points, before the vessel is ordered, because ordering first and discovering the path later is how retrofit budgets break.

Phased upgrade versus a single event

A phased upgrade adds capacity in steps rather than in one event: a first tank tied in during a short window, a second later as load grows, or a single tank brought on-line in stages so the plant never carries the full risk of one big cut-over. The trade is schedule and cost against risk concentration.

Phasing wins when the load is itself growing, because you match the storage to the demand curve instead of paying for the end-state volume up front, and it wins when a single window cannot safely cover the whole tie-in. It loses when the repeated windows and the repeated bypasses cost more than the risk they remove. The decision is a risk calculation, not a preference, and it belongs in the plan before procurement.

The commissioning plan for a live change

Commissioning a live change is not the same as commissioning a new plant. The new leg is filled and balanced while the old leg still carries load, so the plan names which flows move when, what the building management system should see at each step, and which alarms are expected versus which mean stop.

The commissioning plan is built around the risk register: loss of cooling is the top risk, so the first check is that the bypass holds and the live load is safe before any new water is introduced; leakage into the data hall is the second, so every new joint is tested and watched. You document each step as a hold point with a sign-off, because a live change that is not witnessed is a change nobody can defend if the load trips.

When a retrofit is not viable

A retrofit is not always the right answer. If the existing loop has no isolation valves worth the name, no spare connection point, or a control system that cannot absorb a partial configuration, the bypass and the changeover may be more disruptive than a planned replacement of the affected section.

The economic test is similar: when the structure cannot take the weight, when the access does not exist and cannot be made, or when the running plant has so little spare capacity that no safe window exists, replacement of the plant or a dedicated new build alongside it is the honest recommendation. The point of the survey and the risk register is to surface that answer early, before steel is ordered against a tie-in that cannot be made.

Live Changeover Risk Register: Risk, Trigger, Mitigation

Risk

Trigger

Mitigation

Loss of cooling

Bypass cannot hold plant load

Size bypass to live demand; witness test before window

Leak into data hall

New joint fails under pressure

Pressure-test each joint; monitor during and after tie-in

Window overrun

Tie-in exceeds safe isolation time

Keep window shorter than bypass-plus-buffer limit; hold rollback

Structural overload

Tank weight exceeds floor rating

Confirm load path in survey before ordering vessel

Control upset

BMS sees unexpected configuration

Name expected alarms; stage flow moves the BMS absorbs

Access failure

Tank cannot reach proposed location

Verify door, stair and lift path in the survey

Retrofit Sequence: Step, Action, Cooling Impact

Step

Action

Cooling impact

1 Survey

Map actual loop, loads, valves, structure

None; plant stays in service

2 Bypass

Build and pressure-test temporary path

None until valves close

3 Isolate

Close valves around the cut section

Flow diverts through bypass

4 Tie-in

Cut, weld and connect new leg

Plant on bypass; time-boxed

5 Test

Prove new leg holds pressure and flow

Live load still on bypass

6 Reverse

Restore main, remove bypass

Full normal operation resumed

7 Commission

Balance and witness hold points

None; documented handover

Project Reference

No verified project record matches this retrofit topic in the library. Rather than present an invented case, this article relies on the technical analysis, the selection framework and the live-changeover risk table carried below to support a retrofit decision.

About Center Enamel

Center Enamel (Shijiazhuang Zhengzhong Technology Co., Ltd.) designs and fabricates bolted steel storage tanks, welded steel tanks, fusion-bonded epoxy tanks, galvanized tanks, stainless steel vessels, pressure vessels, aluminum geodesic dome roofs and floating roofs. The company has more than 30 years of production experience and has supplied equipment to projects in more than 100 countries. Aluminum geodesic dome roofs are designed and fabricated to API 650 Appendix G, AWWA D108, the Aluminum Design Manual, ASCE 7 and IBC, and the company developed and exported a 72 m diameter aluminum dome roof in 2023. Engineering, fabrication, coating, quality control and site service are handled as one scope, which is why the drawings, the calculation basis and the delivered hardware can be reconciled against each other.

Frequently Asked Questions

Can you add a buffer tank to an operating data centre?

Yes, provided the existing loop can be isolated and bypassed. A buffer or thermal storage tank is tied in on a valved branch while the main continues to serve the load, so the hall keeps cooling through the changeover. The limiting factor is not the tank but the window in which you can cut and rejoin the live main without exceeding the plant's spare capacity, so the survey and the bypass design decide whether it is feasible.

How is a retrofit done without shutting down cooling?

By building a temporary, fully valved bypass around the section you are cutting, tested before the window opens, so flow diverts and chilled water keeps reaching the coils while the main is open. You close the isolating valves, make the tie-in inside a window shorter than the time the plant can run on bypass plus residual buffer, then prove the new leg before removing the bypass. A rollback path is kept in case the new joint does not seal.

What survey is needed before a retrofit?

Survey the actual routing of the supply and return mains, the pump capacities, the valve lineup that allows isolation, available connection points with their pressures, and the spare floor or roof load path. You also need the operating envelope: the minimum flow the live plant tolerates while isolated, today's control sequence, and the alarm and trip points. The deliverable is a constraint map, not a revised drawing.

How is the changeover sequenced?

Sequence the work as isolate, bypass, cut and tie-in, test, then reverse, with each step a witnessed hold point. The changeover window is set by how long the plant runs on the bypass and its residual buffer, not by the welder's estimate. Flow is moved in stages the building management system can absorb, expected alarms are named in advance, and the new leg holds pressure and flow before the bypass is removed.

What is the phased upgrade option?

Phased upgrade adds capacity in steps: one tank in a short window, another as load grows, or a single tank brought on-line in stages so the plant never carries the full risk of one large cut-over. It matches storage to a growing demand curve and reduces risk concentration, at the cost of repeated windows and bypasses. It is chosen by a risk calculation, not preference, and belongs in the plan before procurement.

What space is needed for a retrofit?

Space for the tank and, more often, the structural path to it. The tank is a heavy water-filled object, so the floor or roof load rating at the proposed location must take the full operating weight. Access is the other limit: the tank must be brought through the existing openings or lifted to the roof without an exclusion zone over live equipment, and the survey must confirm both the resting place and the path to it.

When should you replace rather than retrofit?

When the existing loop has no usable isolation, no spare connection, or a control system that cannot absorb a partial configuration; when the structure cannot take the weight or the access cannot be made; or when the running plant has so little spare capacity that no safe window exists. In those cases replacing the affected section or building new alongside is the honest call, which the survey and risk register should surface before steel is ordered.

 

A live retrofit is governed by the changeover, not by the tank. Survey the loop as it actually is, design a bypass that keeps cooling through the cut, route the new pipework inside the constraints of the occupied plant, and confirm the structure and the access before anything is ordered. The tank is the easy part; the tie-in is the project.

 

Commission against a risk register that names loss of cooling as the failure that matters most, witness each hold point, and keep a rollback path open until the new leg has proven itself. Send us the existing loop drawings, the operating envelope and the proposed location, and we will return a retrofit plan with the bypass, the sequence and the commissioning steps defined, so the capacity you add is capacity you can prove without an outage you cannot afford.

Talk To Our Engineering Team

Send the duty, the volume basis, the governing code and the site constraints, and our engineering team will come back with a specification that a supplier can price and a reviewer can approve. We will also tell you plainly when a different tank type or a different scope is the better answer.

Chat with us