Pump stations: questions and answers
founder and CEO of Cold Energy (NCT) · Updated October 30, 2026
This page is for engineers who design, buy or run the pump station of a cooling system: pumps, tanks, make-up and the glycol loop. These are the questions customers, design engineers and operators ask us. Figures come from our EH-APM006 design guide for engineers and the articles on this site; the exact figure for your project comes from a selection run on your duty point.
Questions on this page (17)
- How many pumps does a pump station need, and how much standby?
- Should the pumps have a variable frequency drive or a soft starter?
- Primary-secondary or variable primary flow: which scheme should I choose?
- Does having many pump stations in the scheme make the system less reliable?
- What is a glycol make-up station for?
- How should glycol be monitored, and can it be kept in a galvanised tank?
- Why choose a factory-built pump station over assembling the pumps on site?
- Can a pump station be installed outdoors?
- Will a sectional tank freeze outdoors?
- What data do you need to select a pump station?
- Should the elevation of the pipework be included in the pump head?
- Why does the design use a DN200 main when DN150 passes on velocity?
- Can I put the cooling tower pumps on the roof next to the tower to keep the pipes short?
- Can pumps of different sizes run in parallel?
- Why does a VFD on a pump save so little?
- Where should the expansion tank go in a closed loop?
- What test pressure should be used for a pump station on site?
How many pumps does a pump station need, and how much standby?
At least one duty pump and one standby pump, each sized for full flow, with automatic changeover. That is the standard Cold Energy (NCT) pump skid. Medium and large variable-flow loops use 2+1: three pumps of the same size. A primary loop serving several chillers gets one pump per machine plus a shared standby. The rule is the same in every case: losing any single pump must not take cooling away from the site. More in the design guides for engineers
Should the pumps have a variable frequency drive or a soft starter?
A VFD pays back where the flow varies: a secondary loop with two-way valves at the consumers. With the differential pressure sensor at the most remote branch, it cuts pump energy by about 30% compared with an unregulated pump. If the load is nearly constant and the consumers have three-way valves, a control panel with a soft starter is enough. Run the payback on your own load profile. More in the article "How to cut industrial cooling costs"
Primary-secondary or variable primary flow: which scheme should I choose?
Both work; the difference is in the controls. In primary-secondary, each chiller has its own constant-flow pump and a secondary VFD station serves the network. A decoupler sized for 110–115% of the largest chiller's flow, with no check valve, takes the flow difference. Variable primary flow needs a minimum-flow bypass for the largest machine and slow-acting valves: evaporator flow must never drop below the datasheet minimum, typically 40–50% of nominal.
Does having many pump stations in the scheme make the system less reliable?
No, as long as each station has its own standby and can run on its own. A separate station per loop is normal: open-loop and closed-loop equipment must never be mixed. Reliability is lost at the controls interface. A station must hold its duty if the link to the control room drops, the standby must start before the chiller's flow switch trips, and safety interlocks are hard-wired, not sent over the bus. More in the article on turnkey chiller plants
What is a glycol make-up station for?
It keeps the glycol loop under pressure with ready-mixed solution at the same concentration, not with water. When pressure drops, a pressure switch starts the pump and solution flows from the tank into the loop. Topping up with mains water erodes protection unnoticed: a 60 m³ loop on 40% propylene glycol falls to 34% after three top-ups of 3 m³ of water. Its freezing point rises from −21 °C to about −16 °C. NCT-SGL stations come with 250 to 1,000 litre tanks.
How should glycol be monitored, and can it be kept in a galvanised tank?
Measure concentration with a refractometer scaled for your type of glycol: at start-up, every autumn and after each top-up. Check pH (target 8–10) at least once a year and iron once a quarter. Without monitoring, the solution oxidises within 3–5 years and starts attacking metal. Galvanised steel and glycol do not mix: zinc reacts with the inhibitors, a precipitate forms and protection is depleted. Glycol tanks should be polyethylene or stainless steel.
Why choose a factory-built pump station over assembling the pumps on site?
Defects are found at the factory, not at commissioning. An NCT pump station arrives as one skid: pumps, piping, valves and control panel assembled, pressure-tested at 0.675 MPa for at least 30 minutes, with the skid's internal losses already included in pump selection. On site there are three connections: suction, discharge and power to the panel. Site assembly makes sense when a retrofit keeps working pumps, or when the skid will not fit through existing openings.
Can a pump station be installed outdoors?
Yes, in an insulated container. The NCT pump skid goes into a Sever ("North") enclosure with heating, ventilation, lighting and fire suppression, designed to run at −60 °C. Instead of a pump room, the design provides a pad, pipe entries and power for the container's own systems. A glycol make-up station stays indoors: it is built for +5 to +40 °C. Plain water in outdoor pipes is protected only by heat tracing or draining.
Will a sectional tank freeze outdoors?
Not with factory insulation and an immersion heater. An 8 × 5 × 3.6 m tank at −28 °C stays at +4 °C with a heater of about 1.7 kW; without insulation it would need 25.6 kW. The heater protects an idle tank, not the loop, and runs only through a thermostat and a level switch: without water it burns out. Outdoor pipes get their own heat tracing. The best heating for a tank is a warm room that the tower drains into by gravity.
What data do you need to select a pump station?
Maximum flow in m³/h and the available head, calculated over the most demanding circuit. Then the fluid and glycol concentration, temperature, open or closed loop, redundancy scheme, control panel type, installation location, and limits on footprint and noise. Guessing the head is the main cause of failures at start-up: the pump cannot deliver the flow and the chiller trips on its flow switch. More in the design guides for engineers
Should the elevation of the pipework be included in the pump head?
Not in a closed loop: the water column in the supply is balanced by the column in the return. The head is the sum of pipe friction, fittings, equipment, valves and strainers at their cleaning threshold. For a typical 1,000 kW secondary loop that is about 16 m. In an open cooling tower loop the lift does count: from the minimum level in the tank or basin to the water distribution system, plus inlet pressure at the distribution, 15–20 kPa in standard NCT schemes.
Why does the design use a DN200 main when DN150 passes on velocity?
Pipe size is chosen by friction loss per metre, not by whether the flow fits. The working range is 100–300 Pa/m. At 172 m³/h (1,000 kW, 7/12 °C), DN150 gives 2.7 m/s and 533 Pa/m, while DN200 gives 1.42 m/s and 101 Pa/m. Over a 240 m main, the difference is about 13 m of extra pump head, paid for in every hour of operation. With glycol the losses are higher still.
Can I put the cooling tower pumps on the roof next to the tower to keep the pipes short?
Better not. NPSH may pass at the design point but fail in the worst case: warm water, low level, a clogged strainer, one pump running after its neighbour fails. A pump above the water level also will not refill itself after draining. Open-loop pumps go below the minimum water level, with flooded suction. On the suction side, only pipe, a valve and a coarse screen of 4.8–6.4 mm; the fine filter goes on the discharge.
Can pumps of different sizes run in parallel?
No. On a common header the stronger pump overpowers the weaker one: its check valve closes and it spins with no flow, heating up. Parallel pumps must be the same size, each with its own check valve. Check the motor at the far right of the curve: in our design guide example, when one pump fails the remaining one runs at 151% of its design flow, while the system gets only 75% of its flow.
Why does a VFD on a pump save so little?
Most often because of where the differential pressure sensor sits. Mounted on the station headers, it makes the pump hold full head at any load, and the drive saves three times less. Put the sensor on the most demanding branch. In our example at 172 m³/h, that means 21,600 kWh a year, against 27,100 with the sensor at the pumps and 30,700 with no drive. In an open loop, static lift limits savings: the tower needs the same lift at any flow.
Where should the expansion tank go in a closed loop?
At the suction of the circulation pumps, where pressure does not change when the pumps start, so loop pressure can only rise. A tank on the discharge side, with 25–30 m of pump head, pulls the top of the loop below atmospheric pressure and the loop draws in air. Seal the valve between tank and loop: a closed valve turns the tank into decoration. Put the pre-charge, fill pressure and relief valve setting on the drawing; factory pre-charge almost never matches the calculated value.
What test pressure should be used for a pump station on site?
Under Russian codes (SP 73.13330), the system is tested at 1.5 times working pressure, but no less than 0.2 MPa at the lowest point, held for 5 minutes with a drop of no more than 0.02 MPa. The NCT pump skid has already been pressure-tested at the factory at 0.675 MPa for at least 30 minutes. Chillers and equipment rated for lower pressure are isolated during the test, and expansion tanks are disconnected. State this in the general notes of the design.