Cooling plant controls: 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 controls of a chiller plant: cooling tower, chiller and pump panels, sensors and the link to the BMS or SCADA. These are the questions customers, design engineers and plant operators ask us. The figures come from our design guide for engineers EH-APM007 and our resources; the setpoints for your site come from the plant's sequence of operation.
Questions on this page (20)
- Why do we need a plant master panel if every chiller has its own controller?
- Can several cooling towers be controlled from one control cabinet?
- Which protocol should connect the plant to the BMS: Modbus or BACnet?
- How do cooling tower, pump or chiller panel signals reach the control room?
- Can the flow switch be wired to the plant controller and passed to the chiller over Modbus?
- Which level sensors belong in a cooling tower basin, and what if the float sticks?
- Do you hand over the controller program, setpoints and a backup?
- How much energy do plant controls save?
- How does the control system share the load between several chillers?
- Why does the second chiller start and then stop again right away?
- Why does the plant fall short of capacity when the chillers are healthy?
- Why is the cooling tower water setpoint floating when the process wants a fixed one?
- Can a Thermax absorption chiller be controlled from the plant DCS over Modbus?
- Can the panel have two power feeds with automatic transfer?
- Why does a chiller plant need its own flow meter, and will a clamp-on meter do?
- Do plant temperature and pressure sensors need legal metrological verification?
- How much spare I/O should the plant controller have?
- How should the design specify trend logging?
- Which form of internal separation should the control panel have?
- What proves that a control panel meets the requirements?
Why do we need a plant master panel if every chiller has its own controller?
So that chillers, pumps and cooling towers run as one plant. A chiller controller protects its own machine and holds its own setpoint. The master panel handles what one machine cannot see: how many chillers run, which one leads, rotation by run hours, the common header setpoint, pump standby, the tower bypass and fans, and the plant common alarm. It also collects data for the BMS. Each machine's safeties stay in its own panel. More in the article "Turnkey chiller plant contractor"
Can several cooling towers be controlled from one control cabinet?
Yes, but a different layout is more reliable: one panel per tower plus a plant master panel on top. If everything sits on one controller, its failure stops every tower at once, and the cabinet gets bulky. A tower's own panel keeps that tower running even without the master; the master shares the load and collects alarms. Which panel is the master and which signals pass between panels is decided at the design stage. More in our design guides for engineers
Which protocol should connect the plant to the BMS: Modbus or BACnet?
Whichever your BMS or SCADA already uses, so ask before you choose equipment. Industrial sites mostly use Modbus TCP, and our master panel sends data over it by default. BACnet/IP is common in buildings: offices, airports, pharma. A Modbus-to-BACnet gateway costs less than native BACnet but loses change-of-value subscriptions and alarms as protocol events. Plants running PROFINET usually take Modbus TCP into their own PLC or through a gateway. What you agree on is the register map, not the protocol name.
How do cooling tower, pump or chiller panel signals reach the control room?
Data goes over a digital protocol; anything that must stop equipment goes by hardwire. NCT panels send data over Modbus RTU, Modbus TCP or OPC UA, chosen in the questionnaire. The bus carries temperatures, pressures, level, drive speeds, run hours, alarm codes and setpoints. Hardwired: run enable, common alarm and the fire alarm signal. The BMS displays, logs and adjusts setpoints within set limits, but does not control the process: if the link drops, the plant keeps running on its last setpoints.
Can the flow switch be wired to the plant controller and passed to the chiller over Modbus?
No. Flow switches, differential pressure switches and pump auxiliary contacts are hardwired straight into the chiller's safety circuit. Over a wire the stop command takes tens of milliseconds; through a gateway and SCADA it takes seconds, and if the gateway hangs it never arrives. The chiller then keeps running with no water through the evaporator. The bus carries data and setpoints, never safeties. Each chiller gets a flow switch on its own pipe, not on the common header. More in our design guides for engineers
Which level sensors belong in a cooling tower basin, and what if the float sticks?
For control, use a hydrostatic 4–20 mA probe in a stilling tube: it gives a continuous level, and make-up works on two thresholds. Floats and electrodes give only discrete points, so a sticking float is replaced with a probe. Low-low level needs its own device, which stops the pumps and locks out the basin heater. High-high level sits 50–100 mm below the overflow and closes make-up independently of the controller. The overflow itself has no valves and is never blocked.
Do you hand over the controller program, setpoints and a backup?
Yes, and it belongs in the contract. We do not issue a separate software design: the controller program file is handed over electronically, and a duplicate copy is an option. The contract states who receives the file, in what format, and who may change the logic after start-up, because the warranty depends on it. The sequence of operation comes with a setpoint table: value, range and who may change it. The controller keeps the alarm and event log for 6 months; longer history lives on the SCADA server.
How much energy do plant controls save?
In our practice, proper cooling tower controls save 25–35% of electricity, and controls with variable speed drives pay back in 2–3 years. Without drives, tower fans run all-or-nothing and overspend up to 50%. On pumps, savings depend on where the differential pressure sensor sits: a 30 kW pump at 70% speed instead of 85% saves about 48,600 kWh over 6,000 hours. Each degree of higher evaporating temperature saves the chiller about 2–3%. More in "13 mistakes when choosing a cooling tower"
How does the control system share the load between several chillers?
Through staging and rotation. Staging decides how many chillers run: the next one starts on load or return water temperature after a delay, and stops at a lower threshold after a longer delay, for example start at 90% load, stop at 80%. Rotation evens out run hours: once a day the machine with the fewest hours becomes lead, and roles change only on stopped machines. Chillers of different size or age are staged by efficiency, without rotation. A chiller in alarm drops out of the sequence.
Why does the second chiller start and then stop again right away?
Start and stop share one threshold, or there are no delays. Separate them: for two 500 kW chillers, start the second above 450 kW for longer than 8 minutes, stop it below 400 kW for longer than 15 minutes, with a minimum run time of 10 minutes. The start delay must be longer than one full circulation of the loop volume. If the loop holds too little water, tuning will not help: add a buffer tank sized against the chiller's smallest capacity step.
Why does the plant fall short of capacity when the chillers are healthy?
Most often it is low delta-T syndrome: water returns colder than designed. At 2.5 K instead of 5 K, the same 86 m³/h carries 250 kW instead of 500. Pumps hit their limit, the controls start extra chillers, and all of them run part-loaded. Usual causes are three-way valves at consumers, open bypasses, fouled heat exchangers and oversized control valves. Without a plant flow meter and temperature sensors you cannot see it; two or three days of trend data reveal the cause.
Why is the cooling tower water setpoint floating when the process wants a fixed one?
So that in cool hours the tower delivers colder water and unloads the chiller. A floating setpoint is the wet-bulb temperature plus about 3.9 K. It is limited from below by the lowest temperature the consumer accepts: the condenser datasheet for a chiller, the process specification for production. If the process needs a fixed temperature, the setpoint is fixed and the savings come only from fan drives. More in our design guides for engineers
Can a Thermax absorption chiller be controlled from the plant DCS over Modbus?
Reading data, yes; control, no. The Modbus RTU (RS-485) port on a Thermax machine is read-only. Remote start and stop go as a discrete signal to the terminals of the machine panel switched to remote mode. Plan this in the design from the start. Check the model specification for which permissives the machine issues itself (to pumps and the cooling tower) and what the plant panel controls. More in our design guides for engineers
Can the panel have two power feeds with automatic transfer?
Yes; whether you need it depends on the power reliability category, set per plant layer. The controller, operator panel and control circuits run on a 1–3 kVA UPS with 15–30 minutes of autonomy, so the plant rides through the transfer and reports a power loss. Transfer on the power section is needed when pumps and compressors are critical loads. Afterwards the plant restarts itself in steps: pumps and fans, then compressors one by one. Two feeds from one transformer are still a single point of failure.
Why does a chiller plant need its own flow meter, and will a clamp-on meter do?
It lets you stage chillers on load, meter cooling and prove design capacity. Cooling takes 10 to 40% of a plant's electricity, and the site meter cannot separate it afterwards. Use an inline meter on the return, with 10 D of straight pipe upstream and 5 D downstream, and the control valve after the meter. A clamp-on ultrasonic meter on poor hydraulics is off by tens of percent. No room for straight runs: rearrange the assembly or record in writing that flow is not measured.
Do plant temperature and pressure sensors need legal metrological verification?
No, unless their readings are used for billing or they are part of the safety instrumented system of a hazardous facility. Billing and safety channels are marked separately in the design and do need verification. Process sensors only need calibration: a Pt100 at the ice point must read 100.00 ohms, and the class B tolerance at 0 °C is ±0.30 K. Design thermowells so a sensor can be pulled without draining the loop. More in our design guides for engineers
How much spare I/O should the plant controller have?
20% of each input and output type, and no fewer than two points per type. Spare capacity below 15% is usually used up during commissioning, and after that the first new signal means a new module and a panel shutdown. The difference in hardware is one or two expansion modules. Reserve the cabinet the same way: 20% of terminals, a free stretch of DIN rail, 20% of cable entries and a 24 V power supply with 30% headroom.
How should the design specify trend logging?
As a list: which points are logged, at what interval and for how long. For key points, ask for at least 12 months of raw values with CSV export: 120 points once a minute come to about 757 MB a year. With such a log, short cycling, low delta-T, heat exchanger fouling and sensor drift are found in minutes without a site visit. Long-term history sits on the SCADA server, with one NTP time source and an alarm log time-stamped to the second.
Which form of internal separation should the control panel have?
Choose it by the cost of downtime. If one pump must be serviced while the others keep running, use at least Form 3b under IEC 61439-2: one pump's contactor can be replaced with the plant in operation. If the plant can be stopped for repairs, Form 2 is enough. A higher form means a larger panel and worse heat removal, so it is chosen together with the panel's heat calculation and cooling method.
What proves that a control panel meets the requirements?
A panel datasheet, a test report to IEC 61439 and, in Russia, a registered EAEU declaration of conformity to technical regulations TR CU 004/2011 and 020/2011. At acceptance, the declaration and report numbers are checked, and the EAC mark is on the nameplate or in the datasheet. The panel itself is tested at the factory before shipment, and on site every point is checked against its instrument. More in the article "Turnkey chiller plant contractor"