Heat pumps · Chillers · Sludge drying systems
Industrial cooling · Process

Chiller Plant Control: Four Sensors That Decide Savings

Chiller-plant optimization is often described as a control strategy: reset the chilled-water temperature, fit variable-speed drives, stage the machines. These measures work only when the measurements that drive them are correct. A reset loop fed by a sensor in the wrong header, or a differential-pressure sensor at the pump discharge, can save nothing and may increase energy use.

This article looks at four measurements that decide most of the saving: chilled-water supply temperature, space humidity, condenser approach and pump differential pressure. It then uses an 800 kW plant to show the order of magnitude.

Control starts with measurement

A control loop can only be as good as its sensor. The four measurements below are not the only points in a plant, but they are the ones that set the operating envelope. If one is wrong, the logic that depends on it will be wrong in a way that is hard to see on a dashboard.

Chilled-water reset: humidity sets the floor

Raising the chilled-water supply temperature improves chiller efficiency by roughly 3–5% per K, because the evaporating temperature rises and the compressor pressure ratio falls. The limit is the terminal’s ability to dehumidify. A conventional fan coil designed for 7 °C supply needs a coil surface below the room dew point, around 12–14 °C, to condense water. Raise the supply to 10 °C and the coil may be above the dew point; the room becomes humid and uncomfortable.

The reset must therefore be constrained by the wettest zone, not by outdoor temperature alone. A workable sequence reads the highest relative humidity among the critical zones. If humidity exceeds the set point, lower the supply temperature; if it stays below the set point for a defined period, raise the supply in steps of no more than 0.5 K. The dead band should be wide enough to prevent hunting.

Chilled-water supply Indicative chiller COP When it is allowed
7 °C 5.0 Humid weather, dehumidification priority
9 °C 5.3–5.5 Humidity has margin
11 °C 5.6–5.8 Dry climate, shoulder season

The sensor itself should sit in a straight run of the supply header, at least five pipe diameters downstream of a bend or valve, with the probe fully immersed. A strap-on sensor on the outside of an insulated pipe reads the insulation, not the water. Accuracy should be ±0.2 K or better, and the sensor should be calibrated against a reference at commissioning.

Condenser approach: the tower is half the bill

The condenser-water temperature sets the condensing temperature and therefore the compressor power. Colder condenser water saves chiller energy, but making it colder requires more tower airflow and fan power. There is a total-power optimum, not a minimum temperature.

Approach is the difference between the cooling-tower leaving water and the outdoor wet-bulb. A typical design is 3–5 K. A smaller approach needs more fill, more airflow or both. The control should set the condenser-water target from the wet-bulb, with a minimum approach, and stage the tower fans in groups rather than running all fans at low speed.

The chiller’s sensitivity is about 2–3% per K of condenser-water temperature. Tower fan power rises approximately with the cube of airflow. Once the fans are near full speed, further cooling may cost more than the chiller saves. A practical band is wet-bulb plus 4 K; below that, measure the total plant power rather than assume a saving.

There is also a lower limit. Too cold a condenser supply can reduce the pressure difference across the compressor, impair lubrication and cause surge, especially in a centrifugal machine. Many units require 15–18 °C minimum condenser-water inlet. In winter, bypass or stop some cells to hold that limit.

Staging: compare total kW, not loads

Most plants run in part load, and the goal is to keep the running machines near their efficient band, typically 60–80% of rated capacity, rather than spreading the load evenly across all machines. Equal loading looks fair but often runs every machine at low efficiency.

A better sequence compares the total power of the current combination with the total power of the next combination. That requires a power meter on each chiller and a part-load curve in the controller. The controller should also enforce a minimum run time, typically 20–30 minutes, and a minimum stop time to avoid cycling.

Plant load Machines running Load per machine Reason
Below 200 kW 1 Below 100% Avoid multiple machines at low load
200–400 kW 2 50–100% Near the efficient band
400–600 kW 3 67–100% Keep margin for control
Above 600 kW 4 Above 75% Full-load operation

For an 800 kW plant with four 200 kW machines, the table is illustrative. The actual thresholds should come from the measured part-load curves. Machines of different ages or types should be prioritized by efficiency, with the least efficient unit as the trim machine. Uniform wear is a secondary objective; it should not override the energy order without a life-cycle calculation.

Pump differential: put the sensor at the worst branch

Variable-speed pump control typically uses differential pressure at the most unfavorable branch, not at the pump discharge. A sensor at the pump sees the pump head and cannot detect what the terminals need; it will over-pressurize the system at low load.

Two control strategies are common. Constant differential pressure holds a fixed head at the worst branch; it is fast and simple but over-supplies at low load. Proportional differential pressure reduces the set point as flow falls; it saves more but needs a reliable flow or valve-position signal. Differential temperature control holds the supply-return difference; it reflects the actual heat transfer but is sensitive to sensor accuracy.

Whatever the strategy, protect the evaporator with a minimum-flow limit, typically 50–60% of design. At lower flow, heat transfer deteriorates and the evaporator can freeze. A differential-pressure switch or a flow meter should trip the chiller before the limit is reached. On multiple pumps, fit non-return valves so a slow pump cannot be driven backward.

An 800 kW example

Take an 800 kW plant running 2,400 hours a year at an average load of 55%. The baseline is 7 °C constant supply, condenser water at wet-bulb plus 6 K, four machines equally loaded, and fixed-speed pumps.

Chilled-water reset: for about 800 shoulder-season hours, supply can rise to 9 °C, and chiller COP rises from 5.0 to 5.4. The saving is about 800 × 0.55 × 800 × (1/5.0 − 1/5.4) ≈ 5,200 kWh. The number is modest because the reset window is limited.

Condenser approach: moving from wet-bulb plus 6 K to plus 4 K raises chiller COP by about 5% but adds about 3,000 kWh of fan energy. The chiller saving is about 800 × 0.55 × 2,400 × (1/5.0 − 1/5.25) ≈ 10,100 kWh; net saving about 7,100 kWh.

Staging: moving from equal loading to efficient-band loading raises machine efficiency by about 4%, saving about 800 × 0.55 × 2,400 × 0.04 ÷ 5.0 ≈ 8,400 kWh.

Pump variable speed: pumps are about 20% of plant energy; a 30% average speed reduction saves roughly 10,000 kWh or more, depending on the pump curve and the minimum-flow limit.

Measure Annual saving, indicative Main constraint
Chilled-water reset About 5,200 kWh Terminal humidity
Condenser approach About 7,100 kWh net Tower fan energy and chiller minimum
Staging About 8,400 kWh Cycling and minimum run time
Pump variable speed About 10,000 kWh Minimum evaporator flow

The four measures together are about 8–12% of plant energy in a well-controlled design. A supplier claiming 30% should be asked to show the baseline; a 30% claim often compares against a deliberately poor fixed-speed plant, or counts only the chiller and ignores pumps and towers.

Common questions

Is a higher chilled-water temperature always better?

No. It is limited by the terminal’s dehumidification. If humidity rises, the room becomes uncomfortable. Reset from the wettest zone, in steps of 0.5 K, and keep a dead band.

Is colder condenser water always better?

No. Below a certain point, the tower fan energy exceeds the chiller saving, and the chiller may reach its minimum condenser-water limit. Control to total plant power, not to the coldest possible water.

Why not run all chillers at equal load?

Equal loading can put every machine in a low-efficiency band. Stage by total power and by the measured part-load curve, with a minimum run time to prevent cycling.

Where should the differential-pressure sensor go?

At the most unfavorable branch, not at the pump discharge. A pump-discharge sensor cannot see terminal demand and will over-pressurize at low load.

What is the minimum evaporator flow?

Typically 50–60% of design, depending on the manufacturer. Below that, heat transfer deteriorates and freezing risk rises. Use a flow switch or differential-pressure switch as a hard limit.

How should savings be verified?

Use the whole-plant energy per unit of cooling, measured over the same load and weather, not a single chiller’s COP. A quarterly or annual baseline is more reliable than a two-week test.

Where should the chilled-water sensors go?

Supply and return sensors should be in wells, downstream of the pump and away from elbows. The differential-pressure sensor should span the critical branch, not the pump header. Stratification and dead legs give false readings.

How often should sensors be calibrated?

At commissioning, then annually for critical points. Compare with a portable reference. A drifting sensor can waste more energy than a chiller upgrade, and it is invisible in the trend.

Further technical reading

For staging, see Chiller Plant Group Control. For PLC points, see Chiller PLC Control. For BAS integration, see Chiller Plant to BAS. For part-load staging, see Why Chillers Are More Efficient at Part Load.

Related reading

Chiller plant group control · Chiller water treatment · Air-cooled vs water-cooled