A chiller plant is a system, not a collection of machines. When each chiller, pump and cooling tower follows its own setpoint, the plant can run with three chillers at low load while one could carry the load at high efficiency. The pumps may run at full speed with a small temperature difference, and the cooling towers may fight the chillers for the least energy.
Group control exists to make the plant behave as one system: select the number of chillers from load, select the combination from efficiency, and trim pump and tower output from temperature difference. The objective is not “automatic start/stop”; it is minimum plant kW for the required cooling.
1. Four problems group control must solve
Load estimation. The plant needs a reliable estimate of current cooling load. This can come from return temperature, supply-return difference and flow, or from a heat meter. If load is wrong, every later decision is wrong.
Staging. How many chillers should run, when should the next start, and when should one stop? Too frequent staging damages compressors; too slow staging lets supply temperature swing.
Hydraulic coordination. Chilled-water pumps, condenser-water pumps, cooling towers and chillers must change together. A chiller that unloads while its pump remains at full speed wastes power and may reduce the temperature difference below design.
Efficiency optimization. Different chillers have different part-load curves. Group control should select the combination that meets the load at the lowest total kW, not simply the next chiller in a fixed order.
| Problem | Input | Action | Objective |
|---|---|---|---|
| Load estimation | Return temperature, ΔT, flow | Calculate real-time cooling | Accurate staging |
| Staging | Load, chiller capacity | Start/stop chillers | Stable, not frequent |
| Hydraulics | ΔT, differential pressure | Trim pumps and towers | Match chiller output |
| Optimization | Part-load curves | Choose best combination | Lowest plant kW |
2. Staging logic
Add a chiller when the measured load exceeds the capacity of the running set, or when return temperature stays above setpoint for a defined period. Use a delay to avoid reacting to a short-lived load spike.
Remove a chiller when load falls to a level that the remaining chillers can carry, and when the remaining units will not be pushed above a high load factor. Check the individual load factor, not only the plant total, so one unit is not left at 95% while another is at 30%.
Rotate by accumulated running hours. Start the unit with the fewest hours and rotate on a weekly or monthly basis. This equalizes wear and makes maintenance planning easier.
Limit starts. Compressor starting current and mechanical shock are significant. Limit starts to about 4–6 per hour per chiller and enforce minimum run and minimum stop times. Stability is more important than a perfectly flat supply temperature.
| Condition | Evidence | Action | Constraint |
|---|---|---|---|
| Load rising | Return temperature above setpoint | Add one chiller | Confirm with delay |
| Load falling | Remaining units can carry load | Remove one chiller | Check individual load factor |
| Uneven hours | Runtime accumulation | Start the least-used unit | Rotate at stable load |
| Frequent starts | Start count per hour | Lock out | Minimum run/stop |
3. Where the savings come from
Use a five-chiller plant with 500 kW each at a 1,200 kW load. Three chillers at 400 kW each run at 80% load, which is usually a good efficiency point. Two chillers would each need 120% and could not hold the setpoint; four would run at 60% each and likely use more total kW. Group control selects three, not “the next machine in sequence”.
Chilled-water pump turndown. If the supply-return difference is below design, the pump is moving more water than the load requires. Reducing speed until ΔT returns to design cuts pump power. Pump power varies roughly with the cube of speed: at 80% speed, power is about 51%.
Condenser-water optimization. Lower condenser-water temperature reduces condenser pressure and improves chiller efficiency, but the cooling-tower fans use more energy. Group control finds the combination that minimizes chiller plus tower power, not the lowest possible condenser temperature.
| Source | Mechanism | Typical saving |
|---|---|---|
| Staging | Chillers operate near their efficient band | 5–10% |
| Chilled-water pump turndown | ΔT control reduces speed | Pump power down 40%+ |
| Condenser optimization | Balance chiller and tower power | 3–8% |
| Combined | Plant-wide coordination | 10–20% |
The ranges are typical, not guarantees. The result depends on the original control practice, load profile and sensor accuracy. A plant already operated with manual load-based staging has less upside than one that runs a fixed number of machines.
4. From group control to a smart plant
Load prediction. Use historical data, weather forecast and occupancy schedule to predict load for the next few hours. Predictive control can start or stop a chiller before the temperature drifts, reducing staging frequency.
Efficiency optimization. Build a part-load model for each chiller and calculate the total kW of each feasible combination. Select the lowest-kW combination that meets the load and reserve requirement. This is more precise than a fixed sequence but depends on accurate models and sensors.
Monitoring and diagnostics. Record plant kW, chiller kW, pump kW and tower kW against a baseline. A gradual rise in kW at the same load indicates fouling, sensor drift or a control sequence that is no longer being followed.
Interface to BAS. The group controller should accept a plant-level demand limit or setpoint from the BAS and report status and energy back. The BAS coordinates the whole building; the group controller optimizes the central plant. The boundary should be explicit.
5. Common errors
Controlling chillers but not pumps. Pumps can account for a significant share of plant energy. Optimizing chillers while leaving pumps at full speed gives away much of the benefit.
Trusting uncalibrated sensors. Group control depends on temperature, flow and pressure. A drifted sensor can cause the wrong number of chillers to run for months. Calibrate critical sensors quarterly and consider redundant sensors for the plant differential pressure.
Staging too aggressively. Frequent starts damage compressors and increase maintenance. Set minimum run/stop times and a staging delay before tuning efficiency.
Ignoring shoulder seasons. Low-load, low-wet-bulb conditions offer the largest optimization window, yet many plants keep the summer sequence all year. Review the sequence for spring and autumn.
| Check | Criterion | Frequency |
|---|---|---|
| Sensor accuracy | Compared with a calibrated reference | Quarterly |
| Starts per hour | ≤4–6 per chiller | Continuous |
| Chilled-water ΔT | Close to design | Continuous |
| Runtime balance | Units within a reasonable band | Monthly |
| Plant kW | Compared with baseline | Monthly |
Stage the implementation. A group-control project does not have to be completed in one step. Start with monitoring and data collection, then add staging and hydraulic coordination, then move to load prediction and optimization. This reduces risk and produces savings earlier.
6. Three mistakes in group-control projects
Staging on return temperature alone. Return temperature responds slowly and is affected by flow. A plant that stages only on return temperature will add a chiller too late on a fast load rise and remove one too early during a short lull. Use supply temperature, return temperature and flow together, and require the condition to hold for the add delay.
Minimum run time longer than the load cycle. A 30-minute minimum run is safe for the compressor but can be longer than the interval between load peaks. The plant then runs more chillers than needed because it cannot remove one. Set the minimum run from the compressor manufacturer’s requirement, not from a generic rule, and check it against the plant’s load profile.
Commissioning only at design load. A group-control sequence that works at 1,200 kW may fail at 400 kW, where one chiller is cycling and the pumps are at minimum speed. Test at 40%, 60% and 80% of design load, and record the staging points, pump speed and tower fan status at each step. The low-load test is the one that reveals unstable staging.
| Mistake | Symptom | Correction |
|---|---|---|
| Return-only staging | Late add, early remove | Use supply, return and flow; hold condition for the delay |
| Long minimum run | More chillers running than needed | Set from compressor requirement; check against load profile |
| Design-load-only test | Unstable at low load | Test at 40%, 60% and 80% load |
Common questions
What is the difference between group control and a chiller’s own controller?
A chiller controller manages one machine. Group control manages the whole plant: it decides how many chillers run, which combination is most efficient, and how pumps and towers should follow. That system-level decision is outside a single chiller controller.
How much energy can group control save?
Typically 10–20% plant energy when moving from manual operation to coordinated control. If the plant already has basic sequencing, the additional saving is smaller. Pump turndown is often the largest single contributor.
Why do chillers start too often?
Staging delay is too short, minimum run/stop times are missing, or the load estimate is noisy. Add a confirmation delay and enforce minimum times; do not chase temperature precision at the cost of compressor life.
Which sensors does group control need?
Chilled-water supply and return temperature, flow, differential pressure, condenser-water temperature and individual equipment power. Sensor drift causes wrong staging, so calibration and redundancy matter.
How should shoulder-season operation change?
Lower condenser-water temperature, reduce the number of running chillers and allow pumps to run at lower speed. The shoulder season often offers the largest savings because loads are low and cooling towers can produce colder water.
Does group control need to interface with BAS?
Usually yes. The group controller optimizes the central plant; the BAS coordinates the building. The interface should pass a demand limit or plant setpoint down and report status and energy up.
How many chillers should run?
Run the fewest that meet the load with a safe margin, and rotate lead/lag. Avoid running one chiller at very low load and another at part load. The staging sequence should use measured capacity, not a fixed schedule.
How do I handle pump staging?
Use a differential-pressure setpoint and a pump curve, not a fixed number of pumps. Variable-speed pumps should follow the critical branch. Stage pumps with hysteresis and a minimum run time.
Related reading
- Chiller Part-Load Staging
- Chiller Plant and BAS Integration
- Chilled-Water Pumps and Cooling Towers
- Chiller PLC Control Integration
LONGCOM can review plant staging, pump control, tower optimization and BAS interfaces for a chiller plant. For a system review, contact 400-878-0280 or 13002001280.
Related reading
Chiller plant sensors · Chiller water treatment · Chiller retrofit payback