Chiller-plant BAS integration fails in a predictable pattern: the point list does not match at commissioning, then the communication works but the staging logic does not act, then the plant runs but the lead-lag staging is half a step late and capacity swings.
The root cause is usually an unclear split of duties. Chiller-plant control has two layers. The device layer handles interlocks, protection and start-stop, with millisecond-to-second response. The plant layer handles staging, set-point reset and schedules, with minute-level response. If both are written into one layer, one of them will be inadequate.
This guide follows five sections: division of duties, point list, protocol, strategies and acceptance. It applies to conventional plants based on centrifugal and screw chillers.
Divide the duties: PLC for devices, BAS for the plant
If this line is not drawn at the start, every later issue becomes a patch.
| Layer | Hardware | Functions | Response |
|---|---|---|---|
| Device | Chiller controller or plant PLC | Start-stop, freeze protection, flow interlock, fault stop, pump and valve sequence | Milliseconds to seconds |
| Plant | BAS or standalone plant controller | Staging, supply-temperature reset, rotation, schedule, alarm summary, energy logging | Minutes |
The device layer must close the loop locally. A chilled-water flow failure must stop the compressor within 5 seconds; it cannot wait for a BAS polling cycle. Flow switches, freeze switches and overload protection must be hard-wired to the PLC or chiller controller, not carried only over communication.
The plant layer sends intent, not action. BAS issues “start one more chiller” or “reset supply to 8 °C”; the PLC decides which valve opens first and how long the delay is. If communication fails, the plant can still run safely at its last state.
The common mistake is to reverse the layers to save a PLC: BAS directly commands chiller start-stop and valves. Communication delay of 2–3 seconds means the compressor starts before the valve is open, the flow switch trips repeatedly, and two flow switches burn out in a year.
Point list: count before design
A plant with four chillers and 4,000 kW total typically has the following points.
| Equipment | Points | Typical content |
|---|---|---|
| Chiller, each | 8 AI + 4 DI + 6 DO | Supply/return temperature, evaporator and condenser pressure, current, status, fault, remote start-stop |
| Chilled-water pump, each | 2 AI + 3 DI + 2 DO | Frequency, current, run, fault, auto/manual, start-stop |
| Condenser pump, each | 2 AI + 3 DI + 2 DO | Same as above |
| Cooling tower, each | 1 AI + 3 DI + 2 DO | Fan frequency, run, fault, level, start-stop |
| Header sensors | 8 AI | Supply/return temperature, flow, differential pressure, outdoor temperature and humidity |
| Motorized valves | 2 DO each | Open/close command; position feedback adds 2 DI |
Total is about 150–180 points. This number must be fixed at design stage because it drives three things: gateway count (a Modbus gateway typically handles 32 slaves and a 16-bit register map; 150 points is more stable on two gateways), polling cycle (more points means a longer scan), and cost (about CNY 200–400 per point including cable, module and programming).
| Plant size | Chillers | Points | Gateways | Polling cycle |
|---|---|---|---|---|
| Below 1,000 kW | 1–2 | 40–70 | 1 | 1 s |
| 1,000–3,000 kW | 2–4 | 80–150 | 1–2 | 2 s |
| 3,000–8,000 kW | 4–6 | 150–260 | 2–3 | 3 s |
| Above 8,000 kW | 6+ | 260+ | 3+ | 5 s |
Check the polling arithmetic. Modbus RTU at 9,600 bps, one register read is about 15 bytes round trip; 150 points is 2,250 bytes, plus frame gaps and checks, about 2.5 seconds per scan. A 1-second refresh is not achievable on RTU; use Modbus TCP or BACnet IP.
Protocol: four combinations
| Protocol | Physical layer | Rate | Use | Note |
|---|---|---|---|---|
| BACnet MS/TP | RS-485 | 9.6–76.8 kbps | Controller and device layer | Token passing; latency grows with nodes |
| BACnet IP | Ethernet | 10/100 Mbps | BAS backbone | Requires IT VLAN coordination |
| Modbus RTU | RS-485 | 9.6–115.2 kbps | Chillers and meters | No standard object model; point list defines meaning |
| Modbus TCP | Ethernet | 10/100 Mbps | Chillers and cross-system | Port 502; firewall rule required |
Most domestic chillers provide Modbus RTU or TCP; imported brands often support BACnet. If the BAS backbone is BACnet, a gateway is needed to map Modbus registers to BACnet objects.
Gateway selection has three checks: point capacity with 30% spare; write support, because a read-only gateway cannot run staging; and exception-code pass-through, otherwise a communication fault appears only as “offline” and no one can tell a broken wire from a non-responding slave.
Four plant strategies
Staging. Two criteria are common: cooling load (needs a heat meter; accurate but lagging) and return temperature or header temperature difference (simple but flow-dependent). Practice uses “temperature difference plus delay”: if the return temperature is 1 K above set point for 15 minutes, add a chiller; if it is 1 K below for 20 minutes, remove one. Too short a delay causes frequent cycling; too long lags the load.
Rotation. Start the unit with the lowest cumulative run hours, so no machine is always running and the rest idle. Rotate weekly or monthly, and switch during low load.
Supply-temperature reset. Reset supply temperature from outdoor temperature or the worst terminal differential pressure. At low load, raise supply from 7 °C to 9 °C, which improves chiller efficiency by 6–10%. This is the largest saving, but the terminals must be able to meet the load at the higher temperature.
Sequence. Start cooling tower, condenser pump, chilled-water pump, then chiller, with 30–60 seconds between steps; stop in reverse. The sequence lives in the PLC; BAS sends one “start plant” command.
| Strategy | Criterion | Typical setting | Saving |
|---|---|---|---|
| Staging | Return-temperature deviation | ±1 K, 15–20 min delay | 5–8% |
| Rotation | Cumulative run hours | Weekly or monthly | Life extension |
| Supply reset | Outdoor temperature or terminal pressure | 7–9 °C band | 6–10% |
| Sequence | Fixed timing | 30–60 s per step | Fewer faults |
Alarm levels and debounce
If every alarm is one red dot, operators learn to ignore it. Use three levels. Level 1 (emergency): chiller fault stop, chilled-water flow failure, mechanical room leak, electrical overload. Send SMS or a call immediately. Level 2 (important): high condensing pressure, low oil pressure differential, sustained water-temperature excursion. Push to the app, respond within four hours. Level 3 (advisory): filter differential pressure, service interval, momentary communication loss. Mark on the screen only.
Debounce is mandatory. A 3-second communication loss that reports “offline” dozens of times a day leads operators to disable alarms. Report only after three consecutive failed polls, and clear one minute after recovery.
Acceptance: point-to-point is not optional
“We can see the data” is not acceptance. Do three things. Point-to-point: force each point and confirm the physical device responds; 150 points takes two people a day and finds about 80% of point-list errors. Fault simulation: disconnect a flow switch and check that BAS raises a level-1 alarm within 10 seconds; disconnect communication and check that the plant switches to local operation. Trend logging: record 72 hours and check interval stability, missing points and time-stamp consistency. Time synchronization failure invalidates later energy analysis, so unify clocks during commissioning.
| Item | Method | Pass criterion |
|---|---|---|
| Point-to-point | Force each point | Action matches point list, no mismatch |
| Communication loss | Disconnect cable | Alarm within 10 s; local operation continues |
| Staging logic | Load simulation | Chiller count and set point follow design |
| Trend data | 72 hours | No missing points; interval error under 5% |
| Time sync | Clock check | Devices within 1 s |
Boundaries
The point counts, polling cycles and strategy parameters apply to conventional central plants. A project must follow the chiller communication protocol, BAS brand and plant size. A small plant with one or two chillers may run without a PLC if hard-wired interlocks provide the fallback. A plant with more than three chillers should use a PLC for sequence and failover, because rewriting BAS logic for every change is expensive and hard to verify.
Common questions
Can BAS start and stop chillers without a PLC?
For a small plant with two chillers, yes, provided hard-wired interlock protection is retained. Above three chillers, use a PLC; otherwise sequence and failover logic sits in BAS and every change needs reprogramming and revalidation.
Modbus or BACnet?
Follow the BAS backbone. If BAS is BACnet and the chillers are Modbus, add a gateway; if both are Modbus, connect directly. Neither protocol is inherently better; the gateway mapping and documentation decide maintainability.
Who produces the point list?
The plant integrator, reviewed by the BAS contractor. The common problem is that the chiller manufacturer’s list differs from site, for example a register offset or reversed byte order. Verify against the actual controller before commissioning.
Will the plant stop if communication fails?
Not if device-level interlocks are local. Define the failure strategy: hold last state, run on schedule, or switch to manual. The PLC should keep the plant safe without BAS.
Why is staging always late?
Three causes: long polling, long delay setting, or a temperature sensor in a dead-water zone. Check sensor location first, then the delay, then the protocol.
Why does BAS cooling capacity differ from the heat meter?
BAS often calculates flow × temperature difference; a 0.2 K temperature error gives about 5% deviation. Use a heat meter when accuracy matters, not a calculated value.
What points are needed?
Status, command, alarm, setpoint, valve position, flow and temperature. Count the spares. A BAS interface without point list is a promise.
How do I test the interface?
Point by point, with the chiller and the BAS. Simulate a failure and confirm the response. Record the result.
Related reading
- Commercial Complex Chiller Plant: Diversity and Staging
- Why Chillers Are More Efficient at Part Load
- Water Chiller Commissioning
- Chiller Electrical Wiring and Soft Starters
Related reading
Chiller PLC integration · Chiller plant sensors · Chiller plant group control