Chiller plant sequencing and optimization field guide

Plant sequencing, and why the control decides the bill

Chiller plant sequencing is the control logic that decides how many chillers run, which ones, and at what chilled-water and condenser-water setpoints, as the building load moves through the day and the year. A central plant is several chillers, several pumps, and a set of cooling-tower cells, and any of them can carry a given load in more than one combination. The sequence picks the combination. That choice, made hundreds of times a day by the controls, is where most of the plant's energy is spent or saved.

The hardware gets the attention because it is what you can touch. A high-efficiency chiller with a variable-speed drive is a real piece of capital, and its rated kW per ton is on the submittal. But a 0.55 kW per ton machine run in a dumb sequence will post 0.9 or worse at the plant meter, and no amount of chiller efficiency buys that back. The sequence is the cheapest thing in the mechanical room to change and the most expensive thing to get wrong.

This guide is the operating side: the staging, the setpoint reset, and running the plant as one system. Bringing the plant online and proving the chiller is a separate job, covered in the chiller plant startup and commissioning guide. How the water actually moves through the chillers and out to the coils, primary-secondary versus variable primary flow, is covered in the chilled-water pumping guide. Both are referenced here where the sequence touches them, not repeated.

Why sequencing matters more than the equipment

On most commercial buildings the chiller plant is the single largest electrical load, and it runs at part load nearly all of its hours. So the plant's annual energy is decided less by what it does on the design-day afternoon than by what it does at 40 to 60 percent load on a mild Tuesday, and that is pure sequencing territory. The machine that was selected for its full-load number lives somewhere else entirely.

Run the wrong number of chillers and you pay twice. Stage on a second machine too early and you split a load that one chiller could have carried efficiently across two running at poor part load, dragging both up the kW per ton curve. Hold the chilled-water and condenser setpoints at the design minimum all year and you make every chiller work at maximum lift even when the weather would let it coast. Published work on multi-chiller optimization commonly reports 20 to 40 percent plant energy savings against conventional control, and advanced sequences another 10 to 15 percent, which tells you how much conventional control was leaving on the table.

The blunt version: the controls, not the hardware, decide the efficiency. You can spec the best chiller on the market and still run a wasteful plant, and you can take an older plant and cut its bill hard without replacing a single machine, by fixing the sequence. The exact savings depend on the equipment, the load profile, and the climate, so treat those percentages as the range the field reports, not a promise for your plant.

What is kW per ton?

kW per ton is the plant's efficiency: the electrical power it draws divided by the cooling it delivers, in kilowatts per ton of refrigeration, where a ton is 12,000 BTU per hour. Lower is better. A single water-cooled chiller might rate near 0.5 to 0.6 kW per ton at full-load design conditions, with variable-speed machines lower, but the number that pays the bill is the plant total, not the chiller alone.

Plant kW per ton sums every motor it takes to make the cooling: the compressors, the chilled-water pumps, the condenser-water pumps, and the cooling-tower fans, divided by the tons leaving the plant. This matters because the optimizations trade one motor against another. Drop the condenser water colder and the compressor draws less, but the tower fans draw more. Optimize for the chiller alone and you can drive the total up. Always measure and judge the plant total.

Tons come from the chilled-water flow and the delta-T across the plant, tons equals gallons per minute times delta-T divided by 24. So an honest kW per ton needs an accurate flow meter and accurate supply and return temperatures, plus a true plant power reading. Get any of those wrong and the efficiency number is fiction, which is why the sensors get their own section later. The target is whatever the design and the equipment can actually reach at the conditions you are in, not a single number carried in your head.

kW/ton

Total plant power divided by tons of cooling; the plant efficiency metric, lower is better

Ton

12,000 BTU per hour of cooling, the unit of plant and chiller capacity

Plant total

Compressors plus chilled-water and condenser pumps plus tower fans, the kW that actually pays the bill

How do you stage chillers to the load?

You stage to the real load, which means running the fewest machines that can carry it at good efficiency, and adding the next chiller only when the running ones are about to run out of capacity, not before. The mistake that costs the most energy is staging on too early, splitting a load two or three lightly loaded chillers when one or two well-loaded machines would do it for less total kW per ton.

The staging point is the load at which adding or dropping a chiller actually lowers the plant kW per ton. Below that crossover, the running machines plus one more is more efficient than the running machines alone working harder; above it, the extra machine just adds pump and compressor losses. The cleanest sequences compute that crossover from the measured efficiency curves rather than a fixed percentage, because the point shifts with the condenser-water temperature and the load. A staging table that names which chillers run in each stage is project-specific and belongs in the sequence of operation.

Guard the transitions with timers and deadbands so the plant does not short-cycle a chiller on and off around the staging point. Add on a sustained demand the running machines cannot meet, and shed on a sustained drop with margin, so a passing cloud does not start a chiller. The chiller's own startup, flow proving, and protection through these transitions is commissioning work, covered in the startup guide; the sequence decides when to ask for the machine, the chiller controls decide whether it is safe to start.

Lead-lag rotation and runtime

Lead-lag is which machine starts first and which follows. The lead chiller carries the base load and runs the most hours; the lag chillers stage on as the load climbs. On a plant with interchangeable machines, the sequence rotates the lead so the run hours even out across the fleet instead of wearing one chiller to death while its twin sits cold.

Equalizing runtime is the point of the rotation. ASHRAE RP-1711, the central-plant control research being folded into Guideline 36 by addenda, describes identical parallel equipment, chillers, pumps, and tower cells, being lead-lag alternated so the machine with the most accumulated hours becomes the last to stage and the one with the fewest becomes the lead. That spreads the maintenance, keeps a standby machine exercised so it actually runs when called, and avoids the failure where the lag chiller has not turned over in months and will not start on the hot afternoon you finally need it.

Where the machines are not interchangeable, a large and a small chiller, or a variable-speed and a fixed-speed, the order is not about runtime at all. You lead with whichever machine is most efficient for the load you are in, often the variable-speed chiller at low load and the larger machine as the load climbs. The sequence has to know the difference, because rotating dissimilar machines blindly throws away the efficiency the mix was bought to deliver.

The part-load sweet spot

A chiller is not most efficient at full load. It has a part-load sweet spot, commonly somewhere in the 40 to 70 percent range, where the kW per ton bottoms out, and the plant spends most of its hours near there. The EPA and the manufacturer part-load data both put the bulk of operating hours in that band, which is exactly why part-load efficiency, not the full-load rating, decides the annual bill.

This is the argument for running fewer machines loaded rather than many lightly loaded. Two chillers at 60 percent each will usually beat three at 40 percent, because each lightly loaded machine still drags its pumps, its share of the tower, and its own fixed losses while making less cooling. The part-load curve is why: walk a chiller down from full load and its kW per ton improves for a while, then turns and climbs again as the load gets too small to load the compressor well. Staging is the art of keeping every running machine near the bottom of that curve.

The catch is that the bottom of the curve moves. A variable-speed centrifugal at low lift can hold a good kW per ton far lower than a fixed-speed machine, which surges and loses efficiency as it throttles. So the sweet spot is a property of the specific machine at the specific conditions, not a universal 50 percent. Pull the part-load data for the actual chillers and let it set the staging points, rather than assuming the textbook band.

The two setpoints that drive efficiency

Two setpoints decide how hard every chiller works: the chilled-water supply temperature the plant makes, and the condenser-water temperature it rejects heat to. The gap between them is the lift, and the compressor power tracks the lift. Hold both at their design minimum all year and you make every chiller pull maximum lift even when the load is light and the weather is mild, which is the most common way a perfectly good plant runs inefficiently.

The design conditions are a worst-case afternoon, not a typical hour. The chilled water might be drawn for 44 degrees F supply and the condenser loop for a hot, humid design wet-bulb, because that is the corner the plant has to cover. But the plant rarely lives in that corner. Raise the chilled-water setpoint when the load and humidity allow, and let the condenser water run colder when the tower can make it cheaply, and you shrink the lift for most of the year without ever touching the equipment.

Both setpoints have floors and both have side effects, which the next two sections take in turn. Reset is not a license to push a setpoint wherever the math says; it is a controlled move inside limits the equipment and the building set. The point here is only this: the setpoints are levers, not fixed facts, and a plant that treats them as fixed is leaving the easiest savings on the table.

What is chilled-water reset?

Chilled-water reset is raising the chilled-water supply temperature setpoint above the design minimum when the load and the humidity allow it. Every degree the supply water comes up is less lift on the compressor and less chiller energy, because the evaporator can run warmer. On a mild day a plant designed for 44 degrees F supply might give good service at 48 or 50, and the chillers thank you for it.

The watch-out is the coils and dehumidification. Warmer chilled water means a warmer, weaker coil, and a coil that cannot pull enough moisture out of the air will let space humidity climb even while the dry-bulb holds. A common guard is to stop raising the setpoint, or reset it back down, once the outdoor dew point or humidity ratio crosses a threshold, with figures around a 57 degree F dew point or a 0.009 humidity ratio cited in the literature, but the building's actual humidity requirement and the coil selection set the real limit. Coils have to be sized for the reset temperature, or the warmer water cannot make the load.

Reset off the real constraint, not the calendar. The better sequences raise the chilled-water setpoint until the most-demanding coil's valve is nearly wide open, then hold, so the plant runs as warm as the worst coil in the building permits and no warmer. That ties the chiller energy to the actual load and keeps a hidden coil or a humidity-sensitive space from going uncovered. Push the setpoint too far and you trade a little chiller savings for a comfort or mold call, which is a bad trade.

Condenser-water reset and the tower tradeoff

Condenser-water reset lowers the temperature of the water coming back from the cooling tower toward the chiller, which drops the lift and the compressor power. As a rough figure, the field commonly sees roughly 1 to 2 percent change in chiller efficiency for each degree F of condenser-water temperature, so colder condenser water is real money at the compressor. The tower makes colder water by running its fans harder, and that is where the tradeoff lives.

Drop the condenser-water setpoint and the chiller kW falls while the tower fan kW rises. Add those two curves and the total plant power falls to a minimum at some optimum condenser-water temperature, then climbs again as the fans chase water that is approaching the wet-bulb, where each additional degree costs more fan energy than it saves at the compressor. The optimum is not a fixed setpoint. It moves with the load and the wet-bulb, which is why the best strategies reset the condenser water continuously to hold the lowest combined chiller-plus-tower kW rather than a single number.

There is a hard floor underneath all of it. Every chiller has a minimum entering-condenser-water temperature and a minimum lift it needs to keep oil moving and, on a centrifugal, to stay out of surge, and at high load very cold condenser water can stack refrigerant in the condenser and cost capacity. Reset below that floor to chase efficiency and you risk the machine. The manufacturer sets the minimum, and the reset has to respect it. Hedge the optimum and the floor to the specific chiller and tower on the job.

Condenser-water reset

Lowering the condenser-water setpoint to cut lift and chiller kW, traded against rising tower fan kW

Approach (tower)

Leaving condenser water minus the ambient wet-bulb; the tower cannot make water colder than the wet-bulb

Refrigerant stacking

Refrigerant pooling in the condenser at high load and very low condenser-water temperature, costing capacity

The lift, and why cutting it is the target

Lift is the difference between the condensing temperature and the evaporating temperature, or equivalently between the two pressures, and it is the work the compressor has to do. Make colder water at the condenser, warmer water at the evaporator, and you shrink the lift from both ends. Almost every chiller-side optimization is really a way of cutting the lift: chilled-water reset raises the low side, condenser-water reset lowers the high side, and a variable-speed drive lets the machine ride the smaller lift efficiently instead of throttling against it.

The reason lift is the master variable is thermodynamic. A vapor-compression machine spends its energy bridging the pressure gap, so the kW per ton at any load is set largely by how big that gap is. A chiller carrying half its load at low lift on a cool morning can post a far better kW per ton than the same machine at the same load on a hot afternoon, purely because the lift collapsed. The load did not change. The lift did.

So when you tune a plant, watch the lift as the score. Trend the condensing and evaporating temperatures together, and the kW per ton will track the gap between them. The two setpoint resets, the variable-speed drive, and the waterside economizer are all moves to keep the lift as small as the equipment, the load, and the floors allow. The floor matters as much as the target: cut the lift below the chiller's minimum and you trade efficiency for surge or poor lubrication, so cut it to the limit, not past it.

What is the low-delta-T problem?

Low delta-T is when the chilled water comes back to the plant colder than design, so the temperature rise across the load is smaller than it should be. Drawn for a 14 degree F rise and seeing only 7 or 8, the plant has it. The only way the plant can move the design heat on a shrunken delta-T is to pump more water, so the flow climbs, the pumps draw more, and the plant ends up staging chillers to make flow rather than to make cooling. The kW per ton goes to ruin even though the building load never grew.

This is the most common chronic disease on chilled-water plants, and from the sequencing chair it shows up as a plant that runs too many machines for the cooling it is delivering. The instinct is to chase it with the controls at the plant, staging on another chiller because flow is short. That is the wrong end of the problem. The cause is out at the coils and valves: three-way valves blending cold water to the return, control valves with poor authority, fouled coils, and oversized coils at part load, all returning water nearly as cold as it left.

You cannot pump or stage your way out of a coil problem, and a sequence that tries just spends more energy holding the delta-T flat. The diagnosis and the fix, valve authority, coil cleanliness, two-way valves, and the decoupler behavior that signals it, are pumping and balancing work covered in the chilled-water pumping guide. The sequencing job is to recognize the signature, stop the plant from masking it with extra flow, and trend the delta-T so the problem is visible before it is expensive.

Variable primary flow and the sequence

How the plant pumps its water changes what the sequence has to manage. Primary-secondary runs a constant-flow loop through the chillers and a variable-flow loop to the building, joined by a decoupler, and the sequence stages off the decoupler flow direction: deficit flow says stage on, surplus says a chiller can come off. Variable primary flow uses one set of variable-speed pumps through both the chillers and the coils, which saves pump energy but asks more of the controls.

In a variable-primary plant the sequence carries a job the constant-primary plants never had: holding each running chiller above its minimum evaporator flow as the building throttles down. A modulating minimum-flow bypass valve near the chillers opens when demand falls below that floor, recirculating just enough water to keep the machine safe. Staging is harder too, because all the chillers share one variable loop, so adding or dropping a machine redistributes flow through the others, and the sequence has to manage the bypass and the rate of flow change through every transition without tripping a running chiller.

Which configuration is on the job was decided long before the sequence was written, and the choice and its tradeoffs are the subject of the chilled-water pumping guide. What the sequence owes either one is the same: stage to the load, protect the minimum flow, and prove the transitions across the real load range, not just at design. The pumping guide covers the bypass sizing, the decoupler, and the minimum-flow data; the sequence is where that hardware gets driven.

Pump staging and differential-pressure reset

Pump energy is part of the plant kW per ton, so the sequence has to drive the pumps as deliberately as the chillers. Variable-speed chilled-water pumps ride a differential-pressure setpoint: a sensor reads the pressure across supply and return, ideally out at the most demanding coil, and the drive speeds up or slows down to hold it as the coil valves open and close. Because pump power tracks roughly the cube of speed, slowing the pumps at part load is where the affinity-law savings come from, and the plant lives at part load.

Holding a fixed differential-pressure setpoint leaves most of that saving on the floor. Differential-pressure reset lowers the setpoint until the most-open coil valve is nearly wide open, so the pumps make only the head the system actually needs at the moment, then no more. Set the static setpoint too high, or put the sensor at the pump instead of at the far load, and the pumps run harder than the building requires for the life of the plant.

Stage the pumps with the chillers, and on a headered plant bring pumps on and off to keep the running pumps near their own efficient point rather than running one pump flat out or three barely loaded. The detail of where the sensor sits, how the setpoint is found during balancing, and the affinity-law math is pumping and balancing work covered in the chilled-water pumping guide. The sequence's part is to reset the pressure down to the real demand and to stage pumps to the load, not to a fixed schedule.

Cooling-tower fan staging and approach

The cooling tower rejects the plant's heat, and its fans are the other half of the condenser-water tradeoff. The sequence stages and speeds the tower fans to hold the condenser-water setpoint, and how it does that decides whether the savings from colder condenser water survive the fan energy it took to make it. Variable-speed tower fans, staged and modulated together across all the running cells, are far more efficient than cycling a few cells full-on and full-off, because fan power also falls roughly with the cube of speed.

Approach is the tower's limit. The tower cannot make condenser water colder than the ambient wet-bulb, and the gap between the leaving water and the wet-bulb is the approach. As the setpoint chases the wet-bulb, the approach shrinks and the fans have to work exponentially harder for each additional degree, which is exactly where the chiller-versus-tower tradeoff turns. The optimum condenser-water setpoint sits where the falling chiller kW and the rising fan kW sum to the lowest plant total, and it moves with the wet-bulb and the load.

Run all available cells with the fans modulating slowly rather than fewer cells at full speed, because spreading the heat across more tower surface gives a closer approach for less total fan energy and more even water distribution. The exact fan staging, the cell isolation, and the basin and water-treatment behavior depend on the tower and belong with the tower commissioning work. From the sequencing chair, the tower fans and the condenser-water reset are one optimization, not two, and they have to be tuned together against the plant total. Hedge the fan strategy to the specific tower.

Integrating the waterside economizer

A waterside economizer makes chilled water with the cooling tower and a heat exchanger, no compressor, whenever the wet-bulb is low enough that the tower can produce water cold enough to cool the building directly. A plate-and-frame heat exchanger sits between the condenser-water loop and the chilled-water loop, and on a cold day the tower water passing through it cools the building's return water for almost nothing but pump and fan power. This is free cooling, and in a cool climate it can cover many hours of the year.

The economizer earns its keep only if the sequence integrates it instead of treating it as a separate mode. ASHRAE 90.1 and the energy codes generally require integrated operation: the economizer must be able to provide partial cooling even while the chillers are still running, so the heat exchanger takes whatever load the tower can make and the chillers trim the rest, rather than waiting for conditions cold enough to carry the whole building before it turns on. A non-integrated economizer that only runs when it can do 100 percent of the load throws away most of the available free-cooling hours.

Commission the changeover, because that is where economizer plants fail in practice. The sequence has to bring the heat exchanger in and back out cleanly as the wet-bulb crosses the threshold, without dumping the load, short-cycling the chillers, or hunting between modes on a marginal day. The classic enable point is around a wet-bulb in the 40s, with ASHRAE 90.1 calling for economizing at or below roughly 50 degrees F dry-bulb and 45 degrees F wet-bulb, but the design conditions and the chilled-water setpoint set the real changeover. Confirm it against the adopted code edition and the design.

Optimal start, stop, and the schedule

The cheapest cooling is the cooling you do not make. A plant running into an empty building at full setpoint is burning energy for no one, and the schedule is the first thing to check on any plant that runs more hours than the building is occupied. Stop the plant, or coast it back, when the load goes away, and do not pre-cool space that nobody will use.

Optimal start is the controls calculating how early the plant has to come up to hit setpoint at occupancy, based on the indoor and outdoor conditions and how the building responded yesterday, instead of starting at a fixed clock time with a fat safety margin. On a mild morning the plant may need only a short head start; on a hot, humid one it needs longer. Letting the controls learn that, rather than starting at 5 a.m. every day to be safe, recovers real hours of plant runtime over a cooling season.

Pre-cooling can also be used on purpose, pulling the space and the building mass down before a peak or before the utility's expensive hours, then coasting through the peak on the stored coolth. That is a demand-and-rate play more than an efficiency one, and whether it pays depends on the tariff and the building. The plain win that applies to every plant is simpler: match the run schedule to real occupancy, and let optimal start trim the warm-up to what the day actually needs.

The BAS runs the sequence: trust, but verify

The building automation system, the DDC controllers and the supervisory logic, is what actually executes the sequence hour to hour. Every staging decision, every setpoint reset, every pump and fan command runs as code in that system, against the points it can read: the flows, the temperatures, the power, the valve positions, the run statuses. The sequence is only as good as the programming, and the programming is only as good as the points feeding it.

Trust the controls to run it, but verify that what runs matches the sequence of operation the engineer wrote. The gap between the spec and the installed program is wide on many plants. Resets that were specified but never enabled, staging timers left at the controls vendor's defaults, a condenser-water reset that was drawn but commented out, a setpoint hard-coded where a reset was intended. None of that shows on a walk-through. It shows when you read the program against the sequence and watch the points move.

Make the verification a record, not a memory. Capture the as-running sequence, the live setpoints, the staging points, and the reset schedules, and keep them where the next engineer can find them, because the operator who inherits the plant has no other way to know what the controls are actually doing. A field tool like FieldOS is for exactly this: logging the sequence as it runs, the setpoints in force, and the deviations you find, so the plant's control intent is documented instead of living only in the head of whoever last touched the BAS.

Trend the plant to find the bad sequence

A bad sequence hides in plain sight until you trend it. Log the plant kW, the kW per ton, the flows, the supply and return temperatures, the condenser-water temperature, the wet-bulb, and the run status of every machine, and the data shows you the sequence misbehaving in ways no spot reading ever will. The plant running three chillers at 35 percent each, the condenser water pinned at design minimum on a cold night, the delta-T sagging through the afternoon, all of it is obvious in a trend and invisible on a gauge.

Read the trends as a system, against the lift and the load. Overlay the plant kW per ton on the load and the condenser-water temperature and you can see whether the staging and the resets are doing their job or fighting it. A kW per ton that climbs as the load drops is the signature of staging on too early or reset that never moves. A delta-T that falls under load points straight at the low-delta-T problem. The analytics turn a vague sense that the plant runs hard into a specific finding you can fix.

Trending is also how you know a fix held. Make a staging or reset change, then watch the trend for a week to confirm the kW per ton actually fell and nothing else broke. Capturing those trends and the kW per ton over time in a field tool like FieldOS, alongside the changes you made, turns plant tuning into something with evidence behind it instead of an argument about whether last month felt cooler. The plant that is trended is the plant that gets optimized; the one that is not drifts.

Manual overrides and failed sensors that defeat the sequence

The fastest way to kill a good sequence is a hand-off-auto switch left in hand. Operators override the controls to solve a today problem, a warm zone, a nuisance alarm, a chiller that would not start, and the override outlives the problem by years. A pump forced on, a valve commanded open, a chiller locked out, a setpoint pinned at a manual value, each one defeats a piece of the sequence, and the plant runs around the workaround burning energy nobody can account for.

Find them by reading the points for what is in manual versus auto, and by looking for values that never move when the sequence says they should. A condenser-water setpoint that sits at one number through every weather is either a reset that was never enabled or a manual override, and only the controls will tell you which. The classic find is a plant that was optimized at commissioning and quietly walked back, one override at a time, until it runs the way it did before the work was done.

Failed and drifting sensors do the same damage more subtly, because the sequence believes them. A flow meter reading high makes the plant think it is moving more water than it is, so kW per ton looks better than reality and the staging is off. A stuck temperature sensor freezes a reset. A miswired valve feedback hides a coil that is wide open. The sequence cannot be smarter than its inputs, so part of optimizing a plant is hunting the overrides and the bad sensors that are defeating it, and clearing them before you trust a single trend.

Commissioning and retro-commissioning the sequence

Verifying that the sequence does what the spec says is its own task, separate from starting the chillers. The functional test drives the plant through its real operating range and confirms the staging adds and sheds chillers at the right points, the resets move with the load and the weather, the economizer changes over cleanly, and the minimum flow holds through every transition. A sequence that was never tested across a load swing is where a plant first fails to hold the building, or first reveals it has been wasting energy all along.

Test the sequence the hard way, by making it act, not by reading the program and assuming. Force a load change and watch the plant stage. Drive the wet-bulb input down and confirm the condenser-water reset and the economizer respond. Drop the chilled-water demand and confirm the bypass holds minimum flow before any chiller trips. The chiller's own startup, flow proving, and safety testing through these moves is the subject of the chiller plant startup and commissioning guide; here the focus is the control logic riding on top of proven machines.

Retro-commissioning is the same work on a plant that has been running for years, and it is the highest-payback control job most buildings have. The plant has drifted: overrides have crept in, sensors have wandered, resets have been disabled, the sequence has been edited by hands that did not read the original intent. Retro-commissioning measures the plant as found, finds the gap against the design intent, and walks it back to a sequence that stages to the load and resets the setpoints. Tie the findings into the broader commissioning record the way the startup guide describes.

Measuring the savings and the sensors behind them

If you change a sequence, measure what it bought, because plant tuning that cannot show a number is plant tuning nobody will fund again. Measurement and verification on a plant means a before-and-after on the kW per ton, normalized for load and weather so you are comparing like with like, not a cool June against a hot July. Trend the plant total kW and the tons through a range of conditions before the change, make the change, and trend the same range after. The savings live in the comparison, not in a single afternoon reading.

Every number in that comparison rests on the sensors, and the sequence is only as good as the data it runs on. An accurate chilled-water flow meter, calibrated supply and return temperature sensors, a true plant power measurement, and a sound wet-bulb reading are what make the kW per ton real and the resets correct. A flow meter off by 10 percent puts the kW per ton off by 10 percent and corrupts every staging decision built on it. A drifted temperature sensor poisons the delta-T and the reset together.

So calibrate the sensors that the sequence and the measurement depend on, and check them on a schedule, not once at startup. The temperature sensors against a known reference, the flow meter against its installation requirements and a second method where you can, the power metering against a portable meter. Hedge the M&V approach and the acceptable tolerances to the project and the measurement standard in force; the principle that does not change is that an optimized plant is built on calibrated data, and an uncalibrated plant is guessing.

What to document

The record that matters is the one that says what the sequence is supposed to do and at what setpoints, so the next engineer can tell drift from design and the operator can tell a real problem from where the plant started. Capture the staging points, the reset schedules and their limits, the lead-lag scheme, the minimum flows, the design and operating kW per ton, and the sensor calibration dates, and keep them where they can be found. Without that, every optimization gets relearned from scratch, and every override becomes permanent because nobody knows what the plant was meant to do.

Common mistakes

• Staging too many chillers at low part load, splitting the load across lightly loaded machines instead of running fewer, well-loaded ones.
• Holding the chilled-water and condenser-water setpoints at the design minimum all year, so every chiller pulls maximum lift even on mild days.
• Ignoring a low-delta-T problem and staging chillers to make flow instead of fixing the coils and valves that cause it.
• Driving the condenser-water reset below the chiller's minimum lift or entering-water floor, risking surge, poor lubrication, or refrigerant stacking.
• Manual overrides and hand-off-auto switches left in place that defeat the staging, the resets, and the minimum-flow protection.
• No trending, so the bad sequence stays invisible and a fix can never be proven to have held.
• Uncalibrated flow and temperature sensors, so the kW per ton is fiction and every staging and reset decision built on it is wrong.
• Optimizing for the chiller alone and driving up the tower fan and pump energy until the plant total gets worse, not better.

Field checklist

Standards and references

ASHRAE Standard 90.1 is where the plant-controls requirements live: chiller minimum efficiency in full-load kW per ton and in IPLV, chilled-water and condenser-water reset on plants above a size threshold, variable-speed pumping and pump-power limits, and the integrated waterside-economizer requirement with its changeover conditions. The thresholds and the exact requirements move between editions, so confirm them against the edition the project and the jurisdiction have adopted, along with the local energy code, which may be based on 90.1 or the IECC.

The numbers that govern the chiller protection are not in a code at all. The chiller and cooling-tower manufacturer's part-load data, the minimum and maximum evaporator flow, the minimum lift and entering-condenser-water temperature, and the surge map set the floors the optimization has to respect, and that data wins over any rule of thumb. The condenser-water reset optimum, the part-load sweet spot, and the staging points are properties of the specific equipment at the specific conditions, so hedge them to the machines on the job.

ASHRAE RP-1711, the central-plant research being folded into Guideline 36 by addenda (the published Guideline 36 itself scopes airside and waterside terminals), describes high-performance control sequences for chilled-water plants, including the lead-lag and runtime-equalization logic and the project-specific staging table. ASHRAE Guideline 0 and the commissioning specification frame the functional and retro-commissioning process, and the project's sequence of operation is the document that sets the actual staging, setpoints, and reset intent for the plant. Cite the standard that controls the point, let the manufacturer's data govern the chiller and the tower, and let the project sequence govern the logic. Above all, stage to the real load at the efficient part load, reset the chilled-water and condenser setpoints to cut the lift, and fix low delta-T and trend the plant.

Units, terms, and synonyms

Plant sequencing mixes refrigeration, hydronic, and controls vocabulary, and the same idea reads differently across a sequence of operation, a chiller cut sheet, and a controls submittal. The terms below travel across the whole optimization conversation. Efficiency is kW per ton in US practice and COP, the dimensionless ratio of cooling to power, in metric. Flow is gallons per minute, GPM, or liters per second. Temperature reset shows up as CHWST reset and CWST or ECWT reset on different drawings for the same moves.

Plant sequencing

The control logic that decides how many chillers run, which ones, and at what setpoints, as the load moves

kW/ton

Total plant power per ton of cooling, the efficiency metric, lower is better

Staging

Matching the number of running chillers and pumps to the load, adding the next only when needed

Lead-lag

Which machine starts first and which follows, rotated to equalize runtime on like equipment

Lift

The gap between condensing and evaporating temperature or pressure, the work the compressor does

Chilled-water reset (CHWST)

Raising the chilled-water supply setpoint when load and humidity allow, to cut chiller energy

Condenser-water reset (CWST / ECWT)

Lowering the condenser-water setpoint to cut lift, traded against rising tower fan power

Low delta-T

Return water colder than design, shrinking the rise and forcing extra flow and chiller energy

Variable primary flow (VPF)

One variable-speed pump set through the chillers and coils, with a minimum-flow bypass protecting the evaporator

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is a factory authorized chiller startup?
• What is primary-secondary chilled water pumping?
• What is centrifugal chiller surge?
• What are the types of chillers?
• What is the difference between a batch plant and a drum plant?