Equipment Staging Sequencing

The pump control lesson closed on a loose end: most real systems have more than one pump. A primary plus a standby at the least, often two or three pumps piped in parallel that the building automation system brings on and takes off as demand rises and falls. This page is about that decision — the staging sequence. Pumps are the worked example here, but the logic is the same one a BMS uses to stage boilers, chillers, and cooling-tower cells: when you have several identical units, how does it decide how many to run, and which ones?

Why a Plant Runs Several Identical Units

Before the staging question makes sense, name why the plant has more than one pump at all. A single pump sized for the building's peak load is oversized for nearly every hour that isn't the peak. A variable-speed pump rides that out gracefully for a while — slow down, follow the building, that's the whole pump-control story — but only down to a point. Every VFD has a minimum speed, and below the flow that minimum produces you're back to the problems variable-speed pumping was meant to solve: pressure dumped across throttled valves, and at the extreme the deadhead the pump-control page ended on. One pump can only turn down so far.

Split the duty across several smaller pumps and the turndown problem softens. At light load you run one pump at a sensible point on its curve; as demand climbs you bring on a second, then a third. And there's a second reason that has nothing to do with efficiency: redundancy. A plant built N+1 — one more pump than the peak load needs — keeps running when a pump fails or gets pulled for service. The building doesn't care that a bearing let go at 2 a.m.; the standby picks up the load and a technician deals with it in daylight.

So picture three identical pumps on a common header — a common suction main feeding all three, a common discharge main collecting what they deliver. That's the plant the rest of this page works with.

Now the BMS has two questions to answer, continuously, all day: how many of those three pumps should be running right now, and which ones? Take them in order.

Staging Up and Down — How Many to Run

The sequence needs a signal that stands in for "how hard is the building working." On a variable-speed plant the natural one is already in hand: the speed reference the DP loop is producing (the pump-control page's whole subject). When the running pumps have been pinned near full speed for a while, they're out of room — the building wants more than they can give, and it's time to add a pump. When the speed reference has been sitting low, the running pumps are loafing, and one of them can probably be shut off.

That turns into two thresholds: a stage-up point and a stage-down point. The important detail is that they are not the same number. The stage-down point sits well below the stage-up point, and the gap between them is a deliberate deadband. If a single threshold did both jobs, a demand hovering right at that line would add a pump, which drops the load on each running pump, which pushes demand back below the line, which drops the pump, which pushes demand back up — the sequence would hunt, cycling a pump on and off indefinitely. The deadband means the conditions that bring a pump on are clearly different from the conditions that take one off, so the plant settles.

Two timers finish the picture. A crossed threshold doesn't act immediately — a stage delay requires the demand to stay across the line for a sustained period before the sequence acts, so a brief swing (a single zone slamming open, a momentary pressure dip) doesn't churn the whole plant. And once a stage change does happen, a minimum stage time blocks the next change for a while. A pump that just started shouldn't be asked to stop sixty seconds later; short-cycling is hard on the motor windings, hard on the starter, and hard on the pump itself.

The widget below has the three pumps from the schematic. Drag the demand slider and watch the sequence work — pumps stage on and off at the thresholds, the stage delay counts down before each change, and the minimum stage time holds off changes that come too fast. Try dragging the slider up and down quickly: the sequence refuses to chase you.

The widget is sped up for the demonstration — its 2-second stage delay and 4-second minimum stage time stand in for real-plant values on the order of several minutes each.

Two things are worth pulling out. Walk the demand slider slowly across the point where the second pump comes on, then slowly back down — the pump doesn't leave at the same place it arrived. That space you have to travel back through is the deadband. And jump straight to design day (the peak-load condition the plant was sized for): the pumps don't all snap on at once. Each stage change waits out its delay, and the minimum stage time spaces them — the plant comes up to full capacity deliberately, not all in one lurch.

Lead/Lag and Rotation — Which Ones

With how many answered, the sequence still has to pick which physical pumps. The one that runs first is the lead; the next to come on is the lag (with three pumps, first lag and second lag). On its own that's just an ordering — but it matters more than it looks, because of wear.

If the lead designation is nailed to one physical pump, that pump runs every hour the plant runs. After a year it has logged thousands of hours while its siblings have logged almost none. The lead wears out — bearings, seals, the mechanical parts that only age under running — and the "redundant" standby pumps haven't earned their keep. So a good sequence rotates the lead. The simplest version swaps lead on a fixed schedule, say weekly. The better version rotates to equalize runtime: the BMS tracks accumulated run-hours per pump and hands the lead to whichever pump has the fewest, so the whole set wears down together instead of one at a time.

Rotation also keeps the standby honest. A pump that never runs isn't being preserved — it's deteriorating a different way, with bearings taking a set and seals drying out, so that the day it's finally called on is the day you discover it doesn't work. (Sequences often add a standby exercise for exactly this — running an idle pump briefly on a schedule to keep it limber.) And rotation rides alongside the failure case: when the lead faults — trips, loses its drive, throws a flow alarm — the sequence has to promote a standby immediately and without fuss. Redundancy is only real if the swap is automatic.

None of this is specific to pumps. A central plant stages lead/lag boilers and lead/lag chillers on the identical pattern — the staging signal is plant heating or cooling load instead of pump speed, and the timers are tuned to the equipment, but "how many, and which ones, rotated for even wear" is the same sequence. The widget below runs the plant forward a week at a time so the rotation question stands on its own — it assumes a light-load stretch where just the lead pump runs each week, which isolates which pump logs the hours from how many run at once (the first widget's question). A peak week would stage two or three pumps on together, and runtime equalization would spread those hours the same way, across whichever pumps ran. Pick a strategy, step the weeks, and watch the runtime bars.

Run a dozen weeks on fixed lead and the spread tells the story — one pump carries everything, the other two are cold. Then switch to runtime-equalized and keep stepping: the lead jumps to the lowest-hour pump each week, and the bars pull back together. Same three pumps, same duty — the only difference is a sequence that spreads the wear instead of concentrating it.

Tying It Together

This closes the loop the pump-control page opened. Load piping, VFDs, and pump control built up one pump answering the building — throttled loads, a drive that follows a speed reference, a DP loop deciding what that reference should be. Staging is the layer above: when one pump isn't enough, or isn't wanted, a sequence decides how many run and rotates which ones carry the hours. The deadband, the stage delays, the minimum stage time, the runtime equalization — none of it is exotic, but all of it is the difference between a plant that settles and one that hunts itself to an early grave.

Staging is one kind of sequence. The same BMS runs others this page hasn't touched — resetting setpoints against outside-air temperature, shifting a building between occupied and unoccupied modes, easing through a morning warm-up without every piece of equipment slamming on at once. Those are sequencing topics for pages of their own. Closer to home, keeping a pump from running off the end of its curve — end-of-curve protection — belongs to the same family of safeties a real staging sequence carries; the pump-control page met its worst case as the deadhead. Staging is where the plant stops being a collection of equipment and starts being a system.