BMS in HVAC: Controls, Points and Sequences

Ask ten facilities teams what their HVAC control system is and nine will name a brand. Ask them for the sequence of operation for their main air handling unit and most cannot produce a document. That gap is expensive, because the brand decides almost nothing about how the building behaves and the sequence decides almost everything. A mid-range controller running a well-written, well-tuned sequence will beat a premium controller running whatever the commissioning engineer left in it on the last Friday of the project. This guide is about the part that matters: the loops, the sequences and the discipline of keeping them honest.

The message up front: the deliverable that determines how your building performs for the next twenty years is a written sequence of operation, verified against the installed plant and handed over in a form an operator can read. Hardware gets replaced, graphics get redesigned, the sequence is the institutional memory of how the building is supposed to work. If it is not written down, it does not exist, and the building will quietly drift back to whatever burns the most energy while keeping the complaints away.

1. What an HVAC control system actually is

Strip away the marketing and an HVAC control system is four things working in a loop. Sensors measure a condition. A controller compares that measurement against a target. Logic decides what to change. Actuators, valves, dampers, variable speed drives and stage contactors, change it. Then the sensor measures again. That loop runs continuously, per unit of plant, and the quality of a building's comfort and energy performance is almost entirely determined by how well those loops are written and tuned.

The BMS sits above the loops. It hosts, supervises, schedules, records and presents them. It does not usually do the second-by-second control itself: that lives in field controllers close to the plant, which is why a network outage should not stop an air handling unit from holding supply air temperature. For the wider architecture and the supervisory layer, start with the complete guide to building management systems and the companion piece on building automation systems explained.

One boundary worth setting early. This article is about control logic, not the input and output list that logic runs on. Point naming, point counts, sensor and actuator selection and field device specification are a discipline of their own, covered in the sibling piece on BMS controls, points lists and field devices. I reference points where a sequence depends on one, and otherwise leave that ground to it.

2. Control fundamentals: setpoint, deadband, and the shape of a loop

Before any sequence makes sense, four ideas have to be solid. These four explain most of what operators find mysterious about their own buildings.

• Setpoint: the target the loop is trying to hold. Supply air temperature 13 degrees. Static pressure 250 pascals. A setpoint is a decision, not a fact, and every setpoint in a building should be traceable to a reason somebody can state out loud.
• Deadband: the band around the setpoint inside which the controller deliberately does nothing. A zone with a 22 degree cooling setpoint and a 20 degree heating setpoint has a two degree deadband where neither operates. Deadband is not sloppiness, it is the mechanism that stops heating and cooling fighting each other, and narrowing it to look responsive is one of the most expensive tuning mistakes made in occupied buildings.
• Hysteresis and differential: for staged equipment, the gap between the value that starts a stage and the value that stops it. Without it a compressor or a boiler short cycles itself to death. Differentials and minimum on and off timers are protection, not tuning nuisance.
• Direct versus reverse acting: whether the output rises or falls as the measured value rises. Getting this backwards produces a loop that drives hard to the wrong end and stays there. A trivial error that survives commissioning more often than it should.

3. PID loops, and why badly tuned loops hunt

Most modulating control in HVAC is proportional, integral and derivative control, usually just proportional and integral in practice. Badly tuned loops are behind a large share of comfort complaints and mechanical wear, so it is worth understanding in plain language.

• Proportional responds to how far the measurement is from setpoint right now. Larger error, larger output. Proportional alone always settles slightly off setpoint, because the output needed to hold the target requires residual error to produce it.
• Integral accumulates error over time and pushes the output until that offset is gone. Integral is what actually gets you to setpoint. It is also what destabilises a loop when it is too aggressive, because accumulated error keeps driving after the measurement has arrived.
• Derivative responds to rate of change, damping overshoot. In HVAC it is often left out, because most thermal processes are slow and the sensor noise derivative amplifies is real.

Hunting, where a valve or damper cycles endlessly between extremes and the measurement oscillates without settling, has a short list of causes and I would check them in this order. Integral gain too high for a slow thermal process. Proportional band too narrow, so a small error produces full-travel output. An oversized valve or damper, where the first ten percent of travel delivers most of the capacity and the loop has no usable resolution. A sensor in a poor location reading a value the loop cannot influence, for example a supply air sensor too close to the coil in stratified air. And loop interaction, where two loops are both chasing the same physical outcome.

4. Reset schedules: the cheapest energy measure in the building

A reset schedule moves a setpoint automatically in response to another variable, instead of holding it at a fixed worst-case value all year. It is the highest return control change available in most buildings, because the setpoints installed at commissioning are almost always sized for a design day that occurs a handful of hours a year.

The resets that consistently earn their place:

• Supply air temperature reset: raise the air handling unit supply temperature when cooling demand across the zones is low. Less cooling energy, less reheat, higher chilled water return temperature. The constraint is humidity: in a climate such as the Gulf this reset must be bounded by a dewpoint or space humidity limit, or you trade chiller kilowatt hours for mould.
• Duct static pressure reset: lower the fan static setpoint until the most open VAV damper is near, but not at, fully open. Fan power falls roughly with the cube of flow, so this is the largest single fan saving available and it needs no hardware, only damper position feedback.
• Chilled water temperature reset: raise chilled water supply temperature at part load to improve chiller efficiency. The trade is that coils need more flow and pump energy rises, so evaluate it at plant level, not chiller level.
• Differential pressure reset: the same logic as static pressure reset, applied to variable speed pumps using valve position feedback.
• Heating water weather compensation: lower flow temperature as outside air temperature rises, along a curve. Standard practice in temperate climates, frequently flattened in the field because someone had a cold call and the quickest fix was to raise the whole curve.

Resets are where the analytics layer starts to pay for itself, because a reset that is silently failing looks identical to one that is working unless somebody checks the relationship between the two variables. That case is made in the piece on BMS energy optimisation, FDD and analytics.

5. Scheduling, optimum start and stop

Scheduling is the crudest and most effective control in the building: plant that is off consumes nothing. Every BMS supports occupancy schedules, holiday calendars, exception days and after hours overrides, and in almost every building I have reviewed at least one major schedule was wrong.

Optimum start is the refinement. Instead of starting plant a fixed time before occupancy, the controller learns how long the building actually takes to reach setpoint given current space and outside conditions, and starts only as early as needed. Optimum stop is the mirror: shut plant down before the end of occupancy and coast on thermal mass, on the basis that nobody notices a half degree drift in the final twenty minutes.

Two honest notes. Optimum start algorithms learn, so they need a few weeks of consistent operation to be useful, and they are thrown off by manual overrides and by occupancy changes nobody told the BMS about. Optimum stop is the feature most often disabled first, because it generates complaints if tuned even slightly too aggressively. Set it conservatively, prove it, then tighten.

6. AHU supply air control: the sequence written out

The air handling unit is where most of a building's control complexity lives, so it is worth writing a full sequence rather than describing one. Take a variable air volume unit with heating and cooling coils, a mixing section with return, outside and relief dampers, and variable speed supply and return fans. The table below is the shape a sequence of operation should take: states, conditions, actions, and the interlocks that override everything.

Four details in that table matter more than they look. The cooling side is one sequenced output, not separate economiser and chilled water loops, because two independent loops chasing the same supply air temperature will fight. Heating requires the cooling output proven at zero for a delay before it is enabled, because that interlock is the main defence against simultaneous heating and cooling. Freeze protection is hard-wired, not software, because a controller that is offline or in a strange state must still protect the coil. And minimum outside air is written as a floor every other mode respects, because ventilation is a code obligation, not an energy variable to be optimised away. ASHRAE publishes the ventilation and building energy standards most of this is derived from, and it is worth having the current versions on hand: ASHRAE . For the equivalent ventilation and building services guidance used widely outside North America, CIBSE is the usual reference.

7. VAV box control and reheat

The air handling unit conditions air to a common supply temperature. The VAV box decides how much each zone gets, and in many designs whether to reheat it. The box sequence is short, but it is where simultaneous heating and cooling is most often created.

A cooling only box is simple. The zone temperature loop produces a cooling demand, the demand maps to an airflow setpoint between the minimum and maximum flow limits, and the damper modulates to hold that measured flow. Note the two nested loops: temperature sets flow, flow positions the damper. Boxes that drive damper position directly from temperature, with no flow measurement, are cheaper and considerably harder to balance.

A box with reheat adds a heating stage below the cooling range, and the ordering of that range is the whole argument. The sequence I would specify, coldest zone condition to warmest:

• Heating: airflow held at the heating minimum, reheat coil modulates to hold zone heating setpoint. Flow is not increased until the reheat coil is at full output, and then only to a bounded heating maximum.
• Deadband: airflow at the ventilation minimum, reheat off, cooling damper at minimum. Nothing happens, deliberately, and this state should be where a well-designed zone spends most of its time.
• Cooling: reheat proven off, airflow modulates from the ventilation minimum to the cooling maximum.

The interlock that must be written explicitly: reheat is disabled whenever airflow exceeds the heating maximum, and cooling airflow is not increased while the reheat valve is open. Without that line, a zone with an optimistic minimum flow sits there reheating air the central plant just paid to cool, twelve hours a day, for years, and nothing on the graphics page looks wrong.

8. Chilled water plant sequencing and staging

Central plant control is a staging problem: how many chillers, pumps and towers to run, at what setpoints, and when to add or drop a stage. Written badly it produces short cycling, low delta T and chillers running where they are least efficient. The elements I would expect to see written down:

• Plant enable: the conditions that start the plant at all. An occupancy schedule combined with a cooling demand from the air side and an outside air lockout, rather than a chiller running because it is nine in the morning.
• Lead and lag selection: which chiller starts first, with runtime equalisation or a fixed efficiency-based order, plus a manual override for maintenance. Rotating on equal runtime and rotating to favour the most efficient machine are different strategies, both defensible, but they must be chosen rather than inherited.
• Stage up: add a chiller when the running machines are near full load and supply temperature has been above setpoint by a margin for a sustained period. Staging on proven inability to hold setpoint is more robust than staging on load percentage alone.
• Stage down: drop a chiller when the remaining machines can carry the load with margin, after a minimum stage time. Minimum on and off times, and a stage-down deadband well clear of the stage-up point, are what prevent oscillation between two and three machines.
• Pump and isolation valve interlocks: prove flow before the chiller loads, sequence the isolation valve open before the pump starts, and keep the pump running through the chiller's shutdown period.
• Reset strategies: chilled water supply temperature reset bounded by the worst-case coil and dehumidification requirement, and pump differential pressure reset until the index coil valve is near fully open.

The measure I would put on any chilled water plant dashboard is delta T, alongside plant kilowatts per ton delivered. Low delta T, where return water comes back barely warmer than it left, means the plant is moving far more water than the cooling delivered requires, and it is almost always a control problem at the coils rather than a plant problem: valves not closing, three-way bypasses left in place, or supply air setpoints so low that valves sit wide open. For the mechanical side of keeping that plant healthy, see boiler and chiller preventive maintenance.

9. Condenser water and cooling tower control

The condenser side holds the most unclaimed efficiency, because it is less visible than the chilled water side and its setpoints are frequently fixed at a value chosen during commissioning and never revisited.

The sequence has three moving parts. Tower fan speed modulates to hold condenser water supply temperature at setpoint. That setpoint should itself be reset against outside wet bulb temperature, typically wet bulb plus an approach, bounded below by the chiller manufacturer's minimum entering condenser water temperature. And cell staging decides how many cells are wetted: more cells at lower fan speed is usually more efficient than fewer cells at high speed, because fan power falls steeply with speed while heat rejection surface is free once the cell is open.

The trade-off that has to be written in: lowering condenser water temperature improves chiller efficiency but costs tower fan energy, and below a certain point the fan energy exceeds the chiller saving. That crossover is plant specific. A sequence that says only "maintain condenser water at 29 degrees" has given up on the optimisation; one that chases the lowest possible temperature overspends on fans. The correct form is a reset curve with a stated lower bound and a note explaining where the bound came from.

10. Heating plant and weather compensation

Heating plant sequencing follows the same staging logic as chilled water, with its own protections. Boiler enable on a heating demand and an outside air lockout. Lead and lag with runtime rotation. Stage up on inability to hold flow temperature, stage down on load with minimum run timers. Return water temperature protection on non-condensing boilers, and on condensing boilers the opposite concern: keep return temperature low enough to actually condense, which means resisting the temptation to raise flow temperature to solve a distribution problem.

Weather compensation is the heating equivalent of supply air reset: flow temperature falls along a curve as outside air temperature rises. Two failure modes are near universal. The curve gets raised globally because of a handful of cold complaints that were really a balancing or emitter problem in one area, which spends energy across the whole building to fix one room. And the curve is never rechecked after an envelope or occupancy change, so a building that was reinsulated is still heated to the original curve.

In hot climates the heating discussion shrinks to reheat, domestic hot water and dehumidification, and the logic that matters most is the interlock keeping reheat from running against the cooling plant. In the Gulf that deserves more attention than a weather curve ever will.

11. Demand controlled ventilation with CO2

Ventilating a space for full design occupancy when it is a third full wastes energy conditioning air nobody is breathing. Demand controlled ventilation measures a proxy for occupancy, usually carbon dioxide concentration, and modulates outside air between a base minimum and the design maximum.

A workable sequence: measure zone or return air CO2, and reset the outside air flow setpoint from a base ventilation rate at low CO2 up to the design rate as CO2 approaches its limit. The base rate is not zero, because it covers the building-related ventilation component that exists regardless of occupancy. Where several zones share an air handling unit the outside air must satisfy the worst-case zone, which is why single-zone applications such as lecture theatres, meeting rooms and auditoria deliver far more benefit than open plan offices on a shared unit.

12. What a good sequence of operation document contains

The sequence of operation has the longest useful life of anything produced on a controls project, and it is routinely the worst executed. A usable sequence document, per system, contains:

• System description and schematic, with sensors and control devices shown in position.
• A setpoint schedule: every setpoint, its handover value, its adjustable range, and who may change it. Not "supply air temperature as required".
• Modes of operation: off, warm-up, cool-down, occupied, unoccupied, setback and any special modes, with entry and exit conditions written as testable statements.
• The control narrative per mode: what modulates, in what order, against which measurement, written so a competent engineer who has never seen the building could reproduce the logic.
• Safeties and interlocks: freeze protection, static limits, fire and smoke interface, flow proving, and which of these are hard-wired rather than software.
• Alarm schedule: every alarm, its trigger, priority, delay and the expected operator response. An alarm with no defined response will be acknowledged and ignored.
• Reset strategies: each written as a curve or table with stated bounds and the reason for them.
• Trend schedule: which points are logged, at what interval, for how long. This is what makes the building diagnosable later, and it is nearly always omitted.
• A change log: dated entries for every sequence or setpoint change made after handover, by whom and why.

The sequence belongs in the specification before tender, not reverse engineered from the installed controller afterwards. The clause structure that makes that enforceable is in BMS specification clauses, and the verification side, including how graphics should reflect the sequence rather than decorate it, in BMS commissioning, graphics and operator dashboards.

13. Common control faults and what to check

Most HVAC complaints that arrive as "the BMS is not working" are one of a small set of control faults. This table is the triage list I would give an operator before they call the controls contractor.

Notice how many of these are diagnosed from a trend rather than a site visit, and how many resolve to a setpoint or an interlock rather than a hardware failure. That is the normal distribution of BMS faults. Mechanical maintenance remains essential and largely separate: the filter, belt, coil and fan tasks that keep plant capable of responding to the sequence at all are covered in preventive maintenance for HVAC systems, and I would not try to tune a loop on a unit with a blocked coil.

14. The honest section: where the waste actually is

Everything above describes how HVAC control is supposed to work. What I find in operating buildings is consistently less impressive, and the pattern repeats across sectors and climates. Four causes account for most avoidable HVAC energy waste, and none of them need new equipment to fix.

• Simultaneous heating and cooling. Reheat running against central cooling, perimeter heating fighting an overcooled zone, a chilled water valve passing while the heating coil modulates. Invisible on a graphics page because both loops report satisfied, trivially visible on a trend of the two outputs together. This is the largest single waste in most buildings with reheat, and it is an interlock problem, not a plant problem.
• Sequences left in manual override. Someone forced a valve open, put a fan into hand or fixed a setpoint to stop a complaint at two in the morning, and never reverted it. I have reviewed buildings with dozens of points in override, some years old, each silently disabling part of a sequence that was commissioned correctly. Override is a maintenance tool that has quietly become a permanent operating mode.
• Sensors that drifted years ago. A supply air sensor reading two degrees low makes the plant overcool continuously while every screen shows setpoint achieved. A humidity sensor past its calibration life makes economiser and dehumidification decisions on fiction. None of this alarms, because the reading is plausible. Calibration is the least glamorous and highest yield activity in controls.
• Schedules never updated after occupancy changed. A floor changed use, a shift pattern changed, a tenant left, a building went hybrid, and the schedules stayed as they were at handover. Plant runs for occupancy that no longer exists, and because the building is comfortable nobody raises it.

The uncomfortable conclusion is that most buildings do not need a control system upgrade. They need someone to spend two weeks with the trends, the override list, the calibration records and the original setpoint schedule, and put the building back to the way it was specified. That work is cheaper than new hardware and consistently delivers more. It is also hard to sell, because there is no product attached to it, which is precisely why it does not get done.

Continuous monitoring earns its keep by catching these four automatically rather than waiting for a review. Override lists, simultaneous heating and cooling detection, sensor plausibility checks and schedule versus actual runtime comparison are standard fault detection rules, and they are why analytics and energy management systems pay back when pointed at control faults rather than used as reporting tools. The same applies to the newer IoT and real-time monitoring layer: extra sensing is only worth the money if it closes a loop somebody acts on.

The idea to walk away with

An HVAC control system is a written sequence of operation executed by tuned loops on calibrated sensors, and every part of that sentence is load bearing. The sequence is design intent made explicit. The tuning makes it stable enough to leave alone. The calibration makes it true. Remove any one and the building will still appear to work, will keep occupants reasonably comfortable, and will quietly spend far more energy than it needs to for years, with nothing on any screen indicating a problem.

The corollary is that the highest return work in building controls is almost never a technology purchase. It is writing the sequence down properly, verifying it against the plant, clearing the overrides, calibrating the sensors and reviewing schedules against how the building is used now. That is a fortnight of competent engineering, repeatable annually, and it outperforms most capital projects in the same building.

Final thoughts

If you take one action from this guide, ask for the sequence of operation for your largest air handling unit and read it. If it does not exist, that is your project. If it exists but does not state entry and exit conditions, setpoint ranges, interlocks and alarm responses, it is a brochure rather than a sequence. And if it exists and is good, compare it against the live setpoints and the override list, because the gap between the document and the running building is where your energy has been going.

Controls work rewards patience over product. The loops are simple, the sequences are knowable, and the faults are diagnosable from data you already collect. What is scarce is the discipline to write the logic down, test it honestly and keep it maintained as the building changes. Teams that build that discipline get a building that performs close to its design for its whole life. Teams that do not get a building that drifts, one override and one stale sensor at a time, until somebody proposes replacing capable equipment to solve a problem that was never mechanical.

Related reading: Building management systems: a complete guide, BMS controls, points lists and field devices, BMS commissioning, graphics and operator dashboards, BMS energy optimisation, FDD and analytics, Preventive maintenance for HVAC systems, BMS specification clauses.