HVAC / PLC
Equipment
- Temperatures and pressures
- Heat transfer fluid and air flow rates
- Pump and fan drive frequencies
- Valve positions
- Operating setpoints and statuses
SM SYSTEMS · HVAC INTELLIGENCE
How AI controls HVAC
Data → system model → forecast → optimisation → command → building response.
HVAC Intelligence analyses the actual state of the building, forecasts demand and calculates the optimal equipment operating mode.
Decisions are sent to the BMS or controllers, and the system’s actual response is fed back into the model.
The system does more than observe what is happening. It calculates what needs to happen next.
Based in Ukraine. We deliver physical HVAC projects across Ukraine, while remote HVAC Intelligence services are available for buildings across Europe.
Closed-loop control
From measurement to a physical change in operation01HVACEquipmentThe building and its engineering systems generate the actual load.02DATAMeasurementsTemperatures, flow rates, power and statuses are timestamped and checked for quality.03MODELSystem relationshipsThe model links equipment, circuits, zones and constraints.04FORECASTFuture stateThe building’s response to weather, occupancy and possible commands is calculated.05OPTIMIZATIONOperating mode selectionPermissible operating modes are compared in terms of energy, comfort and load.06CONTROLBMS / PLC commandAgreed setpoints are sent to the controls, which execute the commands.07HVACBuilding responseNew measurements confirm the result or indicate deviations.
RESULT → DATAThe actual result feeds into the next calculation.
01. The system sees the actual state of the HVAC system
A single sensor explains nothing on its own. HVAC Intelligence works with the entire system simultaneously.
METERING
Energy
- Energy consumption
- Heating and cooling output
- COP / EER
- Operating history
BUILDING / WEATHER
Building and environment
- Outdoor air temperature
- Weather forecast
- CO₂
- Building occupancy
Unified system state
The available data points depend on the equipment and metering in place. COP / EER are calculated when sufficient power measurements are available, not from temperature alone.
02. Data becomes a digital system model
Temperature or power alone is not enough. The model describes what is installed, how circuits are connected and how the building responds to changes in operation.
01
Heating and cooling sources
02
Transfer and storage
03
Distribution within the building
04
Controlled system
Shared constraints: capacity, minimum flow rates, permissible temperatures and time between starts.
AI does not work with a collection of sensors. It works with a model of the engineering system.
03. The system predicts what happens next
Predictive HVAC control accounts for future demand: operation can be adjusted before the temperature moves outside the agreed limits.
Thermal Load
Heating demand
How much heat the building needs.
Cooling Load
Cooling demand
How much cooling the zones will need.
Indoor Temperature
Indoor temperature
How indoor conditions will change over time.
Energy Demand
Electrical load
How much power the equipment will require.
Weather Impact
Weather impact
Outdoor temperature and solar gains.
Building Inertia
Thermal inertia
How long stored heat or cooling will last.
Cooling demandMeasurementsForecast
MeasurementsNow+40 min+80 min
An illustration of the principle, not actual building telemetry.
BMS / PLCCollects data and executes local algorithms, schedules and safety functions.
HVAC IntelligenceCompares the predicted response of the entire system to different commands and recalculates the operating mode.
Predictive control complements the BMS rather than replacing its functions. The capabilities of a particular BMS depend on the software installed.
04. AI compares possible operating modes and selects the optimal one
Consider two identical chillers with separate pumps. The decision depends not on a single load percentage, but on future demand and the total energy consumption of the circuit.
CHILLER OPTIMIZATION · EXAMPLE CONTROL LOGIC
- The outdoor temperature is falling.
- Lower cooling demand is expected in 40 minutes.
- In this example, the efficiency map shows a better EER for one chiller at 80% load.
- The operation of both pumps is also included in the calculation.
Current load
Cooling demand 110%Chiller 1
Running
Pump 1Running
Chiller 2
Running
Pump 2Running
2 chillers + 2 pumps. Current demand equals 110% of one chiller’s rated capacity.
In 40 min · forecast
Cooling demand 80%Chiller 1
Running
Pump 1Running
Chiller 2
Running
Pump 2Running
2 chillers at 40% each + 2 pumps. A comparison scenario at lower cooling demand.
In 40 min · forecast
Cooling demand 80%Chiller 1
Running
Pump 1Running
Chiller 2
Standby
Pump 2Standby
1 chiller at 80% + 1 pump. One unit meets the same forecast demand, provided the operating constraints allow it.
- Switch off the second chiller once the load reduction is confirmed.
- Set the first chiller to 80% and adjust the pump group accordingly.
- Match the flow temperature to zone requirements.
- Maintain indoor conditions and verify the actual result.
The change affects the operating mode of the whole system, not just one unit.
55% + 55% and 80% represent different times and different levels of demand. Energy use is compared at the same load: here, 40% + 40% versus 80% + standby. This is not a savings calculation; equipment performance maps, metering and checks on operating constraints are required.
The same logic. Different engineering circuits.
Examples of permissible strategies. The specific scope of control depends on the site’s connections and configuration.
HEAT PUMP OPTIMIZATION
Prepare the system for an overnight drop in temperature
Weather forecast + buffer tank state + thermal inertia.
- Raise the buffer tank temperature in advance within permissible limits.
- Use the building’s thermal inertia and adjust the heating curve.
- Reduce peak demand for backup electric heating.
- Cold weather forecast
- Heat storage
- Night-time load
Storing heat in advance makes sense if the benefit outweighs the additional heat losses and heating costs.
AIR HANDLING UNIT
Adjust the air supply to actual occupancy
Zone occupancy + CO₂ + minimum air flow rates.
- Adjust the fan output.
- Match supply air heating or cooling to the reduced air flow rate.
- Monitor CO₂ and maintain the minimum required ventilation rate.
- Fewer people
- Lower air flow
- CO₂ control
Reducing air flow must not compromise hygiene requirements, air balance or process extract ventilation.
WHOLE BUILDING
Match HVAC operation to available electrical capacity
HVAC + solar generation + battery + power limit.
- Forecast simultaneous demand.
- Shift some heating and cooling operation to periods when energy is available.
- Coordinate thermal storage and the battery through the available controls.
- Load forecast
- Energy coordination
- Peak limiting
Battery coordination is only possible with the appropriate integration, permissions and agreed limits.
05. AI adjusts actual HVAC parameters
Through agreed BMS / PLC points. The available commands depend on protocols, access rights and existing controls.
Temperatures
Temperatures
Equipment Load
Equipment loading
Flow
Flow rates
Schedule
Schedule
Energy
Energy
Operating Strategy
Operating strategy
AI does more than display a recommendation to an engineer. It can automatically adjust permitted setpoints and operating modes.
06. Every decision is checked against the system’s response
After an operating mode change, the system observes how the building actually responds. If the response differs from the prediction, the model is refined.
- 01Decision
- 02Command
- 03Physical outcome
- 04New data
- 05Model refinement
- 06Next decision
PlanPredicted temperature, power and response time.
ActualMeasured parameters and confirmed equipment status.
DeviationData checks, model refinement and recalculation.
The system gradually develops a better understanding of the specific site.
What the building owner gains
Energy consumption
Less unnecessary equipment operation.
Peak power demand
Fewer simultaneous peak loads.
Efficiency
Operation closer to efficient operating points.
Stability
Fewer abrupt changeovers and conflicts between systems.
Fewer manual adjustments. A transparent decision history. Accumulated data for further optimisation.
We do not optimise a single unit. We optimise the whole system.
Chiller + pumpA better operating point for the chiller may increase the pump group’s energy consumption.
Setpoint + comfortChanging the flow temperature affects indoor conditions and the load on other circuits.
Shutdown + backupStopping a unit may transfer its load to other equipment.
Overall outcome: energy + comfort + load + equipment service life.
Energy optimisation is assessed using measurements and comparable conditions, not just a change in setpoint.
REAL SYSTEM
How it works in practice
Yaroslav Smolko demonstrates HVAC Intelligence on a real engineering circuit: data, decision logic and changes to system operating modes.
The video could not be played. Open the file using the link below.
How the system is connected
- 01
Data
We collect data from the BMS and equipment.
- 02
Model
We create a model of the control loop.
- 03
Control
We connect the optimisation module to the BMS / PLC.
- 04
Optimization
We activate the agreed control strategy and refine the model based on actual operation.
AI controls. Existing controls protect.
- The equipment’s factory protection functions remain in place.
- AI operates only within defined engineering limits.
- If the AI layer becomes unavailable, the controls revert to a pre-agreed standard operating mode.
The fallback procedure is tested during integration setup.
Could your HVAC system be more efficient?
Show us your system, equipment and controls. We will determine how to build an intelligent control loop for it.
HVAC Intelligence turns building data into decisions, and decisions into real changes in equipment operation.
Technical context: model predictive control (MPC): IEA EBC / Berkeley Lab ↗