1. Introduction
Indoor air quality (IAQ) has become an increasingly important part of modern building automation.
Temperature alone does not determine whether a room is comfortable or healthy. A room can have an acceptable temperature while still experiencing elevated CO₂ levels, insufficient ventilation or poor air-quality conditions.
KNX provides the infrastructure to monitor and control many of the variables involved:
- CO₂
- Temperature
- Humidity
- Presence
- Air-quality sensors
- Ventilation
- Fans
- Dampers
- HVAC operating modes
Artificial Intelligence can add another layer by learning how these variables interact and predicting how indoor conditions are likely to change.
Instead of simply reacting when CO₂ exceeds a fixed threshold, an AI-enabled KNX system can attempt to answer:
How will indoor air quality change over the next few minutes, and what ventilation strategy can maintain good IAQ without unnecessary energy consumption?
This creates a more intelligent approach to ventilation control.
2. What Is Indoor Air Quality?
Indoor Air Quality describes the condition of the air inside a building.
Important parameters can include:
- CO₂ concentration
- Temperature
- Relative humidity
- Volatile organic compounds (VOCs)
- Particulate matter
- Airborne pollutants
- Ventilation rate
The importance of each parameter depends on the building type and application.
For many KNX projects, CO₂ is particularly useful because it can provide an indirect indication of occupancy and ventilation requirements.
However, CO₂ should not be treated as a complete measurement of indoor air quality.
3. Why KNX Is Suitable for IAQ Automation
KNX can connect the different components required for intelligent IAQ control.
A typical installation may include:
CO₂ Sensor
Temperature Sensor
Humidity Sensor
Presence Sensor
↓
KNX
↓
Ventilation
HVAC
Fans
Dampers
↓
Indoor EnvironmentThis provides a closed control loop.
The building measures its environment, applies control logic and adjusts ventilation or HVAC operation.
AI can enhance this process by analyzing historical and real-time data.
Learn more about : KNX + AI for Predictive HVAC Control
4. CO₂ as an Occupancy and Ventilation Indicator
CO₂ concentration often changes as people occupy a room.
For example:
Room Empty
↓
Low CO₂
↓
People Enter
↓
CO₂ Increases
↓
Ventilation IncreasesThis makes CO₂ valuable for demand-controlled ventilation.
However, the relationship is not instantaneous.
CO₂ concentration depends on:
- Number of occupants
- Room volume
- Ventilation rate
- Outdoor CO₂
- Air distribution
- Duration of occupancy
AI can learn these relationships rather than relying exclusively on fixed thresholds.
5. Traditional CO₂ Control vs AI-Based Control
A conventional strategy might use:
If CO₂ > threshold → increase ventilation.
This is simple and reliable.
An AI-based strategy can consider:
- Current CO₂
- Rate of CO₂ increase
- Occupancy
- Room size
- Historical patterns
- Ventilation response
- Meeting schedule
- Outdoor conditions
For example:
Current CO₂: 850 ppm
Rate of increase: High
Occupancy: Increasing
Ventilation: 40%AI may predict that the room will soon exceed the desired range and increase ventilation earlier.
The goal is to avoid waiting until poor air quality has already developed.
6. Predicting CO₂ Levels
One of the most useful applications of AI is CO₂ prediction.
Instead of asking:
What is the CO₂ level now?
the system can ask:
What is the CO₂ level likely to be in 10 or 20 minutes?
A simplified model might consider:
Current CO₂
+
Occupancy
+
Room volume
+
Ventilation rate
+
Historical behaviour
↓
AI Prediction
↓
Future CO₂This enables predictive ventilation.
7. Occupancy and IAQ
Occupancy has a major influence on indoor air quality.
Consider a meeting room.
At 13:45:
Occupancy: 2
At 14:00:
Occupancy: 12
At 14:15:
Occupancy: 12
A ventilation system operating at a fixed level may not respond efficiently to these changes.
KNX presence and occupancy information can provide additional context.
AI can combine this information with CO₂ behaviour to estimate future ventilation demand.
8. Combining Multiple Sensors
A single sensor rarely provides the complete picture.
A more intelligent IAQ system can combine:
- CO₂
- Temperature
- Humidity
- Presence
- VOC
- PM2.5
- Outdoor air conditions
For example:
CO₂ ↑
+
Occupancy ↑
+
Temperature Stable
+
Humidity ↑This provides more context than CO₂ alone.
AI can use these relationships to distinguish different environmental conditions.
9. AI-Based Ventilation Optimization
Ventilation has two competing objectives:
Maintain good indoor air quality
while
Avoiding unnecessary energy consumption.
Increasing outdoor-air ventilation can improve IAQ, but conditioning that air can require additional heating or cooling energy.
AI can therefore optimize the balance.
For example:
Poor IAQ
↓
Increase ventilation
↓
IAQ improves
↓
AI evaluates demand
↓
Reduce ventilation when appropriateThe system continuously adapts to actual conditions.
10. Ventilation Based on Rate of Change
Traditional control often focuses on the current CO₂ value.
AI can also consider its rate of change.
For example:
CO₂ = 750 ppm
may look perfectly acceptable.
But if CO₂ is increasing rapidly:
650 → 700 → 750 → 800 ppmthe system can recognize that conditions are changing.
AI may predict that the room will soon require additional ventilation.
This allows the system to act proactively.
11. Meeting Room Example
Consider a meeting room with:
- KNX CO₂ sensor
- Presence detection
- Temperature sensor
- KNX ventilation control
At 09:50 the room is empty.
At 10:00, a meeting begins.
The AI system recognizes the typical pattern:
Occupancy ↑
CO₂ ↑Based on historical behaviour, it predicts increasing CO₂.
Instead of waiting for the concentration to exceed a threshold, ventilation can increase progressively.
The result can be:
- Better air quality
- Fewer sudden ventilation changes
- Reduced risk of excessive CO₂
- Potentially lower energy consumption
12. AI and Demand-Controlled Ventilation
Demand-controlled ventilation adjusts ventilation according to actual requirements.
KNX can provide the control interface.
AI can make the demand prediction more sophisticated.
Possible inputs include:
- Occupancy
- CO₂
- Room schedule
- Historical occupancy
- Outdoor temperature
- Outdoor air quality
- HVAC operating conditions
The result can be a ventilation strategy that is more responsive to actual building usage.
13. Outdoor Air Quality
Indoor ventilation decisions should also consider outdoor conditions.
For example, increasing outdoor-air intake is not always desirable if outdoor air quality is poor.
A more advanced system can consider:
Indoor Air Quality
+
Outdoor Air Quality
+
Occupancy
+
Ventilation Demand
↓
AI Decision SupportThis can help determine the appropriate operating strategy.
External air-quality data can be integrated alongside KNX data through the building’s broader data platform.
14. Temperature and IAQ Together
Ventilation affects more than air quality.
Introducing outdoor air can influence:
- Temperature
- Humidity
- Heating demand
- Cooling demand
Therefore, IAQ control should not operate independently from HVAC control.
For example:
Increasing ventilation may improve CO₂ levels but significantly increase cooling demand.
AI can evaluate both effects.
This creates a more holistic building optimization strategy.
15. Humidity Optimization
Humidity can also influence indoor comfort and building performance.
A KNX humidity sensor can provide:
- Relative humidity
- Temperature
- Environmental trends
AI can identify patterns between:
- Occupancy
- Humidity
- Ventilation
- Outdoor conditions
This can help determine when ventilation or other environmental controls should respond.
16. VOC and Other Air-Quality Sensors
CO₂ is not the only relevant parameter.
Depending on the project, KNX-compatible or integrated sensors can provide information about:
- VOCs
- Particulate matter
- Formaldehyde
- Other air-quality indicators
AI can correlate these signals with occupancy, activities and ventilation.
For example:
VOC concentration increases rapidly when a particular room is occupied.
The system can learn the pattern and identify unusual behaviour.
17. Detecting Abnormal IAQ Behaviour
AI can also detect situations that do not match normal behaviour.
Suppose:
Occupancy: Normal
Ventilation: Normal
CO₂: Increasing unusually fastThe system may identify:
Possible ventilation performance issue.
Potential causes could include:
- Fan problem
- Damper problem
- Filter restriction
- Incorrect airflow
- Sensor issue
This connects IAQ optimization with the AI-Based Fault Detection & Diagnostics article already covered in the series.
The distinction remains important:
- FDD focuses on diagnosing the fault.
- This article focuses on maintaining IAQ efficiently.
18. Learning Room-Specific Behaviour
Different rooms behave differently.
A small meeting room may experience rapid CO₂ changes.
A large open office may change much more slowly.
AI can learn these room-specific characteristics.
For example:
Meeting Room A
CO₂ response: Fast
Meeting Room B
CO₂ response: Moderate
Open Office
CO₂ response: SlowThis allows more appropriate control strategies.
19. Predictive Ventilation
The ultimate goal is not simply to react to bad air quality.
It is to anticipate it.
A predictive system could recognize:
This room normally reaches elevated CO₂ approximately 12 minutes after occupancy reaches 10 people.
It can then prepare ventilation accordingly.
The control sequence becomes:
Occupancy prediction
→ CO₂ prediction
→ Ventilation prediction
→ Control action
This is a more intelligent approach than fixed threshold control.
20. AI-Based IAQ Optimization in Schools
Schools are an interesting application.
Classrooms can experience rapid changes in occupancy and CO₂.
A typical pattern may be:
Class starts
↓
Occupancy increases
↓
CO₂ increases
↓
Ventilation increases
↓
Class ends
↓
Occupancy decreases
↓
Ventilation decreasesAI can learn classroom schedules and historical behaviour.
The system can therefore anticipate periods of high ventilation demand.
21. AI-Based IAQ Optimization in Offices
Office buildings often have highly variable occupancy.
One room may be empty for several hours while another is heavily occupied.
AI can identify:
- Frequently occupied rooms
- Typical occupancy periods
- CO₂ patterns
- Ventilation demand
- Unusual room behaviour
This enables ventilation to follow actual usage rather than a fixed schedule.
Learn more about : AI-Based Occupancy Detection in KNX Buildings
22. AI-Based IAQ Optimization in Hotels
Hotel rooms present another use case.
A room may be:
- Unoccupied
- Occupied
- Being cleaned
- Temporarily occupied
- Permanently occupied by a guest
KNX room automation can provide information about room state.
AI can combine:
- Occupancy
- Temperature
- CO₂
- HVAC state
- Room schedules
to optimize environmental conditions.
23. AI-Based IAQ Optimization in Hospitals
Healthcare environments have more complex requirements.
Different areas may have specific ventilation and environmental requirements.
AI should therefore not override defined safety or clinical requirements.
Instead, it can assist with:
- Monitoring
- Trend analysis
- Anomaly identification
- Predictive insights
- Energy optimization within approved operating limits
Critical requirements should always take priority over optimization.
24. AI and Comfort
Good IAQ is part of occupant comfort.
A system should not optimize CO₂ while creating unacceptable:
- Temperature
- Humidity
- Air movement
conditions.
Therefore, AI can consider multiple objectives:
IAQ
+
Temperature
+
Humidity
+
Occupancy
+
Energy
↓
OptimizationThe result is a more balanced control strategy.
25. AI-Based Ventilation Scheduling
Schedules can provide useful information.
For example:
08:00 – Office occupancy begins
12:30 – Lunch period
13:30 – Occupancy increases
17:30 – Occupancy decreasesAI can combine schedules with actual sensor data.
This prevents the building from blindly following a schedule when reality is different.
For example:
A scheduled meeting is cancelled.
Occupancy remains zero.
The system can adapt instead of conditioning the room unnecessarily.
26. AI and Energy Trade-Offs
IAQ optimization should not be treated as an energy-saving exercise alone.
The primary requirement is to maintain appropriate indoor environmental conditions.
AI can optimize within defined comfort and operational limits.
For example:
IAQ Requirement
↓
Minimum ventilation
↓
AI optimization
↓
Energy-efficient operationThis prevents energy optimization from becoming the only objective.
27. KNX + AI IAQ Architecture
A practical architecture could look like this:
KNX Sensors
│
┌───────────┼───────────┐
│ │ │
CO₂ Temperature Presence
│ │ │
└───────────┼───────────┘
↓
KNX/IP Layer
↓
Data Platform
↓
AI Engine
↓
IAQ Prediction & Analysis
↓
Control Recommendation
↓
KNX
↓
Ventilation / HVACThe AI layer does not need to replace the KNX control system.
It can provide optimization information to the existing automation architecture.
Learn more about : KNX AI Architecture: From Sensors to Artificial Intelligence
28. Edge AI vs Cloud AI for IAQ
IAQ data can be processed locally or in the cloud.
Edge AI
KNX
↓
Local Gateway
↓
AI
↓
KNXAdvantages:
- Low latency
- Local operation
- Reduced cloud dependency
- Better data control
Cloud AI
KNX
↓
Gateway
↓
Cloud
↓
AI
↓
KNXAdvantages:
- Large-scale analytics
- Centralized models
- Multi-building comparison
A hybrid approach can combine both.
29. Privacy Considerations
IAQ systems can indirectly reveal occupancy patterns.
For example:
- When rooms are occupied
- How long meetings last
- Which spaces are heavily used
Therefore, occupancy-related data should be handled carefully.
Good practices include:
- Minimize unnecessary personal data
- Apply access controls
- Secure data transmission
- Define retention policies
- Use aggregated information where possible
The goal is to improve the building without unnecessarily tracking individuals.
30. AI Should Complement Existing KNX Control
A reliable architecture should continue operating even if the AI service becomes unavailable.
For example:
Normal Operation
↓
KNX Control
+
AI OptimizationIf AI becomes unavailable:
AI Offline
↓
Normal KNX Control ContinuesThis provides a robust fallback.
AI should enhance the automation system rather than becoming a single point of failure.
31. Practical Implementation Roadmap
Step 1 — Identify IAQ Objectives
Define:
- Required IAQ parameters
- Comfort targets
- Operating limits
- Energy objectives
Step 2 — Identify KNX Data
Map:
- CO₂
- Temperature
- Humidity
- Presence
- HVAC
- Ventilation
Step 3 — Collect Historical Data
Store time-series information.
Step 4 — Understand Normal Behaviour
Identify typical patterns for each room.
Step 5 — Build CO₂ Prediction
Start with short-term prediction.
Step 6 — Add Ventilation Optimization
Use predictions to improve ventilation control.
Step 7 — Add Multiple Environmental Variables
Include temperature, humidity and other available information.
Step 8 — Validate Performance
Compare AI recommendations against actual results.
Step 9 — Introduce Automated Optimization
Only after sufficient validation should AI recommendations influence automatic control.
32. Start With One Room
A practical pilot does not need an entire building.
Choose one:
- Meeting room
- Classroom
- Office
- Training room
Collect:
- CO₂
- Temperature
- Occupancy
- Ventilation status
Then measure:
- CO₂ response
- Occupancy patterns
- Ventilation response
- Comfort
- Energy impact
Once the model proves useful, expand it to other zones.
33. KPIs for AI-Based IAQ Optimization
Useful measurements include:
Air quality
- Time within target CO₂ range
- CO₂ peak frequency
- CO₂ recovery time
Comfort
- Temperature deviation
- Humidity deviation
- Occupant complaints
Energy
- Ventilation energy
- HVAC energy
- Runtime
AI performance
- Prediction accuracy
- False predictions
- Response time
The objective is to improve IAQ without unnecessary system operation.
34. Common Implementation Mistakes
Mistake 1 — Using CO₂ Alone
CO₂ is useful but does not represent every aspect of IAQ.
Mistake 2 — Using Fixed Thresholds for Everything
Different rooms and conditions can require different strategies.
Mistake 3 — Ignoring Occupancy
Occupancy strongly affects CO₂ generation.
Mistake 4 — Ignoring HVAC Interaction
Ventilation affects heating and cooling demand.
Mistake 5 — Optimizing Energy Above IAQ
Air quality and operational requirements must remain within defined limits.
Mistake 6 — Poor Sensor Placement
Incorrect sensor placement can produce misleading information.
Mistake 7 — Giving AI Unlimited Control
AI should operate within defined engineering constraints.
35. The Future of AI-Based IAQ
Future KNX buildings could move from reactive ventilation toward predictive environmental management.
Instead of:
CO₂ is high → increase ventilation.
the system could operate more like:
Occupancy is increasing, CO₂ is rising rapidly and this room historically reaches its IAQ limit within 10 minutes. Increase ventilation gradually now.
This is a significant change.
The building is no longer simply reacting to a threshold.
It is anticipating environmental conditions.
36. Conclusion
Indoor air quality is an important application area for intelligent building automation.
KNX already provides the sensors, communication and control infrastructure needed to monitor and influence the indoor environment.
AI can add predictive capabilities by learning relationships between:
- Occupancy
- CO₂
- Temperature
- Humidity
- Ventilation
- HVAC
- Historical behaviour
The resulting architecture can move from:
Measure → Threshold → React
toward:
Measure → Understand → Predict → Optimize → Control
The objective is not simply to increase ventilation.
It is to maintain appropriate indoor environmental conditions while avoiding unnecessary energy consumption and equipment operation.
For KNX professionals, this creates another practical application for AI: turning conventional demand-controlled ventilation into a more predictive and adaptive building system.
The smartest ventilation system is not the one that runs the most. It is the one that understands when ventilation is actually needed.
Learn More From Related Articles
KNX + AI: How Artificial Intelligence Is Transforming Smart Buildings
