Sergio Maria Capanelli's speech during Chillventa 2018, discussing the fast-paced process of automation in HVAC/R systems generated by the Internet of Things
4. Connectivity Internet of things
Internet of Things (IoT) is the inter-networking of physical devices (also referred to
as “connected devices” and “smart devices”), like vehicles, buildings, and other
items embedded with electronics, software, sensors, actuators, and network
connectivity which enable these objects to collect and exchange data.
http://www.cns-me.com/en/ict-service-details/57/iot-and-smart-cities
5. IoT Services
Descriptive analytics:
describe the current and past situation, converting the data into information
Preventive analytics:
understand what might happen in the future
Prescriptive analytics:
operative/strategic solutions are proposed by the system on the basis of the
analysis performed
Proactive analytics:
automatic implementation of the proposed actions
6. Machine learning
How machine learning works:
Training and Calibration: collected data are used to learn how machines perform on field.
Test and Results: trained algorithm is tested using new data as input. The output is the
“baseline” definition.
Training and Calibration
hour of the day, outdoor temperature/humidity, EEV
position, fan flow, water temperature, compressor
speed, regulation set-point, etc.
Dataset
Test and Results
Input data
Trained
algorithm
7. IoT in HVAC systems
Case study on Preventive Analytics:
show and improve the real operation efficiency1 of Heat Pumps by means of machine learning and
A.I. tools
1regardless the influences due to heat pump geographical location, seasonal and environmental aspects, weather conditions,
etc.
Fans/Pumps
status, speed
Machine learning
and A.I.
0
15
30
45
60
6 12 18 24
Heat Pump
#16 is under-
performing!
Water
temperatures
Compressor
Status
Refrigerant
charge,
subcooling
EEV
status,
superheat
8. IoT in HVAC systems
Predictive maintenance
Actions that were previously taken in
response to an emergency can now
be planned in advance!
Attention!!! The
consumption is
higher than
expected!!!
9. IoT in HVAC systems
Performance optimization
The baseline definition of a trained machine learning algorithm helps in fine-tuning.
Succesful fine-tuning!
Power consumption is
much lower than
baseline.
10. IoT in HVAC systems
Benchmarks between different units
Comparison between different units of the same model installed in different
plants. The normalized power consumption allows the complete analysis.
Unit HVAC2 is
performing
better than
HVAC1.
11. IoT in refrigeration systems
Case study:
show and improve the defrost status of supermarket showcases by means of machine learning and
A.I. tools.
Valve opening
Machine learning
and A.I.
The defrost in
showcase #9 is
anomalous!
Air off
temperature
Evaporator
temperature
Setpoint
temperature
12. IoT in refrigeration systems
ScoreShowcases Height and color: quality of defrost; Length: duration of defrost
13. IoT in refrigeration systems
• The defrost score indicated that
there was an anomaly in the
system.
• The operator realized that the
probe had been wrongly installed.
• After relocation, the score
became green.
Warning!
Please check defrost
duration and defrost
probe positioning!
15. IoT in refrigeration systems
Some punctual red scores
indicated that the lenght of
defrost was not enough for
severe conditions.
Warning!
Severe conditions
occurred twice in two
weeks! Please check
defrost duration!
17. Conclusions
Connectivity of different devices and machine
learning techniques are being applied to HVAC/R
systems leading to promising results.
The implementation of IoT in HVAC/R systems will
result in important benefits such as energy saving
and prevention of failures.
The fast development of IoT in different sectors
will help to face the challenges that facilitate the
improvement of IoT services in HVAC/R systems.
18. Thank you for your attention
sergio.capanelli@carel.com