occupancy based hvac controlpeople.cs.pitt.edu/~mosse/courses/cs3720/final_occupancy...motivations...
TRANSCRIPT
Occupancy Based HVAC ControlCS3720 Final Presentation
Keren Ye, Debarun Das, and Xiaoyu Liang
Outline
Motivations
Method
Experiments & Evaluation
Conclusion
Future Scopes
Motivations
● Lack of user comprehension of the behavior of HVAC systems○ Improvement of Occupant Comfort needed
● Largest consumer of energy in buildings are the HVAC systems○ 26.1% and 53.4% of total energy usage in residential and commercial
applications, respectively○ Energy Consumption Saving
● Usage of new, more efficient ventilation technologies are expensive○ Need to use existing technology more efficiently
Motivations -- Contd.Occupancy Based HVAC systems
● Energy usage can be reduced by reducing heating and cooling when occupants are away [1].● The number of occupants in a room affect the room temperature [2].
Occupancy Counting [3]
● Determine the total number of people in a room at a given time. ● Sensor based methods
○ Need to deploy multiple special sensors within the building○ Accuracy is questionable
● Camera based methods○ Capability of achieving high accuracy○ Require significant computation resources ○ Is often criticized of the privacy concerns
1. Elahe Soltanaghaei and Kamin Whitehouse: Practical occupancy detection for programmable and smart thermostats.2. J.S. Sage-Lauck and D.J. Sailor: Evaluation of phase change materials for improving thermal comfort in a super-insulated residential building
3. Kemal Akkaya, et.al.: IoT-based Occupancy Monitoring Techniques for Energy-Efficient Smart Buildings
Raspberry Pi 3, Model B
Arduino UNO
Sensors
● Raspberry Pi camera● HC-SR04 Ultrasonic Sensor Distance Module● DHT11 temperature / humidity sensor
Actuators
● SG90 Micro Servo Motor - I am the FAN● Arduino LCD module
Method - Hardwares
Pi Camera
Distance Sensor
Temperature& Humidity
Sensor
Fan
LCD
Method - SoftwaresOpenFace
● Amos, Brandon, Bartosz Ludwiczuk, and Mahadev Satyanarayanan. "Openface: A general-purpose face recognition library with mobile applications." CMU School of Computer Science (2016).
OpenCV
● Bradski, Gary B. Open source computer vision library. Springer, 2004.
Method - Motion Detection
Detect the motion of passing through the door
● Potential changes of the number of people in the room
Activate image processing module
Method - Person Detection & OccupancyPerson-detection and counting
● Per-frame person (face) detection● Count & vote on consecutive frames
2
Method - Person Detection & Occupancy
Face detection
● OpenCV face detection● OpenFace face detection
Method - Person Detection & Occupancy
Approaches
● Majority Vote approach● Clustering using facial features
Clustering based counting
Method - Decision Making
Aggregation of sensor data
● Occupancy data from Raspberry Pi module● Temperature & humidity information
Decision making
● HVAC control depends on:○ Current Temperature○ Number of Occupants
Assumptions● Insulated. Outdoor Temperature ignored● Avg mass of Human body = 70 kg● Avg Specific Heat of Human body = 3470 J/(kg °C)● Temperature Range = 18°C - 30°C● Equation
Sleep Activity
HVAC Cooling Rate (°C/hr)
Avg Temp. (°C)
Std. Dev
Temp. In Range %
5 33.17 6.08 54.17
10 25.67 3.47 100.00
15 24.63 4.93 83.33
20 21.92 6.74 66.67
Light Activity
HVAC Cooling Rate
(°C/hr)
Avg Temp. (°
C)
Std. Dev
Temp. In Range %
5 43.42 17.86 50.00
10 26.75 3.12 95.83
15 24.05 5.44 83.33
20 22.59 7.14 66.67
Moderate Activity
HVAC Cooling Rate
(°C/hr)
Avg Temp. (°C)
Std. Dev
Temp. In Range %
5 56.04 34.20 45.83
10 35.42 14.76 54.17
15 23.13 6.84 41.67
20 16.67 6.31 33.33
Experiments & EvaluationMotion detection
● Energy consumption
Image processing module
● Accuracy of the counting
Experiments & Evaluation - Motion Detection Module
Strategies to activate image processing module
● Without motion detection - activate periodically● With motion detection - delayed activation once human motion is detected
Energy consumption of the Raspberry Pi
● Assumes the energy consumption is proportional to CPU utilization
Experiments & Evaluation - Motion Detection Module
Experiments & Evaluation - Image processing module
Accuracy of the image processing module
● Web images○ Downloaded from the web○ Capture them by camera
● Real-time video data○ System deployed in Room 6410
Metrics to evaluate performance of detection
Web images
Real-time video data
Experiments & Evaluation - Image Processing Module
Accuracy of the image processing module
● Web images (avg: 11.1 people)○ Downloaded from the web○ Capture them by camera
● Real-time video data (avg: 2.1)○ System deployed in Room 6410
Web images Real-time video
Methods Accuracy RMSE Accuracy RMSE
Majority vote 12.5% 3.37 36.11% 1.11
Clustering 0 4.14 NA NA
Web images
Real-time video data
Conclusion
● HVAC Control based on occupancy
● Face Detection Algorithm to count the no. of occupants○ Majority Vote○ Triggered by Distance Sensor
● HVAC Control - Selecting rate and time of cooling based on ○ No. of Occupants○ Current Temperature
Future Scopes
● Improvement in accuracy of face detection algorithm
● Application of HVAC Control rules○ Study of actual changes in room temperature○ Inclusion of more parameters for HVAC control
● Usage of user feedback for thermal comfort
Thank you!