Department of Building, Civil, and Environmental Engineering Concordia University

1
Simulation Based Optimization ApproachApplication
to Ventilation System Design and Operation in
Office Environment
Presented by Liang (Grace) Zhou Supervised by
Dr. Fariborz Haghighat
Department of Building, Civil, and Environmental
Engineering Concordia University
2
Office Buildings Energy Related Issues in Canada
Total 8,543.3 PJ/yr (2004)
Total 505.4 Mt of CO2e (2004)
Total 1,171.2 PJ/yr Space heating and cooling
683 PJ/yr
Source Natural Resources Canada (NRCan) Energy
Use Data Handbook Tables, (2005)
3
Office Workplace Health and Productivity
Insufficient Ventilation RatesSick Building
Syndrome (Allergies, asthma, respiratory
diseases, airborne infection, etc.)
Predominant complaints Its too hot and too
cold in the office, simultaneously.
(International Facility Management Association,
2004)
4
Remedy Ventilation Design/Operation Optimization
Better Indoor Built Environment
5
Office Workplace---CFD Model (Airpak)
Phase I Indoor Built Environment Evaluation
RNG K-eTurbulence Model (Yakhot and Orszag, 1986)
6
Optimization method applied into building research
Phase II Optimization Scheme
Gradient-based optimization method
Gradient-free optimization method

• Generalized pattern search
• GenOpt for building design
  (Wetter , 2004)
• Genetic algorithm (GA)
• chiller energy usage optimization
  (Chow et al. 2002)
• building thermal design and control
  (Wright et al. 2002)
  
• adaptive HVAC system control
  (Lu et al. 2005)
• green building design optimization
  (Wang et al., 2005)
• low energy dwellings life cycle cost
  optimization (Verbeeck and Hens,
  2006)
• Advantages
• convergence to global near-optimal
• applicable to problem with discontinuous
  variables
• noise tolerant
• cooperative with black-box models

7
Aggregate and Weight Metrics into One Index
Objective Function for Fitness Evaluation in GA
i the number of occupants Maximize comfort
level Wiaq Wfan Wcooling 0 Maximize air
quality Wtc Wfan Wcooling 0 Maximum
energy efficiency Wtc Wiaq 0 In following
section Wtc 0.5, Wiaq 0.1, Wfan 1,
Wcooling 0.5 Penalty Term (PT) set to 1 when
local velocity, Equivalent T and Thead-to-ankle
exceed criteria
CFD too expensive to be invoked in the GA
optimization search An alternative simplified
model for system response approximation
8
System Response Approximation
CFD

• PMV
• ev
• Ecooling

ANN Training

• PMV
• ev
• Ecooling

Database
8
9
Summary of Numerical Approach
Numerical optimization
10
CFD simulation with swirl floor and ceiling
diffusers
CFD Results Validation
Experiment results provided by Cermak, R.
(2005), Technical University of Denmark
Tsupply 20 C Vsupply 80L/s Power of heat
source 584 W (22.5W/m2) Twall 25 C CO2
emission rate 24mL/s CO2 concentration in
supply air 400 ppm
11
Airflow pattern from swirl floor diffusers
CFD Results Validation
Momentum method with simplified geometry
Tsupply 20 C Vsupply 20L/s per
diffuser Throw 0.8m RNG k-eturbulence model 1
million meshed cells y ranges from 30-120
12
Simulation results with swirl diffusers in
underfloor system
CFD Results Validation
Velocity profiles
Normalized CO2 concentration profiles
Temperature profiles
Solid line measurements dotted line predictions
13
CFD case design from Latin Hypercube Sampling
Design Space for Database
14
ANN Model Testing (30 cases not being used for
training)
Validation of ANN Model
15
GA parameters
GA Search for Near Optimal Solution
16
Improved Ceiling System (varying inner surface
temperature)
Near Optimal Solution
MS1- ceiling system with 2 swirl diffusers in
office, MS2- ceiling system with 1 swirl diffuser
in officeInternal heat load density 25 W/m2,
CO2 emission rate 1.0L/min
17
Improved UFAD (varying inner surface temperature)
Near Optimal Solution
UFAD1- floor system with 4 swirl diffusers in
office UFAD2- floor system with 2 swirl
diffusers in office Internal heat load density
25 W/m2, CO2 emission rate 1.0L/min
18
Optimization Results
Demo case Improvement with optimal solution
Environment controlInternal heat load density
25W/m2, Tsurface 30c, CO2 emission rate
1.0L/s
19
Temperature, Airflow and PMV
Original Underfloor System
Tsupply 20c Vsupply 100L/s
Improved Underfloor System
Tsupply 19c Vsupply 80L/s
20
CO2 Concentration Near Occupant
Original Underfloor System
Improved Underfloor System
21
Temperature, Airflow and PMV
Original Ceiling System
Tsupply 18.5c Vsupply 100L/s
Improved Ceiling System
Tsupply 20c Vsupply 80L/s
22
CO2 Concentration Near Occupant
Original Ceiling System
Improved Ceiling System
23
Future Work Whole Building Design Optimization
Advanced Building Design
24
For Two Generic Office Buildings
8
25
Why CFD-ANN-GA for Office Built Environment
Closing Discussion

• CFD holistic information about comfort, air
  quality, and ventilation energy use in a
  climatized office environment
• ANN for system response approximation speed up
  the fitness evaluation in GA optimization search
  (cutting down the 17 hours CPU time per CFD case
  to a time scale of minutes)
• The current optimization scheme efficient in
  searching for the near optimal configurations and
  operation states of ventilation system in office
  environment---desirable comfort and IAQ without
  sacrificing energy cost

Future Work Survey on Applications of Available
Advanced Building StrategiesTrnsys-ANN-GA for
Ultra-low Energy Design in Two Generic Office
Buildings
26
Thank you!
27
Validation of Optimization Approach ANN Based GA
28
Convergence of ANN Training/Validation/Test
100 sample data sets
1hidden layer with 6 hidden neurons
29
3-D Surface of Rosenbrocks Function
ANN Approximation of Rosenbrocks Function
Rosenbrocks Function
30
Searching for Minimum by Applying GA
Elite count 2 crossover fraction 0.8
mutation fraction 0.2
Applying GA to ANN Approximation of the Function
Applying GA to Original Function
31
Demo case design underfloor system with 4 swirl
diffusers
32
Objective indices in this work
Literature review
33
(No Transcript)