Rick Kramer –Jos van Schijndel – Henk Schellen
Rick Kramer –Jos van Schijndel – Henk Schellen
Contents
• Introduction
• Inverse modeling concept
• Optimization procedure
• Tested models
PAGE 121-10-2013
• Tested models
• Validation
• Conclusions
Department of the Built Environment, unit BPS
Introduction: problem statement
Three problems regarding current modeling and simulation approach:
i. Long simulation run timeDue to long simulation period (In climate change projects: years 2000 - 2099 with time step 1h), combined with detailed physical models.
PAGE 221-10-2013
ii. Detailed modeling of the buildings requires much effortThe monumental buildings are old and protected: blueprints are hard to findand destructive methods to obtain building material properties are often not allowed.
iii. No easy characterization of the indoor climate nor an easy characterization of the energy performance possible
Department of the Built Environment, unit BPS
Introduction: objective
Objective:
The successful application of inverse modeling on a simplified hygrothermal building model.
PAGE 321-10-2013Department of the Built Environment, unit BPS
Inverse modeling concept
PAGE 421-10-2013Department of the Built Environment, unit BPS
Inverse modeling concept
measured data
simulation
PAGE 521-10-2013Department of the Built Environment, unit BPS
optimize model
Optimization algorithms
• Optimization algorithms:
• Global Search (local solver from many points) ~hrs
• Pattern search (direct search / gradient free) ~min
• Genetic Algorithm (stochastic / least accurate) ~sec
• Fmincon (gradient based, local solver) ~sec
PAGE 621-10-2013
• Findings:
• First use GA to narrow down solution space efficiently, then pattern search for thorough
searching, and fmincon for fine-tuning or checking.
Department of the Built Environment, unit BPS
State Space: fast optimization
• Global optimization algorithm PatternSearch:
⇒ Direct search type that is suitable for discontinuous and non-smooth solution spaces
⇒ In combination with state space model: fast optimization process
Department of the Built Environment, unit BPS PAGE 721-10-2013
Every dot represents a 1 yearbuilding simulation
Thermal models
Literature review on simplified building models:Kramer, R., van Schijndel, J., & Schellen, H. (2012). Simplified thermal and hygric building models: A literature review. Frontiers of Architectural Research, 1(4), 318–325.
PAGE 821-10-2013Department of the Built Environment, unit BPS
Thermal models
Literature review on simplified building models:Kramer, R., van Schijndel, J., & Schellen, H. (2012). Simplified thermal and hygric building models: A literature review. Frontiers of Architectural Research, 1(4), 318–325.
PAGE 921-10-2013Department of the Built Environment, unit BPS
Used building: Castle of Amerongen
• Castle of Amerongen (Amerongen, Netherlands)
PAGE 1021-10-2013
*22
Department of the Built Environment, unit BPS
King’s chamber
Thermal models: performance
King’s chamber
model # MSE [°C2] MAE [°C] FIT [%]
1 n.a. n.a. n.a.
2a n.a. n.a. n.a.
2b - - -
3a 4.05* 1.72* 54.57*
3b 0.86 0.72 79.00
PAGE 1121-10-2013
3b 0.86 0.72 79.00
4a 0.63 0.61 82.11
4b 0.91 0.74 78.43
5a 0.95 0.76 84.35
5b 1.00 0.78 83.94
6a 0.96 0.76 84.29
6b 0.69 0.65 86.68
* without fixed temperature
Department of the Built Environment, unit BPS
Thermal models: best thermal model
PAGE 1221-10-2013
• States (3rd order)
• Envelope (Cw)
• Indoor air (Ci)
• Interior part (Cint)
Department of the Built Environment, unit BPS
Thermal models: best thermal model
PAGE 1321-10-2013
• Inputs
• Temperature outdoor (Te)
• Constant temperature (Tfixed), e.g. ground contact
• Solar irradiation
Department of the Built Environment, unit BPS
Thermal models: best thermal model
PAGE 1421-10-2013
• Physical interpretation of Gfast
• Transmission through glazing
• Heat loss by ventilation / infiltration
Department of the Built Environment, unit BPS
Hygric models
PAGE 1521-10-2013Department of the Built Environment, unit BPS
Hygric models: performance
King’s chamber
model # MSE [Pa2] MAE [Pa] FIT [%]
1 4712 56 82.24
PAGE 1621-10-2013
2 4440 51 82.76
3 3023 45 85.77
4 3016 45 85.79
Department of the Built Environment, unit BPS
Hygric models: best hygric model
PAGE 1721-10-2013
• Specifications of model:• 2 states (2nd order): envelope, air;• 4 parameters (parameter is Gx/Cx);• 1 input (Pe).
Department of the Built Environment, unit BPS
Validation
The validation consists of:
• Multiple buildings assessment (1 of 10 in this presentation)
• Residual analysis
• Uncertainty analysis (see *)
• Sensitivity analysis (see *)
PAGE 1821-10-2013Department of the Built Environment, unit BPS
* Kramer, R. (2012). From Castle To Binary Code: the application of inverse modeling for the prediction and characterization of indoor climates and energy performances. University of Technology Eindhoven.
Validation: Saint Bavo’s cathedral
Saint Bavo’s Cathedral (Gent, Belgium)
PAGE 1921-10-2013Department of the Built Environment, unit BPS
South transept
Validation: Saint Bavo’s cathedral
Short-term dynamics:Long-term dynamics:
PAGE 2021-10-2013Department of the Built Environment, unit BPS
Validation: Saint Bavo’s cathedral
Residuals’ (ɛ) probability distribution:
ɛ = Tisim - Timeas
PAGE 2121-10-2013Department of the Built Environment, unit BPS
Validation: Saint Bavo’s cathedral
Residual analysis:
• whiteness test shows correlation between residuals
• Too much auto-correlation = missing inputs or missing error model
• An error model is lacking so auto-correlation is meaningless
PAGE 2221-10-2013Department of the Built Environment, unit BPS
• An error model is lacking so auto-correlation is meaningless
• independence test shows correlation between residuals and inputs
• Too much cross-correlation = influence of inputs modelled incorrectly
Validation: Saint Bavo’s cathedral
Residual analysis: independence test shows correlation between residuals and inputs
PAGE 2321-10-2013Department of the Built Environment, unit BPS
1. State Space modeling is very suitable for inverse modeling of free-floating monumental buildings.
2. Developed hygrothermal model performs excellent for most cases (Goodness of Fit > 80%).
Conclusions
3. Identified parameters are effective, not apparent values and are a ratio Gx/Cx: verification is difficult.
PAGE 2421-10-2013Department of the Built Environment, unit BPS
Thank you!
Rick Kramer, MSc
Tel +31 40 247 5613
Jos van Schijndel, PhD
Henk Schellen, PhD
Validation: Castle Keukenhof
• Castle Keukenhof’s loft (Lisse, Netherlands)
PAGE 2621-10-2013Department of the Built Environment, unit BPS
loft
Validation: Castle Keukenhof
PAGE 2721-10-2013Department of the Built Environment, unit BPS
Rain leakage?
Modeling approach for solar irradiance even suitable for lofts (tilted roofs)
Validation: Castle of Amerongen
• Castle of Amerongen’s washing room (Amerongen, Netherlands)
PAGE 2821-10-2013Department of the Built Environment, unit BPS
washing room
Validation: Castle of Amerongen
No fixed vapour pressure node
Physical interpretation:
Washing room is adjacent to canal
PAGE 2921-10-2013
With fixed vapour pressure node
Department of the Built Environment, unit BPS
Washing room is adjacent to canal
Increased moisture level possibly due to:
- moisture flow through walls- recent flooding
Thank you!
Rick Kramer, MSc
Tel +31 40 247 5613
Jos van Schijndel, PhD
Henk Schellen, PhD