PP - modelEAU

Transcription

PP - modelEAU

28/04/2013
BIOMATH
Department of Applied Mathematics,
Biometrics and Process Control
Integration is a central theme
(Reflections on the SAIA “Specialist” Group)
Peter Vanrolleghem
01-jan-04
Watermatex 2004, Beijing, Nov 3-5 2004
UGent-BIOMATH, Coupure 653, 9000 Gent, Belgium (e-mail [email protected])
Water…
I begynnelsen skapade Gud himmel och jord.
Jorden var öde och tom, och mörker var över
djupet Och Guds Ande svävade över vattnet
djupet.
vattnet.
(Genesis 1:1-2)
In the beginning God created the heaven and the
earth. And the earth was without form, and void;
and darkness was upon the face of the deep
deep.
And the Spirit of God moved upon the face of the
waters. (Genesis 1:1-2)
Docentföreläsning Ulf Jeppsson, 15 Dec 2003
Peter A. Vanrolleghem - 2
BIOMATH
1
28/04/2013
Reflexioner om vatten
Industriell Elekttroteknik och Automation
Noah’s ark flood (5500 B.C.)
© Ulf Jeppsson
BIOMATH
Lunds universitet
Peter A.IEA,
Vanrolleghem
-3
Docentföreläsning 15 december 2003
Global “Bluewater” Resources
Distribution of Global
Fresh Water & Salt Water
Distribution of Global Fresh Water Only
(2.5% of Global Water)
2.5%
0.9%
2.5%
Fresh water
30%
Total water
97.5%
Salt
water
69%
0.3%
This is the proportion of
the world’s fresh water
that is renewable
Glaciers/permanent snow cover
Other (incl soil moisture, swamp water etc)
Fresh groundwater (10.5m ck)
Freshwater lakes and river flows ((93k ck)
a global network for water professionals
BIOMATH
2
28/04/2013
Fresh and salt water
Jeppsson (2003)
BIOMATH
Systems Analysis & Integrated Assessment
= Divide & Conquer
• Integration is a key word  Generalization
• Reflections…
 Watermatex2004  PVRdr ds dt
BIOMATH
3
28/04/2013
Outline
•
•
•
•
•
•
IWRM : Integrated Water Resources Management
Within-fence modelling (complete WWTP)
Ecological Risk Assessment/Management
Plug&Play modelling
Distributed Virtual Experimentation
Concluding remarks
BIOMATH
From global to local… (and back)
Point sources versus non-point sources
NASA
BIOMATH
4
28/04/2013
Worldwide Water Use by
Sector
a global network for water professionals
BIOMATH
BIOMATH
5
28/04/2013
Diffuse pollution modelling
Nil case study for dynamic pesticide fate modelling
Systems Analysis
BIOMATH
Data collation (GIS-layers)
Integration
BIOMATH
6
28/04/2013
HUMEUR
• HUMEUR = HUMUS for EU (2006 ?)
• HUMUS = Hydrologic Unit Model of the US (1993)
•
•
•
•
Nutrients/Pesticides
Catchment-scale
Non-point source focus
SWAT model (Soil & Water Assessment Tool,
SWAT-model
Tool USDA )
• HUMUNGOUS™ = HUMUS for the World (????)
BIOMATH
Data collation
•
•
•
•
Problematic in Europe
No common database structure/format
Different levels of quality/detail
JRC
–
–
–
–
–
Joint Research Centre, DG Environment
First European database by 2006
SWAT-use oriented
Uncertainty aspects are considered
Information on point pollution is major issue (unexpectedly)
BIOMATH
7
28/04/2013
Outline
•
•
•
•
•
•
Within-fence modelling (complete WWTP)
Plug&Play modelling
Concluding remarks
BIOMATH
Within-Fence modelling
BIOMATH
8
28/04/2013
Benchmark Simulation Model No. 2
Influent
wastewater
Primaryy
clarifier
AS
Secondary
clarifier
Thick.
AD
ADM1
BIOMATH
Model interfacing
• Principles
– Elemental balancing (C, H, O, N, P, COD, charge…)
– System-specific
S
ifi b
behaviour
h i
off components
must be carefully reflected upon by domain specialists
(e.g. nitrifiers in waste sludge ending up in digester)
• 3 approaches
– Supermodel (Dold et al., 2003)
– One
One-to-one
to one interfaces (Vanrolleghem et al
al., 2004)
– Plant Wide Model Interface (Ayesa et al., 2004)
BIOMATH
9
28/04/2013
Supermodel
• All components are considered in each subsystem
• Behaviour must be described in each subsystem
• Not scalable when more subsystems are considered
ASP
AD
ASM1
components
ASM1
components
ADM1
components
ADM1
components
BIOMATH
One-to-one interfaces
• For each model combination, 2 interfaces needed
• Nx(N-1)/2 interfaces for N models
• Generalized method to create interfaces is proposed
ASM1
ADM1
Converter
ASM1-ADM1
ADM1
Converter
ADM1-ASM1
ASM1
BIOMATH
10
28/04/2013
Plant-Wide Model Interface
C
C
o
o
n
n
v
v
e ADM1-like e
(PWMI) r
r
t
t
e
e
r
r
PWM compon
nents vector
C
o
n
v
e
r
t
e
r
PWM compon
nents vector
C
o
n
v
ASM1-like
ASM1
like
e
(PWMI)
r
t
e
r
PWM compon
nents vector
PWM compon
nents vector
• Bus: models have a “wrapper” to one interface vector
• 2xN interfaces, but if new model: work to be redone
• Generalized method to create interfaces proposed
BIOMATH
Role of SAIA-IWA ?
• Support Benchmarking Task Group
– Test case for interface development
• Some model upgrading may be useful
– E.g. pH-modelling in all units
– Elemental composition in all models
• Time to revise the IWA model suite ? (SBML !)
– Gernaey et al. (2004): deficiencies in ASM1
– Gernaey et al. (2004): errors in current reports
• Define IWA standard interfaces
– Agree on underlying assumptions
BIOMATH
11
28/04/2013
Outline
•
•
•
•
•
•
Within-the-fence modelling (complete WWTP)
Plug&Play modelling
Concluding remarks
BIOMATH
Ecological risk assessment
Exposure Analysis
Effects Analysis
Atmosferic
Deposition
Treated
Discharges
WWTP
Erosion
& Runoff
Untreated discharges
Predicted Environmental
Concentration (PEC)
Laboratory- and fieldstudies
(toxicity tests)
Predicted No Effect
Concentration (PNEC)
YES, potential risk
BIOMATH
12
28/04/2013
Ecological risk assessment/management
• Risk analysis of chemical substances:
Risk Assessment
Scientific part
?
Precautionary
principle
Risk management
Risk communication
Risk Management
Decision-making,
policy
?
BIOMATH
Include variability & uncertainty
X mg/L
Point estimate
Make more
realistic
Probabilistic approach
PEC
spatial + temporal + other
variability & uncertainty
BIOMATH
13
28/04/2013
Reduce spatial variability
PEC
temporal + other
spatial + temporal + other
Reduce
spatial variability
BIOMATH
Probabilistic risk assessment
Exposure
Environmental
Concentration
Distribution
(ECD)
Effects
Species
Sensitivity
Distribution
(SSD)
Daphnia
% risk
Pimephales
BIOMATH
14
28/04/2013
Ecological risk assessment/management
• Detailed information can be made available now
– Geo-referenced
– Variability
V i bili and
dU
Uncertainty
i
(I
(Imprecise
i probabilities)
b bili i )
• How does the manager deal with this ?
– Summary statistics, indicators
• Where is the divide between
– Assessment
– Management
• Where does the precautionary principle belong ?
BIOMATH
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
Deterministic
Risk Assessment
Exposure
modelling/
monitoring
PP
PEC
Effects
assessment
Verdonck et al.
UPEM conference,
Kollekolle, June 7-9 2004
PP
PNEC
PP
Risk quotient
Risk managementPP
BIOMATH
15
28/04/2013
.. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ..
‘Old’ EU
Deterministic
Risk Assessment
Current EU
Risk Assessment
Probabilistic
Risk Assessment
Exposure
modelling/
monitoring
Exposure
modelling/
monitoring
Exposure
modelling/
monitoring
Effects
assessment
ECD
SSD
PP
Effects
assessment
PP
PP
ECD
PP
PEC
PNEC
PP
Risk quotient
PEC
Effects
assessment
SSD
PP
PNEC
PP
Risk quotient
Risk probability
+ confidence int.
PP
Risk managementPP Risk managementPP Risk management
BIOMATH
Outline
•
•
•
•
•
•
Plug&Play modelling
Concluding remarks
BIOMATH
16
28/04/2013
Plug-&-play Modelling
• OpenMI (Blind, Watermatex2004): Software coupling
• Peter Reichert (Watermatex 2004)
“A standard interface between simulation programs
and systems analysis software”
• Pasky Pascual (Beijings News Plaza, 3-4 Nov 04)
”Modellers shouldn’t be coding anymore”
• New back-end of WEST (Tornado)
software
f
platform
f
with defined API
with plug-&-play capability for tools
set of available tools
extensible by “coders”
BIOMATH
T o rn a d o
Plug&Play
Solvers in
Tornado
M SLE
D AE M odel
C a lc V a r
. ..
...
S o lv e
Root
I n te g
B ro yd e n
O p tim
S cen
Sens
AB 2
P r a x is
C ro s s
AB 3
S im p l e x
G rid
CI
P la in
MC
N e ld e r
CVT
R ic h a rd s o n
IH S
M il n e
AB 4
IH S
M o d i fie d E u le r
CVODE
PR
RK4
E u le r
R K 4ASC
LSODA
R osenb ro c k
M i d p o in t
P lu g in s
BIOMATH
17
28/04/2013
Outline
•
•
•
•
•
•
Plug&Play modelling
Concluding remarks
BIOMATH
Complex virtual experimentation
• Virtual experimentation
= model-based studies
– For
F environmental
i
l systems: Si
Simulation
l i d
determines
i
time
i
• The studies we undertake:
– Always take maximum a weekend to calculate…
– Become more complex according to Moore’s law (x 1.8/yr)
• How to speed up ?
BIOMATH
18
28/04/2013
Typical model:
~ 200 Differential equations
~ 1000 Constants/Parameters
~ 1000 Algebraic equations
BIOMATH
Scenario analysis:
~ 4 Parameters
~ 10 Samples
 104 Combinations
 10,000 Simulations
BIOMATH
19
28/04/2013
CD4WC project
•
Cost-effective development of urban water systems
for Water Framework Directive compliance
– EU
EU-project
j t
– Integrated study of sewer-WWTP-river system
– Methodology for evaluation of design/upgrade scenarios
•
Simulations of WWTP options
–
–
–
–
5 climatic conditions
3 plant
l t sizes
i
(3
(3.000,
000 30
30.000,
000 300
300.000)
000)
20 options
100 Monte Carlo shots (LHS) for uncertainty propagation
= 30.000 simulations
BIOMATH
Computational burden in modelling
•
•
•
•
Scenario analysis
 multiple independent simulations
Sensitivity analysis  multiple independent simulations
M t Carlo
Monte
C l analysis
l i  multiple independent simulations
Optimization :
Genetic Algorithm
Steepest Descent
• Experiment Design
 multiple independent simulations
 multiple dependent simulations
 multiple independent sensitivity
analyses
y
for different p
parameters
• Robustness Optimization (Walters, Watermatex2004)
= Optimization of sensitivity analyses
 multiple independent sensitivity
analyses for different parameters
BIOMATH
20
28/04/2013
• No split-up of model in submodels that are then
simulated over multiple calculation nodes
• Independent virtual experiments (e.g. simulations,
sensitivity analyses) can be distributed
• Different approaches to reach this:
– Grid technology
– Clustering
– WDVE (West Distributed Virtual Experimentation)
BIOMATH
Grid technology
• World-wide network
of computational & storage nodes
• Virtual organizations
BIOMATH
21
28/04/2013
SuperMUSE Cluster (US EPA, Athens, Georgia)
BIOMATH
SuperMUSE Cluster (US EPA, Athens, Georgia)
• Supercomputer for Model Uncertainty
and Sensitivity Evaluation
• Babendreier et al. (2002)
• Cluster-based solution
• Linux/Windows mix
• Goal: 384 client PCs, ~1000 GHz
• No other use of PCs possible
BIOMATH
22
28/04/2013
Idle computers in your offices …
BIOMATH
• Applied to
– WEST
– SWAT
• For
Master
Reichert !?
Application
Front-end
Dispatcher
Comm. Layer
Middleware
– Exp. Design
– Monte Carlo
Middleware
Comm. Layer
Acceptor
Middleware
Comm. Layer
Acceptor
Middleware
Comm. Layer
Acceptor
Application
Back-end
Application
Back-end
Application
Back-end
Slave
Slave
Slave
BIOMATH
23
28/04/2013
Take home
(Things to keep you awake ?)
• Characteristics of the Specialist Group on
Systems Analysis and Integrated Assessment (SAIA)
–
–
–
–
–
A Generalist Group, not a Specialist Group
Galilei : last person to know everything about everything
Now: a molecular layer of knowledge above the whole sea
Specialist Group:
Know everything about something
SAIA:
Know something about everything
BIOMATH
Take home
(Things to keep you awake ?)
• Possible tasks of Specialist Group on
Systems Analysis and Integrated Assessment (SAIA)
–
–
–
–
–
Don’t become a “Specialist” group, but “Generalist” group
Remain multidomain in interest
Support Task Groups (BSM, Calibration guidelines)
Direct Task Group on Modelling ASPs
Be at the nexus of policy-law-science
policy law science
BIOMATH
24

PP - modelEAU

Transcription

Similar documents

Peter Ting (Hong Kong 1951 - ) Pair of Lotus Stand Candle Holders

Cast all your on God because he cares for you. Cast all your on God

The Sidings Low Moor, Rillington, Malton, YO17 8JX

st. pe ter week ata glanc e

Receiver`s Application for Supplemental Relief Under Order Holding

to find more detailed information on the partners working on this

Twin M3 11/09

Prayer to St Peter Chanel - Society of Mary, Marists in the US

SPCS PTA - St. Peter Catholic School