From Multi-Parametric Programming Theory to MPC-on-a-chip...
Transcript of From Multi-Parametric Programming Theory to MPC-on-a-chip...
From Multi-Parametric
Programming Theory to
MPC-on-a-chip Multi-scale
Systems Applications
Stratos Pistikopoulos
FOCAPO 2012 / CPC VIII
Acknowledgements Funding
EPSRC - GR/T02560/01, EP/E047017, EP/E054285/1
EU - MOBILE, PRISM, PROMATCH, DIAMANTE, HY2SEPS
CPSE Industrial Consortium, KAUST
Air Products
People J. Acevedo, V. Dua, V. Sakizlis, P. Dua, N. Bozinis, N. Faisca
Kostas Kouramas, Christos Panos, Luis Dominguez, Anna Vöelker, Harish Khajuria, Pedro Rivotti, Alexandra Krieger, Romain Lambert, Eleni Pefani, Matina Zavitsanou, Martina Wittmann-Hoghlbein
John Perkins, Manfred Morari, Frank Doyle, Berc Rustem, Michael Georgiadis
Imperial & ParOS R&D Teams
Outline
Key concepts & historical overview
Recent developments in multi-parametric
programming and mp-MPC
MPC-on-a-chip applications
Concluding remarks & future outlook
Outline
Key concepts & historical overview
Recent developments in multi-parametric
programming and mp-MPC
MPC-on-a-chip applications
Concluding remarks & future outlook
What is On-line Optimization?
MODEL/OPTIMIZER
SYSTEM
Data -
Measurements
Control
Actions
What is Multi-parametric Programming?
Given: a performance criterion to minimize/maximize
a vector of constraints
a vector of parameters
s
n
u
u
x
xug
xufxz
R
R
0),( s.t.
),(min)(
What is Multi-parametric Programming?
Given: a performance criterion to minimize/maximize
a vector of constraints
a vector of parameters
Obtain: the performance criterion and the optimization
variables as a function of the parameters
the regions in the space of parameters where these functions remain valid
s
n
u
u
x
xug
xufxz
R
R
0),( s.t.
),(min)(
Multi-parametric programming
s
n
u
u
x
xug
xufxz
R
R
0),( s.t.
),(min)(
)(xu
(2) Critical Regions
(1) Optimal look-up function
Obtain optimal solution u(x) as a
function of the parameters x
Multi-parametric programming
1001001010
0
0
0
0
8
121
20
13
00
10
00
01
14
228
45
11
.
83min21 ,
21
2
1
2
1
21uu
xx
x
x
u
u
uu
st
Problem Formulation
Multi-parametric programmingCritical Regions
-10 -8 -6 -4 -2 0 2 4 6 8 10-100
-80
-60
-40
-20
0
20
40
60
80
100
x1
x2
4 Feasible Region Fragments
CR001
CR002
CR003
CR004
x2
x1
Multi-parametric programmingMulti-parametric Solution
100
10
65.8
10
01
11.01
8.9
8.11
06.00
05.00
100
5
5.7
10
01
045.01
13
0
01
00
100
100
10
5.7
71.6
65.8
10
10
01
045.01
031.01
115.01
5.7
5.5
03.026.0
03.073.0
100
100
10
5
71.6
10
10
01
01
031.01
67.14
67.1
033.1
033.0
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
x
xif
x
x
x
x
ifx
x
x
x
ifx
x
x
x
ifx
x
U
Multi-parametric programming
Only 4 optimization problems solved!
100100,1010
0
0
0
0
8
121
20
13
00
10
00
01
14
228
45
11
.
83min
21
2
1
2
1
21
xx
x
x
u
u
st
uuu
-10 -8 -6 -4 -2 0 2 4 6 8 10-100
-80
-60
-40
-20
0
20
40
60
80
100
x1
x2
4 Feasible Region Fragments
CR001
CR002
CR003
CR004
100
10
65385.8
10
01
115385.01
80769.9
8462.11
0641.00
05128.00
100
5
5.7
10
01
0454545.01
13
0
01
00
100
100
10
5.7
71875.6
65385.8
10
10
01
0454545.01
03125.01
115385.01
5.7
5.5
03333.026667.0
0333.07333.0
100
100
10
5
71875.6
10
10
01
01
03125.01
6667.14
6667.1
0333.1
0333.0
2
1
2
1
2
1
2
1
2
1
2
1
2
1
2
1
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
U
if
if
if
if
On-line Optimization via off-line
Optimization
System
State
Control
Actions
OPTIMIZER
SYSTEM
POP
PARAMETRIC PROFILE
SYSTEM
System
State
Control
Actions
Function Evaluation!
Multi-parametric/Explicit Model
Predictive Control
Compute the optimal sequence of manipulated inputs which minimizes
On-line re-planning: Receding Horizon Control
tracking error = output – reference
subject to constraints on inputs and outputs
Compute the optimal sequence of manipulated inputs which minimizes
On-line re-planning: Receding Horizon Control
Multi-parametric/Explicit Model
Predictive Control
Solve a QP at each time interval
Multi-parametric Programming Approach
State variables Parameters
Control variables Optimization variables
MPC Multi-Parametric Programming
problem
Control variables F(State variables)
Multi-parametric Quadratic Program
Explicit Control Law
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
x1
x2
CR0
CR1
CR2
2065.07083.07059.02
2065.07083.07059.02
2065.0
2065.0
7083.07059.0
7083.07059.08585.68355.6
t
t
tt
t
xif
xif
xifx
u
)( 1,2j 22
0064.0
0609.0
9909.01722.0
0861.07326.0s.t
01.0min))((
||
|||1
|2|2
1
0
2
|||, |1|
t
PtxJ
tttjt
tjttjttjt
tt
T
tt
j
tjttjt
T
tjtuu tttt
xxu
uxx
xxuxx
Multi-parametric Controllers
SYSTEM
Parametric Controller
Optimization Model
(2) Critical Regions
(1) Optimal look-up function
MeasurementsControl Action
Input Disturbances
System Outputs
Explicit Control Law
Eliminate expensive, on-line computations
Valuable insights !
MPC-on-a-chip!
Key milestones-Historical Overview
Number of publications
2002 Automatica paper ~ 580 citations
Multi-parametric programming – until 1992 mostly
analysis & linear models
Multi-parametric/explicit MPC – post-2002 much
wider attention
Multi-Parametric
Programming
Multi-Parametric
MPC &
applications
Pre-1999 >100 0
Post-1999 ~70 250+
AIChE J.,Perspective (2009)
Patented Technology
Improved Process Control
European Patent No EP1399784, 2004
Process Control Using Co-ordinate Space
United States Patent No US7433743, 2008
Outline
Key concepts & historical overview
Recent developments in multi-parametric
programming and mp-MPCModel reduction/approximation
mp-NLP & explicit nonlinear mp-MPC
mp-MILP
Robust explicit mp-MPC
State estimation and mp-MPC
Framework for mp-MPC
A framework for multi-parametric
programming & MPC (Pistikopoulos 2008, 2009)
‘High-Fidelity’ Dynamic Model
Model Reduction Techniques
System Identification
Modelling/ Simulation
Identification/ Approximation
Model-Based Control & Validation
Closed-Loop
Control System Validation
Extraction of Parametric Controllers
u = u ( x(θ) )
‘Approximate Model’
Multi-Parametric Programming (POP)
Model Reduction/Approximation
Replace discrete dynamical
System with a set
of affine algebraic models
N-step ahead prediction-enables use of Linear MPC routines
Model Reduction/Approximation
N-step-ahead approximation based on initial conditions
(measurements) and sequence of controls (constant
control vector parameterization). Set of affine algebraic
models
For all j point over the time horizon - approximations are
constructed as follows
t
y
Direct ApproachStrategy:
Discretize state and controls via Orthogonal Collocation Techniques
Multi-parametric Nonlinear Programming Problem (mp-NLP)
Quadratic Approximation Based
Solve sequence of mp-QP‟s
Nonlinear Sensitivity based
Solve sequence of NLP‟s
Partition state space recursively
Approximate
Multi-parametric Nonlinear Dynamic Optimization Problem
mp-NLP Algorithms for Explicit NMPC
Key features:
• Characterizes the parameter space using NLP sensitivity information and linearization of the constraints.
NLP Sensitivity Based (NMPC mp-NLP)
Quadratic Approximation based (General mp-NLP)
Two implementations for the characterization of the Parameter space
v(x) v0
(x) = 0 (M0)1 N0 + (x x0)+(||x | |)
(x) 0
• Characterizes the parameter space by sub-partitioning CRs where the QA approximation provides “poor” solutions.
x0 v* x0 v*
Validity of approximation:
(x) = O(||x||) (x)/||x|| → 0 as x → 0.
mp-NLP Algorithms for Explicit NMPC
Key Advantage: Fast implementation of the control laws
• State-of-the art multi-parametric solvers (e.g. mp-QP)
• Straightforward characterization of critical regions
• Complexity reduction through region merging
• Extension to address hybrid systems
mp-NLP Algorithms for Explicit NMPC
Decompose mp-MINLP into two sub-problemsStrategy:
Multiparametric Mixed-Integer Nonlinear Programming
Primal sub-problem (mp-NLP)
Master sub-problem (MINLP)
y = y*x = f()
Iterate until master sub-problem is infeasible
Approximate
via mp-QPs
Characterize feasible regionPre-processing Simplicial
Approximation
Step 1
Step 2 mp-MILP
Applications
• Pro-active Scheduling under price, demand and processing time uncertainty (seee poster & paper)
• Explicit Model Predictive Control of Hybrid Systems: Control actions as optimization variables, states as parameters, input and model disturbances as parameters
• Integration of scheduling & MPC
Explicit Solution of the general mp-MILP Problem
Hybrid Approach - Two-Stage Method for
mp-MILP1
1 Wittmann-Hohlbein, Pistikopoulos (2011)
Stage 1 – Reformulation
Partially robust RIM-mp-MILP* model;
Solutions are immunized against all immeasurable parameters and complicating
constraint matrix uncertainty
Stage 2 – Solution
Suitable multi-parametric programming algorithms (e.g. Faisca et al. (2009))
Optimal partially robust solution; Upper bound on optimal objective function value
*objective function coefficient and
right hand side vector uncertainty
Global Optimization of mp-MILP1
Challenges in Global Optimization of mp-MILP Problems:
• Comparison of parametric profiles, not scalar values
• High computational requirements
Multi-Parametric Global Optimization:
• Adaptation of strategies from the deterministic case to multi-parametric framework: Parametric B&B procedure
• Globally optimal solution is a piecewise affine function over polyhedral convex critical regions
Can we find “good solutions” of an mp-MILP problem with less effort?
1 Wittmann-Hohlbein, Pistikopoulos; JOGO, submitted , 2011
Constraint matrix uncertainty poses major challenge mp-MINLP
33
Robust Explicit mp-MPC Famous control problem: Dynamic Systems with Model Uncertainties
(Mayne, Rawlings, Rao & Scokaert, 2000)
10
inputs control :
states system :
NuuU
u
x
:
:
,
0,0,
0,0,
ijijijijij
ijijijijij
mnij
nnij
bbbbb
aaaaa
bBaA
RR
Parametric Uncertain System
Uncertainty due to modelling, identification errors,
measurement errors etc.
Constraints represent safety, operational constraints
It is very critical that the system does not violate them
Immunize against uncertainty
xx
uuu
Mx
dDuCx
Bu + WθkAxx
PxxRuuQxxxV
N
kk
kkk
N
k
NNkkkkU
0
maxmin
1
1
0
)(min)(
maxmin
Exogenous Disturbance
Robust Explicit mp-MPC
Robustification – robust reformulation step (Ben-Tal &
Nemirovski, 2000; Floudas& Co-workers, 2004-2007)
Dynamic Programming framework to Robust MPC
Novel Multi-parametric Programming algorithm to constrained Dynamic Programming (Faísca, Kouramas,
Saraiva, Rustem & Pistikopoulos, 2008) Small mp-QP at each stage
No need for global optimization
Main idea:
Step 1. Formulate the dynamics that govern the estimation error
Step 2. Use these dynamics to find the set that bounds the estimation error
Step 3. Incorporate the bounding set into the controller to „robustify‟ against the estimation error
MHE & mp-MPC
Te S
1 1,T T Te f e w
0
12 2 2
,0 0
1
1
* * *0 0 0
1
0
min
s.t. (actual system),
(
ˆ ˆ (estimated system step 1.3),
nominal system),
ˆ , ,
MPC MPC
MPC MPC MPCMPCk
MPC
k k k
N N
N k kQ RPx uk k
k k k k
k k k
N f
k
x x u
x Ax Bu Gw
x Ax Bu
u u K x x
x Ax Bu t
u K x
x
,U X=
X=
S
0 0
1 1
ˆ, 1... 1, , ,
ˆ ˆis mRPI of .
XMPC
k k k k t
k N x x
x x A BK x x
X
S
S S Ex S S
Moving Horizon Estimation (MHE)
Model-based state estimator
Obtains current state estimate xT
Main advantage: incorporates system constraints
MHE is dual to MPC: backwards MPC
1 11 1
| |
122 2 21 2
| | |ˆˆ ,
1
* *
| 1| 1 1| 1 1| 1
ˆ ˆ ˆ ˆmin
ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆs.t. , , ,
ˆ ˆ (
,
T N T T N T
T TT T
T N T T N T T N T N T T N k kQ RPx Wk T N k T N
k k k k k k k k k k
T N T T N T T N T T N T
x x Y x cbU w v
x Ax Bu Gw y Cx v x w v
x Ax Bu Gw
X W W
WO
smoothed update of arrival cost)
A framework for multi-parametric
programming & MPC (Pistikopoulos 2008, 2009)
‘High-Fidelity’ Dynamic Model
Model Reduction Techniques
System Identification
Modelling/ Simulation
Identification/ Approximation
Model-Based Control & Validation
Closed-Loop
Control System Validation
Extraction of Parametric Controllers
u = u ( x(θ) )
‘Approximate Model’
Multi-Parametric Programming (POP)
‘High-Fidelity’ Dynamic Model
Model Reduction Techniques
System Identification
Modelling/ Simulation
Identification/ Approximation
Model-Based Control & Validation
Closed-Loop
Control System Validation
Extraction of Parametric Controllers
u = u ( x(θ) )
‘Approximate Model’
Multi-Parametric Programming
(POP)
REAL SYSTEM EMBEDDED CONTROLLEROn-line Embedded
Control:
Off-line Robust Explicit Control Design:
A framework for multi-parametric
programming and MPC (Pistikopoulos 2010)
Outline
Key concepts & historical overview
Recent developments in multi-parametric
programming and mp-MPC
MPC-on-a-chip applications PSA system
Fuel Cell system
Biomedical systems
PSA system and the cycle
Impurities
BED 1 FEED DEP 1 DEP 2 DEP 3 Bd Pu PE 1 PE 2 REPRES
BED 2 PE 2 REPRES FEED DEP 1 DEP 2 DEP 3 Bd Pu PE 1
BED 3 Bd Pu PE 1 PE 2 REPRES FEED DEP 1 DEP 2 DEP 3
BED 4 DEP 1 DEP 2 DEP 3 Bd Pu PE 1 PE 2 REPRES FEED
Time
FeedOff gas
Pure product
REPRES
DEP 1
DEP 2
DEP 3
A framework for multi-parametric
programming and mp-MPC for PSA
‘High Fidelity’ PSA Model (PDAE)
Extraction of explicit MPC controllers u = u(x(θ))
System Identification
‘Approximate’ Model
Multi-Parametric Programming
In-silico closed loop controller validation
Modeling & Simulation
Model Based Control & Validation
MATLABPOP Toolbox
Modelling - internal Bed
Mass balance
2
2
)1())1((Z
iC
DZibt
iQ
bpZ
iUC
t
iC
pbb
Radial effects neglected
Transport properties independent of state variables
Axial mass dispersion (Wakao and Funazkri, 1978), velocity dependent
neglected
Species Accumulation
Bulk fluid convection
Mass transfer with adsorbent
Dispersion in axial direction
Energy balance
Energy accumulation in gas phase Energy accumulation in solid phase
Energy convection Heat of adsorption Heat dispersion
Lumped energy balance on gas and solid phase
Radial effects neglected
Specific heat, transport properties independent of state variables
Axial mass dispersion (Wakao et.al., 1978), velocity dependent
neglected
2
2
)()1())1((
)1()1())1((
111
11
Z
T
iH
t
iQ
pbt
iC
RTpbbZ
TU
t
T
pC spbt
T
iQpbt
T
pbb
NCOMP
i
NCOMP
i
NCOMP
i
ip
NCOMP
i
v
NCOMP
i
iv
CC
CCC
i
ii
Energy accumulation in adsorbed phase
Momentum balance & adsorption
characteristics
UU
pd
ii
MWi
C
U
pdZ
P
NCOMP
31
175.1
23
21150
iiLDFi QQK
t
Qi
*
RT
HKK i
i iexp
ia
ii
Q
iQ
RTiCiKia
iQ
iQ NCOMP
1max
*
1max
* Nitta et.al. (1984), Ribeiro et.al. (2008), multisite Langmuir adsorption isotherm (multi-component mixture)
LDF Rate expression
Ergun‟s equation, steady
state pressure drop
Valve Equation (for boundary
conditions)
v
p
critical
High
V
critical
High
LowLowHigh
V
C
CP
OtherwiseP
PC
PP
Pif
P
PPC
U
1
2
1
2
1
P = PHigh if gas leaving the bed
= PLow if gas entering the bed
Chou and Huang (1994), Nilchan and Pantelides (1998)
•Prictical constant since Cp
and Cv are assumed constant
•For REPRES and DEP Cp
and Cv calculated
at yH2= 0.7, yCH4
= 0.3
•For blowdown and purge
(off gas) Cp and Cv calculated
at yH2= 0.5, yCH4
= 0.5
Constraints - Boundary conditionsZ = 0 Z = L
),( Q FEEDTFEEDPSLPMUA PRODUCTPP
RT
iFeedPY
iC 0
Z
iC
FeedTT 0
Z
TZ = 0
Z = L
Z = 0 Z = L
i iC
CODEPi iCUU
PURGEVCCODEPPPURGEPValvefU ,,
iiCODEPCRT
iCODEPPC
iC
CODEPTT
0
Z
iC
0
Z
T
Z = 0
Z = L
Feed
Step
Purge
Step
A boundary condition for each process step
Base case system
BED 1 FEED DEP 1 DEP 2 DEP 3 Bd Pu PE 1 PE 2 REPRES
BED 2 PE 2 REPRES FEED DEP 1 DEP 2 DEP 3 Bd Pu PE 1
BED 3 Bd Pu PE 1 PE 2 REPRES FEED DEP 1 DEP 2 DEP 3
BED 4 DEP 1 DEP 2 DEP 3 Bd Pu PE 1 PE 2 REPRES FEED
FeedOff gas
99.99 % H2
REPRES
DEP 1
DEP 2
DEP 3
Number of Beds 4 AdsorbentActivated
Carbon
Feed pressure 7 bars Bed length 1 m
Blowdown
pressure
1.01325
barsBed diameter 0.12 m
Bed Porosity 0.4Feed
temperature303.15 K
Feed
Composition
70 % H2,
30 %CH4
Feed flow rate 8.0 SLPM
REPRESPEPEPuBdDEPDEPDEPFEED ttttttttt 21321 Adsorption time
Objective and process variables Changes in adsorption time effects purity the
most
Adsorption time – Manipulated variable
Purity – Controlled variable
Fast tracking of H2 purity to the set point
99.99%
Regulate changes in adsorption time
Avoid bed saturation
Avoid high fluid inlet velocities as it causes
mechanical damage
Hard constraints on adsorption time has to be
satisfied for safe and economical operation
A framework for multi-parametric
programming and MPC
‘High Fidelity’ PSA Model (PDAE)
Extraction of explicit MPC controllers u = u(x(θ))
System Identification
‘Approximate’ Model
Multi-Parametric Programming
In-silico closed loop controller validation
Modeling & Simulation
Model Based Control & Validation
MATLABPOP Toolbox
System Identification - Approximation
PDAE model not suitable for current model
based control approaches
Process model approximations are
needed
Input – Adsorption time
Output – H2 purity
Sampling time – 1 PSA cycle
Input signal design for system perturbation
Random pulse employed for persistent
excitation
Maximum amplitude decided by hit and
trial studies
Pulse duration (constant) calculation
based on closed loop response
System identification
Model fit to the input
output data above by an
8th order state space system
kk
kkk
Cxy
BuAxx
1
A framework for multi-parametric
programming and MPC
‘High Fidelity’ PSA Model (PDAE)
Extraction of explicit MPC controllers u = u(x(θ))
System Identification
‘Approximate’ Model
Multi-Parametric Programming
In-silico closed loop controller validation
Modeling & Simulation
Model Based Control & Validation
MATLABPOP Toolbox
MPC Formulation for PSA
1
..
min
1
1
0
'1
1
'
k
highklow
mismatchkk
kkk
k
M
kk
R
kk
N
k
R
kku
y
uuu
yCxy
BuAxx
ts
uRuyyQyyZ y = hydrogen purity at the end of adsorption stage
u = adsorption time, sec
N = 4, M = 2 Q = 1
2 optimization variables u0, u1
Optimal R based on the closed loop response
Constraints on u
Low u: low adsorption time/cycle time, fast PSA cycles
More ON/OFFs of the switch valves per unit time
Extra wear and tear of manipulative variable hardware
Fast loading-unloading of adsorbent leading to its degradation
High u: high adsorption time/cycle time, long PSA cycles
Risk of over saturation, or irreversible adsorption of adsorbent
(1) Critical Regions(2) Optimal Look-up Function
Measurements
Control Action
Input Disturbances
System Outputs
EXPLICIT/MULTI-PARAMETRIC MPC CONTROLLER
MPC on a chip
Explicit Control Law Eliminate expensive, on-line computations
Valuable insights!
mp-MPC for PSA control
Explicit/Multi-Parametric MPC Design
Critical Regions from POP software
Solve the mp-optimization problem for all values of the parameters to obtain the
explicit control laws (u = D1x + u0) and the corresponding critical region
maps (D2x.≤ q).
A framework for multi-parametric
programming and MPC
‘High Fidelity’ PSA Model (PDAE)
Extraction of explicit MPC controllers u = u(x(θ))
System Identification
‘Approximate’ Model
Multi-Parametric Programming
In-silico closed loop controller validation
Modeling & Simulation
Model Based Control & Validation
MATLABPOP Toolbox
MPC Vs PID
Step Disturbance in PSA feed rate – 10 % of Design
ControllerResponse time
(Cycles)
Average ∆U
(Seconds)Maximum ∆U (Seconds)
mp-MPC 13 0.74 1.8
PID 25 0.84 5.09
Impulse Disturbance in PSA feed rate – 35 % of Design
mp-MPC 7 0.75 1.6
PID 5 4.72 12.12
Open Loop 9
Impulse Disturbance in PSA feed rate – 54 % of Design
mp-MPC 7 1.77 4.18
PID1 4 17.11 32.29
PID2 5 9.44 21.16
Open Loop 10
MPC Vs PID
Outline
Key concepts & historical overview
Recent developments in multi-parametric
programming and mp-MPC
MPC-on-a-chip applications PSA system
Fuel Cell system
Biomedical systems
PEM Fuel Cell System
PI
PI
PI
H2O
Water
MassFlow
MassFlow
MassFlow
TE
TE
TE
PT
A
K
PDT
PTTE
TE PT
TE PT
M
TE TE
PT
VENT
VENT
Hydrator
Hydrator
RadiatorFilter
Electronic
Load
N2
H2
Air
PEM Fuel Cell System
Develop 1kW PEM fuel cell system
Collect data for the PEM fuel cell, fan, hydrogen storage
Design controller for the integrated system
mair
Vfan
mcool
PEM Fuel Cell System
Tst
λO2
Tamb
Ist
u: mair,Vfan, mcool
d: Tamb,Ist
y: Tst ,λO2
θ: xt , Tamb,Ist , Tst ,Tst,sp
PEM Fuel Cell System - Controller Design
Nominal MPC Controller
1
0
1
1,,
)()()()()()(minM
k
R
kk
TR
k
R
NN
R
NN
N
k
R
kkk
TR
kkuyx
uuRuuyyPyyyyQRyyJ
Subject to: 1
1
t t t
t t
x Ax Bu
y Cx
Optimized PID Controller
Robust MPC Controller Include in the controller design the model error
u: mair,Vfan, mcool
d: Tamb,Ist
y: Tst ,λO2
θ: xt , Tamb,Ist , Tst ,Tst,sp
PEM fuel cell system
Dynamic model
Ideal and uniformly distributed gases
The fuel and the oxidant are humidified
No liquid can go into the membrane because it is waterproof
Uniform temperature in the fuel cell stack
Simplified mathematical models for humidifier, radiator and pump
Controller evaluation (closed-loop simulation)
Incorporate controller into high fidelity model and perform computational
studies
MATLAB+
POP Software
yt
ut+1ut+1
yt
Incorporate controller into the PEM Fuel Cell System - perform experiments
mair
mcool
PEM Fuel Cell System
Tst
λO2
Tamb
Ist
Unit Specifications
Fuel Cell : 1.2kW
Anode Flow : 5..10 lt/min
Cathode Flow : 8..16 lt/min
Operating Temperature : 65 – 75 °C
Ambient Pressure
Control Strategy
Start-up Operation
Heat-up Stage : Control of coolant loop
Nominal Operation
Control Variables :
Mass Flow Rate of Hydrogen & Air
Humidity via Hydrators temperature
Cooling system via pump regulation
Known Disturbance : Current
(2) Critical Regions
(1) Optimal look-up function
PEM Fuel Cell System
mH2
mAir
mcool
TYHydrators
Vfan
Tst
HTst
PEM Fuel Cell System
Outline
Key concepts & historical overview
Recent developments in multi-parametric
programming and mp-MPC
MPC-on-a-chip applications PSA system
Fuel Cell system
Biomedical systems
Type 1 diabetesMaintain blood glucose concentration within the normal range
by optimising insulin delivery
Model predictive control problem
Hyper
Normal
Hypo
AnaesthesiaProvide hypnosis, analgesia and muscle relaxation while
maintaining the vital functions
Multiple input multiple output model predictive control
Acute Myeloid LeukaemiaProvide optimal chemotherapy dose to minimise the cancer cells
While keeping normal cells above a minimum level
Scheduling Problem
Muscle Relaxation
ERC MOBILEDevelopment of models and model based control and optimisation algorithms for biomedical systems
Individual Patient
High Fidelity Model
Model Reduction
Set points
Individual constraints
Patient
Measurement
device/ State Estimation
OPTIMAL DRUG DELIVERY/DOSAGE
MEASUREMENTS/
STATES
Disturbances
Model Development
mp-MPC
Model Predictive Control/ Optimisation
-37 -36 -35 -34 -3313.5
14
14.5
15
15.5
16
16.5
x1
x 2
Optimal control
law/ trajectory
Optimal Scheduling
mp-MPC on a Chip
Optimal dose
1st Cycle 2nd Cycle 3rd Cycle 4th Cycle
Cancer cells
Normal cells
drug 1
drug 2
Framework towards optimal drug delivery systems
Mathematical Modelling
Model Development
Individual Patient
High Fidelity Model
Ve
in
Arte
ry
Skin
Lung
Brain
Heart
Muscle
Skeleton
Liver
Kidney
Adipose
Gut
Spleen
Pancreas
Pharmacokinetics
Cell Cycle
Pharmacodynamics
Effe
ct
Efficacy
Individual variability
Potency
Concentration
C50
E0
Emax
E50
0 10 20 30 40 50 60 7040
60
80
100
120
140
160
180
200
220
240
gluc
ose(
mg/
dl)
time(hr)
day2 day3
35g 100g 35g 40g 90g 50g 50g 80g 20g
day1
Diabetes Type I
Anaesthesia
Leukaemia
Glucose Profile
Anaesthetic concentrations
Cell population profiles
Step 1: The sensor measures the glucose concentration from
the patient
Step 2: The sensor then inputs the data to the controller which analyses it and implements the
algorithm
Step 3: After analyzing the data the controller then signals
the pump to carry out the required action
Step 4: The Insulin Pump delivers the required dose to
the patient intravenously
Controller
Sensor
Patient
Insulin Pump
12
3 4
ERC MOBILE
Outline
Key concepts & historical overview
Recent developments in multi-parametric
programming and mp-MPC
MPC-on-a-chip applications
Concluding remarks & future outlook
MPC-on-a-chip technology –
Reflections (10 years since 2002 Automatica paper appeared .. )
Scientific/academic impact ?
Application/industrial impact ?
MPC-on-a-chip technology –
Reflections (10 years since 2002 Automatica paper appeared .. )
Scientific/academic impact ? HIGH – many
un-resolved issues ..
Application/industrial impact ? Limited –
not panacea to all MPC solutions ..
MPC-on-a-chip – Perspectives
Application types for Multi-parametric
Programming & MPC
Type 1 - Large scale and expensive industrial
processes with slow/medium dynamics
Type 2 - Medium scale and cost industrial
processes with medium/fast dynamics
Type 3 - Small scale and inexpensive
processes/equipment with medium/fast dynamics
MPC-on-a-chip – Perspectives
Type 1 – Large scale and expensive
industrial processes with slow/medium
dynamics
MPC-on-a-chip – Perspectives
Type 1 - Large scale and expensive
industrial processes with slow/medium
dynamics
Control hardware/software availability
MPC implementation mainly via online
optimization
Explicit MPC can play a role for low level
process control
Hybrid (on-line + off-line) approach possible –accelerate on-line dynamic optimization step
MPC-on-a-chip – Perspectives
Type 2 – medium scale and cost
industrial processes with medium/fast
dynamics
HEX
Column
LIN
Air
Product - GAN
Waste
Reboiler/condenser
MPC-on-a-chip – Perspectives
Type 2 – medium scale and cost
industrial processes with medium/fast
dynamics
Limited Control hardware/software availability
Online optimization/MPC usually prohibitive
Multi-parametric MPC ideal – proved in
previous applications (Air Separation,
Automotive)
MPC-on-a-chip – Perspectives
Type 3 – small scale and inexpensive
processes/equipment with medium/fast
dynamics
Patient
MPC-on-a-ChipMechanical Pump
Glucose Sensor
MPC-on-a-chip – Perspectives
Type 3 – small scale and inexpensive
processes/equipment with medium/fast
dynamics
Available control hardware/software limited -
not suitable for online MPC
Multi-parametric MPC technology ideal/
essential
MPC-on-a-Chip part of embedded (all-in-one)
system
Suitable for new technologies (FPGA, wireless)
From Multi-Parametric
Programming Theory to
MPC-on-a-chip Multi-scale
Systems Applications
Stratos Pistikopoulos
FOCAPO 2012 / CPC VIII