probral kl july2007 - engenharia elétrica · projects: sapien , ct -energ and finep -inova energy...
TRANSCRIPT
International Research CooperationPROBRAL 2007-2008 CAPES/DAAD
University of Brasília/Brazil – University of Kaiserslautern/Germany
“Networked Control with Distributed Processing for Building Automation
in an Ambient Intelligence Framework”
Adolfo BauchspiessAdolfo Bauchspiess
KaiserslauternKaiserslautern, 23.7.2006, 23.7.2006
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
ContentsContents
�� Brazil Brazil -- BrasBrasíília lia –– UnBUnB
�� PROBRALPROBRAL
�� Some on going UnBSome on going UnB--ProjectsProjects
�� Energy Saving with Fuzzy Distributed ControlEnergy Saving with Fuzzy Distributed Control
�� Wire Less Building AutomationWire Less Building Automation
�� Conclusions & PerspectivesConclusions & Perspectives
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
BRAZILBRAZIL•• PelPeléé -- SoccerSoccer•• CarnavalCarnaval –– SambaSamba•• Rio de JaneiroRio de Janeiro
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
BrazilBrazil
Brasília15°45´S47°52´W
~ 8.5 Mio Km2
~ 190 Mio Inh.~ 8.000Km Coast
BRInC
Equator
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
BrasBrasíílialia –– Capital City of Capital City of BrazilBrazil• 1960 – President J. Kubitschek, Architekt O. Niemeyer • 1.000m over the see level• ~ 2 Million Inhabitants (Federal District)• Highest income rate in south america• Paranoá Lake
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
UniversidadeUniversidade de de BrasBrasíílialia(Oscar Niemeyer, 1961)(Oscar Niemeyer, 1961)
1Km – Instituto Central de Ciências
FT
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
TechnologicalTechnological FacultyFaculty
4 4 DepartmentsDepartments•• ElectricalElectricalEng. Eng. -- ~200 ~200 Beginners/YearBeginners/Year•• MechanicalMechanicalEng.Eng.•• Civil Eng. Civil Eng. •• Forest Eng.Forest Eng.
6 6 DegreesDegrees••ElectricalElectricalEng.Eng.••MechatronicMechatronicEng.Eng.••CommunicationCommunicationNetworkNetworkEng.Eng.••MechanicalMechanicalEng.Eng.••Civil Eng.Civil Eng.••Forest Eng.Forest Eng.
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
ElectricalElectrical EngineeringEngineering DepartmentDepartment
GRAVGRAVAssociateAssociateProf. Dr. Adolfo Bauchspiess Prof. Dr. Adolfo Bauchspiess AdjointAdjoint Prof. Dr. Prof. Dr. GeovanyGeovanyde Arade Araúújo Borgesjo BorgesAdjointAdjoint Prof. Dr. João Prof. Dr. João YoshiyukiYoshiyukiIshiharaIshiharaAdjointAdjoint Prof. Dr. Marco A. F. Egito CoelhoProf. Dr. Marco A. F. Egito Coelho
Areas organized in research groups,GRAV – Grupo de Robótica, Automação e Visão Computacional
LAVSI – Laboratório de Automação, Visão e Sistemas InteligentesLARA – Laboratório de Robótica e Automação
43+ 43+ lecturerslecturersin 5 in 5 areasareas: : ControlControlandandAutomationAutomation, , TelecomTelecom, , ElectronicsElectronics, , PowerPowerSystemsSystems&& CommunicationCommunicationNetworksNetworks..
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
PROBRALPROBRALPROPROgrama de Cooperagrama de Cooperaççãoão Internacional Internacional
BRBRasilasil--ALALemanhaemanha
�� Only supports Interchange of ResearchersOnly supports Interchange of Researchers
�� Each side must have local research ($) supportEach side must have local research ($) support
�� ~ 2 Profs. + 2 Ph.D candidates / Year~ 2 Profs. + 2 Ph.D candidates / Year
�� 2 (+ 1) Years2 (+ 1) Years
“Networked Controlwith Distributed Processingfor Building Automation
in an Ambient IntelligenceFramework”
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Some onSome on--going UnBgoing UnB--ProjectsProjects
�� LEARnLEARn
�� ExpansionExpansion--CARCARAHCARCARAH
�� ......
�� Related with PROBRAL:Related with PROBRAL:
•• SAPIEnSAPIEn
•• CTCT--EnergEnerg
•• FINEPFINEP--INOVAINOVA
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
LEARnLEARnLaboratório de Ensino de Automação Remota.
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
UAV Inspection of Power UAV Inspection of Power Transmission LinesTransmission Lines
ProjectProject CarcarahCarcarah
--Expansion Expansion LtdaLtda
--ANEELANEEL
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Projects: Projects: SAPIEnSAPIEn, CT, CT--EnergEnergand FINEPand FINEP--INOVAINOVA
EnergyEnergy--Saving Approach:Saving Approach:
ModelModel--Based HVAC ControlBased HVAC Control
( )
444 3444 21444 3444 21relatedenergyrelatedcomfort
h
iu
Tu
Th
iR
cp
ikikikikyikyJ ∑∑−
=∆
=
++++∆+∆+−+=1
01
2 )()()()()( uQuuQu
Predictive Cost Function:Predictive Cost Function:
matricesweighting
iablevardmanipulateu
referencey
iablevarcontrolledy
horizoncontrolh
horizonpredictionh
Where
uu
R
c
p
−−
−−−
−
∆ QQ ,
:
Considers comfort and energy saving. Needs model!Considers comfort and energy saving. Needs model!
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
DistributedDistributed ParameterParameter ModellingModelling
�� Transport PhenomenaTransport Phenomena
�� Ex. Heat FlowEx. Heat Flow
Wind velocity simulation in Modelica, by Felgner, ASIM2002
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
DynamicDynamic SystemSystem IdentificationIdentification
u - input (manipulated variable)y – output (controlled variable)w – measurable disturbancev – non-measurable disturbance
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
FirstFirst--Principles Identification (KLPrinciples Identification (KL--2006)2006)
15 20 25 30 35 40 45
0
5
10
15
20
t/hours
Identification data set of March 29-31, 2006
h on/off
-c on/off
Gs/50 W/m2Tw/°C
To/°C
Typical identification data set
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
a) Linear MISO approacha) Linear MISO approach
0 105
0
5
10
15
20
25Heat
(sec)
0 5 10 155
-300
-250
-200
-150
-100
-50
0Cool
(sec)
0 5 10 155
0
0.2
0.4
0.6
0.8
1
1.2
1.4Outside Temp
(sec)
x 105x 105 x 105
12 14 16 18 20 22 24 26 28 30 3217
18
19
20
21
22
23
24
t/hours
u=[Th Tc Tw Tv Gs] 6th order MISO model
To
y
DuCx
EeBuAxx
+=++=
y
&
-1 0 1
-1
-0.5
0
0.5
1
From Th
-1 0 1
-1
0
1
From Tc
-1 0 1
-1
0
1
From Tw
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
b) b) FirstFirst--Principles Structured Principles Structured IdentificationIdentification
2R1C analogy of the vicinity heat transfer 2R1C-based disturbance model
)1)(1(
)(.
)1)(1(
)(.
)1)(1(
)(.
+++
+++
++=
saas
sTsaK
saas
sTsaK
saas
sGsaKq
w
ww
v
vv
r
srdHeat flow due to disturbances:
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
FirstFirst--principles structured thermal modelprinciples structured thermal model
^
qh
Tv
Vicinity
S2
S1
1/C
RoomThermal
Capacitance
Gs
Radiation
Tw
Outdoor
1s
Intg
55 Heater Temp.
h
Heat
aw.s+1
1
Gw
av.s+1
1
Gv
Kr
av.s+1
Gr
ah.s+1
1
Gh
ac.s+1
1
Gc
Kc
Kh
Kw
Kv
c
Cool
4 Chiller Temp.
Tc
Th
qdTiv qv
Tiw
qr
qw
qc
qu
Tih
Tic
To
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Heat flow obtained with the Heat flow obtained with the firstfirst--principles structured identificationprinciples structured identification
12 14 16 18 20 22 24 26 28 30 32-6
-4
-2
0
2
4
6x 10
-3
t/hours
Heat flow parcels (assuming C=1)
qr
qv
qwqh
qc
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
12 14 16 18 20 22 24 26 28 30 32
18
20
22
24Results - Training data - March 29-31, 2006
12 14 16 18 20 22 24 26 28 30 32
18
20
22
24
Tc/°C - Tc est/°C - Validation - March 24-27, 2006
12 14 16 18 20 22 24 26 28 30 32
0
5
10
15
20Heat, Cool, Weather and Solar Radiation - March 24-27, 2006
Tc
Tcest
Tc
Tcest
h
-c
Ts
Gs*0.05
Results of the first-principles structured identification
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
PROBRALPROBRAL--RelatedRelated
SAPIEnSAPIEn––EnergyEnergySavingSavingbyby FuzzyFuzzyDistributedDistributedControlControl
�� EnergyEnergy SavingSaving
�� ThermalThermal ComfortComfort
�� BuildingBuilding AutomationAutomation
�� FAPFAP--DF DF suportsuport
Campus UnB (2007) 995 window air conditioners Campus UnB (2007) 995 window air conditioners >R$1.000.000,00/Year (1>R$1.000.000,00/Year (1€€ ≈≈ 2,56 R$)2,56 R$)
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Experimental Experimental EnvironmentEnvironment
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
INTERFACE GRINTERFACE GRÁÁFICA FICA ActionViewActionView®®
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
TELA DE ALARMETELA DE ALARME
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
SIMULATIONSIMULATIONSimulinkSimulink®®
MIMO process:MIMO process:
�� 4 rooms4 rooms
�� 3 air conditioners3 air conditioners
ref sq1
ref sq
T_ext2
T_ext1
T_v ic1
T_v ic2
T_v ic3
T4
Room 4
extern disturb
T_v ic2
T_v ic4
Cool_on_of f
Tc3
T3
Room 3
extern disturb
T_v ic1
T_v ic3
T_v ic4
Cool_on_of f
Tc2
T2
Room 2
extern disturb
T_v ic4
T_v ic2
Cool_on_of f
Tc1
T1
Room 1
Rad_S2
Rad_S1
OPOper.
Point1
OPOper.Point
-1
62.18
577.4
KWh / ISE
ref
dif Ext
y
PWM
KWh/ISE
Fuzzy-I AWUpCtrl3
ref
dif Ext
y
PWM
KWh/ISE
Fuzzy-I AWUpCtrl2
ref
dif Ext
y
PWM
KWh/ISE
Fuzzy-I AWUpCtrl1
Fuzzy
ref
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Room ModelRoom Model
2
T2
1
Tc2
S2
1/C2
RoomThermal
Capacitance
1s
Intg
Tw Tiw
Gw2
Tf Tiv
Gv3
Tf Tiv
Gv2
Tf Tiv
Gv1
Gss qr
Gr2
Gss qr
Gc2
Kc
Kw
Kv
Kv
Kv
4
Chil ler Temp.
5
Cool_on_off
4
T_vic4
3
T_vic3
2
T_vic1
1
externdisturb
Tc
qd
Tiv
Tiw
qr
qcTic
T2
Tiv
qv
qw
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
ControllerController
2
KWh/ISE
1
PWM
==~
predialSpin
Switch
S2
Prod.
InOut
Pass if Work Hour1
InOut
Pass if Work Hour PWM1
-K-
Ki11s
Integ.
1s
I0
3
y
2
difExt
1
ref
e
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Fuzzy Rule BasisFuzzy Rule Basis
1) IF (erro = p) AND (setpoint = set) AND (Difext = d ) THEN (output = p)
2) IF (erro = n) THEN (output = zero)
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
SimulationSimulationResultsResults
500 600 700 800 900 1000 1100 1200 130020
22
24
26
28
°C
Room Temperatures - Air conditioner's internal on-off control
500 600 700 800 900 1000 1100 1200 130023
24
25
26
27
28
°C
Room Temperatures - MIMO Fuzzy-I Control
500 600 700 800 900 1000 1100 1200 13000
10
20
30
°C
PWM heat flow signal - MIMO Fuzzy-I Control
T1
T2
T3T4
"ref"
T1
T2
T3T4
ref
Tc1
Tc2Tc3
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
DistributedDistributed buildingbuilding controllercontroller PrototypePrototypeProgrammable Logic Controller
KMC 7000 (~R$300,-)
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Developing Room
On-Off 16-09-2006
Fuzzy Control 14-09-2006
22,0024,00
26,0028,00
30,0032,00
34,0036,00
38,0008
:00
08:3
5
09:1
0
09:4
5
10:2
0
11:0
0
12:0
5
12:4
0
13:1
5
13:5
0
14:2
5
15:0
0
15:3
5
16:1
0
16:4
5
17:2
0
17:5
5
ºC
Setpoint
T. Interna
T. Externa
Inf
Sup
22,00
24,0026,00
28,00
30,0032,00
34,00
36,0038,00
40,00
08:0
0
08:3
5
09:1
0
09:4
5
10:2
0
11:0
0
12:0
5
12:4
0
13:1
5
13:5
0
14:2
5
15:0
0
15:3
5
16:1
0
16:4
5
17:2
0
17:5
5
ºC
S etpoint
T. Interna
T. Ex terna
Inf
S up~30% saving!
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Measured and Reconstructed Signals Measured and Reconstructed Signals -- MultiMulti--Room Experiment (June 2007)Room Experiment (June 2007)
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Intelligent Building Automation Intelligent Building Automation Wire Less Prototype Wire Less Prototype –– LAVSILAVSI--LARA/ENELARA/ENE--UnBUnB
Singular Singular SystemsSystems((DescriptorDescriptor SystemsSystems))
João João YoshiyukiYoshiyuki IshiharaIshihara
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
OutlineOutline
�� IntroductionIntroduction
�� WhatWhat is a singular is a singular systemsystem((descriptordescriptor systemsystem)?)?
�� ExamplesExamples of of descriptordescriptor modelingmodeling
�� Some Some resultsresults: : stabilitystability, , filteringfiltering
�� FutureFuture workwork relatedrelated to PROBRALto PROBRAL
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
WhatWhat is a singular is a singular systemsystem ((descriptordescriptor)?)?
( ) ( ) ( )( ) 0,, =tutxtxf &
( ) ( ) ( )tButAxtxE +=&Linear case
In particular:
( ) ( ) ( )tButAxtx +=&
Implicit equations
Singular equations
Usual state-space equations
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
WhatWhat is a singular is a singular systemsystem ((descriptordescriptor)?)?
( ) ( ) ( )tButAxtxE +=&
Another characterization
Can be re-written as:
( ) ( ) ( )( ) ( )tJutHx
tGutFxtx
+=+=
0
&
Differential-algebraic equations
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
DescriptorDescriptor modelmodel
( )( )( )( )
( )( )( )( )
( )
( ) [ ] ( )txty
tv
tv
tv
tv
tI
R
C
tv
tv
tv
tIL
S
R
C
L
R
C
L
0100
1
0
0
0
1110
100
000/1
0010
0000
0000
0100
000
=
+
−=
&
&
&
&
( ) ( )( ) ( )
( ) ( )( ) ( ) ( ) ( )tvtvtvtv
tvtRI
tICdt
tdv
tvdt
tdIL
SRCL
R
C
L
+++=+−=
=
=
0
0
1
RLC circuit equations
descriptor equations
( ) ( ) ( )tButAxtxE +=&
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
ExampleExample: : RobotRobot withwith RestrictedRestricted MotionMotion
Minimização de energia para robôs com restrição Trabalho de mestrado no LAC - Aluno: Cauê PeresOrientador: Paulo Sérgio P. da SilvaAno - 2003
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
RobotRobot withwith RestrictedRestricted MotionMotion
uSx
F
FDK
I
xM
IT
+
−−=
0
0
00
00
000
00
00
0
0
000
3
0
3
&
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Example: Perturbation analysis of large systems
( ) ( ) ( )kBukAxkEx +=+1
( ) ( ) ( ) ( )( ) ( ) ( ) ( )kuBkxAkxAkx
kuBkxAkxAkx
22221212
12121111
1
1
++=+++=+
ε
( ) ( ) ( ) ( )( ) ( ) ( )kuBkxAkxA
kuBkxAkxAkx
2222121
12121111
0
1
++=++=+
Large systems: small parameters are set to zero
Simplified model
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Example: interconnected systems
( ) ( ) ( )( ) ( ) ( )kJukHky
kGukFkE
+=+=+
ξξξ
1
( ) ( ) ( )( ) ( ) ( )kwDkxCkz
kwBkxAkx
iiiii
iiiii
+=+=+1
( ) ( ) ( ) ( )( ) ( ) ( ) ( )kRwkQukPzky
kNykMukKzkw
++=++=
Large systems: collection of interconnected subsystems
Overall subsystem equation
( ) ( ) ( )( ) ( ) ( )kDwkCxkz
kBwkAxkx
+=+=+1
Interconnections: overall input u(k) and overall output y(k)
Obs: A state-space representation may not exist for some interconnected systems
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
A system with no state equations
33 RRR iiv +−=
[ ]xy
u
x
xx
C
010
0
1
0
0
0
301
011
100
000
000
00
3
=
=
+
−+
&
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Example 2: system with no state equations
Agricultural production growing curve
Logistic function
3 parts:* Exponential growth* Linear growth* Logaritmic growth
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Example 2: system with no state equations
The agricultural production growing curve is solution of the following singular system
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Transistor model
( ) ( ) ( )
[ ]xRy
i
vxttutxtx
C
E
C
0
,0 ,1
0
01
10
00
0
=
=≥
+
=
&
Impulsive Responses
In the state space formulation we have information loss(Impulsive solution)
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Example: Filtering of non causal system
( ) ( ) ( )( ) ( )kHxkz
kGukFxkEx
=+=+1
( ) ( ) ( ) ( )[ ]TTTT pkzkzpkzkx +−=
=
001000
000100H
T
G
=
100000
010000
+−
+
+−
+
=
=
22
22
1010
10
01
0101
100000
010000
001000
000100
,
000000
000000
001000
000100
000010
000001
FE
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Pros Pros andand Cons for Cons for descriptordescriptor modelingmodeling
�� RiccatiRiccati equationsequations are are relatedrelated to to HamiltonianHamiltonianextendedextended pencilspencils((whichwhich are in are in descriptordescriptor formform))
�� StateState--space systems are not closed space systems are not closed under PID feedbackunder PID feedback
�� Any rational transfer function can be Any rational transfer function can be represented by a descriptor systemrepresented by a descriptor system
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Pros Pros andand Cons for Cons for descriptordescriptor modelingmodeling
�� It is a It is a veryvery natural natural wayway to to modelmodel processprocessdynamicsdynamics..
�� It It refersrefers muchmuch more to more to thethe physicalphysical behaviorbehaviorof of thethe systemsystem andand givesgives more more physicalphysicalinsight.insight.
�� TheThe interpretationinterpretation of of resultsresults is is alsoalso more more simplesimple thanthan in case of in case of thethe more abstract more abstract statestate spacespace modelsmodels..
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
AplicationAplication: : LiquidLiquid LevelLevel ControlControl SystemSystem
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
q
A
Ah
h
RRA
RR
RA
RARA
h
h
+
+−
−=
2
1
2
1
212
21
12
1111
2
1
4
14
3
1
11
&
&
q
q
q
h
h
R-
R-
-
-
q
q
h
h
A
A
+
=
0
0
41
43
010
011
1100
0100
0000
0000
000
000
2
1
2
1
2
1
2
1
2
1
2
1
&
&
&
&
State Space Model
Descriptor Model
AplicationAplication: : LiquidLiquid LevelLevel ControlControl SystemSystem
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
q
q
h
h
R-
R-
-KK
-KK
q
q
h
h
A
A
−−−−
=
2
1
2
1
2
1
24
114
124
314
3
2
1
2
1
2
1
010
011
11
01
0000
0000
000
000
&
&
&
&
s xpp
pp
h
h
RRA
KRR
RA
KRA
KR
RA
KR
h
h
≡
++−−
−+−=
2221
1211
2
1
212
221
12
1
11
21
11
11
2
1
4
44
4
44
34
4
34
&
&
State feedback[ ]
−=−=
2
121 h
hKKxKq ss
[ ][ ]Tdd qqhhKKxKq 212121 00−=−=
LiquidLiquid LevelLevel ControlControl SystemSystem -- ClosedClosed looploop
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
DescriptorDescriptor versionsversions
0,0 ≥=++ PEECCPAEPEA TTTT
Example Stability: Classical Result in the Literature
The extension of state space resultsto descriptor systems are not direct.
0=++ CCPAPA TT
0,0 ≥=++ PCECEPAEPEA TTTT
0,0 ≥==++ EXXECCXAXA TTTT
0,0)()( 00 ≥=++++ PCCAQEPEQEPEA TTT
( ) ( ) ( )( ) ( ) ( )tButAxty
tButAxtx
+=+=
&
( ) ( ) ( )( ) ( ) ( )tButAxty
tButAxtxE
+=+=
&
(Ishihara et al., 2002)
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Kalman type Filter for Singular Systems
11111
11
+++++
++
+=+=
iiiii
iiiiii
vKxHz
wGxFxE
The filtered estimates recursion
[ ][ ]
[ ]
−
= +
−
++
+
+
+
+
+
++++
I
z
x
HE
K
G
I
I
FI
H
E
K
GF
I
I
RS
SQ
IP
I
Px i
ii
Ti
Ti
Ti
Ti
Ti
i
i
i
ii
iTi
ii
iiT
iiii
0
00
0
0ˆ
0
0
0
0
0
0
0
0
ˆ 1
|
1
11
1
1
1
1
1
|
1|11|1
=
++ 11
covi
Ti
ii
i
i
RS
SQ
v
w
(Ishihara et al., 2005)
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Kalman type filter for Singular systems
( ) 11
1111
1
|11|1 ][ −+
−+++
−+++ ++= ii
Tii
Tiiiii
Tiii HRHEFPFQEP
The filtered estimates recursion
( ) 11111|1|
1|11|11|1 ˆˆ +
−++++
−+++++ ++= ii
Tiiiiii
Tiiiii
Tiiiii zRHPxFFPFQEPx
[ ][ ]
[ ]
−
= +
++
+
+
+
+
+
++++
I
z
x
HE
K
G
I
I
FI
H
E
K
GF
I
I
RS
SQ
IP
I
Px i
ii
Ti
Ti
Ti
Ti
Ti
i
i
i
ii
iTi
ii
iiT
iiii
0
00
0
0ˆ
0
0
0
0
0
0
0
0
ˆ 1
|
11
1
1
1
1
1
|
1|11|1
Can be rewritten as
IKIGS iii === ,,0 If
TheThe a priori a priori filterfilter is is notnot equivalentequivalent to to thethe a posteriori a posteriori filterfilter
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
RobustRobust FilteringFiltering
�� SystemSystem withwith uncertaintiesuncertainties ::
iiiii
iiiiiii
vxHHz
wxFFxEE
++=++=+ +++
)(
)()( 111
0000
>
=
Ε
iji
iji
T
i
i
i
i
R
Q
P
v
w
x
v
w
x
δδ
1
,,
1,,1
,,
≤∆
∆=
∆=
∆=
++
i
ihiihi
ieiifi
ifiifi
NMH
NME
NMF
kkkkkk xxx |1|1| ˆ,ˆ ,ˆ −+⇒
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Kalman-type Robust Filter
Robust filter
iiiii
iiiiiii
vxHHz
wxFFxEE
++=++=+ +++
)(
)()( 111
( ) iiifTieiiiif
Tieii
Tiiiii
Tiiiii xNNPINNFFPFQEPx |,1,|,1,
1|11|11|1 ˆ)}ˆˆ](ˆˆˆˆ[{ˆ ++
−+++++ −++=
11111|1
ˆ+
−+++++ ii
Tiii zRHP
( )])ˆ([ˆ
ˆˆˆˆˆ
1,1
,|,1,1,1,
11111
1|1
11|1
+−
+++
+−+++
−+
−++
+++
++=
ieT
ifiiifiT
ieihT
ihi
iiTii
Tiiiii
Tiii
NNPNINNN
HRHEFPFQEP
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Future work related to PROBRALFuture work related to PROBRAL1) 1) DescriptorDescriptor systemssystems withwith MarkovianMarkovian jumpingjumping
( ) ( ) ( )
( ) ( ) iiiii
iiiiii
vKxHz
wGxFxE
θθ
θθθ
+=
+=++ 11
AlreadyAlready donedone: : arrayarray algorithmsalgorithms for for filtersfilters of of MarkovianMarkovian jumpingjumping Linear Linear systemssystems
2) 2) otherother possibilitiespossibilities??
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
Conclusion & PerspectivesConclusion & Perspectives
�� Wire Less Automated Building Wire Less Automated Building as NCS Test Bedas NCS Test Bed
�� Modelling of Thermal ProcessesModelling of Thermal Processes�� Stability Analysis of NCS + Distributed AutomationStability Analysis of NCS + Distributed Automation�� SpecificSpecificHardware x COTSHardware x COTS
for for distributeddistributedautomationautomation
? UnB + KL ?? UnB + KL ?
CAPES/DAAD PROBRAL UnB/KL CAPES/DAAD PROBRAL UnB/KL ®®A.Bauchspiess 23.7.2007A.Bauchspiess 23.7.2007
DankeDanke! Obrigado!! Obrigado!
WirWir freuenfreuenuns uns aufauf diedie ZusammenarbeitZusammenarbeit!!
[email protected]@unb.brwww.www.lavsilavsi..eneene.unb..unb.brbr
www.www.eneene.unb..unb.br/adolfobr/adolfo