A Ch Characteri acteriza zation tion of All ll Retr trof ofit - - PowerPoint PPT Presentation

a ch characteri acteriza zation tion
SMART_READER_LITE
LIVE PREVIEW

A Ch Characteri acteriza zation tion of All ll Retr trof ofit - - PowerPoint PPT Presentation

American erican Contr ntrol ol Conf nfer eren ence ce 2018 18 FrB19 B19 Large-Scal cale e Systems ems A Ch Characteri acteriza zation tion of All ll Retr trof ofit it Co Contr ntroller ollers June e 29t 9th, h, 201 018


slide-1
SLIDE 1

A Ch Characteri acteriza zation tion

  • f All

ll Retr trof

  • fit

it Co Contr ntroller

  • llers

June e 29t 9th, h, 201 018 Hampe mpei Sasa sahar hara, a, Takayuk yuki i Is Ishiza zaki, i, Masak saki i In Inoue, Tomono nori ri Sada damoto moto, , Jun-ic ichi Im Imura

Tokyo te tech., h., Ja Japan

American erican Contr ntrol

  • l Conf

nfer eren ence ce 2018 18 FrB19 B19 Large-Scal cale e Systems ems

slide-2
SLIDE 2
  • 1. Intr

troduc

  • duction

tion 2.

  • 2. Ret

etrof

  • fit

it Co Contr trol

  • l

Outl tline ine

  • De

Definitions itions

  • Ret

etrof

  • fit

t Co Contr trolle

  • ller
  • Performa

rmance nce Analysi sis

  • 3. Nume

meric ical l Exa xamp mple (Power r Syst stem) m)

slide-3
SLIDE 3

[1] M. Rotk tkowi

  • witz

tz and S. Lall ll “A characterization of convex problems in decentralized control,” IEEE Trans

  • ans. Autom
  • m.

. Contr trol,

  • l, 2005.

5.

Co Cont ntrol

  • l of
  • f La

Large-Sc Scale ale Sy Syst stems 1/15

15 Tradition ditionally ally Trea eated ted Difficulty iculty:

spar arsity sity cons nstr traints/ aints/ comm mmunica unication tion delay

informa

  • rmation

tion co constr train aints ts Rough ughly ly,

QI QI

tracta actable le (convex) x)

(quadr uadratic ic invarian riance) e) design sign of su sub-co contr ntroller

  • llers

s in a large-scale scale sy syst stem em

  • sca

calability ity issue

  • model

el unavaila ailabil bility ity

  • exi

xist sten ence ce of multipl tiple e oper erator tors Rem emain aining ng Issues es

slide-4
SLIDE 4

centr ntraliz ized ed sy synthesiz thesizer er

Scal calabili bility ty Iss ssue ue

2/15 15 under der QI QI contr ntroller

  • llers

s are e sy synthesiz thesized ed in a centr traliz alized ed manner ner

no not scala alable le wit ith h the he siz ize of

  • f the

he pla lant nt

des esign gn

How

  • w to
  • resolv

solve e thi his is issue ue?

slide-5
SLIDE 5

Model del Unavaila ailabili bility ty

3/15 15

How w to to synth thesi esize e a c a contr trol

  • ller

ler wi with th th the e ac acce cess

  • nly to

to th the e par artia tial l model el informa

  • rmation?

tion?

entir tire e model: el: unavaila ailable le partia tial model: del: avail ilable le cons nsider ider a differ erent ent situa uation tion

  • too
  • large for model

deling ing

  • secur

curity ity of privac acy

?

des esign gn

slide-6
SLIDE 6

Under der Multi ltiple ple Oper perator tors

4/15 15 su suppose: pose: multiple tiple oper erato tors

  • ther

her oper erator

  • rs

design sign independen pendently tly

  • ther’s control actions may vary

?

How w to to sy synth thesi esize e a co a contr trol

  • ller

ler th that t ca can dea eal wi with th th the e possibly var arying g co contr trol

  • ller

lers?

?

?

slide-7
SLIDE 7

Exa xample: mple: Power er Syst stem em

5/15 15

  • multiple

ltiple ISOs Os (inde ndepend pendent ent system tem oper erator) tor)

?

  • too large

e to capt ptur ure e the entir ire syste stem m model del

  • other

r ISOs Os may chang nge their ir control

  • l polic

licy

you

Objectiv jective: e: impr mprove e the contr trol

  • l perform
  • rman

ance ce by attac achi hing ng (retr trof

  • fitting

tting) ) a contr troll

  • ller

er only y wi with h the parti tial al model del

slide-8
SLIDE 8

Retr etrof

  • fit

it Contr trol

  • l

Ex Existin isting: g: entir ire e system stem model del novel: el: local al model el inf nforma

  • rmation

tion

retrofit it contr ntrol:

  • l: imp

mproving ving loca cal l perf rformance

  • rmance

6/15 15

[1] T. Ishiz izak aki, , et al., ., “Retrofit Control: Localization of Controller Design and Implementation,” Automa

  • matic

tica, 2018. 8.

Ou Our proposal

  • posal

today’s talk: necessary and sufficient structure for al all ret etrof

  • fit

t co contr trol

  • ller

lers

slide-9
SLIDE 9
  • 1. Intr

troduc

  • duction

tion 2.

  • 2. Ret

etrof

  • fit

it Co Contr trol

  • l

Outl tline ine

  • De

Definitions itions

  • Ret

etrof

  • fit

t Co Contr trolle

  • ller
  • Per

erforma rmance nce Anal alysis sis

  • 3. Nume

meric ical l Exa xamp mple (Power r Syst stem) m)

slide-10
SLIDE 10

Def efini inition tion: : Retr etrof

  • fit

it Contr troll

  • ller

er

ca case 1: e 1: ca case 2: e 2:

is any element ment s.t. t.

De Definition ition

the syste stem m for any s.t. t.

Env.

: the inter terconnected connected system em 7/15 15 is stable le

is called lled a retrof

  • fit

it contr troll

  • ller

er if it stabili bilizes es is stable

slide-11
SLIDE 11

Equ quiv ivalent alent Conditio dition

: inter ercon connection nection signals nals : measur surement ment, , contr ntrol

The heor

  • rem

em

is a r a ret etrof

  • fit

it co contr troll

  • ller

er an and

8/15 15

: as assumed ed to to be sta e stable

slide-12
SLIDE 12

Equ quiv ivalent alent Conditio dition

: inter ercon connection nection signals nals : measur surement ment, , contr ntrol

The heor

  • rem

em

is a r a ret etrof

  • fit

it co contr troll

  • ller

er an and

8/15 15

: as assumed ed to to be sta e stable

slide-13
SLIDE 13

Equ quiv ivalent alent Conditio dition

: inter ercon connection nection signals nals : measur surement ment, , contr ntrol

The heor

  • rem

em

is a r a ret etrof

  • fit

it co contr troll

  • ller

er an and

8/15 15

: as assumed ed to to be sta e stable

slide-14
SLIDE 14

Equ quiv ivalent alent Conditio dition

: inter ercon connection nection signals nals : measur surement ment, , contr ntrol

The heor

  • rem

em

is a r a ret etrof

  • fit

it co contr troll

  • ller

er an and

8/15 15

: as assumed ed to to be sta e stable

slide-15
SLIDE 15

Equ quiv ivalent alent Conditio dition

: inter ercon connection nection signals nals : measur surement ment, , contr ntrol

The heor

  • rem

em

is a r a ret etrof

  • fit

it co contr troll

  • ller

er an and

loop p transf ansfer er matrix rix : arbitr itrary ary 8/15 15

: as assumed ed to to be sta e stable

becau cause se

co constr straine ained d Youl ula par arame ameteriza terization tion

slide-16
SLIDE 16

Two Cla lasses sses of Retr etrof

  • fit

it

: retr trof

  • fit

it contr trolle ler

  • 1. Ou

Output put rectif ctifying ying retr trof

  • fit

it contr ntroller

  • ller
  • 2. Input

put rectifying ctifying retr trof

  • fit

it contr ntroller

  • ller

9/15 15 rather ther gener eral al hard to solve

sim imple le struct uctur ure, , easy asy to

  • desi

sign gn

slide-17
SLIDE 17

Output tput Recti ectifying fying Retr etr. . Ctr trl. l.

Ou Output put Rectif ctifying ying Retr trof

  • fit

it Contr troller

  • ller

Assu sumpti mption

  • n

can n be fed back

Theo eorem em

All outpu put t rectifying ctifying retr trof

  • fit

it contr ntroller

  • llers

s have e the st structu ucture

( )

: stabil biliz izes es 10/15 /15 (loca cally lly st stabilizing bilizing contr ntroller)

  • ller)
slide-18
SLIDE 18

Des esign ign Criteri iterion

  • n

11/15 /15

resu sultant ltant con

  • ntr

trol

  • l system

stem

: outp tput-rectif ctifying ing retrof

  • fit

t cont ntroll

  • ller

er

slide-19
SLIDE 19

Des esign ign Criteri iterion

  • n

11/15 /15

bot

  • th

h sho houl uld d be taken en in into

  • account

ccount resu sultant ltant con

  • ntr

trol

  • l system

stem

slide-20
SLIDE 20
  • 1. Intr

troduc

  • duction

tion 2.

  • 2. Ret

etrof

  • fit

it Co Contr trol

  • l

Outl tline ine

  • De

Definitions itions

  • Ret

etrof

  • fit

t Co Contr trolle

  • ller
  • Performa

rmance nce Analysi sis

  • 3. Nume

meric ical l Exa xamp mple (Power r Syst stem) m)

slide-21
SLIDE 21

IEEJ EJ EAST30 ST30

12/15 /15

新仙台火 力 女川原子 力 仙台火 力 原町火 力 福島第一原 子力 福島第二原 子力 新地 火力 鹿島共同 火力 鹿島火 力 南横浜 火力

500kV power line 275kV power line

Thermal Hydro Nuclear

PV PV-Inte tegrated ted IEE EEJ EA EAST ST30 30 mode del

(The e Institut titute e of Electri ctrical cal Enginee ineers of Japa pan) n)

Fukus ushi hima ma

slide-22
SLIDE 22

Par arameter ameters

# of gen. n.: : 30 whole

  • le dim.:

.: 390

  • initi

tial al deviati tion

  • n at

t th the e 2nd

nd generator

tor

13/15 /15

  • objec

ectiv tive: e: suppress essing ng pea eak val alue

PV: : 30% of total tal demand and

slide-23
SLIDE 23

Resu esults(1) lts(1)

14/15 /15 ignoring

  • ring env.

(infinite inite-bus) us) free ee response sponse peak: ak: 0.09 0904 04 [Hz]

freq eque uenc ncy deviation tion des esta tabi bili lized ed

slide-24
SLIDE 24

Resu esults lts (2 (2)

Retr trof

  • fit

it Contr trol

  • l

(low

  • w gain)

n) peak ak: : 0.0753 0753 [Hz]

improvement ement: : 1. 1.59 59 [d [dB] sta tability ity is p pres eser erved ed freq eque uenc ncy deviation tion

14/15 /15 free ee response sponse peak: ak: 0.09 0904 04 [Hz]

slide-25
SLIDE 25

Resu esults lts (3 (3)

imp mprovement: ement: 5. 5.06 06 [d [dB]

peak ak: : 0.0505 0505 [Hz] 14/15 /15 free ee response sponse peak: ak: 0.09 0904 04 [Hz]

freq eque uenc ncy deviation tion

Retr trof

  • fit

it Contr trol

  • l

(high gh gain) in)

slide-26
SLIDE 26

Conc nclusi lusion

  • n

15/15 /15

Than ank k you for your r kind d att tten ention tion

Pr Prob

  • blem

lem:

Retr trof

  • fit

it Co Cont ntrol

  • l

novel el co contr ntrol

  • l th

theor eory for lar arge-scal cale e syste tems

  • high

h ef effec ectiv tivene eness ss