• Keine Ergebnisse gefunden

Subtree decomposition for multistage stochastic programs

N/A
N/A
Protected

Academic year: 2022

Aktie "Subtree decomposition for multistage stochastic programs"

Copied!
9
0
0

Wird geladen.... (Jetzt Volltext ansehen)

Volltext

(1)

!" # $ %"&

' ()

* +" $ ,(--.) /"

0((--.) 1"

2 -

3 4

# &

#& #& 3 5

6

# 3 5&

2

!

(2)

!

-

- *

7 -

2 $

-2

$

* 8 9"

#& 3 5 :

$4

$

4 ;

$

# &

( (--.) 1"

<: 5"

$

4

$

$

2 =

3

$ # / 1"& 6

-

4 ! , 2 =

9 %

/

>

>

$

4

4 ?

' $

(3)

9

$

84

*

$ 82 2

'

#&

'

#&@

#

#&&

#&

#&

3 $#

&

4

#&

3

#& #&

#& ;

#&

#& #& 8

> Æ #&

> Æ 4

6 #&

8 )

2 2

#& 3

4

#&

> Æ #&

@

#& > Æ

84

@

#

&5

3

6

2 A

6

# & #

2 ) 2 &

B

#&

#&

# #&

&

B

@ 5

#

&

#& #5&

C #&

(4)

%

#&5

B

#B #&#&&@

D#&

D

5

!

B

#&

#!&

E

6

6 @ #&

@#&

5 0 #& @ E

3

#& B

#5&

@E

! B

#B #&

#&& #&

9 B #5&

% F

#&@

#&D #

#&B

#&& 6

! D5

C )

+"

' $

@5 #

& >

4

>

5 2

3 4

3

4

@ #&

3 4

@

#6 $

$ &

F

( >

#&5 >

2 +"

(5)

/

2 (

$

3 $

2 2 - 2

Æ +"

Æ@

B

B

D

½

¾

82 2

2 $

, ; G"

@ 5

#&

D

5

#&B

#&

6 =

2

? E

2 G5" 6

@5

C0'H+/

C 666*!EEE E/I<(,# !I<

&1G%,:-) $

#

$ &

3

# )

$ &

4

#

&

(6)

G

9"

- /!

/! 5G5!1

4

D5

4 #& 4

#& 2

4

3

2

D5

3

(

#

&

D5

$

J K 5"

4

1"

4 1"

5 -

- 5EE

1

3

3!

15!5G

3 5E

C0'H #

L&

5EE

5E

39

5EE

3

'

3 # &#

& D5

# & 5DD

D

(7)

+

5 6 7

4

2 3

1

2 1

5 2 1

6 3 1

7 2

1

4 7

2 3

1

5 6

4 2

35

5 - 5EE1

!

"! #! !

M !1!9E1 !NG15E M

5E !1NEG 9E95E 9/+E

55 !/155 !+5!1 5GEE

55 !/GG% !GN1G %E

5EE 199G 1!+! G1GNE

25 49 100 144 225

0 5 10 15 20 25 30 35 40

iterations

scenarios

tt ot

cs ph

25 49 100 144 225

0 2000 4000 6000 8000 10000 12000

scenarios

solution time (seconds)

tt ot

cs ph

none

3 !

(8)

1

3%

# & @!

3/ 5!

5E

3

@G Æ

#&

6

3 $

$

-

# &

)

0 3 ) )

'2

)

5" AO<PP:,

G%5Q5N5NNG

!" O(<30

AR)5NN+

9" 8 ! "# $% & &

& C8 ,F AS5NN%

%" CP** * 5NN%

/" O, , (--.)

- %9#9&%++Q%NE5NN/

G" O, , : ; )

" #'!5G5NQG%95NN5

+" ( () ( O< *

- ( 5G#5&55NQ

5%+5NN5

1" : ( (--.) ? 0

(9)

N

0 20 40 60 80 100

10 −10 10 −8 10 −6 10 −4 10 −2 10 0 10 2

tt ot cs

ph

iterations

γ

39 5EE

4 5 6 7

1

3 2

44 55 66 77

15 14 13 12 11 10 8 9

26 22

19 27 28 29 31

17 18 20 21 23 24 25 30

16

4 5 6 7

33 22

1

3 2

15 14 14 13 13 12 12 11 11 10 10 88 99

26 22

19 27 28 29 31

17 18 20 21 23 24 25 30

16

44 55 66 77

15 14 13 12 11 10 8 9

3% #

&

81 256 625 1296

10 1 10 2 10 3 10 4 10 5

tt2 cs2 tt4 cs4

none

scenarios

solution time (seconds)

3/

Referenzen

ÄHNLICHE DOKUMENTE

In this paper, we developed a methodology for constructing scenario trees out of a fan of individ- ual scenarios such that the probability distribution and the filtration structure

As a result, the Stochastic Dual Dynamic Programming (SDDP) algorithm can be used to obtain approximations of the corresponding risk-averse recourse functions.. This allows us to

The approaches to scenario reduction in [5, 11] and to scenario tree generation in [21, 14] and [10] make use of probability metrics, i.e., of metric distances on spaces of

Dupaˇ cov´ a, Moment bounds for stochastic programs in particular for re- course problems, in: Distributions with given Marginals and Moment Pro- blems, V.. Dupaˇ cov´ a,

FRANCESCA MAGGIONI † AND GEORG CH. In general, multistage stochastic optimization problems are formulated on the basis of continuous distributions describing the uncertainty.

Gcnerally speaking, it is easier to bound the objective function and its optimal value than to gct bounds on optimal solutions and it is not, easy to cxtend

To test the behavior of both the scenario and nodal versions of the augmented Lagrangian decomposition algorithm on an actual numerical example, we consider an eight

A general decomposition framework for large convex optimization problems based on augmented Lagrangians is described.. The approach is then applied to multistage