Sunteți pe pagina 1din 196

Topics in Gas Network Optimization

Energy Networks Group


Zuse Institute Berlin

CO@Work 2015, ZIB


Outline

1 Nomination Validation

2 Booking Validation

3 PPPP

4 Navi

1
Gas Networks

. Transmit natural gas to industry


and municipal utilities

. Passive elements:
I Pipelines
I Resistors

. Active elements:
I Valves
I Control valves
I Compressors
Open Grid Europe, Germany

2
Validating Nominations

Given: . Detailed description of a gas network


. Nomination specifying amounts of gas flow at
entries and exits

3
Validating Nominations

Given: . Detailed description of a gas network


. Nomination specifying amounts of gas flow at
entries and exits
Task: Find

3
Validating Nominations

Given: . Detailed description of a gas network


. Nomination specifying amounts of gas flow at
entries and exits
Task: Find
1. settings for the active devices
(valves, control valves, compressors)

3
Validating Nominations

Given: . Detailed description of a gas network


. Nomination specifying amounts of gas flow at
entries and exits
Task: Find
1. settings for the active devices
(valves, control valves, compressors)
2. values for the physical parameters of the network

3
Validating Nominations

Given: . Detailed description of a gas network


. Nomination specifying amounts of gas flow at
entries and exits
Task: Find
1. settings for the active devices
(valves, control valves, compressors)
2. values for the physical parameters of the network
that comply with
. gas physics
. legal and technical limitations

3
Validating Nominations

Given: . Detailed description of a gas network


. Nomination specifying amounts of gas flow at
entries and exits human experience
Task: Find
1. settings for the active devices
(valves, control valves, compressors)
2. values for the physical parameters of the network
that comply with
. gas physics
. legal and technical limitations

3
Validating Nominations

Given: . Detailed description of a gas network


. Nomination specifying amounts of gas flow at
entries and exits human experience
Task: Find
1. settings for the active devices
(valves, control valves, compressors)
2. values for the physical parameters of the network
that comply with
. gas physics
. legal and technical limitations
simulation tool

3
Validating Nominations

Given: . Detailed description of a gas network


. Nomination specifying amounts of gas flow at
entries and exits human experience
Task: Find
1. settings for the active devices
(valves, control valves, compressors)
2. values for the physical parameters of the network
that comply with
. gas physics
. legal and technical limitations
simulation tool
We use mathematical optimization to integrate both steps!

3
Using Optimization Rather Than Simulation

Simulation
. allows very accurate gas physics
models
. relies on human experience for
control for active devices
. is thus inappropriate to determine
infeasibility of a nomination

4
Using Optimization Rather Than Simulation

Simulation Optimization
. allows very accurate gas physics . works on simplified models of gas
models physics
. relies on human experience for . automatically finds settings for active
control for active devices devices
. is thus inappropriate to determine . eventually proves infeasibility of an
infeasibility of a nomination infeasible nomination

4
Using Optimization Rather Than Simulation

Simulation Optimization
. allows very accurate gas physics . works on simplified models of gas
models physics
. relies on human experience for . automatically finds settings for active
control for active devices devices
. is thus inappropriate to determine . eventually proves infeasibility of an
infeasibility of a nomination infeasible nomination

Beware: Different solution spaces due to different modeling

Simulation A Optimization A

Simulation B Optimization B

4
Model Components

Gas network is modeled as graph G = (V, A)

Model components:

5
Model Components

Gas network is modeled as graph G = (V, A)

Model components:
. Flow conservation constraints at nodes

5
Model Components

Gas network is modeled as graph G = (V, A)

Model components:
. Flow conservation constraints at nodes
. Model for all kinds of arcs
I Pipes
I Resistors
I Shortcut
I Valves
I Control valves
I Compressors

5
Model Components

Gas network is modeled as graph G = (V, A)

Model components:
. Flow conservation constraints at nodes
. Model for all kinds of arcs
I Pipes → nonlinear
I Resistors → nonlinear
I Shortcut
I Valves
I Control valves
I Compressors

5
Model Components

Gas network is modeled as graph G = (V, A)

Model components:
. Flow conservation constraints at nodes
. Model for all kinds of arcs
I Pipes → nonlinear
I Resistors → nonlinear
I Shortcut
I Valves → discrete switching
I Control valves → discrete switching
I Compressors → discrete switching

5
Model Components

Gas network is modeled as graph G = (V, A)

Model components:
. Flow conservation constraints at nodes
. Model for all kinds of arcs
I Pipes → nonlinear
I Resistors → nonlinear
I Shortcut
I Valves → discrete switching
I Control valves → discrete switching
I Compressors → discrete switching

Mixed-Integer Nonlinear Program

5
Modeling of Pipelines
Weymouth equation
pi qij pj
αij · qij · |qij | = pi2 − βij pj2 i j

with constants
αij diameter, length, temperature
βij height difference of vertices

pi2 − pj2

qij

6
Valves

pi qij pj
i j

7
Valves

pi qij pj
i j

Two states:

Closed
Open

7
Valves

pi qij pj
One decision variable: xij ∈ {0, 1}
i j
xij

Two states:

Closed
Open

7
Valves

pi qij pj
One decision variable: xij ∈ {0, 1}
i j
xij

Two states:

Closed xij = 0 ⇒ qij = 0


Open

7
Valves

pi qij pj
One decision variable: xij ∈ {0, 1}
i j
xij

Two states:

Closed xij = 0 ⇒ qij = 0


Open xij = 1 ⇒ pi = pj

7
Control Valve Constraints

pi qij pj
i j

8
Control Valve Constraints

pi qij pj
i j

Three states:

Closed
Bypass
Active

8
Control Valve Constraints

pi qij pj
Two variables: xijbypass + xijactive ≤1 i j

xijactive , xijbypass

Three states:

Closed
Bypass
Active

8
Control Valve Constraints

pi qij pj
Two variables: xijbypass + xijactive ≤1 i j

xijactive , xijbypass

Three states:

Closed xijbypass = xijactive = 0 ⇒ qij = 0


Bypass
Active

8
Control Valve Constraints

pi qij pj
Two variables: xijbypass + xijactive ≤1 i j

xijactive , xijbypass

Three states:

Closed xijbypass = xijactive = 0 ⇒ qij = 0


Bypass xijbypass = 1 ⇒ pi = pj
Active

8
Control Valve Constraints

pi qij pj
Two variables: xijbypass + xijactive ≤1 i j

xijactive , xijbypass

Three states:

Closed xijbypass = xijactive = 0 ⇒ qij = 0


Bypass xijbypass = 1 ⇒ pi = pj
Active xijactive =1 ⇒ 0 ≤ ∆ ≤ pi − pj ≤ ∆

8
Compressor Station

Compressor Station: pi qij pj


i j

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
70

60

50

Had
40 M1
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
70

60

50

Had
40 M1
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70

different configurations 60

50

Had
40 M1
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70

different configurations 60

50

Had
40 M1
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70

different configurations 60

50

Had
40 M3 M1
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70

different configurations 60

50

Had
40 M3 M1 M1 k M3
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70

different configurations 60

50

Had
40 M3 M1 M1 k M3
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70

different configurations 60

50

Had
40 M3 M1
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
M1  M3
. Compressor station can operate in 70

different configurations 60

50

Had
40 M3 M1
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
M1  M3
. Compressor station can operate in 70

different configurations 60

50

Had
40 M3 M1
30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70 parallel
serial
different configurations 60

50

Had
40

30

20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70 parallel
serial
different configurations 60

50

Had
40

Modeling approaches: 30

. Convex hull of all configurations 20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70

different configurations 60

50

Had
40

Modeling approaches: 30

. Convex hull of all configurations 20

10

0
0 1 2 3 4 5 6 7 8 9 10

9
Compressor Station

Compressor Station: pi qij pj


i j
. Union of single compressor machines
. Closed and bypass state
. Each compressor machine has a
characteristic diagram
. Compressor station can operate in 70 parallel
serial
different configurations 60

50

Had
40

Modeling approaches: 30

. Convex hull of all configurations 20

. Binary variables per configuration 10

0
0 1 2 3 4 5 6 7 8 9 10

9
Subnetwork Operation Modes
Schematischer Stationsaufbau (neu)
. Complex subnetworks can be operated only in specific modes,
e.g. compressor plants
VA_219 VA_221
PI_798 VA_198O0003 80060012008
HEROS-H-S-FI
VA_193 VA_192 VA_198

VA_193O

PORZ_VS-S VA_192I VA_200


(11p/s, 12p/s)

PORZ_VS-H PORZ_VS-H1
(5, 6) (5, 6)

VA_194 VA_195 VA_199


PI_863

PI_799 VA_199O0004 800600120010


VA_205

Zuse Institute Berlin


4 April 20, 2011
Dep. Optimization

10
Subnetwork Operation Modes

Verdichten von Stolberg (M11 o. M12)


. Complex subnetworks can be operated only in specific modes,
e.g. compressor plants
VA_219 VA_221
PI_798 VA_198O0003 80060012008
HEROS-H-S-FI
VA_193 VA_192 VA_198

VA_193O

PORZ_VS-S VA_192I VA_200


(11p/s, 12p/s)

PORZ_VS-H PORZ_VS-H1
(5, 6) (5, 6)

VA_194 VA_195 VA_199


PI_863

PI_799 VA_199O0004 800600120010


VA_205

tute Berlin
5 April 20, 2011
timization

10
Subnetwork Operation Modes
9. Von Scheidt verdichtet nach Paffrath und von Stolbe
Verdichten von Stolberg (M11 o. M12) verdichtet nach Paffrath
. Complex subnetworks can be operated only in specific modes,
e.g. compressor plants
VA_219 VA_221 VA_219 VA_221
PI_798 VA_198O0003 80060012008 PI_798 VA_198O0003 80060012008
HEROS-H-S-FI HEROS-H-S-FI
VA_193 VA_192 VA_198 VA_193 VA_192 VA_198

VA_193O VA_193O

PORZ_VS-S VA_192I VA_200 PORZ_VS-S VA_192I VA_200


(11p/s, 12p/s) (11p/s, 12p/s)

PORZ_VS-H PORZ_VS-H1 PORZ_VS-H PORZ_VS-H1


(5, 6) (5, 6) (5, 6) (5, 6)

VA_194 VA_195 VA_199 VA_194 VA_195 VA_199


PI_863 PI_863
PI_799 VA_199O0004 800600120010 PI_799 VA_199O0004 800600120010
VA_205 VA_205

tute Berlin Zuse Institute Berlin


5 April 20, 2011 11 A
timization Dep. Optimization

10
Subnetwork Operation Modes
9. Von Scheidt verdichtet nach Paffrath und von Stolbe
Verdichten von Stolberg (M11 o. M12) verdichtet nach Paffrath
. Complex subnetworks can be operated only in specific modes,
e.g. compressor plants
VA_219 VA_221 VA_219 VA_221
PI_798 VA_198O0003 80060012008 PI_798 VA_198O0003 80060012008
HEROS-H-S-FI HEROS-H-S-FI
VA_193 VA_192 VA_198 VA_193 VA_192 VA_198

VA_193O VA_193O

on Scheidt verdichtet nach Paffrath PORZ_VS-S


(11p/s, 12p/s)
VA_192I VA_200 PORZ_VS-S
(11p/s, 12p/s)
VA_192I VA_200

PORZ_VS-H PORZ_VS-H1 PORZ_VS-H PORZ_VS-H1


(5, 6) (5, 6) (5, 6) (5, 6)

VA_194 VA_195 VA_199 VA_194 VA_195 VA_199


PI_863 PI_863
PI_799 VA_199O0004 800600120010 PI_799 VA_199O0004 800600120010
VA_205 VA_205

VA_219 VA_221
PI_798 VA_198O0003 80060012008
HEROS-H-S-FI
VA_193 VA_192 VA_198

VA_193O

PORZ_VS-S VA_192I VA_200


(11p/s, 12p/s)

tute Berlin PORZ_VS-H Zuse Institute Berlin


5 (5, 6)
PORZ_VS-H1
April 20, 2011 11 A
timization (5, 6)
Dep. Optimization
VA_194 VA_195 VA_199
PI_863
PI_799 VA_199O0004 800600120010
VA_205

tute Berlin
10 April 20, 2011
timization

10
Subnetwork Operation Modes
9. Von Scheidt verdichtet nach Paffrath und von Stolbe
Verdichten von Stolberg (M11 o. M12) verdichtet nach Paffrath
. Complex subnetworks can be operated only in specific modes,
e.g. compressor plants
VA_219 VA_221 VA_219 VA_221
PI_798 VA_198O0003 80060012008 PI_798 VA_198O0003 80060012008
HEROS-H-S-FI HEROS-H-S-FI
VA_193 VA_192 VA_198 VA_193 VA_192 VA_198

VA_193O
3a. Verdichten von Paffrath nach Stolberg und nach Scheidt
VA_193O

on Scheidt verdichtet nach Paffrath PORZ_VS-S


(11p/s, 12p/s)
VA_192I VA_200
verdichtet aus Paffrath und unverdichtet aus Scheidt nach S
PORZ_VS-S
(11p/s, 12p/s)
VA_192I VA_200

PORZ_VS-H PORZ_VS-H1 PORZ_VS-H PORZ_VS-H1


(5, 6) (5, 6) (5, 6) (5, 6)

VA_194 VA_195 VA_199 VA_194 VA_195 VA_199


PI_863 PI_863
PI_799 VA_199O0004 800600120010 PI_799 VA_199O0004 800600120010
VA_205 VA_205

VA_219 VA_221 VA_219 VA_221


PI_798 VA_198O0003 80060012008 PI_798 VA_198O0003 80060012008
HEROS-H-S-FI HEROS-H-S-FI
VA_193 VA_192 VA_198 VA_193 VA_192 VA_198

VA_193O VA_193O

PORZ_VS-S VA_192I VA_200 PORZ_VS-S VA_192I VA_200


(11p/s, 12p/s) (11p/s, 12p/s)

tute Berlin PORZ_VS-H Zuse Institute Berlin PORZ_VS-H


5 (5, 6)
PORZ_VS-H1
April 20, 2011 11 (5, 6)
PORZ_VS-H1
A
timization (5, 6)
Dep. Optimization (5, 6)

VA_194 VA_195 VA_199 VA_194 VA_195 VA_199


PI_863 PI_863

PI_799 VA_199O0004 800600120010 PI_799 VA_199O0004 800600120010

VA_205 VA_205

tute Berlin Zuse Institute Berlin


10 April 20, 2011 7 A
timization Dep. Optimization

10
Subnetwork Operation Modes
9. Von Scheidt verdichtet nach Paffrath und von Stolbe
Verdichten von Stolberg (M11 o. M12) verdichtet nach Paffrath
. Complex subnetworks can be operated only in specific modes,
e.g. compressor plants
VA_219 VA_221 VA_219 VA_221
PI_798 VA_198O0003 80060012008 PI_798 VA_198O0003 80060012008
HEROS-H-S-FI HEROS-H-S-FI
VA_193 VA_192 VA_198 VA_193 VA_192 VA_198

VA_193O
3a. Verdichten von Paffrath nach Stolberg und nach Scheidt
VA_193O

on Scheidt verdichtet nach Paffrath PORZ_VS-S


(11p/s, 12p/s)
VA_192I VA_200
verdichtet aus Paffrath und unverdichtet aus Scheidt nach S
PORZ_VS-S
(11p/s, 12p/s)
VA_192I VA_200

PORZ_VS-H PORZ_VS-H1 PORZ_VS-H PORZ_VS-H1


(5, 6) (5, 6) (5, 6) (5, 6)

VA_194 VA_195 VA_199 VA_194 VA_195 VA_199


PI_863 PI_863
PI_799 VA_199O0004 800600120010 PI_799 VA_199O0004 800600120010
VA_205 VA_205

VA_219 VA_221 VA_219 VA_221


PI_798 VA_198O0003 80060012008 PI_798 VA_198O0003 80060012008
HEROS-H-S-FI HEROS-H-S-FI
VA_193 VA_192 VA_198 VA_193 VA_192 VA_198

VA_193O VA_193O

PORZ_VS-S VA_192I VA_200 PORZ_VS-S VA_192I VA_200


(11p/s, 12p/s) (11p/s, 12p/s)

tute Berlin PORZ_VS-H Zuse Institute Berlin PORZ_VS-H


5 (5, 6)
PORZ_VS-H1
April 20, 2011 11 (5, 6)
PORZ_VS-H1
A
timization (5, 6)
Dep. Optimization (5, 6)

VA_194 VA_195 VA_199 VA_194 VA_195 VA_199


PI_863 PI_863

PI_799 VA_199O0004 800600120010 PI_799 VA_199O0004 800600120010

VA_205 VA_205

. Choose one of the modes with a binary variable

tute Berlin Zuse Institute Berlin


10 April 20, 2011 7 A
timization Dep. Optimization

10
Subnetwork Operation Modes
9. Von Scheidt verdichtet nach Paffrath und von Stolbe
Verdichten von Stolberg (M11 o. M12) verdichtet nach Paffrath
. Complex subnetworks can be operated only in specific modes,
e.g. compressor plants
VA_219 VA_221 VA_219 VA_221
PI_798 VA_198O0003 80060012008 PI_798 VA_198O0003 80060012008
HEROS-H-S-FI HEROS-H-S-FI
VA_193 VA_192 VA_198 VA_193 VA_192 VA_198

VA_193O
3a. Verdichten von Paffrath nach Stolberg und nach Scheidt
VA_193O

on Scheidt verdichtet nach Paffrath PORZ_VS-S


(11p/s, 12p/s)
VA_192I VA_200
verdichtet aus Paffrath und unverdichtet aus Scheidt nach S
PORZ_VS-S
(11p/s, 12p/s)
VA_192I VA_200

PORZ_VS-H PORZ_VS-H1 PORZ_VS-H PORZ_VS-H1


(5, 6) (5, 6) (5, 6) (5, 6)

VA_194 VA_195 VA_199 VA_194 VA_195 VA_199


PI_863 PI_863
PI_799 VA_199O0004 800600120010 PI_799 VA_199O0004 800600120010
VA_205 VA_205

VA_219 VA_221 VA_219 VA_221


PI_798 VA_198O0003 80060012008 PI_798 VA_198O0003 80060012008
HEROS-H-S-FI HEROS-H-S-FI
VA_193 VA_192 VA_198 VA_193 VA_192 VA_198

VA_193O VA_193O

PORZ_VS-S VA_192I VA_200 PORZ_VS-S VA_192I VA_200


(11p/s, 12p/s) (11p/s, 12p/s)

tute Berlin PORZ_VS-H Zuse Institute Berlin PORZ_VS-H


5 (5, 6)
PORZ_VS-H1
April 20, 2011 11 (5, 6)
PORZ_VS-H1
A
timization (5, 6)
Dep. Optimization (5, 6)

VA_194 VA_195 VA_199 VA_194 VA_195 VA_199


PI_863 PI_863

PI_799 VA_199O0004 800600120010 PI_799 VA_199O0004 800600120010

VA_205 VA_205

. Choose one of the modes with a binary variable


. Couple this variable with the elements’ variables via inequalities
tute Berlin Zuse Institute Berlin
10 April 20, 2011 7 A
timization Dep. Optimization

10
Gas Network: H-Nord

. 661 nodes
I 32 entries
I 142 exits
. 498 pipes
9 resistors
33 valves
26 control valves
7 compressor stations
. 32 cycles

11
A sample solution

Entry
Exit

12
Gas Network: H-Süd

. 1662 nodes
I 47 entries
I 265 exits
. 1136 pipes
45 resistors
224 valves
78 control valves
29 compressor stations
. 175 cycles

13
Gas Network: L-Gas

. 4133 nodes
I 12 entries
I 1001 exits
. 3623 pipes
26 resistors
300 valves
118 control valves
12 compressor stations
. 259 cycles

14
In Practice: PowerNova
The issue: tracking of calorific values

. Nomination given in power, not (mass) flow


. Calorific values only known at entry nodes
. Bilinear equations for mixing of calorific values

1
B. Geißler et al.: “Solving Power-Constrained Gas Transportation Problems using an MIP-based Alternating
Direction Method”, 2014, Optimization Online
15
In Practice: PowerNova
The issue: tracking of calorific values

. Nomination given in power, not (mass) flow


. Calorific values only known at entry nodes
. Bilinear equations for mixing of calorific values

The solution: Iterative approach1

. Guess (average) calorific values


. With fixed calorific values: compute target flow values
. Compute solution, minimizing a penalty term for exit demands
. With fixed flows: compute actual calorific values
. Update target flows and penalty weights in objective

1
B. Geißler et al.: “Solving Power-Constrained Gas Transportation Problems using an MIP-based Alternating
Direction Method”, 2014, Optimization Online
15
In Practice: Flow Commitments

The issue: Modeling a Stackelberg game

. TSO can use non-technical measures to change nomination (leaders)


. Shippers react to these measures with given preference (followers)

16
In Practice: Flow Commitments

The issue: Modeling a Stackelberg game

. TSO can use non-technical measures to change nomination (leaders)


. Shippers react to these measures with given preference (followers)

The solution: Bilevel Optimization

. Model game as Bilevel Optimization Problem


. Leader part is MIP
. Follower part is LP
. Equivalent reformulation as MIP (with complementarity) via KKT
optimality conditions

16
In Practice: Flow Commitments

The issue: Modeling a Stackelberg game

. TSO can use non-technical measures to change nomination (leaders)


. Shippers react to these measures with given preference (followers)

The solution: Bilevel Optimization

. Model game as Bilevel Optimization Problem


. Leader part is MIP
. Follower part is LP
. Equivalent reformulation as MIP (with complementarity) via KKT
optimality conditions

→ More details later in talk about PPPP!

16
Belgium: an Instance from the Literature

. almost tree topology


. 20 vertices
. 29 pipes

17
GasLib: a Library of Gas Network Instances

. realistic benchmark instances


. detailed description of gas network
I 582 nodes
I 278 pipes, 8 resistors
I 54 active elements
. thousands of nominations

gaslib.zib.de

18
Outline

1 Nomination Validation

2 Booking Validation

3 PPPP

4 Navi

18
Virtual Trading Points, Entries, and Exits

19
Virtual Trading Points, Entries, and Exits

N3 N2

X1 VTP N1

X3
X2

19
Virtual Trading Points, Entries, and Exits

N3 N2

X1 VTP N1

X3
X2

capacity: transport gas from node to VTP,


booked independently for entries and exits

19
Capacity Estimation

N1 N2

X1 X2

20
Capacity Estimation

N1 N2
[20]

[10]
X1 X2

Long pipeline with small capacity

20
Capacity Estimation

N1 N2
[20]

[10]
X1 X2

Long pipeline with small capacity


What is the “capacity” of the whole network?

20
Capacity Estimation

N1 N2
[20]

[10]
X1 X2

Long pipeline with small capacity


What is the “capacity” of the whole network?
How much capacity can be booked at the nodes?

20
Capacity Estimation

20 N1 N2
[20]

[10]
20 X1 X2

20
Capacity Estimation

20 N1 N2 20
[20]

[10]
20 X1 X2 20

20
Capacity Estimation

10 N1 N2 10
[20]

[10]
20 X1 X2

20
Capacity Estimation

20 N1 N2
[20]

[10]
X1 X2 20

20
Capacity Estimation

20 N1 N2
[20]

[10]
X1 X2 20

At most 10 units of capacity may be booked at every node!

20
Capacity Estimation

20 N1 N2
[20]

[10]
X1 X2 20

At most 10 units of capacity may be booked at every node!

But up to 20 units could be transported.

20
Capacity Estimation

20 N1 N2
[20]

[10]
X1 X2 20

At most 10 units of capacity may be booked at every node!

But up to 20 units could be transported.

⇒ Distinguish firm and interruptible capacity!

20
Regulatory Authorities

. TSOs must market as much capacity as (technically) possible

21
Regulatory Authorities

. TSOs must market as much capacity as (technically) possible


. Denied requests for more capacity by shippers must be explained

21
Regulatory Authorities

. TSOs must market as much capacity as (technically) possible


. Denied requests for more capacity by shippers must be explained
. Need (proven) infeasibility of worst-case scenarios

21
Regulatory Authorities

. TSOs must market as much capacity as (technically) possible


. Denied requests for more capacity by shippers must be explained
. Need (proven) infeasibility of worst-case scenarios
. Solve Nomination Validation with global solver

21
Booking and Nomination

Booking contracts for firm capacity at entry/exit (interval [l, u])


. additional constraints (bounds on group of entries)
Nomination use of capacity (value v ∈ [l, u])
. balanced (stationary model)

Example with two entries N1 , N2 and exit X


. Booking . Nomination 1 . Nomination 2
N1 [0, 5] N1 3 N1 0
N2 [0, 4] N2 2 N2 4
X [0, 5] X 5 X 4

22
Booking Validation - The Problem

Given
. complete description of network
I pipes
I (control) valves
I compressor stations
. capacity contracts
I entry/exit nodes or zones
I valid for specific dates,
temperatures
I pairwise exclusion
. historical measurements
I hourly demand at exits
I temperature

23
Booking Validation - The Problem

Given Result
. complete description of network . for several
I pipes I contract dates
I (control) valves I reference temperatures
I compressor stations . compute
. capacity contracts I feasibility probability
I entry/exit nodes or zones I (infeasible) nominations
I valid for specific dates,
temperatures
I pairwise exclusion
. historical measurements
I hourly demand at exits
I temperature

23
Approach
Ideally

. check feasibility for every realizable nomination

24
Approach
Ideally

. check feasibility for every realizable nomination

Actually
. generate many probable nominations
I estimate distributions for exit demands
I sample scenarios from distributions
I apply scenario reduction
I worst-case entry completions for each scenario

24
Approach
Ideally

. check feasibility for every realizable nomination

Actually
. generate many probable nominations
I estimate distributions for exit demands
I sample scenarios from distributions
I apply scenario reduction
I worst-case entry completions for each scenario

. check feasibility for each


I solve Nomination Validation
I compute in parallel

24
Approach
Ideally

. check feasibility for every realizable nomination

Actually
. generate many probable nominations
I estimate distributions for exit demands
I sample scenarios from distributions
I apply scenario reduction
I worst-case entry completions for each scenario

. check feasibility for each


I solve Nomination Validation
I compute in parallel

. aggregate results to estimate feasibility probability

24
Historical Exit Demand
·104

8
Gas flow [kWh/h]

0
−15 −10 −5 0 5 10 15 20 25 30
Temperature [ C]

25
Historical Exit Demand
·104
3
Gas flow [kWh/h]

−15 −10 −5 0 5 10 15 20 25 30
Temperature [◦ C]

26
Historical Exit Demand
. separate distribution estimations for
I temperature classes
I workday and weekend

27
Correlation of Exit Demand
. Multivariate Distribution Estimation

0.8

0.6

0.4

0.2

−0.2

−0.4

−0.6

−0.8

−1

28
Scenario Reduction

-1

-2

-3

-3 -2 -1 0 1 2 3

29
MIP Model for Contract-Compliant Nominations

Pn ≥ 0 power at node n
xc ∈ {0, 1} use of contract c
Pnc ≥0 power at node n from contract c

30
MIP Model for Contract-Compliant Nominations

Pn ≥ 0 power at node n
xc ∈ {0, 1} use of contract c
Pnc ≥0 power at node n from contract c

X X
Pn = Pn nomination is balanced
n∈N n∈X

30
MIP Model for Contract-Compliant Nominations

Pn ≥ 0 power at node n
xc ∈ {0, 1} use of contract c
Pnc ≥0 power at node n from contract c

X X
Pn = Pn nomination is balanced
n∈N n∈X
X
Pn = Pnc different contracts contribute at one node
c3n

30
MIP Model for Contract-Compliant Nominations

Pn ≥ 0 power at node n
xc ∈ {0, 1} use of contract c
Pnc ≥0 power at node n from contract c

X X
Pn = Pn nomination is balanced
n∈N n∈X
X
Pn = Pnc different contracts contribute at one node
c3n
X
Lc xc ≤ Pnc ≤ Uc xc only active when used
n∈c

30
MIP Model for Contract-Compliant Nominations

Pn ≥ 0 power at node n
xc ∈ {0, 1} use of contract c
Pnc ≥0 power at node n from contract c

X X
Pn = Pn nomination is balanced
n∈N n∈X
X
Pn = Pnc different contracts contribute at one node
c3n
X
Lc xc ≤ Pnc ≤ Uc xc only active when used
n∈c
xc1 + xc2 ≤ 1 pairwise exclusion of contracts

30
MIP Model for Contract-Compliant Nominations
Substitutable capacity

. some contracts c ∈ Cs are classified substitutable


. predicted behavior according to historical data

31
MIP Model for Contract-Compliant Nominations
Substitutable capacity

. some contracts c ∈ Cs are classified substitutable


. predicted behavior according to historical data

Nonsubstitutable capacity

. other contracts c ∈ Cns are classified nonsubstitutable


. adversial behavior assumed

31
MIP Model for Contract-Compliant Nominations
Substitutable capacity

. some contracts c ∈ Cs are classified substitutable


. predicted behavior according to historical data

Nonsubstitutable capacity

. other contracts c ∈ Cns are classified nonsubstitutable


. adversial behavior assumed
X X
Pn = Pnc + Pnc distinguish between contracts
c∈Cs c∈Cns
X
Sn = Pnc substitute value from statistical scenario
c∈Cs

31
Entry Order
. entry supply hard to predict (nonsubstitutable)
. assume extreme behavior

32
Entry Order
. entry supply hard to predict (nonsubstitutable)
. assume extreme behavior

. entries are filled up in order of (random) preference

N 3 N 1 N4 N5 N2

32
Entry Order
. entry supply hard to predict (nonsubstitutable)
. assume extreme behavior

. entries are filled up in order of (random) preference

N 3 N 1 N4 N5 N2

. encode entry order in objective function

max 5PN3 + 4PN1 + 3PN3 + 2PN5 + PN2

32
Evaluation

Booking

33
Evaluation

Scenario s1
50%

20%
Booking Scenario s2

30%
Scenario s3

33
Evaluation

Nomination n1,1
Scenario s1
Nomination n1,2
50%
Nomination n2,1
20%
Booking Scenario s2
Nomination n2,2

30% Nomination n3,1


Scenario s3
Nomination n3,2

33
Evaluation

Nomination n1,1
Scenario s1
Nomination n1,2
50%
Nomination n2,1
20%
Booking Scenario s2
Nomination n2,2

30% Nomination n3,1


Scenario s3
Nomination n3,2

feasible

33
Evaluation

Nomination n1,1
Scenario s1
Nomination n1,2
50%
Nomination n2,1
20%
Booking Scenario s2
Nomination n2,2

30% Nomination n3,1


Scenario s3
Nomination n3,2

feasible

33
Evaluation

Nomination n1,1
Scenario s1
Nomination n1,2
50%
Nomination n2,1
20%
Booking Scenario s2
Nomination n2,2

30% Nomination n3,1


Scenario s3
Nomination n3,2

feasible infeasible

33
Evaluation

Nomination n1,1
Scenario s1
Nomination n1,2
50%
Nomination n2,1
20%
Booking Scenario s2
Nomination n2,2

30% Nomination n3,1


Scenario s3
Nomination n3,2

feasible infeasible

33
Evaluation

Nomination n1,1
Scenario s1
Nomination n1,2
50%
Nomination n2,1
20%
Booking Scenario s2
Nomination n2,2

30% Nomination n3,1


Scenario s3
Nomination n3,2

feasible infeasible unknown

33
Evaluation

Nomination n1,1
Scenario s1
Nomination n1,2
50%
Nomination n2,1
20%
Booking Scenario s2
Nomination n2,2

30% Nomination n3,1


Scenario s3
Nomination n3,2

feasible infeasible unknown

33
Evaluation

Nomination n1,1
Scenario s1
Nomination n1,2
50%
Nomination n2,1
20%
Booking Scenario s2
Nomination n2,2

30% Nomination n3,1


optimistic: 80%
Scenario s3
pessimistic: 50%
Nomination n3,2

feasible infeasible unknown

33
Evaluating Gas Network Capacities

Evaluating Gas Network Capacities


MOS-SIAM Series on Optimization, 2015
Dagmar Bargmann, Mirko Ebbers, Armin Fügenschuh,
Björn Geißler, Nina Geißler, Ralf Gollmer, Uwe Gotzes,
Christine Hayn, Holger Heitsch, René Henrion,
Benjamin Hiller, Jesco Humpola, Imke Joormann,
Thorsten Koch, Veronika Kühl, Thomas Lehmann,
Ralf Lenz, Hernan Leövey, Alexander Martin,
Radoslava Mirkov, Andris Möller, Antonio Morsi,
Djamal Oucherif, Antje Pelzer, Marc E. Pfetsch,
Lars Schewe, Werner Römisch, Jessica Rövekamp,
Martin Schmidt, Rüdiger Schultz, Robert Schwarz,
Jonas Schweiger, Klaus Spreckelsen, Claudia Stangl,
Marc C. Steinbach, Ansgar Steinkamp,
Isabel Wegner-Specht, Bernhard M. Willert.

34
ForNe team

35
Outline

1 Nomination Validation

2 Booking Validation

3 PPPP

4 Navi

35
The Shipper’s View - Virtual Trading Point

36
The Shipper’s View - Virtual Trading Point

VTP

36
The Shipper’s View - Virtual Trading Point
50 50 50

VTP

. Shippers nominate capacity on entry and exit nodes for the next gas day
w.r.t. their contractually fixed bookings (upper and lower bounds).

36
The Shipper’s View - Virtual Trading Point
50 50 50

VTP

50 150 50

. Shippers nominate capacity on entry and exit nodes for the next gas day
w.r.t. their contractually fixed bookings (upper and lower bounds).

36
The Shipper’s View - Virtual Trading Point
50 50 50

100

50 VTP

200

50 150 50

. Shippers nominate capacity on entry and exit nodes for the next gas day
w.r.t. their contractually fixed bookings (upper and lower bounds).

36
The Shipper’s View - Virtual Trading Point
50 50 50

100 200

50 VTP 100

200

50 150 50

. Shippers nominate capacity on entry and exit nodes for the next gas day
w.r.t. their contractually fixed bookings (upper and lower bounds).

36
The Shipper’s View - Virtual Trading Point
50 50 50

100 200

50 VTP 100

200 200

50 150 50

. Shippers nominate capacity on entry and exit nodes for the next gas day
w.r.t. their contractually fixed bookings (upper and lower bounds).

36
The Shipper’s View - Virtual Trading Point
50 50 50

100 200

50 VTP 100

200 200

50 150 50

. Shippers nominate capacity on entry and exit nodes for the next gas day
w.r.t. their contractually fixed bookings (upper and lower bounds).
. Changes are “always” possible by renominations.
36
The Shipper’s View - Virtual Trading Point
50 50 50

100 200

50 VTP 100

200 200
150

50 150 50

. Shippers nominate capacity on entry and exit nodes for the next gas day
w.r.t. their contractually fixed bookings (upper and lower bounds).
. Changes are “always” possible by renominations.
36
Non-Technical Network Control Measures

TSOs basically sell two different types of capacities:


Firm capacities TSO guarantees transport
Interruptible capacities Best effort but no guarantee

37
Non-Technical Network Control Measures

TSOs basically sell two different types of capacities:


Firm capacities TSO guarantees transport
Interruptible capacities Best effort but no guarantee

Basic Non-Technical Network Control Measures of the TSO


. In- or decrease amount of gas at certain entries or exits using legal
contracts with shippers or other TSO’s.
. Shifts in nominations at market area interconnection points.
. In- or decrease gas flows by buying so-called control energy.

37
Non-Technical Network Control Measures

TSOs basically sell two different types of capacities:


Firm capacities TSO guarantees transport
Interruptible capacities Best effort but no guarantee

Basic Non-Technical Network Control Measures of the TSO


. In- or decrease amount of gas at certain entries or exits using legal
contracts with shippers or other TSO’s.
. Shifts in nominations at market area interconnection points.
. In- or decrease gas flows by buying so-called control energy.

Final demand vector of gas flow is the result of a process of adjustments


by the TSO and possible reactions of shippers in terms of renominations.

37
A Look into the Future

Source: German Advisory Counsil on Global Change (WBGU).

38
How to Extend the Capacity of a Network

The ”natural” way: Physical Extension

. Deployment of new pipes to increase capacity


. Very costly (approximately 1 million euro per km)
. Takes a very long time

39
How to Extend the Capacity of a Network

The ”natural” way: Physical Extension

. Deployment of new pipes to increase capacity


. Very costly (approximately 1 million euro per km)
. Takes a very long time

The ”smart” way: Extension by Non-Technical Measures

. Design new capacity types and contracts


. Giving the TSO more flexibility and possibilities to act
. Allowing a better distribution of the gas in- and outflow
. Avoiding extreme situations

39
A New Contract Type

1
3
2

x: demand values
x: pressure induced “capacity” bounds

40
A New Contract Type

3
2

x: demand values
x: pressure induced “capacity” bounds

40
A new capacity type: fDZK

. New capacity type: fixed dynamically allocable capacity (fDZK)

41
A new capacity type: fDZK

. New capacity type: fixed dynamically allocable capacity (fDZK)


. Sold in contracts to gas power plants (power plant product)

41
A new capacity type: fDZK

. New capacity type: fixed dynamically allocable capacity (fDZK)


. Sold in contracts to gas power plants (power plant product)
. Equivalent to firm capacity (supply is guaranteed), BUT

41
A new capacity type: fDZK

. New capacity type: fixed dynamically allocable capacity (fDZK)


. Sold in contracts to gas power plants (power plant product)
. Equivalent to firm capacity (supply is guaranteed), BUT
. TSO can restrict power station to be supplied by fixed fallback entry

41
A new capacity type: fDZK

. New capacity type: fixed dynamically allocable capacity (fDZK)


. Sold in contracts to gas power plants (power plant product)
. Equivalent to firm capacity (supply is guaranteed), BUT
. TSO can restrict power station to be supplied by fixed fallback entry
. Usage: Restriction holds for the next gas day (6 a.m. to 6 a.m.)
but has to be announced at 3 p.m. the day before

41
A new capacity type: fDZK

. New capacity type: fixed dynamically allocable capacity (fDZK)


. Sold in contracts to gas power plants (power plant product)
. Equivalent to firm capacity (supply is guaranteed), BUT
. TSO can restrict power station to be supplied by fixed fallback entry
. Usage: Restriction holds for the next gas day (6 a.m. to 6 a.m.)
but has to be announced at 3 p.m. the day before

decision about restriction next gas day

06 12 15 18 24 06 12 18 24 06 t

day 0 day 1 day 2

41
Problem Statement

Decide about the restrictions to fallback entries


. for all power stations with a fDZK contract,

decision about restriction next gas day

06 12 15 18 24 06 12 18 24 06 t

day 0 day 1 day 2

42
Problem Statement

Decide about the restrictions to fallback entries


. for all power stations with a fDZK contract,
. in a nondiscriminatory manner,

decision about restriction next gas day

06 12 15 18 24 06 12 18 24 06 t

day 0 day 1 day 2

42
Problem Statement

Decide about the restrictions to fallback entries


. for all power stations with a fDZK contract,
. in a nondiscriminatory manner,
. while enabling safe network operation,

decision about restriction next gas day

06 12 15 18 24 06 12 18 24 06 t

day 0 day 1 day 2

42
Problem Statement

Decide about the restrictions to fallback entries


. for all power stations with a fDZK contract,
. in a nondiscriminatory manner,
. while enabling safe network operation,
. until 3 p.m. the day before the restriction shall become operative.

decision about restriction next gas day

06 12 15 18 24 06 12 18 24 06 t

day 0 day 1 day 2

42
Problem Statement

Decide about the restrictions to fallback entries


. for all power stations with a fDZK contract,
. in a nondiscriminatory manner,
. while enabling safe network operation,
. until 3 p.m. the day before the restriction shall become operative.
Remark: No specification of the technical network operation needed.

decision about restriction next gas day

06 12 15 18 24 06 12 18 24 06 t

day 0 day 1 day 2

42
Base Decision On ...
Network topology

43
Base Decision On ...
Network topology Network state

43
Base Decision On ...
Network topology Network state Bookings/Nominations

50 50 50

100 200

50 100

200 150

50 150 50

43
Base Decision On ...
Network topology Network state Bookings/Nominations

50 50 50

100 200

50 100

200 150

50 150 50

Historical Demands
6,000

5,000

4,000
Flow in kWh

3,000

2,000

1,000

0
2006 2007 2008 2009 2010 2011 2012 2013

Time

43
Base Decision On ...
Network topology Network state Bookings/Nominations

50 50 50

100 200

50 100

200 150

50 150 50

Historical Demands Weather Forecast


6,000

5,000

4,000
Flow in kWh

3,000

2,000

1,000

0
2006 2007 2008 2009 2010 2011 2012 2013

Time

43
Base Decision On ...
Network topology Network state Bookings/Nominations

50 50 50

100 200

50 100

200 150

50 150 50

Historical Demands Weather Forecast


6,000

5,000

4,000
Flow in kWh

...
3,000

2,000

1,000

0
2006 2007 2008 2009 2010 2011 2012 2013

Time

43
Definitions and Notation

. Gas Network
I G = (V , E ) a gas network, entries V+ ⊆ V and exits V− ⊆ V
I K ⊆ V− gas power stations with fDZK contract

44
Definitions and Notation

. Gas Network
I G = (V , E ) a gas network, entries V+ ⊆ V and exits V− ⊆ V
I K ⊆ V− gas power stations with fDZK contract
. Demand Vector
I d : V+ ∪ V− → R a map from entry and exit nodes to flow values
I D set of all demand vectors

44
Definitions and Notation

. Gas Network
I G = (V , E ) a gas network, entries V+ ⊆ V and exits V− ⊆ V
I K ⊆ V− gas power stations with fDZK contract
. Demand Vector
I d : V+ ∪ V− → R a map from entry and exit nodes to flow values
I D set of all demand vectors
. Scenario
I S : {1, . . . , 40} → D a map from time to demand vectors

44
Definitions and Notation

. Gas Network
I G = (V , E ) a gas network, entries V+ ⊆ V and exits V− ⊆ V
I K ⊆ V− gas power stations with fDZK contract
. Demand Vector
I d : V+ ∪ V− → R a map from entry and exit nodes to flow values
I D set of all demand vectors
. Scenario
I S : {1, . . . , 40} → D a map from time to demand vectors
. Prognosis
I (S, Pr ) with
I S := {S1 , S2 , . . . , Sn } a finite set of scenarios
I Pr : S → [0, 1] a probability distribution on S P
n
I pi := Pr (Si ) for all Si ∈ S with pi ∈ [0, 1] and i=1 pi = 1

44
Problem Statement Revisited

Decide about the restrictions to fallback entries


. for all power stations with a fDZK contract,
. in a nondiscriminatory manner,
. while enabling safe network operation,
. until 3 p.m. the day before the restriction shall become operative.
Remark: No specification of the technical network operation needed.

decision about restriction next gas day

06 12 15 18 24 06 12 18 24 06 t

day 0 day 1 day 2

45
Definition: Safe Network Operation
. (V , E ) be a gas network,

46
Definition: Safe Network Operation
. (V , E ) be a gas network,
. (S, Pr ) be a prognosis,

46
Definition: Safe Network Operation
. (V , E ) be a gas network,
. (S, Pr ) be a prognosis,
. K ∈ 2K be a subset of gas power stations,

46
Definition: Safe Network Operation
. (V , E ) be a gas network,
. (S, Pr ) be a prognosis,
. K ∈ 2K be a subset of gas power stations,
. A ∈ A be the initial state of the network,

46
Definition: Safe Network Operation
. (V , E ) be a gas network,
. (S, Pr ) be a prognosis,
. K ∈ 2K be a subset of gas power stations,
. A ∈ A be the initial state of the network,
. f : (G , A, S, 2K ) → {0, 1} be an oracle indicating if network G with
initial state A can be controlled over the next 40 hours by restricting
powerplants K and assuming scenario S,

46
Definition: Safe Network Operation
. (V , E ) be a gas network,
. (S, Pr ) be a prognosis,
. K ∈ 2K be a subset of gas power stations,
. A ∈ A be the initial state of the network,
. f : (G , A, S, 2K ) → {0, 1} be an oracle indicating if network G with
initial state A can be controlled over the next 40 hours by restricting
powerplants K and assuming scenario S,
. and  be a tolerance parameter with 0 ≤  < 1.

46
Definition: Safe Network Operation
. (V , E ) be a gas network,
. (S, Pr ) be a prognosis,
. K ∈ 2K be a subset of gas power stations,
. A ∈ A be the initial state of the network,
. f : (G , A, S, 2K ) → {0, 1} be an oracle indicating if network G with
initial state A can be controlled over the next 40 hours by restricting
powerplants K and assuming scenario S,
. and  be a tolerance parameter with 0 ≤  < 1.
A subset of powerplants K ∈ 2K admits a safe network operation if
restricting them to their fallback entries satisfies
n
X
pi · f (G , A, Si , K ) ≥ 1 − ,
i=1

i.e, it can be operated with probability (1 − ) assuming prognosis (S, Pr ).


46
Mathematical Problem Formulation
Given a gas network G = (V , E ) together with power stations K ⊆ V− ,
an initial network state A ∈ A, a prognosis (S, Pr ), and tolerance value .

The fDZK Problem is to decide if there exists K ∈ 2K admitting a safe


network operation

The fDZK Optimization Problem is to find a smallest subset of power


stations K ∈ 2K w.r.t the cardinality admitting a safe network operation,
i.e.,

min |K |
K ∈2K
Xn
s.t. pi · f (G , A, Si , K ) ≥ 1 − .
i=1

Remark: The crucial part of this definition is the oracle function f .

47
Mathematical Solution Approach

Bookings and Nominations

Historical Demands

Forecasted Temperature
Prognosis-Generation-Tool
Weekday/Weekend

Forecasted Renewables

... p1 p2 pn

S1 S2 ... Sn

Fix Restrictions
Gas Flow Evaluation Game S1 Game S2 ... Game Sn

Result Analyze TSO’s Behaviour in Games

48
Generating a Prognosis

We need to determine a reasonable prognosis capturing the uncertainty of


tomorrow’s demands in order to determine a reasonable solution.

Bookings and Nominations

Historical Demands

Forecasted Temperature
Prognosis-Generation-Tool
Weekday/Weekend

Forecasted Renewables

... p1 p2 pn

S1 S2 ... Sn

49
Mathematical Solution Approach

Bookings and Nominations

Historical Demands

Forecasted Temperature
Prognosis-Generation-Tool
Weekday/Weekend

Forecasted Renewables

... p1 p2 pn

S1 S2 ... Sn

Fix Restrictions
Gas Flow Evaluation Game S1 Game S2 ... Game Sn

Result Analyze TSO’s Behaviour in Games

50
A Game Of Contracts

By using contracts (e.g. fDZK) the TSO can change the demand vectors
within a scenario. Depending on the type of contract he can
. either increase the amount of flow on some entries (e.g. fDZK) or exits
. or decrease the amount of flow (e.g. interruptible capacity).

51
A Game Of Contracts

By using contracts (e.g. fDZK) the TSO can change the demand vectors
within a scenario. Depending on the type of contract he can
. either increase the amount of flow on some entries (e.g. fDZK) or exits
. or decrease the amount of flow (e.g. interruptible capacity).

But we have to take the reactions of the shippers into account:


. Decrease flow on some entry or increase flow on some exit
(If flow balance is increased by TSO’s action).
. Increase flow on some entry or decrease flow on some exit
(If flow balance is decreased by TSO’s action).

51
A Game Of Contracts

By using contracts (e.g. fDZK) the TSO can change the demand vectors
within a scenario. Depending on the type of contract he can
. either increase the amount of flow on some entries (e.g. fDZK) or exits
. or decrease the amount of flow (e.g. interruptible capacity).

But we have to take the reactions of the shippers into account:


. Decrease flow on some entry or increase flow on some exit
(If flow balance is increased by TSO’s action).
. Increase flow on some entry or decrease flow on some exit
(If flow balance is decreased by TSO’s action).

Idea: Model this process as game between TSO and the so-called
“shipper’s union” (SU) as the antagonist.

51
First Variation
3
2

2
2 2

x: demand values
x: pressure induced “capacity” bound

Next turn: TSO

52
First Variation
3
2

2
2 2

1 2

x: demand values
x: pressure induced “capacity” bound

Next turn: SU (Decrease Inflow by 2)

52
First Variation
3
2

1 2

x: demand values
x: pressure induced “capacity” bound

Next turn: TSO

52
First Variation
3
2

2
3 2

1 2

x: demand values
x: pressure induced “capacity” bound

Next turn: SU (Decrease Inflow by 3)

52
First Variation
3
2

2
3

x: demand values
x: pressure induced “capacity” bound

Game Over

52
Second Variation
3
2

2
2 2

x: demand values
x: pressure induced “capacity” bound

Next turn: TSO

53
Second Variation
3
2

2
3 2

x: demand values
x: pressure induced “capacity” bound

Next turn: SU (Decrease Inflow by 1)

53
Second Variation
3
2

2
3 1

x: demand values
x: pressure induced “capacity” bound

Game Over

53
Mathematical Solution Approach

Bookings and Nominations

Historical Demands

Forecasted Temperature
Prognosis-Generation-Tool
Weekday/Weekend

Forecasted Renewables

... p1 p2 pn

S1 S2 ... Sn

Fix Restrictions
Gas Flow Evaluation Game S1 Game S2 ... Game Sn

Result Analyze TSO’s Behaviour in Games

53
Gas Flow Evaluation

. Goal: Given a concrete demand vector, decide a network can safely be


operated during the next 40 hours

54
Gas Flow Evaluation

. Goal: Given a concrete demand vector, decide a network can safely be


operated during the next 40 hours
. No need for explicit control of network elements

54
Gas Flow Evaluation

. Goal: Given a concrete demand vector, decide a network can safely be


operated during the next 40 hours
. No need for explicit control of network elements
. Idea: Representation of macroscopic gas flows

54
Gas Flow Evaluation

. Goal: Given a concrete demand vector, decide a network can safely be


operated during the next 40 hours
. No need for explicit control of network elements
. Idea: Representation of macroscopic gas flows
I How is the gas distributed in the network?

54
Gas Flow Evaluation

. Goal: Given a concrete demand vector, decide a network can safely be


operated during the next 40 hours
. No need for explicit control of network elements
. Idea: Representation of macroscopic gas flows
I How is the gas distributed in the network?
I Are there areas with a high/low amount of gas?

54
Gas Flow Evaluation

. Goal: Given a concrete demand vector, decide a network can safely be


operated during the next 40 hours
. No need for explicit control of network elements
. Idea: Representation of macroscopic gas flows
I How is the gas distributed in the network?
I Are there areas with a high/low amount of gas?
. Needed: A very fast solving model to check many demand situations.

54
Gas Flow Evaluation

. Goal: Given a concrete demand vector, decide a network can safely be


operated during the next 40 hours
. No need for explicit control of network elements
. Idea: Representation of macroscopic gas flows
I How is the gas distributed in the network?
I Are there areas with a high/low amount of gas?
. Needed: A very fast solving model to check many demand situations.
. Plan: Design a coarse, fast solving LP model!
Pipes: Discretize and approximate Euler PDEs.
Compressors: Model as single element with max pressure difference.
Valves: Decide about discrete decisions heuristically.
Optional: Include discrete decisions into game.

54
Mathematical Solution Approach

Bookings and Nominations

Historical Demands

Forecasted Temperature
Prognosis-Generation-Tool
Weekday/Weekend

Forecasted Renewables

... p1 p2 pn

S1 S2 ... Sn

Fix Restrictions
Gas Flow Evaluation Game S1 Game S2 ... Game Sn

Result Analyze TSO’s Behaviour in Games

55
Outline

1 Nomination Validation

2 Booking Validation

3 PPPP

4 Navi

55
What is a Navigation System to us?

Typical Navi Instructions

. Turn left in 400 metres.


. In 300 metres keep left.
. Follow the street for 1.4 kilometres.
. ...

56
What is a Navigation System for a Dispatcher?

Typical Navi Instructions

. ????????????????
. ????????????????
. ????????????????
. ...
57
A First ”Idea”

. Text to be added

58
A Second ”Idea”

. Text to be added

59

S-ar putea să vă placă și