a stochastic generalized nash model for natural gas ... › ... › web › material ›...
TRANSCRIPT
![Page 1: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/1.jpg)
1
A stochastic generalized nash model for natural gas transport systemsgas transport systems
Asgeir Tomasgard
A stochastic generalized nash model for natural gas transport systems
![Page 2: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/2.jpg)
2
Outline
• Optimization models versus equilibriumsS ti l ilib i d l• Spatial equilibrium models
• A stochastic complementarity model
![Page 3: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/3.jpg)
3
Mathematical programming
• Maximization or minimization of a real function by choosing values of variables from within an allowed seta ues o a ab es o a a o ed se
• A large number of problem classes:Li i– Linear programming
– Nonlinear programming– Integer programming
S h i i– Stochastic programming, etc.
![Page 4: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/4.jpg)
4
Optimality conditions
• Optimization problem with inequalities and equalities
• Given that we have a convex problem the first-order conditions of optimality (KKT) is sufficientconditions of optimality (KKT) is sufficient
– Convex feasible set– Concave objective function for maximization / convex objective
function for minimizationfunction for minimization
![Page 5: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/5.jpg)
5
Complementarity problems
• Linear system of equations• Nonlinear system of equations• Nonlinear system of equations• Linear complementarity problem• Nonlinear complementarity problem• Nonlinear program• Finite-dimensional system of variational inequalities
• All of these problems can be generalized to Mixed complementarity problemsp y p
![Page 6: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/6.jpg)
6
The linear complementarityThe linear complementarity problemp• Problem statement
Fi d t h th t• Find a vector x such that
• For a given vector q and matrix M• Denoted: LCP(q,M)
![Page 7: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/7.jpg)
7
Nonlinear complementarityNonlinear complementarity problem (NCP)• Find a vector x such that:
• Applications: – General equilibrium theory of economics, policy design and
analysis game theory mechanics etcanalysis, game theory, mechanics, etc.
![Page 8: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/8.jpg)
8
Mixed Complementarity Problem
• Many practical applications generate problems where some of the variables are nonnegative others aresome of the variables are nonnegative, others are bounded and others are free– To accommodate this flexibility, the MCP is usedy
• Find a vector such that
![Page 9: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/9.jpg)
9
Mixed Complementarity Problem
• Example (KKT for an optimization problem):
• Complementary pairs of variables:p y p– Economics: the price of a commodity and excess supply– Contact mechanics: the contact force between two variables and the
distance between them• MCP appear in study of equilibrium problems
– Numerous applications (economics, engineering and chemistry)
![Page 10: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/10.jpg)
10
Terminology
• A feasible x satisfies the inequalitiesIf z strictly satisfies the inequalities it is called strictly feasible– If z strictly satisfies the inequalities, it is called strictly feasible
• The set of feasible vectors is called its feasible region and is denoted FEA(q,M)and is denoted FEA(q,M)
• A vector x satisfying the complementarity condition is called complementaryp y
• The CP is then to find a x that is both feasible and complementary
![Page 11: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/11.jpg)
11
Complementarity ProblemsComplementarity Problems versus Optimization
– Optimization problems (via KKT-conditions)– Game theory problems (for instance Nash-Cournot games)– Many other problems in engineering and economics
• Theorems and algorithms developed for CP can be applied to a large number of applicationsapplied to a large number of applications
• CPs can include problems where dual prices (Lagrangean multipliers) appear in the primal ( g g p ) pp pformulation
![Page 12: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/12.jpg)
12
Equilibrium Problems• Equilibrium is a stable situation in which forces cancel one
anotherE i l l d d– Economics: supply equals demand
– Chemistry: the forward rate and reverse rate of reaction is equal– Physics: all forces acting on an object are balanced– Game theory: Nash equilibrium (situation where no player has an incentiveGame theory: Nash equilibrium (situation where no player has an incentive
to deviate from his strategy unilaterally)
• Formulation of equilibria:– Normally formulated as MCP or more generally as a Variational Inequality
(VI)Th VI i if i th d l f th t d f ilib i t– The VI is a unifying methodology for the study of equilibrium systems
![Page 13: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/13.jpg)
13
Spatial equilibrium models
• Consider at network withA set of suppliers (I) supplies a– A set of suppliers (I), supplies ai
– A set of markets (J), demands bj
– Transport costs of cij
• Want to find a transportation schedule which minimizes the cost of supplying all markets:
![Page 14: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/14.jpg)
14
Example (cont.)
• Can be interpreted as a market equilibrium problemThe dual multiplier for the supply constraint represents the price in– The dual multiplier for the supply constraint represents the price in the supply markets (wi)
– The dual multiplier for the demand constraint represents the price in the demand markets (p )the demand markets (pj)
• We can then formulate the equilibrium conditions in the following way (LCP):the following way (LCP):
![Page 15: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/15.jpg)
15
Example (cont.)
• So far we have assumed constant demand and supplysupply– Now suppose that demand and supplies are price responsive
• All markets are perfectly competitive
• An associated optimization problem can be used to compute the equilibrium prices and quantities
• Here, formulated as an NCP
![Page 16: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/16.jpg)
16
Example (cont.)
• We would then get the following equilibrium conditions:conditions:
![Page 17: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/17.jpg)
17
CP or optimization problem
• So far the spatial equilibrium model could have been formulated as an optimization problemformulated as an optimization problem
• What if we introduce competition in a spatial• What if we introduce competition in a spatial equilibrium model?
![Page 18: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/18.jpg)
18
Liberalized network industry
• Telecoms, energy, natural gas, railway, mail, ...V ti l ti• Vertical separation– upstream market: network infrastructure (natural monopoly characteristics)– downstream market: sales to customersdownstream market: sales to customers
• Many actors in downstream market:– use network infrastructure (lease, buy access, ...)– provide products, value-added services, ...– compete for customers– often also subsidiary of upstream actoroften also subsidiary of upstream actor
(separated former monopolist)
• Downstream actors act egoistically: make decisions which are best for them – and not for the industry (or social surplus etc.)
![Page 19: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/19.jpg)
19
Formulating and solving CPs
• Several articles explains how GAMS can be used to formulate CPC
– Extension of GAMS for complementarity problems arising in applied economic analysis, Journal of Economic Dynamics and Control, Rutherford, 1995
– Traffic Modeling and Variational Inequalities using GAMS, Dirkse and Ferris, 1997
– Complementarity Problems in GAMS and the PATH solver, Ferris and Munson, 1998
• Solvers: PATH and MILES• In addition: AMPL with KNITROIn addition: AMPL with KNITRO
![Page 20: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/20.jpg)
Example from the North sea gas pipeline network
A stochastic complementarity model
• Decision structure• Notation & model• Generalized Nash Equilibrium• Case results
20
![Page 21: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/21.jpg)
The North-Sea CaseThe North Sea Case
• Production nodes (with gas fields) transportation nodes market• Production nodes (with gas fields), transportation nodes, market nodes
• Roles: large producers, smaller producers (modelled as a competitive fringe) Gassco (independent system operator)fringe), Gassco (independent system operator).
21
![Page 22: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/22.jpg)
Representation of transport networkRepresentation of transport network
22
![Page 23: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/23.jpg)
Transportation market structureTransportation market structure
• Primary market• Primary market– Large producers have capacity booking rights
• Booking rights are more than 2 times the overall capacity• Conflicts resolved using Capacity Allocation Key• Tariff is fixed
• Secondary marketsy– ISO releases any available capacity– Bilateral trades of capacity between players
Price is negotiated– Price is negotiated– A competitive fringe clears the secondary market
23
![Page 24: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/24.jpg)
The decision structureThe decision structure
24
![Page 25: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/25.jpg)
Large producers decision problemLarge producers decision problem
• Each large producer faces a two• Each large producer faces a two-stage stochastic program with recourse
• This is still a one level game because the contingent strategy is laid at the time of booking and notlaid at the time of booking and not changed as a result of the other players bookings.
• Stochastic parameters:– Spot price– TOP volumes
25
![Page 26: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/26.jpg)
Purpose of the analysisPurpose of the analysis
• Investigate the effect of different objectives for the ISO• Investigate the effect of different objectives for the ISO– Max flow, max value and max social surplus
• Analyze the effect of stochasticity
26
![Page 27: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/27.jpg)
Price in the secondary marketPrice in the secondary market
• Demand from a competitive fringe in each field g• Demand from a competitive fringe in each field g– Comes from the profit maximization of the competitive fringe:
• The first order condition of optimality is:• The first order condition of optimality is:
• We assume that the cost function (W()) is quadratic. The inverse demand function can then be formulated as:demand function can then be formulated as:
27
![Page 28: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/28.jpg)
Large producersg pobjective function
Income from the spot markets
Cost of booking in the primary market
Income from the spot marketsand delivery in the TOP contracts
Surplus from trade in the secondary marketfor transportation capacityfor transportation capacity
Production cost
28
![Page 29: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/29.jpg)
Large producersconstraints
29
![Page 30: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/30.jpg)
ISO - objective function (1)ISO objective function (1)
• Maximize flow (MF):
30
![Page 31: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/31.jpg)
ISO – objective function (2)ISO objective function (2)
• Maximize value of flow (MVF):• Maximize value of flow (MVF):
31
![Page 32: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/32.jpg)
ISO - objective function (3)ISO objective function (3)
• Maximize social surplus (MSS):• Maximize social surplus (MSS):
The slope in the linear aggregated supply function is given as:The slope in the linear, aggregated supply function is given as:
32
![Page 33: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/33.jpg)
ISO – some of the constraintsISO some of the constraints
• Conservation of mass for the field nodes• Conservation of mass for the field nodes
• Conservation of mass for the junction nodes
• Conservation of mass for the market nodes
33
![Page 34: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/34.jpg)
ISOISO
• Positive price in the secondary market• Positive price in the secondary market
34
![Page 35: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/35.jpg)
BenchmarkBenchmark
• The ISO schedules production routing and sale in order to maximize• The ISO schedules production, routing and sale in order to maximize the social surplus of all the players in the network
• Objective function:
• Constraints:Constraints:
35
![Page 36: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/36.jpg)
Generalized Nash Equilibrium gamesGeneralized Nash Equilibrium games
36
![Page 37: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/37.jpg)
Common constraintsCommon constraints• The common constraints in our model
– Defined as constraints where decision variables from more than one player appear
37
![Page 38: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/38.jpg)
(Quasi) Variational inequalities(Quasi) Variational inequalities
38
![Page 39: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/39.jpg)
The VI solution• Theorems 4-6 from Harker (1991)
39
![Page 40: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/40.jpg)
The complementarity programThe complementarity programLarge producer n Independent System Operator
(ISO)Competitive fringeg p
• Time 0– Booking capacity– Price in spot market is
unknown– (contingent production and
(ISO)Time 0
– No decision
• Time 0– No decision
– (contingent production and capacity decisions)
–• Time 1
– Price in spot market is known
• Time 1– Routing decision– Sell spare capacity
• Time 1– Production decision
Buy capacity from ISOknown– Production decison
imlemented– Sell surplus capacity– Buy addtional capacity
Sell spare capacity – Buy capacity from ISO and/or large producers
KKT- conditions for n two-stage stochastic programs
KKT - conditions for s deterministic
1st order optimality conditions (wait and see)stochastic programs deterministic
optimization problems (this is a wait and see problem)
(wait and see)
40
![Page 41: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/41.jpg)
Network used in the analysisNetwork used in the analysis
• 2 production nodes• 2 production nodes• 1 junction node• 2 market nodes• 2 large producers (L1 and L2) are present in both production nodes• 1 competitive fringe in each production node
• Model solved by Path (to find VI solution)• Model solved by Path (to find VI solution)– Dual variables for the common constraints
are the same for all players
41
![Page 42: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/42.jpg)
Case 1: ResultsCase 1: Results
• Benchmark 7220,43Diffi l i h fl l i b f VI l i• Difficult to interpret the max flow solution because of VI solution
• MSS gives the largest total surplus in the network• MV gives the largest value of flow
42
MV gives the largest value of flow
![Page 43: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/43.jpg)
Case 1: Changed weighting for MFCase 1: Changed weighting for MF
• Corresponds to a change in currency from (1/100) NOK to EUR• Corresponds to a change in currency from (1/100) NOK to EUR
43
![Page 44: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/44.jpg)
Case 3: The effect of stochasticityCase 3: The effect of stochasticity
• What is the difference between a stochastic and a deterministic• What is the difference between a stochastic and a deterministic setting?– Cost of uncertainty– Wait-and-see solution (WSS)– Expected value of perfect information (EVPI)– Value of stochastic solution (VSS)Value of stochastic solution (VSS)
– 15 scenarios
44
![Page 45: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/45.jpg)
Case 3: stochastic solution and WSSCase 3: stochastic solution and WSS
• The EVPI is large for the MSS formulation• Benchmark is 9008,59
In the WSS solution the distance to the benchmark is only 1 1%– In the WSS solution the distance to the benchmark is only 1.1%
45
![Page 46: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/46.jpg)
Case 3: Expected result of using the expected value solution
• We first solve a deterministic problem where the stochastic parameters were replaced with their expected values (EVP)
• We then fixed the first stage decisions from the EVP solution and solved the stochastic problemstochastic problem
• For the MSS formulation: the stochastic first stage solution is 1,77% better than the EVP first stage solution (in the stochastic problem)
46
![Page 47: A stochastic generalized nash model for natural gas ... › ... › web › material › slides_tomasgard.pdf · A stochastic generalized nash model for natural gas transport systemsgas](https://reader034.vdocument.in/reader034/viewer/2022042402/5f13b24fa6daae4e0a790cfc/html5/thumbnails/47.jpg)
ConclusionsConclusions
• Both the MSS and MVF have meaningful interpretation when finding a• Both the MSS and MVF have meaningful interpretation when finding a VI solution. The max flow formulation only when we scale down the other players objectives!
• The inclusion of stochasticity leads to inefficiencies in the network– Both social surplus and surplus for the large producers are affected– Both social surplus and surplus for the large producers are affected– The flow is higher in the stochastic setting than in the wait and see
solution where booking rights are exercises just before production takes placeplace.
47