pool strategy of an electricity producer with
play

Pool strategy of an electricity producer with endogenous formation - PowerPoint PPT Presentation

Pool strategy of an electricity producer with endogenous formation of clearing prices Antonio J. Conejo, Carlos Ruiz University of Castilla-La Mancha, Spain, 2011 Contents Background and Aim Approach Model Features Model


  1. Pool strategy of an electricity producer with endogenous formation of clearing prices Antonio J. Conejo, Carlos Ruiz University of Castilla-La Mancha, Spain, 2011

  2. Contents • Background and Aim • Approach • Model Features • Model Formulation – Deterministic – Stochastic • Examples • Conclusions 29/04/2011 2

  3. Contents • Background and Aim • Approach • Model Features • Model Formulation – Deterministic – Stochastic • Examples • Conclusions 29/04/2011 3

  4. Background and Aim Strategic power producer • Comparatively large number of generating units • Units distributed throughout the power network 29/04/2011 4

  5. Background and Aim Pool-based electricity market • Cleared once a day, one-day ahead and on a hourly basis • DC representation of the network including first and second Kirchhoff laws • Hourly Locational Marginal Prices (LMPs) 29/04/2011 5

  6. Background and Aim Strategic power producer Best offering strategy to maximize profit Pool-based electricity market 4/29/2011 6

  7. Background and Aim • Considering the market: MPEC formulation • Considering the real-world: Stochastic formulation • Stochastic MPEC! 29/04/2011 7

  8. Contents • Background and Aim • Approach • Model Features • Model Formulation – Deterministic – Stochastic • Examples • Conclusions 29/04/2011 8

  9. Approach Bilevel model: Profit Maximization Upper-Level Offering LMPs subject to curve Dual Variables Social Welfare Maximization Lower-Level (Market Clearing) 4/29/2011 9

  10. Approach • Bilevel model: Optimization problem constrained by other optimization problem (OPcOP)!

  11. OPcOP 11

  12. Approach MPEC: Profit Maximization Upper-Level Offering subject to LMPs curve KKT Conditions 4/29/2011 12

  13. MPEC 13

  14. MPEC 14

  15. Contents • Background and Aim • Approach • Model Features • Model Structure – Deterministic – Stochastic • Examples • Conclusions 29/04/2011 15

  16. Features 1) Strategic offering for a producer in a pool with endogenous formation of LMPs. 2) Uncertainty of demand bids and rival production offers. 3) MPEC approach under multi-period, network- constrained pool clearing. 4) MPEC transformed into an equivalent MILP. 29/04/2011 16

  17. Contents • Background and Aim • Approach • Model Features • Model Formulation – Deterministic – Stochastic • Examples • Conclusions 29/04/2011 17

  18. Deterministic Model Upper- Level → Profit Maximization: Costs - Revenues Ramping Limits Price = Balance dual variable 4/29/2011 18

  19. Deterministic Model Upper- Level → Profit Maximization: Dual variable 29/04/2011 19

  20. Deterministic Model Lower- Level → Market Clearing Maximize Social Welfare Power Balance 29/04/2011 20

  21. Deterministic Model Lower- Level → Market Clearing Price 29/04/2011 21

  22. Deterministic Model Lower- Level → Market Clearing Production / Demand Power Limits Transmission Capacity Limits Angle Limits 4/29/2011 22

  23. Deterministic Model Lower- Level → Market Clearing 29/04/2011 23

  24. Deterministic Model Lower- Level → Market Clearing → KKT conditions 29/04/2011 24

  25. Deterministic Model Lower- Level → Market Clearing → KKT conditions 29/04/2011 25

  26. Deterministic Model MPEC model KKT Lower-Level 29/04/2011 26

  27. Deterministic Model Linearizations The MPEC includes the following non-linearities:    1) The complementarity conditions ( ). 0 a b 0 2) The term in the objective function. 4/29/2011 27

  28. Deterministic Model Linearizations → Complementarity Conditions    0 a b 0 Fortuny-Amat  transformation  a 0  b 0  a uM   b ( 1 u ) M    u 0 , 1 M Large enough constant (but not too large) 4/29/2011 28

  29. Deterministic Model Linearizations → Term: Based on the strong duality theorem and some of the KKT equalities 29/04/2011 29

  30. Contents • Background and Aim • Approach • Model Features • Model Formulation – Deterministic – Stochastic • Examples • Conclusions 29/04/2011 30

  31. Stochastic Model Uncertainty incorporated using a set of scenarios modeling different realizations of: • Consumers’ bids • Rival producers’ offers 29/04/2011 31

  32. Stochastic Model Deterministic model for each scenario S Pairs of production quantities ( ) and P  tib  tn market prices ( ).  Building of the optimal offering curve 4/29/2011 32

  33. Stochastic Model Optimal offering curve 29/04/2011 33

  34. Stochastic Model To ensure that the final offering curves are increasing in price some additional constraints are needed: These constraints link the individual problems increasing the computational complexity of the model. 4/29/2011 34

  35. Stochastic Model Math Structure 4/29/2011 35

  36. Stochastic Model Math Structure 4/29/2011 36

  37. Stochastic Model Math Structure 1. Direct solution: CPLEX, XPRESS 2. Decomposition procedures (Lagrangian Relaxation) 4/29/2011 37

  38. Contents • Background and Aim • Approach • Model Features • Model Formulation – Deterministic – Stochastic • Examples • Conclusions 29/04/2011 38

  39. Examples Six- bus test system→ electricity network 29/04/2011 39

  40. Examples Six- bus test system→ demand curve 29/04/2011 40

  41. Examples Six- bus test system→ generating units 29/04/2011 41

  42. Examples Six- bus test system→ uncongested network results The maximum power flow through lines 2-4, 3-6 and 4-6 are 269.62, 229.44 and 39.6933 MW respectively 4/29/2011 42

  43. Examples Six- bus test system→ uncongested network results 29/04/2011 43

  44. Examples Six- bus test system→ congested network results Capacity of line 3-6 limited to 230 MW: 29/04/2011 44

  45. Examples Six- bus test system→ congested network results Capacity of line 3-6 limited to 230 MW: 29/04/2011 45

  46. Examples Six- bus test system→ congested network results Capacity of line 4-6 limited to 39 MW: 29/04/2011 46

  47. Examples Six- bus test system→ stochastic model • Uncongested network case • 8 equally probable scenarios • They differ on the rival producer offers ( ) and  O tjb   D on the consumer bids ( ) tdk  • Selected to obtain a wide range of prices 29/04/2011 47

  48. Examples Six- bus test system→ stochastic model results 29/04/2011 48

  49. Examples Six- bus test system→ stochastic model results Offering curves for strategic generator 1 29/04/2011 49

  50. Examples Six- bus test system→ stochastic model results Offering curves for strategic generator 1 29/04/2011 50

  51. Examples IEEE One Area Reliability Test System • 24 Nodes • 8 strategic units • 24 non-strategic units • 17 consumers • 24 hours 29/04/2011 51

  52. Examples IEEE One Area Reliability Test System → Results 29/04/2011 52

  53. Examples IEEE One Area Reliability Test System → Results Marginal cost offer Strategic offer 29/04/2011 53

  54. Examples Computational issues • Model solved using CPLEX 11.0.1 under GAMS on a Sun Fire X4600 M2 with 4 processors at 2.60 GHz and 32 GB of RAM. 6-bus 6-bus 6-bus Model IEEE RTS uncongested congested stochastic CPU Time [s] 2.91 5.82 204.77 449.33 4/29/2011 54

  55. Contents • Background and Aim • Approach • Model Features • Model Formulation – Deterministic – Stochastic • Examples • Conclusions 29/04/2011 55

  56. Conclusions • Procedure to derive strategic offers for a power producer in a network constrained pool market. – LMPs are endogenously generated: MPEC approach. – Uncertainty is taken into account. – Resulting MILP problem. • Exercising market power results in higher profit and lower production. • Network congestion can be used to further increase profit. 29/04/2011 56

  57. Thanks for your attention! http://www.uclm.es/area/gsee/web/antonio.htm 29/04/2011 57

  58. Appendix A Computational Issues • Model has been solved using CPLEX 11.0.1 under GAMS on a Sun Fire X4600 M2 with 4 processors at 2.60 GHz and 32 GB of RAM. • The computational times are highly dependent on the values of the linearization constants M . 6-bus 6-bus 6-bus Model IEEE RTS uncongested congested stochastic CPU Time [s] 2.91 5.82 204.77 449.33 4/29/2011 58

  59. Appendix A Computational Issues Heuristic to determine the value of M: 1. Solve a (single-level) market clearing considering that all the producers offer at marginal cost. 2. Obtain the marginal value of each relevant constraint. 3. Compute the value of each relevant constant as:   100    M dual variable value 1 4/29/2011 59

  60. Appendix B Stochastic model 29/04/2011 60

  61. Appendix B Stochastic model 29/04/2011 61

  62. Appendix B Stochastic model 29/04/2011 62

Recommend


More recommend