Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros...
Transcript of Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros...
![Page 1: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/1.jpg)
DTU Electrical Engineering26 February 2019
Data-driven Security-Constrained OPFSpyros ChatzivasileiadisAssociate Professor, DTU
work with: Lejla Halilbasic, Florian Thams, Jeanne Mermet-Guyennet, Andrea Tosatto, Andreas Venzke, Georgia Tsoumpa, Riccardo Zanetti
ENTSO-e, 27-Feb-2019
1
![Page 2: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/2.jpg)
DTU Center of Electric Power and Energy26 February 2019
Relevance to the ENTSO-e studyEuropean Grid 2030 and 2050• Run large-scale studies on a (reduced) ENTSO-e network of 7500 nodes, including
TYNDP• Assess impact of RES penetration and grid reinforcements on the nodal prices
European zonal markets (www.multi-dc.eu)• Cost recovery of AC and HVDC losses: Develop a methodology to integrate both AC
losses and HVDC losses in zonal markets
Redispatching (www.multi-dc.eu) (not in this presentation)• Sharing reserves between areas to avoid redispatching due to N-1/low inertia
Data-driven Security-Constrained OPF• Introduce a new market and operation framework that is scalable• Can handle limitations such as uncertainty, flexible power flows, etc.
2
![Page 3: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/3.jpg)
DTU Center of Electric Power and Energy26 February 2019
Developing the Business-as-Usual Scenario
ENTSO-E TYNDP eHighway2050
2030 Grid+
~2016 Load and installed RES
ENTSO-e statisticalfactsheet 2016
![Page 4: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/4.jpg)
DTU Center of Electric Power and Energy26 February 2019
Developing the Business-as-Usual Scenario
ENTSO-E TYNDP eHighway2050
2030 Grid+
~2016 Load and installed RES
Business-as-Usual2030
Add 2030 projections: (EUCO30)• +20% Load • +440% Offshore Wind • +80% Onshore Wind • +140% PV
ENTSO-e statisticalfactsheet 2016
![Page 5: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/5.jpg)
DTU Center of Electric Power and Energy26 February 2019
The Business-as-Usual Network 2030
• ~7500 nodes • 8900 AC lines • ~1000 transformers• 1269 generators • 140 DC lines • 270 converters
Characteristics
RES installed capacity: 63%
160 Hydro power plants
40 PVs
290 Wind farms
60% of RES in the distribution network
![Page 6: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/6.jpg)
DTU Center of Electric Power and Energy26 February 2019
The Business-as-Usual Network 2030
• ~7500 nodes • 8900 AC lines • ~1000 transformers• 1269 generators • 140 DC lines • 270 converters
Characteristics
BaURES Penetration 49,16%Load Shedding 0,37%
RES installed capacity: 63%
160 Hydro power plants
40 PVs
290 Wind farms
![Page 7: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/7.jpg)
DTU Center of Electric Power and Energy26 February 2019
Best Paths: Scalability Assessment for 2030
AC Repowering HVDC AC & HVDC
![Page 8: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/8.jpg)
DTU Center of Electric Power and Energy26 February 2019
Transmission investments to increase RES penetration by 2030
51
52
53
54
55
56
57
BaU2030 DC Scalability 1 DC Scalability 2 DC Scalability 3 DC Scalability 4
RES
Pen
etra
tion
(%)
• 50 DC lines (150 GW)• 56 AC lines require upgrade (63.62 GW)• 20/56 lines: <20% increase
Highest grid development stage, AC+DC upgrades: 56.06%
Maximum possible RES penetration: 56.47%
Only DC: 55.86%
Only AC: 53.03%
BaU: 52.85%
![Page 9: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/9.jpg)
DTU Center of Electric Power and Energy26 February 2019
Reducing operating costs by 2030
40
41
42
43
44
45
46
47
48
49
50
51
BaU2030 DC Scalability 1 DC Scalability 2 DC Scalability 3 DC Scalability 4
Gen
erat
ion
Cos
t (B
€/a)
BaU
only DC
AC+DC(but with trafo
limits)
Savings:
6.3 B€/a
![Page 10: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/10.jpg)
DTU Center of Electric Power and Energy26 February 2019
DC Upgrade: Congestion Duration Curves
Most of the new DC lines are used up to their limit more than 50% of the time
![Page 11: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/11.jpg)
DTU Center of Electric Power and Energy26 February 2019
Nodal prices in 2030 (work in progress, no upgrades)
11
![Page 12: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/12.jpg)
DTU Center of Electric Power and Energy26 February 2019
Nodal prices 2016 vs 2030 (work in progress)
12
• Increased RES presentation increases the variance of nodal prices
• Nodal prices follow a multi-modal distribution from 3 modes to 4 modes
• Producer surplus decreases
![Page 13: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/13.jpg)
DTU Center of Electric Power and Energy26 February 2019 13
Main Takeaways• Offshore wind capacity only deployable with additional corridors
• Controllable flows are required: DC upgrades have shown substantially better performance
• Optimal grid development: AC & DC AC upgrades enable full potential of DC & vice versa
• Transformer bottlenecks need to be considered
Nodal pricing (preliminary insights)
• France seems to experience often extremely high prices at certain regions
• Prices within the former UCTE synchronous areas 1 and 2 seem to be more uniform
• Increased RES presentation increases the variance of nodal prices
• Nodal prices follow a multi-modal distribution
![Page 14: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/14.jpg)
DTU Center of Electric Power and Energy26 February 2019
Market Integration of HVDC: Pricing losses
14
![Page 15: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/15.jpg)
DTU Center of Electric Power and Energy26 February 2019
• HVDC interconnectors:
• usually longer than AC interconnectors
• often connecting areas belonging to different TSOs (at least in Europe)
• As a result, the losses occurring on HVDC lines are not negligible, and the cost has to be shared among TSOs
• If price difference between areas is small, TSOs cannot recover the cost of HVDC losses, i.e. cost of losses higher than potential revenue
Under construction Operative
Motivation
![Page 16: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/16.jpg)
DTU Center of Electric Power and Energy26 February 2019
• In 2017 the price difference between SE3 and DK1 has been zero for more than 5300 hours (61%), resulting in 1.2 M€ losses.
Source: https://www.nordpoolgroup.com/
NORWAY
SWEDEN
DENMARK
KONTISKAN
Some examples - Denmark
• In 2017 the price difference between DK1 and DK2 has been zero for more than 6400 hours (73%), resulting in 0.8 M€ losses.
• In 2017 the price difference between DK1 and NO2 has been zero for more than 4000 hours (47%), resulting in 3.2 M€ losses.
NORWAY
SWEDEN
DENMARK
SKAGERRAK
NORWAY
SWEDEN
DENMARK
GREAT-BELT
![Page 17: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/17.jpg)
DTU Center of Electric Power and Energy26 February 2019
• In 2017 the price difference between FI andEE has been zero for more than 6600 hours(76%), resulting in 3 M€ losses.
Source: https://www.nordpoolgroup.com/
Some examples - Finland
• In 2017 the price difference between FI andSE3 has been zero for more than 8600 hours(99%), resulting in 3.8 M€ losses.
• For these 5 HVDC interconnectors, losses amounts to 12 M€ per year.
• Considering the number of HVDC interconnectors and all the newprojects, this number is intended to grow significantly.
SWEDEN
FINLAND
ESTONIA
FENNOSKAN
SWEDEN
FINLAND
ESTLINK
ESTONIA
![Page 18: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/18.jpg)
DTU Center of Electric Power and Energy26 February 2019
• Losses are handled in a different way for AC and HVDC lines.
• For HVDC lines:
• To move from the explicit to the implicit method, a loss factor has to be includedin the market clearing algorithm.
• Is it a good idea to introduce a loss factor only for HVDC lines in meshedgrids?
EXPLICIT METHOD IMPLICIT METHOD
WITH LOSSES PRICE DIFFERENCE POWER
EXCHANGESWITHOUT LOSSES
POWER
EXCHANGESTSO
LF
Problem statement
![Page 19: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/19.jpg)
DTU Center of Electric Power and Energy26 February 2019
* Fingrid, Energinet, Statnett, Svenska Kraftnät, Analyses on the effects of implementing implicit grid losses in the Nordic CCR, April 2018
Implicit grid losses - Nordic CCR
LT
LV
EE
FI
DK1SE4
SE3
SE2
SE1
NO1
NO2
NO5NO3
NO4
DK2
• Nordic TSOs, April 2018: Analyses on the effects of implementing implicit grid losses in the Nordic CCR
• All simulations with implicit grid losses show an economic benefit
• One exception is FennoSkan
• The higher the number of lines with implicit losses implemented, the higher the benefit
EE
FI
SE3
SE2
SE1
NO1
NO2
NO5NO3
NO4
![Page 20: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/20.jpg)
DTU Center of Electric Power and Energy26 February 2019
• Simple linear model
• Losses
~~
𝑓𝑓𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻
• Quadratic losses for HVDC converters and HVDC lines
• Losses are considered as an extra load equally shared by the sending and the receiving node
• Linearization of losses for their introduction in the market clearing algorithm
~
𝑃𝑃𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻 𝑃𝑃𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻
HVDC line model
![Page 21: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/21.jpg)
DTU Center of Electric Power and Energy26 February 2019
Priceswithout LF($/MWh)
with LF($/MWh)
Zone 1 10.00 10.00
Zone 2 10.00 10.00
Zone 3 10.00 10.38
Example: impact on prices
~~1 2
3
292 MW
20$/MWh 10$/MWh
Priceswithout LF($/MWh)
with LF($/MWh)
Zone 1 20.00 20.00
Zone 2 10.00 10.00
Zone 3 20.00 20.00
Priceswithout LF($/MWh)
with LF($/MWh)
Zone 1 20.00 20.00
Zone 2 20.00 20.82
Zone 3 20.00 21.61
No CongestionWith Congestion Loss is absorbed in the congestion rent
Inter-TSO compensation2 DC lines:
1-2, 2-3
![Page 22: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/22.jpg)
DTU Center of Electric Power and Energy26 February 2019
AREA 2AREA 1 AREA 3
Test case: IEEE RTS system
![Page 23: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/23.jpg)
DTU Center of Electric Power and Energy26 February 2019
For each area:
• 32 producers and 17 consumers
Flow-based market coupling
• Estimation of the PTDF matrix: marginal variations in one generator at a time
• Emulate how TSOs purchase the required power to cover their losses
• Equilibrium problem: each market participant seeks to maximize its profit
Test case: IEEE RTS system
A. Tosatto, T. Weckesser, S. Chatzivasileiadis, Market Integration of HVDC lines, Submitted. Available: https://arxiv.org/abs/1812.00734
![Page 24: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/24.jpg)
DTU Center of Electric Power and Energy26 February 2019
Loss factor AC losses HVDC losses
NO LF 254.6 MW 4.5 MW
HVDC LF 250.5 MW 2.2 MW
Δ Cost HVDC Δ Cost AC
84.58 $/h 151.39 $/h
NO LOSS FACTOR
391 MW864 MW
79 MW
36.11 $/MWh
36.11 $/MWh
36.11 $/MWh
25 MW
LOSS FACTOR
36.11 $/MWh
36.11 $/MWh
36.11 $/MWh
832 MW 425 MW
70 MW
Economic benefit: 235.97 $/h
Test case 1: some results
Cheap area Cheap area
Expensive area Expensive area
![Page 25: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/25.jpg)
DTU Center of Electric Power and Energy26 February 2019
Loss factor AC losses HVDC losses
NO 182.5 MW 6.9 MW
HVDC 186.9 MW 4.0 MW
Δ Cost HVDC Δ Cost AC
282.66 $/h -484.25 $/h
NO LOSS FACTOR
18 MW332 MW
121 MW
42.24 $/MWh
42.24 $/MWh
42.24 $/MWh
117 MW
LOSS FACTOR
42.24 $/MWh
43.96 $/MWh
44.69 $/MWh
352 MW 7 MW
150 MW
Economic loss: 201.59 $/h
66 MW
Test case 1: some results
Cheap area Cheap area
Expensive area Expensive area
![Page 26: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/26.jpg)
DTU Center of Electric Power and Energy26 February 2019
AC loss factors
Line 2-3
More info: A. Tosatto, T. Weckesser, S. Chatzivasileiadis, Market Integration of HVDC lines, available: https://arxiv.org/abs/1812.00734www.multi-dc.eu
System emulating DK1, DK2, SE, NO
![Page 27: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/27.jpg)
DTU Center of Electric Power and Energy26 February 2019
• The results show that the introduction ofHVDC loss factors for this specificsystem is beneficial for most of thetime.
• However, there are cases where thesocial welfare is decreased (>14%).
• Theory guarantees that this does nothappen with LFs for both AC and DCsystems
• The introduction of AC-LFs double thebenefit.
Economic evaluation
More info: A. Tosatto, T. Weckesser, S. Chatzivasileiadis, Market Integration of HVDC lines, available: https://arxiv.org/abs/1812.00734www.multi-dc.eu
![Page 28: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/28.jpg)
DTU Center of Electric Power and Energy26 February 2019
• Is it a good idea to introduce a loss factor only for HVDC lines in meshed grids?
• The HVDC loss factor can act positively or negatively w.r.t. the amount of system losses depending on the system under investigation.
• If to be introduced in the market clearing algorithm, the recommendation is to consider the losses in both AC and DC systems.
Conclusion
![Page 29: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/29.jpg)
DTU Center of Electric Power and Energy26 February 2019
Data-driven Security-Constrained OPF
29
![Page 30: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/30.jpg)
DTU Center of Electric Power and Energy26 February 2019
The feasible space of power system operations
30
• Nonlinear and nonconvex
AC power flow equations
• Component limits
AC-OPF
![Page 31: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/31.jpg)
DTU Center of Electric Power and Energy26 February 2019 31
• Nonlinear and nonconvex AC
power flow equations
• Component limits
+ Stability limits
The feasible space of power system operations
![Page 32: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/32.jpg)
DTU Center of Electric Power and Energy26 February 2019 32
• Nonlinear and nonconvex AC
power flow equations
• Component limits
+ Other security criteria (e.g., N-1)
The feasible space of power system operations
![Page 33: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/33.jpg)
DTU Center of Electric Power and Energy26 February 2019 33
• Nonlinear and nonconvex AC power
flow equations
• Component limits
+ Uncertainty ξ in nodal power
injections
The feasible space of power system operations
![Page 34: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/34.jpg)
DTU Center of Electric Power and Energy26 February 2019 34
• Nonlinear and nonconvex AC power
flow equations
• Component limits
+ Stability limits
+ Other security criteria (e.g., N-1)
+ Uncertainty ξ in nodal power injections
The feasible space of power system operations
![Page 35: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/35.jpg)
DTU Center of Electric Power and Energy26 February 2019 35
• Nonlinear and nonconvex AC power
flow equations
• Component limits
+ Stability limits
+ Other security criteria (e.g., N-1)
+ Uncertainty ξ in nodal power
injections
The feasible space of power system operations
with prob. “1-ε”
with prob. “ε”
![Page 36: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/36.jpg)
DTU Center of Electric Power and Energy26 February 2019
Operational Challenges
36
• Identifying the boundary of the feasible
operating region
• Incorporating the boundary in an
optimization framework
• Finding the true optimal solution &
maintaining computational efficiency
![Page 37: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/37.jpg)
DTU Center of Electric Power and Energy26 February 2019
How to encode the feasible operating regionfor electricity markets?
37
Security considerations live in
AC space, but market is based
on DC approximations!
![Page 38: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/38.jpg)
DTU Center of Electric Power and Energy26 February 2019
How to encode feasible operating regionfor electricity markets?
38
TSO MarketNTC
Traditionally, TSOs define
Net-Transfer Capacities
Security considerations live in
AC space, but market is based
on DC approximations!
![Page 39: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/39.jpg)
DTU Center of Electric Power and Energy26 February 2019
How to encode feasible operating regionfor electricity markets?
39
TSO MarketNTC
Traditionally, TSOs define
Net-Transfer Capacities
Security considerations live in
AC space, but market is based
on DC approximations!
![Page 40: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/40.jpg)
DTU Center of Electric Power and Energy26 February 2019
Better but reality of power system operations is nonconvex!
40
Improvements with Flow-Based
Market Coupling,
but still a single convex region!
![Page 41: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/41.jpg)
DTU Center of Electric Power and Energy26 February 2019
Better but reality of power system operations is nonconvex!
41
Improvements with Flow-Based
Market Coupling,
but still a single convex region!Figure from: KU Leuven Energy Institute, “EI Fact Sheet: Cross-border Electrcity Trading: Towards Flow-based Market Coupling,” 2015. [Online].Available: http://set.kuleuven.be/ei/factsheets
![Page 42: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/42.jpg)
DTU Center of Electric Power and Energy26 February 2019
Our proposal• Pass the data! (not the line limits)
• TSOs have large amounts of data related to their daily contingency analysis calculations (RSCIs also collect and coordinate these analyses in a regional level)
• Instead of deciding on the interconnection flow limits to form a single convex region, let the market operator decide based on the “secure/insecure” status of an operating point
Benefits• Market players have the true, larger (and non-convex) feasible area at their disposal
possibility to determine a better (true) global optimum
• Uncertainty and security constraints incorporated in the data extremely scalable optimization (MILP)
• Map AC limits to DC-OPF eliminate redispatching• Can apply to both zonal and nodal markets
42
![Page 43: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/43.jpg)
DTU Center of Electric Power and Energy26 February 2019
How does it work?
43
Database of secure and insecure operating
points
{P,Q,V,θ,ζ}
Operating points provided by the TSOs through
simulated and real data
Train a decision tree to classify secure and
insecure regions
Exact reformulation to MILP
![Page 44: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/44.jpg)
DTU Center of Electric Power and Energy26 February 2019 44
Database of stable and unstable OPs
{P,Q,V,θ,ζ}
Decision Tree
Offline security assessment
stable
unstable
Data-driven security-constrained OPF
![Page 45: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/45.jpg)
DTU Center of Electric Power and Energy26 February 2019 45
Database of stable and unstable OPs
{P,Q,V,θ,ζ}
Decision Tree
Offline security assessment
stable
unstable
Data-driven security-constrained OPF
Partitioning the secure operating region
![Page 46: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/46.jpg)
DTU Center of Electric Power and Energy26 February 2019
Data-driven security-constrained OPF
46
Database of stable and unstable OPs
{P,Q,V,θ,ζ}
Decision Tree
Offline security assessment
stable
unstable
Partitioning the secure operating region
![Page 47: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/47.jpg)
DTU Center of Electric Power and Energy26 February 2019
Data-driven security-constrained OPF
47
Database of stable and unstable OPs
{P,Q,V,θ,ζ}
Decision Tree
Offline security assessment
stable
unstable
Partitioning the secure operating region
![Page 48: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/48.jpg)
DTU Center of Electric Power and Energy26 February 2019
Data-driven security-constrained OPF
48
Database of stable and unstable OPs
{P,Q,V,θ,ζ}
Decision Tree
Offline security assessment Optimization
stable
unstable
Integer Programming to incorporate partitions (DT)
• DC-OPF (MILP)
• AC-OPF (MINLP)
• Relaxation (MIQCP, MISOCP)
• Each leaf is a convex region
• FBMC corresponds to the leaf that maps the largest convex region
• Here, each convex region includes security constraints (N-1 and other) we can include N-1 in zonal markets
![Page 49: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/49.jpg)
DTU Center of Electric Power and Energy26 February 2019
We gain ~22% of the feasible space using data and Mixed Integer Programming
49
Largest convex region
covers ~78%
![Page 50: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/50.jpg)
DTU Center of Electric Power and Energy26 February 2019
MIP + convex AC-OPF approximation finds bettersolutions than nonconvex problem!
50
Optimum located at boundary of consideredsecurity region
L. Halilbašić, F. Thams, A. Venzke, S. Chatzivasileiadis, and P. Pinson, ”Data-driven security-constrained AC-OPF for operations and markets,” in 2018 Power Systems Computation Conference (PSCC), 2018.
![Page 51: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/51.jpg)
DTU Center of Electric Power and Energy26 February 2019
Works also for DC-OPF (MILP): Market dispatch is N-1 secure and stable
Eliminate redispatching costs
51
• Data-driven SC-OPF for markets: DC-OPF becomes MILP– But, MILP is already included in market
software (e.g. Euphemia, for block offers, etc.)
– Efficient MILP solvers already existing
Redispatching costs: over 400 Million Euros in a year, just for Germany
![Page 52: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/52.jpg)
DTU Center of Electric Power and Energy26 February 2019
Works also for DC-OPF (MILP): Market dispatch is N-1 secure and stable
all eigenvalues on the left-hand side
52
![Page 53: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/53.jpg)
DTU Center of Electric Power and Energy26 February 2019
Efficient Database Generation
• Modular and highly efficient algorithm
• Can accommodate numerous definitions of power system security (e.g. N-1, N-k, small-signal stability, voltage stability, transient stability, or a combination of them)
• 10-20 times faster than existing state-of-the-art approaches
• Our use case: N-1 security + small-signal stability
• Generated Database for NESTA 162-bus system online available!https://github.com/johnnyDEDK/OPs_Nesta162Bus (>500,000 points)
F. Thams, A. Venzke, R. Eriksson, and S. Chatzivasileiadis, ”Efficient database generation for data-driven security assessment of power systems”. Accepted in IEEE Trans. Power Systems, 2019. https://www.arxiv.org/abs/1806.0107.pdf
53
![Page 54: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/54.jpg)
DTU Center of Electric Power and Energy26 February 2019
Results
Points close to the security boundary (within distance γ)IEEE 14-bus NESTA 162-bus
Brute Force 100% of points in 556.0 min intractableImportance Sampling 100% of points in 37.0 min 901 points in 35.7 hoursProposed Method 100% of points in 3.8 min 183’295 points in 37.1 hours
54
• Further benefits for the decision tree:
• Higher accuracy
• Better classification quality (Matthews correlation coefficient)
Generated Database for NESTA 162-bus system online available! https://github.com/johnnyDEDK/OPs_Nesta162Bus
![Page 55: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/55.jpg)
DTU Center of Electric Power and Energy26 February 2019
Conclusions 1/2• Large-scale simulations on a ~7500 node European grid
– Controllable flows and a combined AC & DC upgrade are necessary to accommodate increased RES
– 2030: nodal price variance increases, multimodal distribution of prices, very-high-price pockets within one zone (e.g. France)
– Work in progress: how will the nodal prices get affected by the planned grid reinforcements?
• HVDC: offers the necessary power flow flexibility, but TSOs cannot recover the cost of losses– Introduction of loss factors: for a guaranteed social welfare increase, loss factors for both
AC and DC lines must be introduced– Alternative: move to a nodal market with a (linearized) AC-OPF [some US ISOs already
run similar algorithms)– Work in progress: how can we have an “LMP” for HVDC flexibility?
55
![Page 56: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/56.jpg)
DTU Center of Electric Power and Energy26 February 2019
Conclusions 2/2• Introduced a new Data-driven Security-Constrained OPF framework
– Tractable and scalable– Can handle any security criteria– Can handle uncertainty– Can handle topology changes and controllable flows by HVDC– Pass the data, not the line limits
• Currently working on (knowledge gaps)– Locational Price for HVDC flexibility (and possible extensions to other topology changes)– Evolution of nodal prices with RES penetration and grid reinforcements– Scalable formulations to include security and uncertainty in OPF
56
![Page 57: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/57.jpg)
DTU Center of Electric Power and Energy26 February 2019
Thank you!www.chatziva.com/publications
References:
L. Halilbašić, F. Thams, A. Venzke, S. Chatzivasileiadis, and P. Pinson, ”Data-driven security-constrained AC-OPF for operations and markets,” in 2018 Power Systems Computation Conference (PSCC), 2018.
F. Thams, L. Halilbašić, P. Pinson, S. Chatzivasileiadis, and R. Eriksson, ”Data-driven security-constrained OPF,” in 10th IREP Symposium – Bulk Power Systems Dynamics and Control, 2017.
F. Thams, A. Venzke, R. Eriksson, and S. Chatzivasileiadis, ”Efficient database generation for data-driven security assessment of power systems”. IEEE Trans. Power Systems, 2019. arXiv: http://arxiv.org/abs/1806.01074.pdf .
A. Tosatto, T. Weckesser, S. Chatzivasileiadis, Market Integration of HVDC lines, submitted. https://arxiv.org/abs/1812.00734
L. Halilbašić, P. Pinson, and S. Chatzivasileiadis, ”Convex relaxations and approximations of chance-constrained AC-OPF problems,” IEEE Transactions on Power Systems, 2018, (in press).
A. Venzke, L. Halilbasic, U. Markovic, G. Hug, S. Chatzivasileiadis., ”Convex relaxations of chance constrained AC optimal power flow,” IEEE Transactions on Power Systems, vol. 33, no. 3, pp. 2829-2841, May 2018.
57
![Page 58: Data-driven Security-Constrained OPF - ChatzivaData-driven Security-Constrained OPF Spyros Chatzivasileiadis Associate Professor, DTU. work with: Lejla Halilbasic, Florian Thams, Jeanne](https://reader030.fdocuments.in/reader030/viewer/2022040611/5ed8cec36714ca7f47689a43/html5/thumbnails/58.jpg)
DTU Center of Electric Power and Energy26 February 2019
How to model redispatching• Different alternatives (cooptimization, running series of load flows, AC-OPF)
• Suggestion: run a linearized AC-OPF– Objective function: minimize costs * (Predisp-Pinit)– Linearize power flow equations around the selected operating point, determined
by the zonal (day-ahead) market, and extract linear sensitivities– Solve a convex optimization problem
– Several approaches available– It has been shown that as long as we are focusing on a solution close to a selected
operating point, linear approximations of AC power flow can offer an acceptable accuracy
58