Market Design Considerations for Scarcity Pricing: A Stochastic ... … · Motivation of Scarcity...

Market Design Considerations forScarcity Pricing:

A Stochastic Equilibrium Framework

Anthony PapavasiliouCenter for Operations Research and Econometrics

Université catholique de Louvain

Joint with Yves Smeers, Gauthier de Maere d’Aertrycke

Workshop on Electricity Systems of the FutureMathematics of Energy Systems

Isaac Newton Institute, Cambridge

1 / 57

Outline

1 ContextMotivation of Scarcity PricingHow Scarcity Pricing WorksModeling Alternative Scarcity Pricing Designs

2 Building Up Towards the Benchmark US Design (SCV)Energy-Only Real-Time MarketEnergy Only in Real Time and Day AheadAdding Uncertainty in Real TimeReserve Capacity

3 A Sketch of the European Design (REP)

4 Belgian Case Study

2 / 57

Outline





3 / 57

A Paradox of Highly Renewable SystemsGas and oil units are (i) the most flexible, and (ii) the leastprofitable

Inv. cost Marg. cost Min load Energy market Profit(e/MWh) (e/MWh) cost (e/MWh) profit (e/MWh) (e/MWh)

Biomass 27.9 5.6 0 35.6 7.7Nuclear 31.8 7.0 0 34.2 2.4

Gas 5.1 50.2 20 0.1 -5Oil 1.7 156.0 20 0 -1.7

4 / 57

Motivation for Scarcity Pricing

Scarcity pricing: a real-time demand for reserve capacity,determined by loss of load probability

introduces a non-volatile real-time price for reserve capacityaffects the real-time price of energy

Definition of flexibility for this talk:Secondary reserve: reaction in a few seconds, fullresponse in 7.5 minutesTertiary reserve: available within 15 minutes

such as can be provided bycombined cycle gas turbinesdemand response

5 / 57

The CREG Scarcity Pricing Studies

First study (2015): How would electricity prices change ifwe introduce ORDC (Hogan, 2005) in the Belgian market?Second study (2016): How does scarcity pricing dependon

Strategic reserveValue of lost loadRestoration of nuclear capacityDay-ahead (instead of month-ahead) clearing

This talk: Third study (2017): Can we take a US-inspireddesign and plug it into the existing European market?ELIA parallel runs (2018): ELIA (Belgian TSO) releasesreport on the simulation of scarcity prices in the Belgianmarket for 2017New scarcity adder incentive (2019): By October 2019,ELIA will be posting adders publicly

6 / 57

Scarcity Pricing Adder Formula

In its simplest form, the scarcity pricing adder is computed as

(VOLL− MC(∑

g

pg)) · LOLP(R),

where MC(∑

g pg) is the incremental cost for meeting anadditional increment in demand, R is the available reserve

More frequent, lower amplitude price spikesPrice spikes can occur even if regulator mitigates bids ofsuppliers in order to mitigate market powerCan co-exist with capacity mechanisms, perceived asno-regret measure for improving the energy-only market

7 / 57

Illustration from Texas: July 30, 2015

8 / 57

Illustration from Texas: July 30, 2015

9 / 57

Focus of this Presentation

Focus of this presentation: in order to back-propagate thescarcity signal

When shouldDo we need a real-time reserve market?Do we need virtual bidding? day-ahead reserve auctionsbe conducted? Before, during, or after the clearing of theenergy market?

10 / 57

A Possible Evolution of the Belgian Market

11 / 57

The Models in the Evolution Chain

Simultaneous DA RT reserve Virtualenergy and reserves market trading

SCV X X XRCV X XRCP XREP

The dilemmas of the market design:

Simultaneous day-ahead clearing of energy and reserve, or Reservefirst (S/R)?

Clearing of energy and reserve in real time, or Energy only (C/E)?

Virtual trading, or Physical trading only (V/P)?

12 / 57

Outline





13 / 57

Energy-Only Real-Time Market

Consumers

14 / 57

Notation

SetsGenerators: GLoads: L

ParametersBid quantity of generators: P+

g

Bid quantity of loads: D+l

Bid price of generators: CgBid price of loads: Vl

DecisionsProduction of generators: pRT

g

Consumption of loads: dRTl

Dual variablesReal-time energy price: λRT

15 / 57

Model

Just a merit-order dispatch model:

max∑l∈L

Vl · dRTl −

∑g∈G

Cg · pRTg

pRTg ≤ P+

g ,g ∈ G

dRTl ≤ D+

l , l ∈ L

(λRT ) :∑g∈G

pg =∑l∈L

dl

pg ,dl ≥ 0,g ∈ G, l ∈ L

16 / 57

Energy-Only in Real Time and Day Ahead

Consumers

17 / 57

Additional Notation

DecisionsDay-ahead energy production of generator: pDA

g

Day-ahead energy consumption of load: dDAl

Dual variablesDay-ahead energy price: λDA

18 / 57

Model

Generator profit maximization:

maxλDA · pDAg + (ΠRT

g − λRT · pDAg )

where ΠRTg = (λRT − Cg) · pRT

g is the real-time profit

Similarly for loads

Market equilibrium: ∑g∈G

pDAg =

∑l∈L

dDAl

19 / 57

Adding Uncertainty in Real Time

Consumers

Random net

injectionProducers

RT energy

market

20 / 57

Additional Notation

SetsSet of uncertain real-time outcomes (e.g. renewable supplyforecast errors, demand forecast errors): Ω

ParametersReal-time profit of agent: ΠRT

g,ω

FunctionsRisk-adjusted profit of random payoff: Rg : RΩ → R

21 / 57

Model


maxλDA · pDAg +Rg(ΠRT

g,ω − λRTω · pDA

g ),

whereΠRT

g,ω = (λRTω − Cg) · pRT

g,ω

Similarly for load maximization

Day-ahead market equilibrium:∑g∈G

pDAg =

∑l∈L

dDAl

22 / 57

Reserve Capacity in Real Time

Consumers

Random net

injectionProducers

RT energy

market

RT reserve

market

ORDC

Real-time market

Virtual traders

Consumers

Producers

DA energy

market

23 / 57

Additional Notation

SetsORDC segments: RL

ParametersORDC segment valuations: V R

lORDC segment capacities: DR

lramp rate: Rg

DecisionsReal-time demand for reserve capacity: dR,RT

l,ω

Real-time supply of reserve capacity: rRTg,ω

Dual variablesReal-time price for reserve capacity: λR,RT

24 / 57

Model

Real-time trading of energy and reserve for outcome ω ∈ Ω:

max∑l∈RL

V Rl · dR,RT

l +∑l∈L

Vl · dl −∑g∈G

Cg · pg

(λRT ) :∑g∈G

pRTg =

∑l∈L

dRTl

(λR,RT ) :∑

g∈G∪L

rRTg =

∑l∈RL

dR,RTl

pRTg ≤ P+

g,ω, rRTg ≤ Rg , pRT

g +rRTg ≤ P+

g,ω, g ∈ G

dl ≤ D+l , r

RTl ≤ Rl , rRT

l ≤ dRTl , l ∈ L

dR,RTl ≤ DR

l , l ∈ RL

pRTg , rRT

g ≥ 0, g ∈ G, dRTl , rRT

l ≥ 0, l ∈ L, dR,RTl ≥ 0, l ∈ RL

25 / 57

Remarks

Suppose that a given generator gis simultaneously offering energy (pRT

g > 0) and reserve(rRT

g > 0)

is not constrained by ramp rate (rRTg < Rg)

We have the following linkage between the energy and reservecapacity price:

λRTω − Cg = λR,RT

ω

This no-arbitrage relationship is the essence of scarcity pricing

26 / 57

Reserve Capacity in Day Ahead

Consumers

Random net

injectionProducers

RT energy

market

RT reserve

market

ORDC

Real-time market

Virtual traders

Consumers

Producers

DA energy

market

DA reserve

market

Day-ahead market

27 / 57

Additional Notation

DecisionsDay-ahead supply of reserve capacity: rDA

g

Dual variablesDay-ahead price for reserve capacity: λR,DA

28 / 57

ModelGenerator profit maximization:

maxλDA · pDAg + λR,DA · rDA

g +

Rg(ΠRTg,ω − λRT

ω · pDAg − λR,RT

ω · rDAg ),

whereΠRT

g,ω = (λRTω − Cg) · pRT

g,ω + λR,RTω · rRT

g,ω

Similarly for load profit maximization

Day-ahead market equilibrium:∑g∈G

pDAg =

∑l∈L

dDAl ,

∑g∈G∪L

rDAg = 0

29 / 57

To Summarize

We have arrived at our first target model: SCVSimultaneous day-ahead clearing of energy and reserveCoordinated trading of energy and reserve in real timeVirtual trading

30 / 57

Back-Propagation of Prices

The first-order conditions with respect to day-ahead energy andreserve decisions yields no-arbitrage conditions that explainhow real-time prices back-propagate to forward markets:

λDA =∑ω∈Ω

qg,ω · λRTω

λR,DA =∑ω∈Ω

qg,ω · λR,RTω

where qg is the risk-neutral probability measure of agent g

31 / 57

US Market

Consumers

Random net

injectionProducers

RT energy

market

RT reserve

market

ORDC

Real-time market

Virtual traders

Consumers

Producers

DA energy

market

DA reserve

market

Day-ahead market

Simultaneous auction

32 / 57

Outline





33 / 57

Belgian Market

Consumers

Random net

injectionProducers

RT energy

market

34 / 57

Moving from Virtual to Physical Trading

It is easy to replace virtual trading (V) with physical trading (P),by introducing physical constraints in the day-ahead model

For example, for generators:

pDAg + rDA

g ≤ P+g

rDAg ≤ Rg

rDAg ≥ 0

35 / 57

Moving from Coordinated Clearing of Real-Time Energyand Reserve to Energy-Only Trading

It is similarly easy to switch from real-time clearing of energyand reserve to energy-only trading by switching betweenco-optimization and merit order dispatch in real time

36 / 57

Moving from Simultaneous Day-Ahead Clearing toReserve First

Qualitatively, we want to capture the difference between thefollowing:

Simultaneous auctioning: system operator co-optimizes,taking into account all the relevant inter-dependencies ofpower production and reserve capacitySequential auctioning: agents determine opportunity costson the basis of possibly inaccurate forecasts of the systemstate for the following day

We formulate the problem as a multistage stochasticequilibrium by nesting risk functions (Philpott, 2016)

37 / 57

Sequence of Events

Decide

DA reserve

State of the

system in the

following day

Decide

DA energyRT dispatch

Revelation of

RT imbalance

Type of day: assessment of the TSO for what quantity of operating reservewill be required for the following day

In line with current effort of ELIA to transition towards dynamic reserve sizingand procurement in the day ahead (De Vos, 2018)

38 / 57

Populating the Tree with Data

Denote a given node as (t , ω), where t is stage and ω isoutcome

No specific random vector is revealed in stage 2, instead thesystem state:

Node (2, 1): Low-risk dayNode (2, 2): Medium-risk dayNode (2, 3): High-risk day

In stage 3, renewable supply P+wind is revealed:

Node (3, 1): 111 MW; node (3, 2): 101 MWNode (3, 3): 156 MW; node (3, 4): 56 MWNode (3, 5): 206 MW; node (3, 6): 6 MW

39 / 57

Some Additional Features of the European Model

For the case study, we introduce some additional features:Two types of reserve (secondary and tertiary) that aresubstitutableInelastic requirements for reserve capacity after activationPenalties on deviations between day-ahead and real-timeenergy production

40 / 57

The European Model

In the following, the European market equilibrium model ispresented from the point of view of generators:

real-time energy marketday-ahead energy exchangeday-ahead reserve capacity auction

Loads are modeled similarlyMarket clearing conditions are added where appropriate

41 / 57

Real-Time Equilibrium in the European Model


(PEG,RTg,ω,ω′) : max

pRT ,sRT ,+,sRT ,−λRTω′ · pRT

g,ω′ − Cg · pRTg,ω′

−ε+g · s

RT ,+g,ω,ω′ − ε

−g · s

RT ,−g,ω,ω′

(αG,RT ,+g,ω,ω′ ) : pRT

g,ω′ ≤ PRT ,+g,ω′ · yg,ω

(αG,RT ,−g,ω,ω′ ) : −pRT

g,ω′ ≤ −PRT ,−g,ω′ · yg,ω

(βG,F ,RTg,ω ) : rF ,DA

g − rF ,RTg,ω ≤ 0

(βG,S,RTg,ω ) : rS,DA

g − rS,RTg,ω ≤ 0

(γG,RT ,+g,ω,ω′ ) : pRT

g,ω′ − pDAg,ω − sRT ,+

g,ω,ω′ ≤ 0

(γG,RT ,−g,ω,ω′ ) : pDA

g,ω − pRTg,ω′ − sRT ,−

g,ω,ω′ ≤ 0

pRTg,ω′ , s

RT ,+g,ω,ω′ , s

RT ,−g,ω,ω′ ≥ 0

42 / 57

A Gap in the Existing EU Balancing Design

In Belgium today, it is clear what balancing serviceproviders need to be able to deliver before activationBut system scarcity is measured by leftover capacity afteractivationThere are plausible arguments for

dropping the constraints βF/S: why should we carryprotection after we have eliminated imbalances?including the constraints βF/S: the end of one imbalanceinterval marks the beginning of a new one

The presence or absence of these constraints has majorimplications for real-time prices

43 / 57

Day-Ahead Energy Exchange in the European Model

(PEG,DA,2g,ω ) : max

y ,pDAλDAω · pDA

g,ω +

R2g(ΠRTg,ω′(y ,p

DA)− λRTω′ · pDA

g )− Kg · yg,ω

(δg,ω) : yg,ω ≤ 1

(αG,DA,+g,ω ) : pDA

g,ω + rF ,DAg + rS,DA

g ≤ PDA,+g · yg,ω

(αG,DA,−g,ω ) : −pDA

g,ω ≤ −PDA,−g · yg,ω

yg,ω,pDAg,ω ≥ 0

44 / 57

Day-Ahead Reserve Auction in the European Model

(PEG,DA1g ) max

rF ,DA,rS,DAλR,F ,DA · rF ,DA

g + λR,S,DA · rS,DAg

+R1g(ΠDAg,ω(rF ,DA, rS,DA))

(βG,F ,DAg ) : rF ,DA

g ≤ RFg

(βG,S,DAg ) : rS,DA

g ≤ RSg

rF ,DAg , rS,DA

g ≥ 0

45 / 57

Outline





46 / 57

Case Study Setup

We simulate the Belgian market for September 2015 -March 2016We assume risk-neutral agentsWe solve the equilibrium problems using a stochasticoptimization equivalent

47 / 57

Energy Price

Design Summary PriceSCV US design 33.29RCV Allow virtual trading 34.36RCP Remove imbalance penalties 34.36

RCP-0.1 Trade real-time reserve 34.36REP-0.1 EU design extreme 2 27.60

REP-0.1-inel. EU design extreme 1 45.42Hist. DA Historical day-ahead 38.87Hist. RT Historical real-time 35.26

48 / 57

Observations

Validation: REP-0.1 and REP-0.1-inelastic (proxies ofBelgian market) envelope the historically observedday-ahead and real-time energy pricesThe requirement of whether or not to hold reserve capacityafter the activation of reserve has a major impact on prices

49 / 57

Secondary Reserve Prices

Design Summary Price DA Price RTSCV US design 15.34 14.57RCV Allow virtual trading 15.78 15.69RCP No imbalance penalties 15.78 15.65

RCP-0.1 Trade real-time reserve 15.79 15.15REP-0.1 EU design extreme 2 1.42 N/A

REP-0.1-inel. EU design extreme 1 26.90 N/AHistorical 9.59 N/A

50 / 57

Tertiary Reserve Prices

Design Summary Price DA Price RTSCV US design 11.27 10.50RCV Allow virtual trading 10.54 10.54RCP No imbalance penalties 10.59 10.52

RCP-0.1 Trade real-time reserve 10.67 10.17REP-0.1 EU design extreme 2 1.42 N/A

REP-0.1-inel. EU design extreme 1 26.90 N/AHistorical 5.27 N/A

51 / 57

Profits

SCV RCV RCP RCP-0.1 REP-0.1 REP-0.1-inel.

G1 6.44 7.37 7.37 7.40 2.59 16.15G2 19.59 20.66 20.68 20.79 15.07 31.80G3 7.02 8.06 8.06 8.09 2.64 19.03G4 10.48 12.04 12.04 12.08 3.84 28.62G5 19.96 21.05 21.07 21.18 15.45 32.26G6 7.23 8.29 8.30 8.32 2.66 19.42G7 20.36 21.43 21.45 21.56 15.82 32.57G8 19.50 20.56 20.58 20.69 14.93 31.67

Profitable plants (normal font): profits above 8.66 e/MWhBreak-even plants (italic font): profits 6.03 - 8.66 e/MWhNon-viable plants (bold font): profits below 6.03 e/MWh

52 / 57

Observations

Removing the requirement of carrying reserve afteractivation (REP-0.1) places 4 out of 8 units in a non-viablefinancial positionThe introduction of a real-time market for reserve capacity(RCP and RCP-0.1) restores 3 of these units to breakingeven, and 1 of them to covering its investment costscomfortably

53 / 57

Making Sense of the ResultsA major difficulty with the absence of a real-time reserve marketis that it becomes difficult to value reserve precisely:

λR,DA = βG,DAg + E[αG,DA

g,ω ]

whereαG,DA

g,ω : ramp rate constraint multiplier

βG,DAg : capacity constraint multiplier

If we are forced to carry the full amount of reserve afteractivation, the scarcity signal is too strong:


g,ω ] + E[βG,RTg,ω′ ]

whereβG,RT

g,ω′ : multiplier associated to requirement of carryingreal-time reserve capacity after activation

54 / 57

Making Sense of the Results

The real-time ORDC automates this calculation in aself-correcting fashion, and arbitrage propagates this price tothe day-ahead market, thereby signaling investment in reservecapacity in case of tight system conditions:


g,ω ] + E[λR,RTω′ ]

55 / 57

Conclusions of the Belgian Case Study

Our recommendations to the Belgian regulatory commission:Introducing a real-time market for reserve capacity is thetop priorityVirtual trading and simultaneous clearing of day-aheadenergy and reserves are less crucial in a risk-neutralsetting

56 / 57

Thank You for Your Attention

For more information:[email protected]

http://uclengiechair.be/

https://perso.uclouvain.be/anthony.papavasiliou/public_html/

57 / 57

mailto:[email protected]

http://uclengiechair.be/

https://perso.uclouvain.be/anthony.papavasiliou/public_html/

Market Design Considerations for Scarcity Pricing: A Stochastic ... … · Motivation of Scarcity...

Documents

Transcript of Market Design Considerations for Scarcity Pricing: A Stochastic ... … · Motivation of Scarcity...