Outline

•Background & Motivation•Reserve Modeling Framework • Types of improvements • COMPETES simulations•Results

Challenges Arising from Wind

Source: Flexibility in 21st Century Power Systems, NREL Report

Quack!

Source: CAISO

Reserve• Operating Reserve is extra capacity (MW) needed

in case of contingency • Loss of a generator• Loss of a transmission line • Sudden change in load• Now: change in renewable energy

40

60

80

100

120

1 6 11 16 21

MW

Hour

Expected Demand

Reserve+Expected Demand

Frequency restoration (mFRR), not automatic frequency

response

Operational Reserve1. Size / Procure

• How much do we need? • E.g., extra 30 MW on-line in every hour

2. Allocate• Who will be scheduled? • Generator B & C will each provide 15

MW3. Activate

• Who will provide the energy if actually needed?

• Deliverability in real time market

Procurement • How much do we need?

• Often called ’reserve requirement’ • Examples

• Capacity of largest generator or transmission line

• X% of demand and Y% of renewables for …• One day• One season

20

40

60

80

100

120

1 6 11 16 21

MW

Hour

Expected DemandReserve Ex %Reserve Ex Fixed

Allocation Who will be scheduled?

• Most US markets • Market based

• Primary• Secondary• Tertiary

• Determined in zones

• Most European markets• Long-term contracts

• Portfolio based• Unit based • Some dispatch

• Determined by country

Source: ENTSO-E

Activation• Who will actually provide reserves if needed?

• Generators change energy output level in balancing• Contract-based• TSO can call on contracted generators to

provide reserve in real time• Market-based (US)• System operator calls on generators

selected in the day-ahead for reserve • Energy must be deliverable

• Transmission constraints might limit deliverability within and between countries

ECN-JHU Current Research Question

What changes to market design will most enhance efficiency in procuring/allocating/activating reserve?

Types of Improvements • Reserve requirement procurement periodo Seasonal

§ Current practice, four seasonal periods assessedo Enhancement: Daily

§ Requirement determined daily • Allocation typeo Contract-based

§ Current practice§ Bi-lateral contracts between TSO and generators

o Enhancement: Market-based§ Procured through co-optimization with energy market

• Amount of coordinationo Independently determined, current practice o Enhancement: Northwest Europe coordinates

Example requirement:3% of demand and 5% of renewable generation

Efficiency of Reserve

• Each axis shows a different improvement to reserve

• Increasing complexity and efficiency moving away from origin

• Star = hypothetical ideal

• Dot = worst case

Reserve Requirement

CoordinationDA & BalancingNo Coordination

Seasonal

Daily

Increasing complexity, efficiency à Thanks to Qingyu Xu

COMPETES Network • 33 node pan-European network• Transmission mimics integrated EU

network with capacity limited by NTC• Future generation +

potential energy storage

• Renewable scenario based on ENSTO-E 2030 Vision 4 of “European Green Revolution”

Model Formulation: Unit Commitment

•Min Operating Cost• Subject to• Generator min & max capacity• Ramp limits• Min up & down times• Transmission line capacity & flow (Net

Transfer Capacity)• Startup & no-load binary constraints /

relaxed formulation

Operational Markets• Day Ahead• Schedules generation for the following day• Inputs: bids & offers, forecast for load and wind • Outputs: prices, schedule (on/off), dispatch• à Reserve allocation phase

• Balancing • Updates schedule to reflect new information• Inputs: new bids & offers, updated forecast• Outputs: prices, fast start schedule, dispatch• à Reserve activation phase

• Was the right amount procured?• Was it allocated to those who could deliver it?

, reserve sizing

Simulations & Sensitivity Analysis• Simulations• Simulated one day-ahead forecast• Followed by 5 real-time ”actual” wind

realizations• Results show mean of 5 simulations • Error bars show minimum and

maximum deviations • Added an extra coordination

component• Due to results found by K. van den

Bergh in [4], we consider coordination in balancing alone with no coordination in day-ahead

K. van den Bergh, [4]

Day-Ahead

Bal. 1

Bal. 2

Bal. 5Bal. 3

Bal. 4

Efficiency of Reserve

• Each axis shows a different improvement to reserve

• Increasing complexity and efficiency moving away from origin

• Star = hypothetical ideal

• Dot = worst case

Seasonal

Daily

Coordination

Rese

rve

Requ

irem

ent

Operating Costs• Example results comparing • Star, ‘ideal case’ (DMC)

• Reserve size based on daily average• Market-based allocation• Coordination in day-ahead and balancing

• Rectangle (DMN)• Reserve size based on daily average• Market-based allocation• No coordination

+54.9%54.0 55.6

0.32-0.290%

+59.3%

EU

NL0%

54.9% =DMNopcosts − DMCopcost

DMCopcost

Deviation of minimum scenario

Deviation of maximum scenario

54.9%:20% load shedding80% generation deviations

Seasonal

Daily

Coordination

Rese

rve

Requ

irem

ent

Results – operating cost% deviations from ‘ideal’ case without load shedding

Seasonal

Daily

Coordination

Rese

rve

Requ

irem

ent

30.9 33.0 -0.21 0.51 -0.23 0.49+0.12% +0.11%+32.1%

+40.9%39.9 41.7 -0.30 0.44

+0.05%0.37-0.33

EU

NL

27.7 31.4

+29.1%

-0.48 2.50 -0.72 1.62

34.3 37.9 -0.20 0.91 -0.66 0.95

+0.67% -0.04%

+0.28%+35.7%

0%

0%

Daily

Coordination

Rese

rve

Requ

irem

ent

Seasonal

CZE, DEW, FRA, GER, ITA, POR, SKO, SPA, SWE, SWI, NOR, BLK

DEN, POL, UKI

NED, BLTBEL, FIN,

IRE -

-

Lowest Cost Solution by CountryWhere is each country is better off?

Results – wind curtailed% deviations from ‘ideal’ case, NL data in MWh

Seasonal

Daily

Coordination

Rese

rve

Requ

irem

ent

1.1 19 -8.2 9.1 -5.3 11.4-3.5% -0.6%+6.2%

+6.6%1.7 20 -7.7 10

-2.7%13-4.8

EU

NL

448 8350

3885 MWh

448 8350

0 MWh 0 MWh

0 MWh3875 MWh

0%

0 MWh

Reserves: Source FuelAll market based simulations showed similar percentages

Storageprod.59%

Coal

2% Lignite1%

Res-e

1%

Nuclear0%

Oil

0%

Wind

0% Sun

0%

Gas

37%

Contracted Reserve CasesReserve Requirement

CoordinationBalancingNo Coordination

Seasonal

Daily

+40.9%

+0.17%+32.1%

0%

+110% +30.0%

• All contracted cases showed higher costso Some cases were

double the cost of the market-based cases§ Some countries faced

significant load shedding

§ Wide difference in operating costs country by country

• Fewer MWh of wind curtailment than ‘ideal’ case when reserves were coordinated in balancing o Additional plants online

meant lower curtailment

ConclusionsThree Suggested Improvements: 1. Difference between daily vs.

seasonal requirement is minimal

2. Coordination in balancing achieves almost all benefit, or can produce better solution

3. Naïve contracts for reserves produce least efficient solution compared to market• Coordination in reserve

allocation & balancing might make up for higher costs

Other ObservationsØMore coordination may

lead to more wind curtailment• Possibly due to location of

reserve within country• Consideration of forecast

uncertainty and wind farm location can reduce curtailment

ØStorage can provide a significant amount of reserve

References [1] S. Kasina, S. Wogrin, and B.F. Hobbs, “A comparison of

unit commitment approximations for generation production costing,” Working Paper, Johns Hopkins University, 2014.

[2] Ö. Özdemir, F. Munoz, J. Ho, and B.F. Hobbs, “Economic Analysis of Transmission with Demand Response and Quadratic Losses by Successive LP,” IEEE Trans. Power Syst., DOI: 10.1109/TPWRS.2015.2427799, in press.

[3] J. Cochran, M. Miller, O. Zinaman, M. Milligan, D. Arent, B. Palmintier, M. O’Malley, S. Mueller, E. Lannoye, A. Tuohy, B. Kujala, M. Sommer, H. Holttinen, J. Kiviluoma, and S. K. Soonee, “Flexibility in 21st Century Power Systems,” Golden, CO, 2014.

[4] K. van den Bergh, R. B. Hytowitz, K. Bruninx, E. Delarue, W. D'haeseleer, and B.F. Hobbs, "Benefits of coordinating sizing, allocation and activation of reserves among market zones," Electric Power Systems Research, 143: 140–148, Feb. 2017.

Thank you!Questions?

Email: hytowitz@jhu.edu

Backup slides

Results – operating cost% deviations from ‘ideal’ case

Seasonal

Daily

Coordination

Rese

rve

Requ

irem

ent

54.3 56.1 -0.06 0.37 -0.08 0.36+0.14% +0.17%+55.4%

+54.9%54.0 55.6 -0.26 0.39

+0.04%0.32-0.29

EU

NL

57.4 61.1

+58.8%

-0.28 2.69 -0.74 1.60

58.0 61.5 -0.34 0.77 -0.66 0.95

+0.87% -0.05%

+0.14%+59.3%

0%

0%

Generation MixTWh difference between (Seasonal/Market/No Coordination) - (Daily/Market/Coordination)←

Mor

e in

‘ide

al c

ase’

Mor

e in

S/M

/NC

case→

Seasonal

Daily

CoordinationRese

rve

Requ

irem

ent

Net Trade (Imports)for market based simulations, (+) = fewer imports, (-) = more imports

Daily Requirement Seasonal RequirementDay-ahead & Balancing

CoordinationNo or Only Balancing

CoordinationDay-ahead & Balancing

CoordinationNo or Only Balancing

CoordinationBEL 0% 0.07% 0.06% -0.01%CZE 0% 0.15% 0.23% 0.14%DEN 0% 0.15% 0.07% 0.12%DEW 0% 0.47% 0.01% 0.64%FIN 0% 0.34% -0.05% 0.04%FRA 0% 0.16% 0.07% 0.19%GER 0% 0.03% 0.26% 0.04%IRE 0% 0.20% 0.26% 0.19%ITA 0% -0.05% -0.09% -0.02%NED 0% 1.27% -0.08% 1.90%POL 0% -0.06% -0.04% -0.08%POR 0% 0.45% 0.60% 0.65%SKO 0% 0.06% -0.58% -0.50%SPA 0% -0.21% -0.19% -0.62%SWE 0% -0.02% 0.13% 0.07%UKI 0% 0.55% -0.04% 0.55%SWI 0% 0.31% 0.29% 0.34%NOR 0% -0.40% 0.41% -0.31%BLK 0% -0.04% -0.08% -0.19%BLT 0% -0.13% -0.09% -0.14%AUS 0% 0.32% 0.09% -0.03%TotalEnergy Traded 0.15% less trade 0.07% less trade 0.15% less trade

Reserve Requirement

05

10152025

Freq

uenc

y

Bin

Winter

010203040

Freq

uenc

y

Bin

Spring

05

10152025

6302 6708 7115 7522 7928 8335 8741 9148 9554 9961

Freq

uenc

y

Bin

Summer

0

10

20

30

Freq

uenc

y

Bin

Fall

Seasonal: 8675 MW Seasonal: 8032 MW

Seasonal: 7625 MW Seasonal: 7870 MW

Four different scenarios (A↔D) considered

31

A B C D

Sizing - - - +

Allocation - - + +

Activation - + + +

“+” = coordinated “-” = uncoordinated

(3) Activation of reserves (real-time)

32

Allocation, activation and net cost savings relative to scenario A.

Conclusions1) Coordinating real-time reserve activation is always beneficial

2) Coordinating reserve sizing & allocation can lead to suboptimal

results (possibly even deteriorated) if network constraints are

neglected

3) Further research deals with including network constraints in

(deterministic) reserve sizing and allocation rules

33

Model Formulation

binary variables

ui......

Commitment within the Netherlands

Day-Ahead

xi......

continuous variables

Model Formulation

relaxedbinary variables

Commitment outside the Netherlands

Day-Ahead

xi......

continuous variables

ui......

binary variables fast start units

ui......

Model Formulation

......

continuous variables

Commitment only within the Netherlands

Balancing

xi......

Fixed variables: line flows,slow units

ui......xi