NREL is a national laboratory of the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy, operated by the Alliance for Sustainable Energy, LLC.

Simplified flexibility parameters for evaluating renewable integration

JRC Workshop on Addressing Flexibility in Energy Models

Paul Denholm

December 5, 2014

2

Approaches to Capacity Expansion Planning

• Traditional load-duration curve approaches

o Screening curves identify least cost mixes based on levelized cost of energy

o Doesn’t incorporate any chronological (time-series) analysis

• Linear and Mixed-Integer optimization

o Finds lowest cost mix based on a life-cycle cost

o Can incorporate chronology in the objective function

o But full year hourly (or sub-hourly) simulations are computationally complex

3

Approaches to Incorporating Time Series in Capacity Expansion Models

• Reduced set of time-periods o Full time-series for a few weeks that (hopefully)

represent the entire year

o Variable generation makes picking “typical” periods challenging

• Time-slice (non chronological) approach o Estimates typical dispatch characteristics in a set of

representative time periods

o Requires establishing parametric relationships for key parameters such as curtailment

o NREL approach in the ReEDS models

4

NREL’s Grid Modeling Tools

5

ReEDS Model

A spatially and temporally resolved model of capacity expansion in the U.S. electric sector.

Designed to explore potential electric-sector growth scenarios in the U.S. out to 2050 under different economic, technology, and policy assumptions.

(Regional Energy Deployment System Model)

6

ReEDS Model: History and Team

Current team is ~10 staff (not all full-time on ReEDS).

Selected studies:

• 2008 20% Wind Vision

• 2012 Renewable Electricity Futures

• 2012 SunShot

• Various RPS, CES, PTC, … analyses

ReEDS has been in use for >10 years, with a steady increase in sophistication and capabilities over that time.

7

What does ReEDS do?

• Each 2-year solve produces a set of new investments and described operation of new and existing fleet.

• Between solves, ReEDS updates: o Existing generator fleet, including retirements

o Existing transmission

o Performance of existing fleet

o Costs/performance of new technologies

o Electricity demand, reserve margin requirements

o Variable renewable capacity values, curtailment, operating reserve requirements

• Skip forward two years, and solve again.

8

Reduced-form Dispatch

Seventeen time-slices: four seasons x four diurnal + one superpeak. Continuous units: minimum turndown, but no startup or shutdown, flat heatrate. Constraints guarantee adequacy requirements and ancillary services.

9

Modeling Framework

Technology cost & performance Resource availability Demand projection Demand-side technologies Grid operations Transmission costs

Black & Veatch

Technology Teams

Flexible Resources

End-Use Electricity

System Operations

Transmission

ABB inc. GridView

(hourly production cost)

rooftop PV penetration

2050 mix of generators

does it balance hourly?

Implications GHG Emissions

Water Use Land Use

Direct Costs

Capacity & Generation 2010-2050

10

Integration with Operational Model

• To supplement ReEDS’ reduced-form unit-commitment model, for the REF analysis we rebuilt ReEDS infrastructure in GridView, a commercial production cost model to test how the ReEDS-projected infrastructure might behave in an hourly dispatch.

• We are now automating the capability, using PLEXOS this time instead of GridView.

11

UC on NREL’s HPC

Peregrine Characteristics: • 11520 Intel Xeon E5-2670 "SandyBridge" cores • 14400 next-generation Intel Xeon "Ivy Bridge"

core • 576 Intel Phi Intel Many Integrated Core (MIC)

core co-processors with 60+ cores each • 32 GB DDR3 1600Mhz memory per node • Peregrine will deliver a peak performance of 1

petaFLOPS

NR

EL P

IX 2

45

80

12

Can we include more chronology in the objective function?

• Fundamental tradeoff between chronology and simplicity. o But can we incorporate full chronological simulations

but avoid many of the key complications?

• Unit commitment • Full storage optimization

• Can reduced form chronological simulations still

provide valuable insights? • And is 1-hour simulation good enough?

13

Renewable Energy Flexibility (REFlex) Model

• Dispatch only model

• Block dispatch by generator type

• Simplify key parameters traditionally captured in unit commitment

o Minimum generation point for thermal generation

o Minimum thermal generation for ramp

• Simplified valley-filling algorithm for storage and DR dispatch

14

What can this approach do?

• Analyze optimal mixes of VG in high penetration scenario

• Examine curtailment • Analyze impact of storage and DR • Run very fast

• What it can’t (probably) do:

o Optimize new conventional generation mix in low VG scenarios

o Basically assumes thermal fleet is relatively static or decaling

15

Example – Curtailment Analysis

• At high penetration, economic limits will be due to curtailment

o Limited coincidence of VG supply and normal demand

o Minimum load constraints on thermal generators

o Thermal generators kept online for operating reserves

16 16

Minimum Generation Levels Limited by Baseload Capacity

Price/Load Relationship in PJM

Below Cost Bids

0

50

100

150

200

250

0 10000 20000 30000 40000 50000 60000

Load (MW)

Wh

ole

sale

Pri

ce (

$/M

Wh

)

0

5

10

15

20

25

30

35

18000 20000 22000 24000 26000Load (MW)

Wh

ole

sale

Pri

ce (

$/M

Wh

)

17 17

Min den depends on VG Mix

17

18

Example – Curtailment as a Function of Penetration

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

50%

20% 30% 40% 50% 60% 70% 80%

Fraction of System Electricity from Solar and Wind

Fra

cti

on

of

VG

Cu

rta

ile

d

0/100

20/80

30/70

40/60

60/40

80/20

Solar / Wind Mix

Reflex < 1 Minutes per run

19

Results from Full UC/ED Model

PLEXOS Simulations (DAUC/SCED) ~5 Hours per run

20

Example – Storage Dispatch

• Full storage optimization is computationally complex

• Valley filling (search) algorithms can be much faster

21 National Renewable Energy Laboratory Innovation for Our Energy Future

REFlex CSP Dispatch

0

10

20

30

40

50

60

70

1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96

Hour

Ge

ne

rati

on

(G

W)

Curtailed Solar

Dispatched CSP

Usable PV

Wind

Conventionals

Load

Non-Dispatched CSP

Dispatched CSP

Dispatch of CSP May 10-13

22 National Renewable Energy Laboratory Innovation for Our Energy Future

PLEXOS CSP Dispatch

Dispatch of CSP in WWSIS-2 Study (July)

0

5,000

10,000

15,000

20,000

25,000

30,000

35,000

40,000

45,000

0

20,000

40,000

60,000

80,000

100,000

120,000

140,000

0 24 48 72 96 120 144 168

CSP

In

flo

w/G

en

era

tio

n (

MW

)

Ne

t Lo

ad (

MW

)

Hour

Net Load with Wind and PV CSP Inflow CSP Generation

23

Example –Energy Storage

0%

5%

10%

15%

20%

25%

30%

35%

40%

20% 30% 40% 50% 60% 70% 80%

Fraction of System Electricity from Wind&Solar

Fra

cti

on

of

VG

Cu

rta

ile

dNo Storage

4 hours

8 hours

12 hours

24 hours

24

Example – Electric Vehicle Charging

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

0:0

0

1:0

0

2:0

0

3:0

0

4:0

0

5:0

0

6:0

0

7:0

0

8:0

0

9:0

0

10

:00

11

:00

12

:00

13

:00

14

:00

15

:00

16

:00

17

:00

18

:00

19

:00

20

:00

21

:00

22

:00

23

:00

Ave

rage

Ch

argi

ng

De

man

d P

er

Ve

hic

le (

kW)

Time (In 5-Minute Intervals)

SUV-40

SUV-20

Sedan-40

Sedan-20

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

0 24 48 72 96

Lo

ad

(M

W)

Hour

Net Load with PV Normal Load Solar PV Output

Vehicle Availability

High PV Impacts

25

Example – Electric Vehicle Charging

Overnight optimized, uncontrolled daytime charging

Optimized overnight, imperfect foresight daytime

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

0 24 48 72 96

Lo

ad

(M

W)

Hour

Net Load with PHEVs & PV Normal Load

Net Load with PV PHEV Charging Profile

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

0 24 48 72 96

Lo

ad

(M

W)

Hour

Net Load with PHEVs & PV Normal Load

Net Load with PV PHEV Charging Profile

26

What else can we ignore?

• Subhourly Dispatch?

• Incorporation of cycling costs into UC/ED process?

27

Example: Western Wind and Solar Integration Study

•Phase 3: Frequency Response and Grid Impact

Phase 2: Cycling Cost and Emissions Impacts

What happens to the transmission grid’s frequency with high penetration of distributed PV at low load?

What happens to the grid when remote transmission lines are highly-loaded to move wind long distances?

From a system perspective, cycling costs are relatively small

Emissions impacts of cycling are relatively small

28

5- minute dispatch

120000

121000

122000

123000

124000

125000

126000

127000

128000

129000

130000

0:00 0:20 0:40 1:00 1:20 1:40 2:00

1-Hour

5-Minute

UC based on this 1-hour ramp rate

This 5-minute ramp may exceed committed capability

29

Methodology

• PLEXOS unit commitment and dispatch modeling

o Day ahead market (hourly)

– Coal and nuclear units committed

o 4 hour ahead market (hourly)

– Better forecasts

– Gas CC and steam units committed

o Real time market (tested hourly vs subhourly)

– Gas CT committed and dispatched

30

WWSIS Core Scenarios

Wind Capacity (MW)

710 to 1,650

140 to 710

110 to 140

70 to 110

10 to 70

PV Cpacity (MW)

76 to 200

51 to 76

29 to 51

10 to 29

0 to 10

CSP Capacity (MW)

199 to 200

142 to 199

105 to 142

84 to 105

64 to 84

Reference 8% wind 3% solar

High Mix 16.5% wind 16.5% solar

31

Consistency between cases

• Constant

o Commitment of non-CT generators

o Planned hourly hydro generation

o Reserve requirements

• Changes

o Interval of real-time dispatch (5-min and hourly tested)

32

2-part heat rate curves

Another area of sensitivity analysis needed….

33

Difference between hourly and 5-min net load

July 25-28

34

Difference between hourly and 5-min net load

5 Minute net load and interpolated hourly net load (load – mind – PV)

35

Difference between hourly and 5-min net load

0

2000

4000

6000

8000

10000

12000

14000

16000

-2.5

-2.2

-1.9

-1.6

-1.3 -1

-0.7

-0.4

-0.1

0.2

0.5

0.8

1.1

1.4

1.7 2

2.3

Mo

re

Fre

qu

en

cy

Percent difference

36

Run Times

• Day Ahead UC ~ 3 days

• 4-Hour Ahead UC ~1 Day

• 5-Minute Dispatch ~ 2 Days

o Approximately 12 times longer than 1-hour dispatch

o What do we get for this increase in run time?

37

Results

• No unserved load

• Some change in unserved reserves

• Very little change in total production cost

• Occasionally significant change in LMPS

38

Unserved load and reserves

• No unserved load in any scenario

• Reserve requirement totals ~40 TW-h

• Unserved reserves

HiMix Reference

RT – hourly resolution 138 MW-h 178 MW-h

RT – 5-minute resolution 263 MW-h

337 MW-h (0.0008%)

39

5-minute resolution dispatch stack

40

Hourly resolution dispatch stack

41

Total production costs

HiMix Reference

RT – hourly resolution $11.03 billion $15.12 billion

RT – 5-minute resolution $11.02 billion

$15.13 billion

Changes in production cost between hourly and 5-min runs are within the range of uncertainty

42

Generation by type

43

Number of starts

44

Curtailment

45

Price differences

Demand response deployments

46

My Opinions

• Full unit commitment is desirable, but unless we get a 100x plus increase in speed we need to simplify the problem

• Capacity expansion problems don’t lend themselves to traditional parallelization

• Simplification of UC with simplified parameters appears to produce reasonable representation of thermal fleet under high VG scenarios

• Simple time-shifting storage optimization appears to produce reasonable results

47

Concluding Thoughts

• Does importance of accurate unit commitment simulation decrease in high renewable scenarios?

o Retirement of long-start units leads to UC being a hour-ahead problem vs. a day-ahead problem

• Increased importance of chronological simulation for demand response, energy storage

48

Flexibility Metrics?

Input Metrics • Ramp Rate • Ramp Rate • Transition Time

• Start-up • Min Up/Down

Time

Output Metrics • Loss of Load

Expectation • Reserve Violations • Unmet ramp

requirement

Outcome Metrics • System

Costs/Benefits • Carrying Capacity

Paul.denholm@nrel.gov