CHALLENGES OF REPRESENTING ELECTRICITY SYSTEM FLEXIBILITY IN - - PowerPoint PPT Presentation

Petten, 04 December 2014

CHALLENGES OF REPRESENTING ELECTRICITY SYSTEM FLEXIBILITY IN ENERGY SYSTEMS MODELS

Vera Silva, EDF R&D

Co-authors: Gregoire Prime, Timothee Hinchliffe, Dominique Lafond, François Rehulka, Miguel Lopez-Botet

| 2

• A system that has sufficient capacity to meet peak load is adequate but

if this capacity is composed mostly of low flexible plants it can experience problems for handling demand and generation variability.

• Flexibility has always been an essential ingredient to handle the

variability and uncertainty in the demand-generation balance. It is required at the operational time scales but needs to be considered at planning stage.

• Assessing the flexibility adequacy will probably emerge as a new task

in power system planning and metrics and models are being developed to help with this task.

• The representation of flexibility at the energy and power systems

planning will help to deliver a system that can handle variability in a cost effective way.

THE PROBLEM OF FLEXIBILITY ADEQUACY

| 3

INTEGRATION OF ELECTRICITY SYSTEM FLEXIBILITY TO ENERGY

SYSTEMS MODELS Goal: to obtain an a long term electricity system expansion solution that ensures a flexible system by solving a problem that includes : 1) the interaction between the energy and the electricity systems 2) the long term uncertainties and 3) the relevant short term operation constraints.

Long term forecasts of demand, commodity prices, etc Reliability and flexibility requirements

Multi-energy system model

Transmission expansion options and candidate generation technologies Generation and Transmission planning Production cost simulations and operation flexibility assessment

Adequate and flexible transmission and generation expansion solution Electricity system model

FlexAssessment Continental Model with Investment loop MADONE

| 4

DIFFERENT APPROACHES TO ADDRESS THE INTEGRATION OF ENERGY

AND ELECTRICITY SYSTEMS PLANNING

Option 1) Representation of electricity system flexibility in the Times model by increasing the simulation granularity and including additional constraints=> MADONE Option 2) Coupling energy system models with electricity system models using a chain of simulation tools with the possibility of back feeding relevant information

• Energy system optimization : Madone
• Electricity system planning : Continental Model with Investment

loop

• Detailed near term flexibility assessment : Continental with

FlexAssessment

| 5

• Perimeter EU27+NO+CH (Europe 29)

 with different levels of detail depending on the country

• Trans-national networks represented

 electricity and gaz, CO2

• Storage capacities

 hydro (one lake per country + hydro-pumping), gaz, CO2

• Pipelines/electricity injections at the frontier of EU29

 NordStream, Southstream, Nabucco, DESERTEC…

• National resource potentials & limits

 wind off-shore: km2(depth, wind speed, distance to coast) X

Capacity density; Wind on-shore km2 ( area potentially available) X Capacity density; solar PV: area available, roofs surfaces; CO2 storage; biomass resources

• Period and simulation time step

 yearly from 2005 to 2010, every 10 years from 2010 to 2050  representation of each year with load curves eg: 24, 288 points

• Outputs

 technology mix & detailed energy balances , energy dependency,

and environmental indicators, balance for electricity (including exchanges), association of energy uses and activities….

275 Mtep / 549 TWh 1339 Mtep / 2114 TWh 258 Mtep / 580 TWh

12/11/2005 EFESE/OSIRIS lot MADONE / Enerbat / EPI / MFEE/ EIFER 4

2/1. 8 0.8/0.7 0.08/0.02 0.65/0.65 1.2/0.47 1.9/0. 9 2.05/1.5 1.9/2.75 3/3.85 0.6/0.6 3.2/1.5 2.5/0.8 1.1/1.2 3.2/2.2 2.4/2.4 0.5/1.4 0.995/2.65 2.3/3.2 2/2 0.5/0.6 0.75/0.75 0.95/0.95 1.98/2.44 1.2/1.3 0.3/0 0.35/0.35 0.75/0.75 2.05/1.65 0.5/0.5 1.5/0.8 0.9/0.6 0.43/0.16 4.24/1.81 1.3/1.5 2.3/2.2 1.75/0.8 0.5/0.5 0.6/0.1 2/1

Capacités d’interconnexion en 2005 (enGW)

TO /FROM

Légende:

MADONE PERFORMS A MULTI-ANNUAL AND MULTI-ENERGY SIMULATION

OF EACH OF THE 29 INTERCONNECTED EUROPEAN COUNTRIES

detailed interm. aggregated

| 6

The model simulates a “realistic” European electricity system, including:

 description of different countries generation mix and key transmission corridors  interconnection capacities between countries  management of water reservoirs and pump storage  a large number of scenarios of climate years represented by demand and variable

generation across the European system => time-synchronise data with hourly (or lower) resolution

 several scenarios of generation availability

Some key challenges of this problem:

 hydro and storage flexibility optimization => stochastic problem  generation scheduling needs to be performed across the whole Europe including

interconnection and key transmission constraints => high performance computing

 analysis of system static and dynamic security => hierarchical approach

CONTINENTAL MODEL WITH INVESTMENT LOOP FOR MODELING THE EUROPEAN INTERCONNECTED ELECTRICITY SYSTEM

| 7

The objective is to obtain the thermal generation mix that ensures that for every new unit the revenues equals its annuitized fixed costs :

 Fixed costs include investment and O&M  Variable costs include start-up and fuel costs

The conventional generation mix is optimized in two iterative steps:

 Load duration curve based heuristic to propose a

candidate solution

 Validation of the heuristic solution solving the

hourly load-generation dispatch => creates a price signal that feeds the investment loop The generation mix needs to respect an adequacy criterion

 Maximum of 3h/year with marginal price = VOLL

CONTINENTAL

Investment loop Demand Variable generation profiles Interconnection constraints Storage Investment costs Commodity prices CO2 price

INPUT DATA Optimal thermal generation mix Production dispatch Production costs Market clearing prices CO2 emissions Hydro stock level paths Interconnection uses OUTPUT

CONTINENTAL MODEL WITH INVESTMENT LOOP: ELECTRICITY

GENERATION INVESTMENT MODEL FOR INTERCONNECTED SYSTEMS

| 8

Reference: Langrene, N., van Ackooij, W., Breant, F., « Dynamic Constraints for Aggregated Units: Formulation and Application », Power Systems, IEEE Transactions on , vol.26, no.3, Aug. 2011

Minimize global production cost for each zone

Unit commitment and economic dispatch minimizes thermal and hydro generation cost over all the scenarios Constraints include primary, secondary and tertiary reserve and generation dynamic ratings Multi area optimization with interconnection constraints represented by NTC

Stochastic hydro-generation scheduling

Maximize the reduction in terms of generation costs using dynamic optimization to obtain the « water value » for each time step Define a set of strategies of the optimal use of hydro reservoirs in order to minimize the global generation cost

Scenario based representation of stochastic parameters :

Large number of annual scenarios of demand, wind and PV generation, water inflows, fuel costs, thermal unit availabilities

Scenarios data Water values

For each dispatch period and for each zone the dispatch solution and the market clearing prices are obtained to access the revenues of generation units

CONTINENTAL MODEL HYDRO-THERMAL GENERATION SCHEDULING

| 9

MADONE

• Bottom-up TIMES model : all 29 interconnected

European countries

• Renewables national resources potentials &

limits detailed: Wind off-shore, wind on-shore, roof

for PV etc…

• Horizon = 2005 to 2050 with a perfect foresight
• f each year
• Time-slices: 24 or 288

24 =Peak and Off-Peak for each month 288 = 2 representative day (Week/W-E, bi-

hourly)/month

• Peak equation: additional demand constraint

With or without renewable contribution

• Deterministic

Or multi-scenarios for one chosen year : testing

with 4 annual scenarios

Continental Model with Investment loop

• EDF R&D’s Elec Production Cost model
• Electricity generation portfolio optimization
• Stochastic simulation of hourly system
• peration

 Demand-generation balancing solved for one

year with hourly resolution

Interconnection constraints included  Stochastic parameters: (T°, hydro, wind, PV and

generation outages)

) ( ) (

max

t Dem t Dem 

OPTION 1 - REPRESENTATION OF FLEXIBILITY IN THE ENERGY MODEL : COMPARISON MADONE-CONTINENTAL

| 10

89 90 89 90 90 91 90 91 91 67 63 65 40 68 71 68 70 63 77 90 106 125 90 86 92 88 96 50 100 150 200 250

Def 5k€_Seuil 5k€_200MW Def 10k€_Seuil 5k€_200MW Def 50k€_Seuil 5k€_200MW Def 10k€_Seuil 5k€_MC1pt Def 5k€_Seuil 5k€_MC2all Def 20k€_Seuil 20k€_200MW Def 20k€_Seuil 10k€_600MW Def 20k€_Seuil 10k€_200MW REF_Def 20k€_Seuil 5k€_200MW

Capacité installée globale : MAD Global (GW)

Base Mid-merit Peak

89 89 89 81 81 81 43 41 41 42 43 40 234 101 34 224 96 10

• 50

100 150 200 250 300 350

288p + peak ss EnR 288p + peak 288p 24p + peak ss EnR 24p + peak 24p

Base Mid-merit Peak

• Consistency of

base capacity needs between the 2 models

• Mid-base capacity

underestimated with TIMES model

• Peak capacity

dependant on peak equation calibration

MADONE (TIMES) CONTINENTAL

COMPARISON OF CONTINENTAL AND MADONE OPTIMAL THERMAL

GENERATION MIX

| 11

• In the tests we made, a multi scenario approach helps reducing the gap for mid-

merit capacity but leads to a larger over-estimation of peak capacity

 Choice of a (limited =4) set of scenario: how to select the right ones?  Calibration of peak equation could be a solution… but largely dependent on the system

studied

88 46 136 288p + peak multiscénario

91 63 96

€ € € € € € € € € € ll € € € € € € REF_Def 20k€_Seuil 5k€_200MW

89 41 101

288p + peak

MULTI SCENARIOS SIMULATIONS

Madone

Continental

| 12

MADONE IS SUITABLE TO PROVIDE A “MERIT ORDER” BETWEEN

TECHNOLOGIES INCLUDING THE MIX AND GEOGRAPHIC DISTRIBUTION OF RENEWABLES

Representing explicitly dynamic constraints in a long-term TIMES large planning model doesn’t seem realistic for the time being Without modeling operation margin and reserve requirements & dynamic constraints the generation dispatch is not accurate

For instance, a peak equation imposes investment in back-up capacity, not its use.

… but the objective is to have the « right » merit order between technologies investment decisions,

« right »= least cost + meeting capacity adequacy & flexibility adequacy system requirements

And then to assess, ex-post, if the generation mix calculated meets the electricity system constraints (Option 2)

Madone Continental model with investment loop

Renewables mix per country: F (cost/potential) with sensitivity to interconnection

Flexibility and adequacy constraints

| 13

OPTION 2: CHAIN OF SIMULATION TOOLS FOR DETAILED FLEXIBILITY

ASSESSMENT OF CONTINENTAL MODEL SCHEDULING SOLUTIONS

Location of VG Load factors (with resolution 1h or lower) VG forecast errors

Flex Assessment

CONTINENTAL Model

Reserves and flexibility adequacy

Economical analysis Dynamic simulation model

Market prices and generation costs Generation load factors Interconnection load factors Generation mix Frequency stability VG curtailment Plant revenues

Investment / hourly dispatch

Investment loop Detailed description of VG Demand time series Investment costs Generation dynamic constraints Fuels costs CO2 price Network transfer capacities Input data

Madone/TIMES model

The model coupling can include multi-annual investment trajectories simulated with Times complemented with annual snapshot simulations with Continental Model

| 14

• 100

200 300 400 500 600 30 60 90 120 150 180 210 240 270 300 330 360

• 100

200 300 400 500 600 30 60 90 120 150 180 210 240 270 300 330 360

• 100
• 100

200 300 400 500 600 30 60 90 120 150 180 210 240 270 300 330 360

Net demand with 40 % VG penetration Net demand with 15 % VG penetration Demand

Difference between 31 weather years (Δmax= 25% daily net energy demand) Difference between 31 weather years (Δmax= 90% of daily net energy demand) Difference between 31 weather years (Δmax= 7% of daily energy demand)

• 100
• 100

200 300 400 500 30 60 90 120 150 180 210 240 270 300 330 360

• 40
• 30
• 20
• 10
• 10

20 30 40 50 60 70 80 90 100

1 year Scenario Week

For one single weather year daily variation of net demand and intra- day variation requires to take consider forecasting errors & margins Intra-day variation of net demand requires to take into account near term flexibilities

Solaire Eolien Fatal Hydraulique Demande DemandeRes

Solar Wind Biomas s Hydro Demand Net Demand

VARIABLE GENERATION IMPACTS DEMAND-GENERATION

BALANCING FROM PLANNING TO OPERATION

Close to 100 scenarios (1 year with hourly resolution) created from synthetic demand, wind and PV data for 31 weather years combined with generation availability

| 15

Variation compared to non VG ref case

GW

• 100
• 100

200 300 400 500 600 700 1000 2000 3000 4000 5000 6000 7000 8000

European net load duration curve European load duration curve

h

GW

• 100
• 100

200 300 400 500 600 700 1000 2000 3000 4000 5000 6000 7000 8000

Peak power needs increase

h

BASE PEAK BASE PEAK

Comparison of the Generation mix transformation with VG Load duration curve approach vs. optimized with the investment loop

Simulation of hourly scheduling leads to a mix with:

• similar base load
• increase of mid-merit plant
• significant increase of flexible peaking plant

compared to a load duration curve investment approach

• 100%
• 50%

0% 50% 100% 150% 200% Base Mid Merit Peak Base generation decreases in the order of the energy provided by VG MID-MERIT MID-MERIT

With flexibility constraints Load duration curve

VARIABLE GENERATION REDUCES THE NEED FOR BASE-LOAD AND

INCREASES THE NEED FOR FLEXIBLE PEAKING PLANT Example with 40% VG penetration in Europe

| 16

Continental provides generation investment solutions that ensure sufficient flexibility to manage annual, seasonal, weekly… to hourly variability However solving a multi-zone hydro-thermal investment and operation

• ptimization, at the European system scale, is a complex problem and

approximations are required:



Different technologies are represented by clusters of identical units (nuclear, coal,CCGT, OCT) with aggregated dynamic constraints => MSG, Min-up and down times, ramp-rates, start and stop times,

• etc. (EU:1640 units and 169 clusters)



The annual optimization is performed by successive optimizations using a sliding window with a perfect foresight of stochastic values (units failure, demand, wind and PV) are known for the duration

• f the window.

In order to simulate short term system operation two layers are added to the approach:



FlexAssessment – tests the robustness of the dispatch solutions considering all aspects of the stochastic behavior of demand and generation.



European synchronous system dynamic model–assessment of the dynamic frequency stability for every dispatch period

INTEGRATION OF A DETAILED FLEXIBILITY ASSESSMENT INTO

CONTINENTAL MODEL

| 17

FLEXASSESSMENT BUILDING BLOCKS AND FLOWCHART

Generation scheduling

• hourly generation

schedule

• for several annual

scenarios

Frequency distribution of

• peration margins
• Upward and downward requirements

for different lead times (eg: day- ahead,1h, 2h, 4h, etc)

Available margins

• Upward and downward available

margins for different lead times

Flexibility assessment : required Vs available operation margins

• Upward and downward flex adequacy for

different lead times represented as a Probability of Insufficient operation margin = f(direction, lead-time)

2-MarginAssessment 3-OPIUM 1- EnhanceDispatch

?

Calculation of operation margin requirements

Lead time

Quantification of the technical flexibility available

| 18

OPIUM generates for each time step t and for each lead time T (1h, 2h, day- ahead, etc) a probabilistic density function of difference between demand and generation in t+T.

Convolution: generation excess generation deficit

Example of distribution of each source of uncertainty for

• ne period

Example of the distribution of demand- generation balance

e.g. n % risk used to define reserve requirements

Probability to need a margin lower than x MW

OPIUM – PROBABILISTIC TOOL FOR THE CALCULATION OF OPERATION

MARGINS AND RESERVE REQUIREMENTS

| 19

Flexibility sources are stacked by decreasing flexibility of each source:

 « spinning component»  Start OCT (gaz or oil)  Start Offline CCGT  Start Offline Coal and nuclear  Load shedding.

Required margin and available margin comparison outputs for each flexibility source a saturation probability.

Proba ( , , , )

saturation flexSource t T way

OCGT upward 1h 01/01-19h

1000 2000 3000 4000 5000 0.1 0.2 0.3 0.4 0.5 0.6

Power (MW) Upward Activation Probability by Technology, for lead time = 2 Head Room Margin OCT Margin CCGT Margin Load Shedding 1 h

FLEXIBILITY ADEQUACY: EXAMPLE OF REQUIRED VS AVAILABLE

UPWARD MARGIN

| 20

flexibility source

annual indicators

number of period with insufficient flexibility mean hourly power deployment (MW) generation headroom 400,8 702,1 quick plant start 1 100,5 150,8 quick plant start 2 0,0 60,1 slow plant start 0,0 0,0 load shedding 0,0 0,0

Flexibility indicators that can be obtained: example for lead time = 1 h, upward direction

Not deployable Insufficient 1H margin probability =0 Means that the probability of deployment

• f flexibility higher than « head room » +

quick plant start 1-2 within in less than 1 h is 0  zero risk to run out to upward margin Annual utilization of different flexibility sources Average 1h upward flexibility deployment = 913 MW

For this example the deterministic scheduling provides sufficient flexibility to cover for all possible scenarios of 1h variability

During 400,8 h we need to resort to OCGT to manage 1h fluctuations

Not desired

INDICATORS REGARDING SHORT TERM FLEXIBILITY ISSUES

| 21

• Renewable capacity is based on the geographic distribution of wind and solar resources. Its

development involves a necessary adaptation of the existing electrical and energy system and drives additional costs.

• The development of renewables increases the need for flexibility and significant work has been

done to model the need for flexibility in planning models.

• Coupling of investment and operation models seems to be the state of the art practice for realistic

size systems. This notion is currently being extended to multi-energy systems.

• EDF R&D is studying the possibility of coupling the energy model MADONE with the electricity

system planning tool Continental Model. Similar type of work has been done by other institutions as :

 Coupling PRIMES – Plexos - DSIM used in the DG study « Energy integration of renewable energy in

Europe « (Kema, Imperial College)

 Coupling TIMES-Plexos (or other) envisaged by the JRC

• Hierarchical approaches permit having a good representation of both the energy and the

electricity system but the solutions obtained are not necessarily optimal.

• Further research is needed either concerning the development of unified models or to define the

type of information exchanges between models if chains of models are used.

CONCLUSIONS AND NEXT STEPS

APPENDIX

| 23

Flexibility is mostly connected with operation decisions and represents the ability of a system to adapt its to both predictable and unpredictable fluctuating conditions, either on the demand or generation side, at different time scales, within economical boundaries. WHAT IS ADEQUACY ? WHAT IS FLEXIBILITY ? Adequacy is connected with the issues of investment decisions and is used as a measure of long term ability of a system to match demand and supply with an accepted level of risk. This is a measure that internalizes the stochastic fluctuations of the aggregate demand and supply.

| 24

ELECTRICITY SYSTEM FUNCTIONS AND FLEXIBILITY

Time scale Domain Elements affected Flexibility sources Close to real time horizon

Seconds Frequency regulation:

Frequency containment reserves (FCR)

Dynamic frequency stability FCR reserve providers Minutes Frequency regulation:

Frequency restoration reserves (FRR)

Frequency FRR reserve providers

Scheduling and dispatch horizon

Minutes to hour Replacement reserves (RR) and balancing Economic dispatch Follow net load variation and FCR and FRR Observability and Forecasting Increase reserves Ramping capability Quick start plant Hours to days Generation scheduling Day-ahead and intra- day markets Generation dispatch Transmission and distribution operation Wind utilisation Observability and Forecasting Efficient market design Scheduling flexibility

Planning horizon

Years Expansion planning Generation adequacy Flexibility adequacy Transmission and distribution reinforcement Optimise generation mix Coordination between generation and network investment

| 25

BUT HOW IS IT BEING SOLVED…

Single optimisation problem minimises investement and

• peration decisions

Generation investement loops connected to production cost models Investement loops + production cost model+

• perational flexibility

assessment Unit construction and Commitment algorithm (J. Ma,

• V. Silva, 2011)

DSIM (Imperial College London, 2012) Plexos with flexibility offline loop (EPRI, UC Dublin) Maximizing future flexibility in electric generation Portfolios (Giraldo & McCalley, 2013) FESTIV (NREL, 2013) CONTINENTAL + FLEXASSESSMENT (EDF R&D, 2014) MEPO-UC (Palmintier, Mort, 2013) CONTINENTAL (EDF R&D, 2009 and 2011) … ect Best solution but difficult to solve => use of simplifications Best solution to identify the critical constraints but not to obtain the optimal investement solution Requires significant simplification on

• peration constraints

| 26

IMPACT OF VARIABLE GENERATION ON THE NET DEMAND

ADDRESSED TO CONVENTIONAL GENERATION

Source: H. Holtinnent et all, « Flexibility in the 21 century power systems», 21st century Power Partnership project

| 27

MADONE: a bottom-up TIMES model

Industry:

• 18 main sectors
• 47 sub-sectors
• 53 energy using technos.

Residential sectors:

• 8 types of dwelling
• 11 energy needs (heating, HW, cook., light…)
• 11heating+hot water technologies
• 3 cooking technos
• 16 other electric appliances

Service sectors:

• 7 main sectors
• 2 types of dwelling for each sector
• 7 energy needs (heating, HW, light, computers…)
• 6 heating+HW technos

Transport sectors:

• Passengers and freight
• 9 transports modes
• 23 transport means

Agriculture:

• 6 energy uses

Oil & solid fuel supply Gas supply (Eurasian area)

• Native gas production
• Pipeline transportation
• Gas storage
• LNG

Biomass-waste supply

• Primary ressources (17 types)
• Conversion technologies (13)
• Final bio-energy products (7)

Uranium supply Non energy uses:

• 6 energy uses

FINAL ENERGY PRIMARY ENERGY

Wind, solar, hydro supply

• availabilty factors

(country + on/off shore + wind speed (9) + hours differenciation)

• Ressources limits:

areas X capacity density hyp. Ex: wind off –shore: km2 available per country according to wind speed (9), distance to coast (2) and depth (3) Ex: solar PV: m2 of roofs available, land available…

Electricty & steam production:

• >50 power generation technos
• Cogeneration: 21 technos
• District heating: 7 technos
• Industrial boilers
• Interconnections among countries

Refineries Industrial transformations

( efficiency rates of direct consumption processes) TRANSFORMATION PROCESSES

Energy sector consumption:

• Grid losses, ancillary conso etc.

= linked to production CO2 storage potentials:

Saline aquifers, DGOF on & off-shore

| 28

Generation reliability model to access generation availability

• Thermal generation uncertainty represented using forced outage rates (ORR) and failure

to synchronize (Ps). A capacity outage probability table (COPT) is built for each dispatch period.

• Hydro generation uncertainty represented using an non-biased normal distribution

Probabilistic model of wind forecast errors

• Forecast uncertainty depends on lead-time and forecasted load factor.
• Wind uncertainty is modeled using empirical distributions
• Shape and dispersion of the distribution depend on the forecasted load

factor

Probabilistic model of PV forecast errors

• Separate representation of small rooftop installations and PV farms
• PV forecast errors are represented by empirical distributions that depend on

the hour of the day and the month of the year

Probabilistic model of demand forecast errors model

• Demand forecast errors are represented by non-biased normal distributions
• The standard deviation of the distribution depends on the hour of the day

and the month of the year

• 3000
• 2000
• 1000

0.05 0.1 0.15 0.2 0.25 0.3 Actual - committed generation (MW) Probability

50 20 40 60 80 100 5 FLF (%IC) LF (%IC) Density

• 20
• 10

10 20 0.02 0.04 0.06 0.08

Deviation (%IC) Density

Hour PV Farms Roof-top PV

OPIUM – PROBABILISTIC TOOL FOR THE CALCULATION OF OPERATION

MARGINS AND RESERVE REQUIREMENTS