The French nuclear Bet Quentin - PowerPoint Presentation

Slide1

The French nuclear Bet

Quentin Perrierperrier@centre-cired.fr

WORKSHOP

The European electricity market and national energy policies

28-29th

september

2017

Ecole

des Mines de Paris

Introduction

Context:

The 58 French nuclear reactors are reaching 40 years old

Retrofitting all of them would cost a 100 billion euros

How many reactors should be retrofitted?

Strong

inertia & necessity to plan in advanceStrong uncertainties: technological cost, power demand, CO2 priceHow can we account for uncertainty?Is there an optimal strategy?If not, what strategy can we define?Optimization model combined with the framework of Robust Decision Making (RDM) developed by Lempert (2006)

Methodology:

Challenges:

Research question:

given the uncertainties and inertia of the power system?

Outline of this presentation

What uncertainties?Presentation of the model: FLORE

Application of RDM and resultsConclusions

What uncertainties?

Cost of retrofitCost of new nuclearPower demandCO2 price

Objective: aim large to examine all plausible values

1. Cost of retrofitted nuclear

Overnight retrofit cost: € 100 billion euros + ?

Cost of capital: what borrowing rate?

Plant availability:

can the decrease be reversed?

Length of lifetime extension:

10 or 20 years?Cost of nuclear waste: between €15 to € 55 billionCost of insurance: negligible?O&M cost: will cost be contained?

To sum up: a retrofit cost from 44 to 95 euros/MWh

2. Cost of new nuclear

Data in the literature goes from 70

€/MWh

to

120€/MWh.

Hinckley

Point C reactor: 92.5 pound/MWh (127 €/MWh)38.5

3.44 568.510.5FlamanvilleOlkiluoto

3. Uncertainty on demand

Official prices from the Quinet report56 euros

en 2020, 100 en 2030 et 200 en 2050

Low trajectory: half these prices

4. Uncertainty on CO2 price

Double dip recession or economic boom?

Electrification of usage and electric vehicle?

The FLORE model

French Linear Optimization for Renewable Expansion

The FLORE model

Minimized total cost, including investment and dispatchHorizon: 2014-2050. Calibrated for France, year 201412 technologies

A detailed representation of hydropowerEndogenous retrofit of nuclear plantsDemand : 6 representative weekC

ode and data are available online at

my GitHub Page

!

TechnologyInvestmentGenerationOnshore windEndogenousEndogenousPVEndogenousEndogenousCoalEndogenousEndogenousCCGTEndogenousEndogenousOCGTEndogenous

EndogenousFuelEndogenousEndogenousHistorical nuclearExogenousEndogenousRetrofitted nuclearEndogenousEndogenousNew nuclearEndogenousEndogenousRun-of-RiverExogenousExogenous

DamsExogenousEndogenousPumped hydroExogenousEndogenous

Application of the Robust Decision Making framework

and results

There is no single optimum

See the interactive model: http://flore.shinyapps.io/modelThe result depends on the choice of uncertain input parametersHow to choose one strategy?Probabilistic approach

: not adapted Justifies the use of RDM. A 5-step methodology

Definition: A

strategy is the choice of which reactors to retrofit

58 reactors => 3.1017 possible strategies

Take

a

representative

sample of possible strategies: Temporal possibilities: 3 groupsShare of retrofitted plants: 3 possibilities per group14 Oldest reactors23 In-between21 Newest reactorsNo retrofitHalf retrofit

FullretrofitNo retrofitHalf retrofitFullretrofit

No retrofitHalf retrofitFullretrofit33 = 27 strategies

For each strategy, testing all plausible values for costs, demand and CO2 prices.

108 ”states of nature” or “plausible futures”.

For each strategy and each plausible future, I compute the regretRegret = Difference between the performance (cost) of the chosen strategy and the performance of what would have been the best strategy in that futureRegret is here the additional cost, the cost of error, the opportunity cost

Candidate strategy: with the lowest upper-quartile regret

The PRIM algorithm (Patient Induction Rule Method) allows to find which clusters of parameters are statistically associated with a strong regret for the candidate strategy S9

Set

of all plausible

futures

Retrofit-friendly futures

Retrofit-adverse

Futureslow demand ANDcost of retrofitted nuclear ≥80 euros/MWhHigh regrets for S9(~vulnerabilities)Low regrets for S9

The initial problem,

multidimensional (108 possibilities), has been simplified to only two clusters

Trade-offs: Switching from one strategy to another reduces regret in one cluster of futures, but increases regret in the other cluster

We can trace an “efficiency frontier” or “low-regret frontier”.

S1, S12, S6, S9 et S18 are

on this

frontier

S1 : full early phase-out (

négawatt)S6, S12 : retrofit 70% S9 : close 14 oldest reactorsS18 : close 7 oldest reactors

No objective probability of each cluster. It ultimately depends on the implicit probabilities of the decision maker.

I compute the expected regret for various implicit probabilities (i.e. preferences)

If you think that retrofit cost above 80 euros/MWh is:likely: pick S1Unlikely: pick S9

Impossible: pick S18

Given current cost estimates: S9 et S18 are two robust strategies !

Comparaison with scenarios from the National Debate on Energy Transition

SOB

DEC

EFF

DIV

Proxy : nuclear production

By considering 27 strategies, we evidence new, interesting strategies

Conclusions

Conclusions

A analytic framework to understand the current economic debate on nuclear, by taking uncertainty into accountWe study 27 strategies, show their trade-offs and the low-regret frontierWe evidence two robust strategies given current cost estimates: closing 7 to 14 of the oldest reactors, and retrofitting all of the rest. Perform better than current official scenariosThis RDM method can be used iteratively in the future, with new data (on costs, demand,

etc).

Thank you for your attention!

Questions?