françois*lévêque,** - diwintroducion* • *a*book*divided*in*4*parts* –...
Post on 24-Feb-2021
0 Views
Preview:
TRANSCRIPT
Controversies on nuclear power
François Lévêque, Professor of economics at Mines
ParisTech
Berlin Conference on Energy and
Electricity Economics , 29 may 2015
IntroducIon • A book divided in 4 parts
– EsImaIng the costs of nuclear power: points of references, sources of uncertainIes
– The risk of major nuclear accidents: calculaIon and percepIon of probabiliIes – Safety regulaIon: an analysis of the American, French and Japanese cases – NaIonal policies and internaIonal governance
• A posiIve economic approach – Understanding phenomena and assessing effects
• A twofold wager – A non parIsan book could be worthwhile for readers – CasIng light on uncertainIes is a good way to make beUer decisions
• Let’s see 4 controversies for illustraIon
The costs escalaIon curse (Controversy 1)
Past construcIon lead-‐Imes
What does econometrics tell us?
• The scale-‐up is the main driver of the increase in the costs. Building larger reactors took more Ime and they turn out to be more expensive
• There is no evidence of learning effects at the industry level
• PosiIve learning effects are condiIonal to the same type of reactor and same constructor (i.e., architect-‐engineer)
• Safety concerns also took part in the cost escalaIon. The reactors with beUer performance in terms of safety indicators were also more expensive
Today, the construcIon cost of new NPPs is very uncertain
€2012/kW
Source: W. d’Haeseleere (2013)
The future costs of new nuclear 1/2
• The cost of new nuclear will not likely stabilize before many years – ConstrucIon of new plants is shibing toward third-‐generaIon plants
– Only FOAK costs are beginning to be known – Technological compeIIon is an addiIonal source of instability
The future costs of new nuclear 2/2 • The future of nuclear power depends on the ability of vendors and constructors to escape from the cost escalaIon past curse – lower overnight costs and construcIon Ime – more standardizaIon and learning effects – the case of China?
• The future of nuclear power depends on the ability of countries to lower the cost of capital for nuclear investments – Stability of safety regulaIon – CO2 price commitments
Controversy 2
• EsImaIng probabiliIes of nuclear accidents from observaIons or experIse?
Major accidents
A small number of observaIons
EsImaIng probabiliIes of nuclear accident from observed frequencies is a nonsense
• Observed frequencies – INES>3: 1,6 10-‐3 per reactor.year – Core meltdowns: 8.3 10-‐4 per reactor.year – INES 7: 2.7 10-‐4 per reactor.year
• Is 0,11 the probability of an INES7 in 2015 on the planet? ([1-‐(1-‐2.7x10-‐4)435]; Poisson distribuIon)
• No! We cannot assume that observed events are representaIve, that reactors (models and locaIons) are idenIcal, that events are independent, that safety is Ime-‐invariant, etc…
What about ProbabilisIc Safety Assessments (PSAs)?
• Knowledge on nuclear accidents is not limited to the observaIon of past accidents • ProbabilisIc Safety Assessments: for instance, the core damage frequency of the UK
EPR is esImated to 10 -‐6 per year and the core damage with large early release frequency to 3.9x10-‐8
• PSAs figures are much more lower than observed frequencies
• PSAs have strong limitaIons – Limited scope (specific iniIaIng events, specific cascade of failures) – They are not designed to obtain a single number and its confidence interval but
to pinpoint local safety weaknesses and remedies – They assumed perfect compliance with safety standards and regulatory
requirements • PSAs aggregate a huge amount of knowledge that can complement observed
frequencies of accidents
Combining observed frequencies and PSAs
Bayesian Poisson Gamma model, Escobar-‐Rangel and Lévêque, Safety Science, (2014)
Fukushima Daiichi: a strong or a small risk revision to make ? (Controversy 3)
How does the Fukushima-‐Daiichi accident change our predicIon of accident? (8 observaIons of core-‐meltdown, even minimal, before Fukushima-‐Daiichi , 11 aber)
A strong Fukushima Daiichi effect?
Poisson ExponenIally Weighted Moving Average (paramètre d’indépendance : 0,82)
Basing public decision on probabiliIes as calculated by experts or as perceived by laymen? (Controversy 4)
• Experimental psychology studies (e.g., D. Kahneman, 2011) show that our percepIon of probabiliIes is biased
• For instance, the probability of a 0,0001 loss is perceived lower than a probability of 1/10.000 (the so-‐called denominator neglect heurisIc)
How to balance probabiliIes as calculated by experts and as perceived by people in decision-‐making? 1/2
• How to take the percepIon biases into consideraIon in esImaIng the nuclear social cost of accident?
• Nuclear accident is a – Rare event, hence perceived probability is overesImated – Ambiguous event, hence our minds select on the highest value of
probability and damages – Dread event, hence we neglect the denominator and focus on the
event itself which leaves a strong footprint • Consequently, whenever decision making is based on perceived
probabili4es – Overinvestment in nuclear safety – Premature phase-‐outs (e.g., German decision aber Fukushima-‐Daiichi) – Distorted choice between alternaIve power technologies (coal or
hydro are perceived less dangerous)
How to balance probabiliIes as calculated by experts and as perceived by people in decision-‐making? 2/2
• Conversely, whenever decision-‐making is based on calculated probabili4es, people may fight against new plants and whenever they succeed investments would have been made for nothing and a huge amount of money would have been lost (e.g., the shut down of the Superphénix reactor in France)
• How to balance perceived probabiliIes and calculated probabiliIes in esImaIng the expected cost of nuclear accident? – InsItuIonal design: NSAs deliver calculaIons and Government and
Congress integrate percepIons through the policy process – QuanIfying risk aversion and probabiliIes biases
To go further…
top related