Fuel Cells, comments
Learning by doing in energy technologies
an RFF workshop
June 17-18, 2003, Washington, with support of
William and Flora Hewlett Foundation and
Energy Foundation
presentation on Fuel Cells by Dr. Ferdinand Panik
comments by Reyer Gerlagh
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
1
Issues
Five questions in invitation:
1. What are benefits of low-cost Fuel Cells?
2. Can we predict/extrapolate the experience curve?
3. How fast can we drive down the experience curve and is
this costly?
4. How to include the experience curve in CGE/CC models
5. Are supporting policies warranted, and what are their
costs?
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
2
1. What are benefits of low-cost Fuel Cells?
” Benefits of FC cars are mostly environmental (lower
emissions of fine particles and CO2)
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
3
2. Can we predict/extrapolate the experience
curve?
” Experience curve (Technology Model) = IRS (CGE)
” It is too hard to disentangle empirically various sources for
IRS: R&D vs. LbD vs. Set up Costs, inter/intra-firm spillovers, intertemporal spill-overs, etc.
But it is also unnecessary. What matters is:
” Free entry/exit results in Average Cost Pricing
” s* = 1–MC/AC, only depends on the IRS-factor
” s* is probably not lower for matured technologies
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
4
3. How fast can we drive down the
experience curve and is this costly?
” Learning Costs for FCs is low (< 1 billion $, 2010?)
” Learning Costs for FC cars is moderate (1-10 billion $,
2020?)
” Learning Costs for H2 distribution and storage is huge
(> 10 billion $, 2050/2100?)
” Succes of FC cars depends on co-evoluation with H2
suppliers (renewables) and other use of H2
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
5
4. Can we include LbD in CGE/CC models
” Yes, when interpreting as IRS (black box)
” Be cautious aggregating IRS and DRS technologies
” What is elasticity of substitution between aggregate
technologies (VES)?
” Equilibria become less stable when IRS technologies are
substitutes
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
6
5. Are supporting policies warranted, and what
are their costs?
” There are 2 possible reasons for stimulating renewables
(e.g. wind energy, solar energy, FCs) vs. fossil fuels
1. Renewables have higher returns to scale or fossil fuel
technologies have diminishing returns due to resource
scarcity (e.g. CO2 constraint) (carbon tax > Pigouvian
level)
2. Elasticity of substitution between renewables and fossil
fuel technologies is too high for renewables to take off on
own steam, but they are superior
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
7
6. on Climate Change policy and FCs
” To constrain climate change, an energy transition is
essential
” A carbon tax / emission permits / standards is the most
simple, and most effective instrument to set this transition
in motion
” Direct stimulation of R&D is also effective
” Buying low-cost FC technology is costly, unless markets
can be exploited for which FCs are competitive at an early
stage
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
8
¾ Benefits of Fuel Cells as Drive Unit
” Fuel Cell cars are mostly substitute for ICE cars, and
offer limited new services
” Fewer moving parts (potentially low costs?)
” Cleaner environment (fine particles?/CO2)
” Disadvantage: Reduced range (meth:2; H2:10)
” Disadvantage: Large fuel tank
” Barrier: Fuel supply and distribution
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
9
Can we extrapolate the experience curve?
ln(costs per unit)
Early phase of rapidly increasing knowledge
Second phase of slowly
increasing knowledge
20%
Matured
10%
0%
ln(cumulative output)
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
10
Comment on the experience curve
” Typical learning curve measures increasing returns to
scale as if time has no impact
ln(ct) = α + β ln(zt)
β<0
TC
costs
żt = yt
P=AC
MC
cumulative output
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
11
Suggestions for the experience curve
” What should the learning curve measure?
1. (Increasing) returns to scale
2. connectedness of production (scale advantages) over
time
” For this, typical learning curve should be extended with a
‘time filter’ to correct for
1. ‘forgetting’ over time
2. increasing/decreasing production costs over time, e.g.
due to increases in wages / other technologies
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
12
LBD Monopolistic Competition Model
Max ∫e–rt (ptyt – ctyt) dt
s.t.
yt = ψ(pt;Nt)
(λt)
ct = e γ t zt–βφ(N) Zt–β(1 – φ(N))
(µt)
żt = yt – δ zt
(ηt)
Aggregation: Yt = Nt yt = χ(pt); Zt = Nt zt
ψ(pt;Nt) = firm’s demand function; ψp(.;.)<0
φ(Nt) = firm’s internal spill-over rate; φ(1) =1, φ′(.)<0
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
13
Model’s results
Monopolistic supply => markup on marginal costs:
pt = (1 + αt)(ct – ηt), where αt = 1/(σ(Nt)–1)
where σ(yt;Nt)=ptψp(pt;Nt)/ψ(pt;Nt)
Free exit: ∫e–rt (ptyt – ctyt) dt ≥ 0
Free entry: ∫e–rt (ptyt – ctyt) dt ≤ βφ(Nt)ct/(1–βφ(Nt))
Steady state: c ≤ p ≤ (1+(r–g)βφ(N)/(1–βφ(N))) c
SS = Almost Average Cost Pricing: c ≤ p ≤ 1.05 c
SS: ct = e (γ–gβ) t (Y0/(g+δ))–β N βφ(N)
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
14
Extended experience curve
” Extend learning curve
ln(ct) = (α + α1ln(N)) + β ln(zt) + γ t
α1>0; β < 0; γ > 0
żt = yt – δ zt
δ>0
” Back-of-the-envelope estimates (high aggregation level)
γ ≈ 0.75 %/yr (efficient wage costs increase)
β ≈ – 0.25 (gy = gz = 3%/yr, ct constant)
” New high-growth technology
gy = gz = 20 %/yr ; gc = – 4.25 %/yr ; gc / gz = 21%
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
15
Reinterpreting the experience curve
ln(costs per unit)
Rapid growth (20%/yr)
Moderate growth (5 %/yr)
20%
Saturated growth (3%/yr)
10%
0%
ln(cumulative output)
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
16
¾ Experience curve implications
1. Matured technology is also subject to IRS (knowledge
costs of replacing fossil fuel technology > 6%)
2. A technology being replaced becomes more expensive
3. Fall in costs will continue as long as new technology
maintains momentum
4. The (simple) experience curve becomes steeper when
new technology receives additional support, and flatter
when new technology looses momentum
5. Choice for future technology is undetermined
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
17
Numerical illustration from DEMETER-2E
Business as Usual
log production costs
2.5
2
1.5
1
0.5
0
-2
0
2
4
6
log cumulative installed capacity
BAU
TAX20
Learning curve for renewable energy sources
BAU versus Carbon-tax policy
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
18
Riding down the FC experience curve
” With currently known technology, costs come down to
—
Costs of FC cars: 2xbenchmark, at 100.000 cars/yr.
—
Costs of FC’s: 1000 $/car (50 kW), at 300.000 cars/yr.
” Assuming Learning Rate of 20%, costs could halve for
every 20-fold increase in sales
—
FC cars become competitive at 2 million cars/yr.
—
Potential for substantial further cost-reduction at
further increasing market share ?
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
19
¾ Costs of riding down the experience curve
” Production Costs per Unit: c = α z –β
” Total Production Costs = c y = α y(1–β)
y
” Cumulative Costs = ∫0 α z –β dz = α y(1–β)/(1–β)
” Total Learning Costs = CC – TPC = β /(1–β) TPC
” Assume Learning Factor β of 20%
—
TLC FCs (100.000 units) = 100 million $
—
TLC FC cars (2 million cars) = 6.5 billion $
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
20
¾ Experience curves in CGE/CC models
” IRS is consistent with endogenous growth models
” IRS tends to increase with the aggregation level
” Source of IRS (set up costs/non-rival knowledge) is less
important
” Practical suggestion: yt = Atη kt α lt(1–α); At+1 = At(1–δ) (λyt)δ
” Alternative: yt = Atη/(1–α) kt α lt(1–α); At+1 = At(1–δ) (λlt)δ
” Implication: long-term equilibrium becomes more sensitive
to policies and parameter variations
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
21
IRS and technology choice
Consumer Indifference Curve
(iso-utility-line)
FC cars
FC cars
Consumer Indifference Curve
(iso-utility-line)
Production
Set
ICE cars
vrije Universiteit
Amsterdam
Production
Set
ICE cars
Reyer Gerlagh, Institute for Environmental Studies
22
Two possible technology futures
FC car
costs per unit
costs per unit
FC car
ICE car
FC subsidy
vrije Universiteit
ICE car
FC subsidy
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
23
14
14
12
12
10
10
($/GJ)
($/GJ)
Numerical illustration from DEMETER-2E
8
6
8
6
4
4
2
2
0
2000
2050
BAU Fossil Fuels
2100
2150
BAU Renew ables
Business as Usual
vrije Universiteit
2200
Amsterdam
0
2000
2050
TAX20 Fossil Fuels
2100
2150
2200
TAX20 Renew ables
Carbon-tax policy
Reyer Gerlagh, Institute for Environmental Studies
24
The future of Fuel-Cell Cars
—
FC cars are a technological improvement (fewer
moving parts)
” FC cars have potential for a break-through
—
FC cars have limited customer advantages (noise)
—
FC cars have serious customer disadvantages (range)
” FC cars will not break-through on their own steam
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
25
¾ Policy Implications neglecting
resource/factor use
1. Both FC and ICE technology may have Increasing
Returns to Scale, that may exceed 20%
2. The ratio MC/AC as the basis for subsidy is possibly the
same for FC and ICE technologies.
3. Subsidizing FC technologies (when in contrast to ICE
technologies) to overcome learning costs is unwarranted
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
26
¾LbD/IRS & Climate Change
” Climate change is main reason for stimulating (a transition
to) FCs.
” When fossil fuels are considered exhaustible, or
cumulative CO2 emissions are constrained because of
Climate Change, ICE may be subject to diminishing
returns compared to FCs
” LbD/IRS warrants extra stimulus for FC
” optimal carbon tax = Pigouvian level + IRS compensation
= win/win
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
27
FC and climate change policies
” Generic carbon tax / emission standards
” Advantage:
—
minimal public information costs (government does not
need to choose between technologies)
—
No costs of public funds
” Disadvantage:
—
Gap between FCs and ICE may be too wide to be
bridged by tax-induced price wedge
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
28
FC and government purchase
” Government purchases of (expensive) FC cars
” Advantage:
—
Reduced uncertainty for both government and
producers => stimulates investments
” Disadvantage:
—
Possibility of lock-in in inferior FC technology
—
Use of (expensive) public funds
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
29
FC and knowledge policy
” Government stimulates sharing knowledge / reduces
fragmentation
” Advantage:
—
directly stimulates producing knowledge
—
lower costs of production
” Disadvantage:
—
Possibility of too monopolistic behavior of ‘owners’ of
technology
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
30
¾ FC policy: walking on two legs
” Is FC car superior to ICE car?
” Development of FC is co-evolution of FC technology and
hydrogen/methanol infrastructure
—
Stimulate development of FCs through stepping stones
(Exploit all niche markets for FCs to spread learning
costs)
—
Stimulate renewables conversion into H2?
” Required policy is transient (but still long-lasting and
costly)
vrije Universiteit
Amsterdam
Reyer Gerlagh, Institute for Environmental Studies
31