Learning-by-Doing vs. Barriers to Cost
Reduction In Photovoltaics:
David vs. Goliath or Bambi vs. Godzilla?
Ian Sue Wing
Boston University
RFF Workshop on Learning-by-Doing in Energy Technologies
June 17-18 2003, Washington D.C.
• Problems in paper’s premises and analysis
• Obscure
– Magnitude of PV’s uphill battle to become competitive
– Nature of barriers to cost reduction
– How LBD might actually reduce the cost of electricity
(COE)
– What else needs to happen for costs to fall
• Why I am not a true believer in PV’s potential
– Cost differential relative to other technologies
– Limits to economic potential on input side
2
• Arrow (1962)
– Experience, Cost Red’n = f (Cumulative Output )
• No data on output of PV electricity (PVE)
– Use cumulative capacity as proxy for output
– Misplaced focus on PV module (i.e. input ) cost
• PV modules: value as an input to generation
– ¢/Wp: cost of technical efficiency of conversion
– ¢/kWh: cost of electricity produced
• Consumers care about cost of output of generation
– Relevant metric: ¢/kWh not ¢/Wp
– Demand = f (PV COE relative to other sources )
3
• Systematic, four-step process of analysis
1. Estimate cost differential b/w. PV, competitive sources of electricity supply
2. Account for components of PVE cost
3. Analyze characteristics of components to get a sense of key drivers of overall cost
• E.g. technological, economic, political forces
4. Develop strategies to reduce PVE cost to competitive levels
• Use margins of adjustment uncovered in step (3)
4
• PVE: high cost, not competitive, subsidized
– >20 ¢/kWh vs. 6 ¢/kWh competitive price
– PV electricity demand ⇒ output subsidies
– Cost differential ⇒ large subsidies to stimulate PVE demand
– Result: low PVE output, demand for PV inputs
• Duke’s assumed growth in demand for capacity
– Where will it come from???
– Predicated on large output subsidies not accounted for
– Welfare conclusions optimistic
– Questionable whether PV module subsidies ⇒ learning, reduced PVE cost
5
Other factors influencing cost
High generation cost
Slow movement down learning curve in PV module manufacturing
Low demand for PV electricity output
Low demand for
PV modules
Subsidies
6
• Spatially heterogeneous resource
– No substitution: Leontief
– Determines PV’s potential supply
• Large technical potential, low economic potential
– Competition with other uses for land surface area
– Geographic variations in land use opportunity costs/constraints
– Influenced by factors spatially uncorrelated with solar resource
– Mix of technical, economic, political constraints
• Barrier to exploiting highest grade solar resource
– COE higher/quantity lower than if sunniest locations covered with PV
– Increases PV’s cost disadvantage
– Degree to which these factors important not known
• Limited solar potential ⇒ PV minor player in global energy future?
7
Source: EIA (2000) 8
Personal and Corporate Income Tax Incentives
Personal Income Tax Incentives Only
Corporate Income Tax Incentives Only
State and utility programs
State programs
Utility programs
Local programs
12
Source: World Energy Council (2002)
• Intermittent, spatially distributed input
– Seasonal, diurnal fluctuations in supply potential
– Generation geographically spread out
• Grid-connected distributed generation = growth market for PV?
• Some reasons why not
– Each generation site must push electrons back into the grid
– Interconnection requires additional equipment
– Capital replication ⇒ dis economies of scale?
– Limited potential for LBD to reduce costs?
• Interconnection issues
– Grid designed around centralized, stable, controllable generation
– Distributed PV do not satisfy these criteria
• Multiplicity of access points + stochastic output fluctuations ⇒ problems of grid control/reliability
– Capital investment in power system control equipment
– PV must be backed by spinning reserves
– Increasing marginal costs of PV as share of electricity supply grows
13
• LBD: no panacea
– As yet limited usefulness as conceptual basis for analysis
• Needed: understanding of role of innovation
– Technical problem-solving ⇒ R&D
– Technical breakthroughs ⇒ “de-bottlenecking”
– Increased supply potential will permit cost-reduction via LBD
• Key problems awaiting solution
– Lowering interconnection costs
– Technical feasibility of grid control w/. distributed PV generation
• Difficulties that may never be overcome
– Limits to solar’s economic potential
– PV’s initial cost disadvantage
• Lesson in underlying economics: complementarity b/w.
R&D, LBD
– No surprise (e.g. Lieberman, 1984)
14
i ( t ), max r ( t ), s ( t )
W
NSB ( t ) =
=
∞
∫
0
NSB ( t ) e − ρ t dt py ( t ) + B ( y ( t ))
Objective: choose capacity additions,
R&D, insolation input to maximize PVE generation’s discounted net social benefit
− V ( i ( t ), x ( t ), h ( t ))
− Z ( s ( t )) − r ( t )
Net social benefit = private revenue + PVE environmental benefit – installation cost – opportunity cost of insolation land surface –
R&D expenditure y ( t ) = F ( k ( t ), s ( t ))
PVE production function (almost Leontief) s ( t ) ≤ s k & ( t ) = i ( t ) − δ k ( t )
Limited potential insolation supply
Capacity evolves w/. installation, depreciation h & ( t ) = M ( r ( t ), h ( t )) x & ( t ) = i ( t )
V ( i ( 0 ), x ( 0 ), h ( 0 )) +
State of knowledge evolves w/. R&D
Experience evolves w/. capacity additions
Z ( s ( 0 )) + r ( 0 ) >> py ( 0 ) + B ( y ( 0 )) 15