The Rapidly Changing Economics of Solar PV Power Stefan Reichelstein1

Anshuman Sahoo2

1 Graduate School of Business Steyer-Taylor Center for Energy Policy and Finance Stanford University 2 Department of Management Science & Engineering Steyer-Taylor Center for Energy Policy and Finance Stanford University

February 2014

Introduction

Solar PV Power: A Tale of Two Industries Downstream:

Solar PV currently accounts for a small share of electric power generated in the U.S. But growth rates have been impressive: 40 GW of solar power generation capacity installed in 2013 40 GW ≈ total cumulative installations between 1970s and 2010 By 2013, global installations of solar PV had reached ≈ 150 GW Solar developers, e.g. SolarCity, are succeeding in the marketplace

Upstream:

Solar panel prices have dropped dramatically in recent years → 80% drop between 2008 and 2012 Massive additions to global PV manufacturing capacity → Entry from Chinese suppliers By all common nancial metrics, solar PV module manufacturers have been "bleeding" nancially, at least until recently 1 / 25

Introduction

Three Central Questions for the Economics of Solar PV Question 1: Is electric power generated by solar PV technology currently cost competitive with that from fossil fuels? Consider two alternative PV technologies and two scales: Crystalline silicon and thin-lm PV technologies Utility scale (1MW+) and commercial scale (0.1 - 1MW) Question 2: Which factors are crucial in determining the cost competitiveness of Solar PV? Question 3: Can we simply extrapolate the recent trajectory of solar panel prices to predict future prices? DOE's Sunshot Initiative: Need for a `breakthrough' in solar PV technology?

2 / 25

Levelized Cost of Electricity

Cost Comparison of Alternative Energy Sources Ongoing public discussion about the eectiveness of renewable energies when compared against coal, natural gas and oil. Statements like: "As it stands, electricity from solar power is X-times more costly than electricity obtained from coal power plants" ricochet all the time in the media. Fundamental challenge for such cost comparisons is that alternative technologies dier vastly in terms of: Upfront capital expenditures Periodic operating costs Applicable tax rules

3 / 25

Levelized Cost of Electricity

A Common Cost Metric: LCOE The 2007-MIT study on "The Future of Coal" provides the following verbal denition: "...the levelized cost of electricity is the constant dollar electricity price that would be required over the life of the plant to cover all operating expenses, payment of debt and accrued interest on initial project expenses, and the payment of an acceptable return to investors" Fundamentally, LCOE is a break-even sales price (per kWh) needed to justify an investment in a particular power generation facility

4 / 25

Levelized Cost of Electricity

LCOE Components

Levelized Cost of Electricity

LCOE = w + f + c · ∆ where

w denotes the time-averaged variable operating cost (in $ per kWh) f denotes the time-averaged xed operating cost (in $ per kWh) c denotes the unit cost of capacity (in $ per kWh) ∆ is the tax factor (in%) Remark → LCOE aligns with the Net Present Value (NPV) of an investment → LCOE is the average price, per kWh, that must be earned so that the present value of all operating cash ows (after tax) is zero 5 / 25

Cost Competitiveness of Solar PV Today (Question 1)

Input Parameters Chosen as `best case' Parameter Derate factor Module Price ($/Wp-DC) BoS price ($/Wp-DC) DC-to-AC capacity factor Degradation factor Fixed O&M costs ($/Wp-DC/yr) Discount rate Useful Life Investment tax credit Depreciation type Tax rate

Utility Silicon

Scenario Utility - Commercial - Commercial Thin Film Silicon Thin Film 85% 0.70

1.10 19.9% 99.65% 0.015 7.50%

1.30 20.6% 99.50% 0.020 7.50%

1.35 17.7% 99.50% 0.023 8.00% 30 years 30% 5-years MACRS schedule 40%

1.55 18.1% 99.25% 0.030 8.00%

6 / 25

Cost Competitiveness of Solar PV Today (Question 1)

LCOE Calculator for Commercial Scale Solar PV The Levelized Cost of Electricity Input Parameters: Solar-Commercial-Silicon Useful life (economic), T Tax Depreciation Method System Price (for solar, enter DC system price), SP (total overnight cost) Solar DC/AC Derate factor (for other technologies, enter 100%) Investment Tax Credit, i Capacity factor, CF System Degradation Factor, xt Fixed O&M Cost, Ft Variable O&M Cost, w (excluding fuel) Fuel Cost Carbon Dioxide Emissions Cost (Allowance Cost) Emissions Cost of Capital, r Effective corporate tax rate, α

30 2 2.05 85.00% 30.00% 17.70% 99.50% 23 0 0 0 0.0000 8.00% 40.00%

years ($/W)

($/kW - yr) ($/kWh) ($/kWh) ($/tCO2e) (kg CO2e/kWh)

LCOE calculation Unit Capacity Cost, c Tax factor, ∆ Average fixed O&M cost, f Average variable O&M cost (including fuel), w Levelized cost of electricity

0.1235 0.7069 0.0156 0.0000 0.1029

($/kWh) ($/kWh) ($/kWh) ($/kWh)

7 / 25

Cost Competitiveness of Solar PV Today (Question 1)

Current LCOE Estimates

Finding: Our parameter values give rise to the following Levelized Cost of Electricity estimates (in ¢ per kWh): Silicon Thin lm

Utility-Scale 7.3 8.1

Commercial-Scale 10.3 11.7

8 / 25

Cost Competitiveness of Solar PV Today (Question 1)

Current LCOE Estimates Comparisons:

Commercial-scale PV already more than competitive in areas with high commercial retail electricity rates

SoCal average commercial retail rates around 12 - 14 ¢/kWh Utility-scale PV ≈ 10% more expensive than electricity generated with natural gas LCOE for NGCC facility about 6.6 ¢/kWh Credits from Renewable Portfolio Standards (in select U.S. states) give additional boost to utility-scale installations Trackers can achieve substantial cost reductions Tradeo: Higher system cost vs. higher capacity factor LCOE reduction of up to 20% for ideal sites Silicon Crystalline and Thin Film are running `neck-and-neck' 9 / 25

Sensitivity Analysis (Question 2)

Location, Location, Location....

Map shows the "eective" hours of sunshine per year.1 Capacity factor is roughly proportional to insolation factor 1 Source: U.S. Department of Energy

10 / 25

Sensitivity Analysis (Question 2)

Sensitivity Analysis

Cutting capacity factor in half eectively doubles LCOE → Permanent disadvantage for northern locations! 11 / 25

Sensitivity Analysis (Question 2)

Intermittency and Time of Day Pricing There is considerable variation across the hours of the day and across the seasons of the year in:

The pattern of intermittent power generation Time-of-Day electricity pricing Basic LCOE calculation appropriate for dispatchable energy sources and/or at pricing structures Basic LCOE does not capture synergies or complementarities between intermittent generation and pricing patterns

12 / 25

Sensitivity Analysis (Question 2)

Solar PV: Generation Pattern (summer)

Example summer generation for a 1kW installation in San Francisco:2 Highest productivity between 10:00 a.m and 5:00 p.m. Peak generation more than 3x average generation Power Generation (W) Solid line depicts average power generation.

Example Power Generation from a Fixed Array Solar PV System: Summer ● 750

● ●

●

●

500

● 250

● ●

● ●

0 0

●

●

●

● 5

●

●

●

●

● 10

15

● 20

●

●

●

● 25

t: Hour Ending 2

Simulated data from NREL PVWatts

13 / 25

Sensitivity Analysis (Question 2)

Small commercial customers: Pricing Pattern (summer) Customers with demand < 500kW often on PG&E A6 schedule

Retail price is $0.44/kWh during summer peak demand periods3 PG&E A6 Time−of−Use Commercial Rates (Summer)

p(t) in $/kWh Solid line depicts average price, p.

0.5

●

●

●

●

●

●

0.4

0.3

● 0.2

●

●

●

●

● ●

●

●

●

●

●

●

● ●

●

●

●

0.1

0.0 0

5

10

15

20

25

t: Hour Ending 3

Data from PG&E A6 rate schedule

14 / 25

Sensitivity Analysis (Question 2)

LCOE Adjustment

To account for the joint eect of intermittency and time-of-day pricing: Recent analysis suggests to append LCOE with a multiplicative adjustment factor: Co-variation coecient → Retain, but adjust, existing LCOEs (Reichelstein & Sahoo, 2013) Co-variation coecient = 1 if either: Power prices are uniform over time, or Power generation is uniform (dispatchable energy source) For small commercial customers in the San Francisco Area: Co-variation coecient equals 1.17

LCOE 1.17 Eectively, a 15% reduction in levelized cost Adjusted LCOE =

15 / 25

Sensitivity Analysis (Question 2)

Impact of Federal Tax Subsidies How crucial are the federal Investment Tax Credit (ITC) and the accelerated tax depreciation allowance (MACRS) for the economic viability of solar power? Finding ITC and MACRS have the following impact on the LCOE estimates for commercial-scale silicon PV (in ¢ per kWh): 30% ITC No ITC

5-year MACRS 10.3 15.5

20-year DB 12.3 17.8

Without either subsidy, tax factor, ∆, jumps from .7 to 1.3 Tax rules beyond 2017: 10% ITC and MACRS → LCOE would increase from 10.3 to 13.7 ¢ per kWh 16 / 25

Future Module Price Reductions (Question 3)

The 80% Learning Curve

Historically, sales prices for solar PV modules have shown a remarkably consistent downward trend Swanson (2011) documents the following classic "80% learning curve": Every time the cumulative volume of modules produced doubles, prices fall by 20%

17 / 25

Future Module Price Reductions (Question 3)

The Last Three Years

Annual installations: 28 GW in 2011, 30 GW in 2012, 39 GW in 2013 Key Question: Growing demand and cost reductions versus Excessive additions to production capacity?

18 / 25

Future Module Price Reductions (Question 3)

Economically Sustainable Module Prices Economically Sustainable Module Prices (ESP) are what module manufacturers would need to charge in order to "break-even" cover their costs and receive a normal return on investment ESPs comprise (i) capacity related costs (ii) manufacturing costs and (iii) period costs (SG&A and R&D) Reichelstein and Sahoo (2014) analyze sample data: 11 module manufacturers for the period 2008-2012 Collective market share of ≈ 26% Infer production costs from rm-level nancial statements Infer production levels and shipments from nancial statements and industry analysts

19 / 25

Future Module Price Reductions (Question 3)

Economically Sustainable Module Prices

Conclusion: ESPs and ASPs close for most of 2008-2010, but substantial gap for 2011-2012 20 / 25

Future Module Price Reductions (Question 3)

Points to Note

Given investments in production capacity and administrative structure, only current manufacturing costs are relevant in the short-run. Our estimate of short-run avoidable cost: ≈70 ¢ per Watt Implication for international trade disputes about "dumping": China ↔ U.S. and Europe Analysis of rm-level data allows us to estimate learning rates for manufacturing- and capacity costs (Sahoo, 2014)

21 / 25

Future Module Price Reductions (Question 3)

Extrapolating ESPs

22 / 25

Future Module Price Reductions (Question 3)

Reading the Tea Leaves

The ESP curve represents economic fundamentals and assumes an industry equilibrium Our model of future cost reductions invokes both cumulative production volume and calendar time as drivers of learning → Roughly a 75% learning curve We expect convergence between ESPs and ASPs in coming years ASPs are predicted not to exceed the ESPs New capacity additions unlikely until ASPs are "on track" to catch up with ESPs New capacity additions would move us further down the ESP curve. → Market demand will determine that trajectory → Public subsidies will drive market demand 23 / 25

Summary

Conclusion

For PV installations ("downstream"), we conclude:

In favorable U.S. locations, commercial-scale solar PV power already cost competitive with retail commercial rates LCOE for utility-scale PV currently 10% above generation cost of natural gas facilities Findings assume both current federal tax subsidies and an ideal geographic location

For module manufacturers ("upstream"), we nd:

Dramatic price reductions for modules in 2011-2012 in large part due to excess module manufacturing capacity At the same time, the cost reductions achieved appear to be more favorable than suggested by the 81% learning trend line Economically sustainable price still near $0.9 per Watt in 2013

For the next ve years, we predict:

ASPs distinctly below the 81% learning trend line Without faster growth in market demand, the goal of 50 ¢ per Watt for modules by 2017 will probably not be met, with a remaining gap of about 15 ¢ per Watt 24 / 25

Questions?

Conclusion

Stefan Reichelstein and Anshuman Sahoo Stanford University Contact: [email protected]

25 / 25