John P. Weyant

Stanford University

EIA 30th Anniversary Conference

Washington, D.C.

April 8, 2008

Three Frontiers in Energy Modeling:Baselines, Technology and Uncertainty

Three Frontiers in Energy Modeling

I. Developing Baselines

II. Representing Technology

III. Incorporating Uncertainty

I. Developing Baselines:Alternative Global Carbon Emission Projections

0

10

20

30

40

50

60

1990 2000 2025 2050 2075 2100

Year

Bil

lio

n M

etr

ic T

on

s

2005 Technologies

High Baseline

Reference Baseline

Low Baseline

550 ppmv Stabilizaion

Projected Probabilistic Rangeof Global Carbon Emissions

0

5

10

15

20

25

30

35

2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100

Ca

rbo

n E

mis

sio

ns

(G

T)

9th Percentile

93th Percentile

Mean

II. Representing Technological ChangeLimitations and Possible Extensions to

Current Methods for Modeling TCCurrent approaches omit important dynamics of technologicalchange. A broader framework for analyzing TC is needed.

Private R&DInvestment

HeterogeneousInnovators

(“Returns to R&D”)

Uncertainty inR&D Returns:

Discontinuous Diffusion

UncertainTax Policy

Path-dependenceand inertia

Innovation &Knowledge

Accumulation

IntersectoralSpillovers

Technological Change:Diffusion and

Market PenetrationINER

TIA

Learning-by-Doing

Public R&DSpillovers

MajorInnovation

IntrasectoralSpillovers

OUTSIDE INFLUENCESINDUCED CHANGE

Total R&D Investment

TechnologicalChange

Returnsto R&D

Tax Policy Spillovers

The VALUE OF DEVELOPINGNEW ENERGY TECHNOLOGY

(Present Discounted Costs to Stabilize the Atmosphere)

Minimum Cost Based on Perfect Where & When

Flexibility Assumption. Actual Cost

Could be An Order of

Magnitude Larger.

BAU(1990)BAU(Tech+)

advanced

technology

550

650

750

$1

$10

$100

$1,000

$10,000

$100,000

Presen

t Discou

nted

Cost, B

illions of 1990 U

S $

Technology Assumption

Steady-State

CO2

Concentration

(ppmv)

Battelle Pacific Northwest Laboratories

Table 1: Technology Assumptions Year 2100

Technology units 1990 Base

Mini-CAM B2

Mini-CAM B2

AT

US Automobiles mpg 18 60 100

Land-based Solar Electricity 1990 c/kWh 61 5.0 5.0

Nuclear Power 1990 c/kWh 5.8 5.7 5.7

Biomass Energy 1990$/gj $7.70 $6.30 $4.00

Hydrogen Production (CH4 feedstock) 1990$/gj $6.00 $6.00 $4.00

Fuel Cell mpg (equiv) 43 60 98

Fossil Fuel Power Plant Efficiency (Coal/Gas) % 33 42/52 60/70

Capture Efficiency % 90 90 90

Carbon Capture Power Penalty (Coal) % 25 15 5

Carbon Capture Power Penalty (Gas) % 13 10 3

Carbon Capture Capital Cost (Coal) % 88 63 5

Carbon Capture Capital Cost (Gas) % 89 72 3

Geologic Disposal (CO2) $/tC 37.0 37.0 23.0

Reducing Cost and Increasing Efficiencyof Photovoltaic Systems

(M. Green, UNSW)

Cost ¯• Cheaper Active Materials

(abundant inorganic or organic)• Lower Fabrication Costs

(low-cost deposition / growth)• Cheaper BOS Components

(substrates, encapsulation, …)

Efficiency ­Reduce the Thermodynamic Losses at Each Stepof the Photon-to-Electron Conversion Process• Light Absorption• Carrier Generation• Carrier Transfer and Separation• Carrier Transport

Cost ¯• Cheaper Active Materials

(abundant inorganic or organic)• Lower Fabrication Costs

(low-cost deposition / growth)• Cheaper BOS Components

(substrates, encapsulation, …)

Efficiency ­Reduce the Thermodynamic Losses at Each Stepof the Photon-to-Electron Conversion Process• Light Absorption• Carrier Generation• Carrier Transfer and Separation• Carrier Transport

Wafer-based (c-Si)Wafer-based (c-Si)

Thin-films (CIGS, CdTe, a-Si, …)Thin-films (CIGS, CdTe, a-Si, …)

“III Generation”concepts

“III Generation”concepts

Inorganic Thin-Film Photovoltaics

High efficiency Materials

Performance enhancement through• optimized geometry• quantum effects

• (Novel) low-cost, abundant, non-toxic,and stable semiconductor materials

• Thin films: low volumes and lowerrequirements for charge transport

• Low-cost deposition processes

Nanoscale morphology

III generation concepts with efficiency limitsbeyond the single junction limit of 31%

intermediate band, up-converters, tandem (n=3)intermediate band, up-converters, tandem (n=3)

hot carrierhot carrier

TPVs, thermionicsTPVs, thermionics

tandem (n=2)tandem (n=2)

multiple exciton generationmultiple exciton generation

31%31%down-convertersdown-converters

68%

III. Incorporating UncertaintyInformation, Foresight & Uncertainty:Three Alternative Sets of Assumptions

Invest StartOperation

StopOperation

State of Energy System

Time

Plan & Build Operate

t1 t2t0

(1) Static, Myopic, or Recursive Dynamic(2) Perfect Foresight (Rationale Expectations)(3) Decision Making Under Uncertainty

Levelized Cost Comparison for Electric Power GenerationWith $200 per Ton Tax on Carbon (2002 Fuel Prices)

0

0.05

0.1

0.15

0.2

0.25

0.3

Nuclear Coal Gas CC Gas CT Solar PV SolarThermal

Wind

Generation Technology

Carbon Tax

Fuel-2002

Variable O&M

Fixed O&M Charges

Capital Charges

$/KWHr(2006$s )

Levelized Cost Comparison for Electric Power GenerationWith $200 per Ton Tax on Carbon ( 2005 Fuel Prices)

0

0.05

0.1

0.15

0.2

0.25

0.3

Nuclear Coal Gas CC Gas CT Solar PV SolarThermal

Wind

Generation Technology

Carbon Tax

Fuel-Oct.2005

Variable O&M

Fixed O&M Charges

Capital Charges

$/KWHr(2006$s)

Interplay BetweenR&D and Investment Decisions

R&DDecision

C0

InvestmentDecision

C1C1 C2

Cost Cost

Time

Assessments of R&D Projects(Erin Baker, et al.)

Purely Organic Solar Cells

.1.3.90Low cost substrate (total < $50/m2)P4

.34

.90

.50

.85

ex1

.04

.5

.3

.9

ex2

.01Total

.25Low cost deposition (total < $50/m2)P3

.5Stability 30 yearsP2

.8Efficiency 15%P1

ex3Funding Trajectory $15M/yr 10 yrsNeed Estimates for

We can reconcile divergent expert judgments through peer review; or run separate scenarios and see how overall policy changes under different expert judgments.

0

5

10

15

20

25

2000 2005 2010 2015 2020 2025 2030 2035

Year

cen

ts/k

Wh

IEA/Duke

Kourvaritakis

Progress Ratios

Learning Curves Extrapolated for Solar PV Energy, with Thin-Film

Technology Transfer in 2023

Projected Probabilistic Rangeof Global Carbon Emissions

0

5

10

15

20

25

30

35

2000 2010 2020 2030 2040 2050 2060 2070 2080 2090 2100

Ca

rbo

n E

mis

sio

ns

(G

T)

9th Percentile

93th Percentile

Mean

Approaches to ModelingUnder Uncertainty

• Stochastic Dynamic Programming

• Stochastic Linear/Non-Linear Programming

• Stochastic Control

• Stochastic Simulation

• Intelligent Stochastic Simulation

• Bounding

• Sensitivity Analysis

• Multi-Dimensional Sensitivity Analysis

• Strategic Scenarios

MARKAL

ImportantParametersand Inputs

Policies ofInterest

DirectedSamplingMethods

Issues ofInterest

Relevant Questions

Answers toQuestions

CEMAC Approach to Modeling & Analysis

(Stochastic Wrapper)

Distribution Distribution

Recommendations for Energy Modelers

• Work Harder on Baselines

• Don’t Ignore Technological Change

• Don’t Ignore Uncertainty

• Don’t Let the Perfect Be the Enemy of the Good and Useful

– i.e., Simple things are a lot better that no things.

Thank You!

Potential Areas of Model Refinement (I)

1. Technology/Technology Change– Invention

– Innovation

– Diffusion

2. Spatial/Temporal Disaggregation

3. Uncertainty– In the World, aka Scenario Uncertainty

– How it Impacts Behavior of Modeled Agents

– Related to Degree of Foresight Assumed

4. Data– Technology, Energy End Uses, Resources

– Institutions

– Economic Output, I/O, Fuel Markets, Trade

Potential Areas of Model Refinement (II)

5. Representation of Market Imperfections

6. Representation of “Non-Rational” Behavior

7. Ability to Analyze “Plausible” Policies

• Standards

• Sectoral Caps

• Remedies for Market Imperfections

8. Macro/Microeconomic Integration

9. Public Finance/Financial Market Integration

10. Marrying Conceptual Structures With Data

Basic Strategies for Developing Models

• Identify All Potential Questions First, Then Design the Model to Help Address Them

• Develop a Flexible Modeling Architecture That Can Be Easily Adapted to New Problems

• Do Both!

Model Development/Assessment Issues:Common Pitfalls in Policy Modeling

• Lack of Focus

– Pick a basic model structure without a set of applications firmly in mind

– Not modifying model in response to new problems

• Mistaking the Model for Reality

– If its not in the model it probably doesn’t exist

– Test alternative assumptions only against the model

– Methodological limitations imply real world restrictions

• Poor Communication of Results

– Overstating strength of results

– Omitting key relevant assumptions/qualifications

Assessments of R&D Projects (Baker, Cont.)Define Investment Level and Technical success

• Example: Advanced Solar; purely organic solar cells• Investment: $15 Million per year, for 10 years.• Technical Success:

– Cost of $50/m2; – efficiency of 15%; – 30 year life time (defined as working at least 75% of

original efficiency after 30 years)• We will define intermediate hurdles:

– Identifying molecules that can achieve efficiency.– Identifying molecules among that group that can

achieve stability.– Hurdles related to the cost of depositing the material

and identifying a low cost substrate.• Then, assess probability of success.

0.00

200.00

400.00

600.00

800.00

1000.00

1200.00

1400.00

1600.00

1800.00

Exajo

ule

s

AIM

DN

E2

10

0

GR

AP

E

MA

RIA

MIn

iCA

M

EP

PA

IIA

SA

IMA

GE

Model

World Primary Energy in 550 ppm Case in 2100

OtherBiomassWindSolarNuclearHydroCoalGasOil

The Energy Resource Mix

Observations Regarding Current Approaches to Modeling Tech. Change

• Current approaches to TC provide a good foundation:– spillovers

– innovation incentives and knowledge capital

– heterogeneous firms and technologies

• Current approaches suggest weak or ambiguous effect of ITC, butunderestimate importance:

– Focus only on R&D-based technological change

» learning-by-doing

» diffusion or imitation by existing technology

– Assume continuous, known returns to R&D function (no surprises or discontinuities)

» No provision for major innovations

» Model only one dimension of technological change (cost)

– Neglect path-dependence and inertia in changing technology dynamics

• Modeling challenge will be to incorporate enough complexity torealistically capture technology dynamics in a meaningful way.

• Policy challenge will be to use insights from models, but qualify findingswith a more complete understanding of technological evolution.