1
Bottom-Up, Top-Down and Non-Convex Recent
Modeling Activities at IIASA-ECS
Leonardo Barreto and Leo Schrattenholzer (Project
Leader) Environmentally Compatible Energy
Strategies Project (ECS) International
Institute for Applied Systems Analysis (IIASA)
IIASA Briefing Washington D.C., March 13, 2002
2
Outline

• IIASA and ECS
• IIASA Modeling Framework
• Combining Top-down and Bottom-up
• Endogenizing non-convex technical change
• Summary

3
IIASA Yesterday and Today

• 1967 Initiative of US President Johnson and SU
  Prime Minister Kossygin
• 1972 Create research center (NGO) as a neutral
  bridge between East and West
• 2000 Analyze, from an NGO perspective,
  sustainability and global change in three fields
• (1) Energy and Technology
• (2) Environment and Natural Resources
• (3) Population and Society

4
ECS Research Themes

• Research focus Global energy and environmental
  interactions
• Technology assessment, learning curves, RD
  effectiveness
• Energy infrastructures
• Analysis of historical driving forces and
  scenarios of future global environment-energy-econ
  omy interactions
• Integrated Assessment of climate change

5
ECS Collaboration

• Since 1981 co-organizer (with Stanford
  University) of the International Energy Workshop
  (IEW)
• Energy Modeling Forum (EMF)
• World Energy Council (WEC)
• Intergovernmental Panel on Climate Change (IPCC)
• International Energy Agency (IEA)
• Commission of the European Community
• National institutions (governmental, industry,
  universities) in Japan, the US, Russia, China,
  and others

6
ECS Achievements (1)

• 1998
• IIASA World Energy Council present major
  findings of a five-year global energy study
• Global Energy Perspectives
• to 17th World Energy Congress of more than 5,000
  energy industry leaders

7
ECS Achievements (2)

• 2000With the Intergovernmental Panel on Climate
  Change (IPCC), IIASAs ECS Project coordinated
  the development of scenarios for the IPCC
• Special Report on Emissions Scenarios
• 2001ECS continued its scenario work with the
  IPCC, Third Assessment Report on Climate Change
  2001 Mitigation
• Contributing to Chapter 2 on Greenhouse Gas
  Emission Mitigation Scenarios and Implications

8
The IIASA Modeling Framework

• Storyline
• Economic Development
• Demographic Projections
• Technological Change
• Environmental Policies
• Energy Intensity

• Common Databases
• CO2DB
• EDGAR etc.

Scenario Generator -Economic and Energy
Development
RAINS Regional Air Pollution Impacts Model
MACRO
Soft Link
MESSAGE
BLS Basic Linked System of National Agricultural
Models
MAGICC Model for the Assessment of GHG-Induced
Climate Change
9
Other Models Used at IIASA-ECS
ERIS Small-scale model with Endogenous Technical
Change -1FLC Learning-by-Doing -2FLC
Learning-by-Doing Learning-by-Searching -
Stochastic two-stage programming version
MERGE Model for Evaluating Regional and Global
effects of GHG Policies
ISPA Stochastic Meta-model for Multi-Objective Po
licy Analysis
MARKAL Market Allocation Model
10
World Regions used by IIASA-ECS
11
MESSAGE-MACRO

• Two-way link between the bottom-up MESSAGE and
  top-down MACRO models
• The link establishes consistency between demand
  (MACRO) and supply (MESSAGE) and thus between
  scenarios
• The models are solved independently
• Nonlinearities are collected in one place
• High flexibility
• High transparency

12
Running MESSAGE-MACRO

Reference GDP
Reference final-energy demand
Scenario Generator
Energy intensities
Conversion
Conversion
Conversion
Final-energy demand
Useful-energy demand
Useful-energy demand
MACRO
MESSAGE
Final energy shadow prices
Final energy demand
Cost functions
Total energy system cost
Conversion

13
GDP Per Capita Growth in Asia 1960-1997
14
The MESSAGE Model

• Includes 400 individual energy conversion and
  end-use technologies
• Formulated for 11 world regions
• Calculates an optimal (least-cost) energy supply
  path, which satisfies a given useful-energy
  demand
• Incorporates technological progress in different
  path-dependent directions according to the
  scenario specification

15
Examining Carbon Scrubbing

• Collaboration with Carnegie Mellon University
  (Prof. Ed Rubin)
• Estimate learning rates for carbon capture
  technologies (CCT)
• Examine the effect of assumptions about learning
  of CCT on IIASAs integrated assessment scenarios

16
Technological Learning and Carbon Scrubbing
20
demand reduction
B2
15
fuel switching
(mainly shifts away from coal
scrubbing and
10
removal, synthetic-
fuels production
B2-550t
scrubbing and
5
removal, power
sector (natural gas)
scrubbing and
0
removal, power
1990
2010
2030
2050
2070
2090
sector (coal)
17
Examining Hydrogen FuturesThe B1-H2 Scenario

• Collaboration with Tokyo Electric Power Company
  (TEPCO)
• Examine prospects for fuel cells and other
  hydrogen-using technologies
• Quantify a long-term sustainable
  hydrogen-economy scenario

18
Hydrogen Production in B1-H2
19
Diffusion of Fuel Cells in B1-H2 Transportation
Sector
20
The MERGE Model

• Collaboration with Stanford University (Prof.
  Alan S. Manne)
• Incorporating learn-by-doing effects into MERGE
  and related models
• Examine the effect of different assumptions about
  technological change on carbon emissions patterns
• Emulation of long-term IIASA-SRES scenarios using
  MERGE (for future projects, e.g. China)

21
Small-Scale Model of Electricity Choices with
Learn-by-doing
22
Comparing Carbon EmissionsIIASAs A1 and
MERGE-LBD
23
MERGE Emulation of B2 Scenario Primary Energy in
China
24
The ERIS Model

• ERIS (Energy Research and Investment Strategy),
  developed at PSI (Switzerland) and IIASA
• Small-scale model with endogenized learning
  curves (learning-by-doing and learning-by-searchin
  g)
• Flexible tool to assess approaches to endogenize
  technological change
• Global, multi-region model with CO2 trading (for
  the time being, only for the power sector)

25
Two-Factor Learning Curves

• RD conceptualized as a factor inducing
  technological learning
• Specific cost as function of cumulative capacity
  (CC) and knowledge stock (KS)

b Learning-by-doing elasticity c
Learning-by-searching elasticity
26
Learning Rates of Energy Technologies
Source McDonald and Schrattenholzer (2001), 42
technologies
27
Endogenizing Learning Curves

• Non-linear non-convex optimization problem
• Different locally optimal solutions can lead to
  quite distinct energy supply paths
• Globally optimal solution Least-cost energy
  system path
• No guarantee of globally optimal solution with
  conventional NLP solvers

28
Solving the Optimization Problem

• Mixed-Integer Programming (MIP) if other
  nonlinearities do not exist
• Guided optimization with conventional NLP
  algorithms (different solvers/starting points)
• Global optimization algorithms (e.g. BARON)

29
Global Power-Sector Carbon Emissions
Multi-Regional Electricity ERIS Model with and
without Endogenized Technological Learning
30
ERIS and Spillovers of Learning

• Multi-regional ERIS endogenizing
  learning-by-doing using MIP approach
• Learning investments in one region lead to
  specific-cost reductions also in others
• With spillovers of learning, deploying a
  technology in a region can thus affect
  technology choices in other regions
• This phenomenon cannot be captured by models with
  exogenous technical change

31
Spillovers of Learning Carbon Emissions in
Non-Annex B
32
Finding Globally Optimal Solutions

• BARON Branch and Reduce Optimization Navigator
  (Sahinidis, University of Illinois, 2000)
• General-purpose global optimization software
• Combines enhanced branch and bound with range
  reduction techniques
• GAMS/BARON

33
Example 4 Locally Optimal SolutionsOptimized
RD Expenditures (Mill. US90) - ERIS
2
1
4
3
Global Optimum
Solar PV
Wind Turbine
34
Globally Optimal RD ExpendituresExample with
two Technologies - ERIS
35
Summary

• IIASA-ECS explores long-term perspectives of the
  global energy system and related policy issues,
  generating new methods, tools and insights to
  support decision-making
• International and interdisciplinary NGO status
  allows providing global insights while playing an
  honest-broker role
• ECS collaborates with different partners and is
  looking forward to extending the network of
  collaborators