Estimating the cost of novel (pre-commercial) systems for CO2 capture - Ed Rubin, Carnegie-Mellon University
1. Estimating the Cost of Novel
(Pre-Commercial) Systems
for CO2 Capture
Edward S. Rubin
Department of Engineering and Public Policy
Department of Mechanical Engineering
Carnegie Mellon University
Pittsburgh, Pennsylvania
Presentation to the
CCS Cost Workshop
International Energy Agency (IEA)
Paris, France
November 6, 2013
2. Characteristics of Novel
(Pre-Commercial) Capture Systems
•
Includes any technology not yet deployed or available
for purchase at a commercial scale
– Current stage of development may range from
concept to large pilot or demonstration project
•
•
Process design details still preliminary or incomplete
•
May require new components and/or materials that are
not yet manufactured or used commercially
Process performance not yet validated at scale, or
under a broad range of conditions
E.S. Rubin, Carnegie Mellon
3. Examples of Pre-Commercial Systems:
Everything beyond Present
Post-combustion (existing, new PC)
Pre-combustion (IGCC)
Chemical
looping
Cost Reduction Benefit
Oxycombustion (new PC)
OTM boiler
CO2 compression (all)
Amine
solvents
Physical
solvents
Cryogenic
oxygen
Present
E.S. Rubin, Carnegie Mellon
Advanced
physical
solvents
Advanced
chemical
solvents
Ammonia
CO2 compression
PBI
membranes
Solid
sorbents
Membrane
systems
ITMs
Biomass cofiring
Ionic liquids
Biological
processes
Metal organic
frameworks
CAR process
Enzymatic
membranes
5+ years
10+ years
15+ years
Time to Commercialization
20+ years
Source: USDOE, 2010
4. Capital Cost per Unit of Capacity
Capital Cost per Unit of Capacity
Typical Cost Trend of a New Technology
Research
Development Demonstration
Deployment
Mature Technology
Research
Development Demonstration
Deployment
Adspted from EPRI
Mature Technology TAG
E.S. Rubin, Carnegie Mellon
Time or Cumulative Capacity
Time or Cumulative Capacity
5. Historical Trends of FGD and SCR
(DeSOx and DeNOx) Costs . . .
8
1982
1976
1980
250
1982
200
1990
1995
150
1975
100
1974
1972 (1000 MW, eff =80-90%)
1968 (200 MW, eff =87%)
50
20
30
40
50
60
70
80
Cumulative World Wet FGD Installed
Capacity (GW)
Source: Rubin et al. 2007
All costs based on a standardized
500 MW coal-fired power plant
(except where noted)
SCR Capital Costs ($/kW), 1997$
120
10
6
5
4
3
2
1980
1990
..
1976 . display the characteristics
seen in the previous1995
slide
1975
1974
1968
1972
(200 MW, eff =87%)
(1000 MW, eff =80-90%)
1
0
0
7
O&M Costs (1997$/MWh)
Capital Costs ($/kW) in 1997$
300
1980 ← First Japan commercial installation on a coal-fired power plant
0
1980 10
0
20
30 40
50 60
70 80
90 100
First
1983 ←1983 German commercial installation
Cumulative World Wet FGD Installed
90110100
100
Capacity (GW)
1989
1989
90
1979
80
1978
↓ First US commercial installation
70
60
1993
1977
1993
1995
1995
2000
2000
50
40
-20
-10
0
10
20
30
40
50
Cumulative World SCR Installed Capacity (GW)
E.S. Rubin, Carnegie Mellon
60
70
80
6. Five Approaches to Cost
Estimation for Novel Systems
•
•
•
•
•
Conventional engineering-economic estimates
Conventional with probabilistic inputs
Application of learning (experience) curves
Expert judgment /elicitations
Don’t ask, don’t tell
E.S. Rubin, Carnegie Mellon
8. Most novel processes use
conventional costing methods
Many cost elements depend on the
technical maturity of the process
E.S. Rubin, Carnegie Mellon
9. Cost estimates for novel processes often
ignore “process contingency cost” guidelines
•
Process Contingency Cost
“factor applied to new technology … to quantify
the uncertainty in the technical performance
and cost of the commercial-scale equipment.”
- EPRI TAG
Process
Contingency
Cost
Technology Status
(% of associated
process capital)
New concept with limited data
40+
Concept with bench-scale data
30-70
Small pilot plant data
20-35
Full-sized modules have been
operated
5-20
Process is used commercially
0-10
E.S. Rubin, Carnegie Mellon
Source: EPRI, 1993; AACE, 2011
•
Most studies of
advanced capture
systems for full-scale
power plants (~500
MW) assume much
smaller values than
these guidelines call
for, e.g.:
•
7%
•
<20% (EPRI, 2011)
•
•
(IEAGHG, 2011)
18%
(USDOE, 2010)
10%
(IECR, 2008)
10. What system are we trying to cost?
•
•
Early commercial deployment? (first of a kind), or
Mature widely-deployed system? (Nth of a kind)
The major difference is usually in the system design
(though some cost factors also may change)
Whichever case is chosen, process contingency cost
depends on the current state of technical
E.S. Rubin, Carnegie Mellon
11. Cost studies also commonly ignore
guidelines for “project contingency cost”
•
Project Contingency Cost
“factor covering the cost of additional equipment
or other costs that would result from a more
detailed design of a definitive project at an
actual site.” - EPRI TAG
EPRI Cost
Classification
Class I
(~AACE Class 5/4)
Class II
(~AACE Class 3)
Class III
(~ AACE Class 3/2)
Class IV
(~AACE Class 1)
Project
Contingency
Design Effort
(% of total process
capital, eng’g &home
office fees, and process
contingency)
Simplified
30–50
Preliminary
15–30
Detailed
10–20
Many Class I-III
studies assume
≤10%
Conclusion:
• The total contingency
cost for novel capture
processes is grossly
under-estimated in
most cost estimates
• (e.g., totals of ~20-30%
rather than ~40-100%)
Finalized
5–10
Source: EPRI, 1993
E.S. Rubin, Carnegie Mellon
•
12. Example of Contingency Cost
Impact on Total Cost
Increase in Capital Cost and COE
(based on a 2-stage membrane capture system)
Process + Project Contingency Cost
for the CO2 capture process alone
Parameter
20%
50%
100%
Capture System Capital Cost
($/kW)
17%
42%
85%
Capture System COE
($/MWh)
8%
20%
40%
Total Plant Capital Cost
($/kW)
5%
13%
26%
Total Plant COE
($/MWh)
4%
11%
18%
E.S. Rubin, Carnegie Mellon
17. One-Factor Learning (Experience)
Curves are the Most Prevalent
Model equation: Ci = a xi –b
where,
Ci = cost to produce the i th unit
xi = cumulative capacity thru period i
b = learning rate exponent
a = coefficient (constant)
Fractional cost reduction for a doubling of
cumulative capacity (or production) is defined as
the learning rate: LR = 1 – 2b
20000
1983
1981
Photovoltaics
10000
5000
USA
Japan
2000
1992
1995
1982
Windmills (USA)
1000
500
1987
1963
Gas turbines (USA)
200
100
10
1000
10000
100
Cumulative MW installed
Source: IIASA, 1996
•
Most appropriate for projecting future cost of a
current technology already commercially deployed
•
Application to novel (pre-commercial) processes
requires careful consideration of the “starting point”
(cost and experience base) for future cost reductions
E.S. Rubin, Carnegie Mellon
RD&D
Commercialization
1980
100000
18. Reported Learning Rates for Power
Plant Technologies Vary a Lot
Histogram of Learning Rates in the Literature
12
3
Mean = 14%
Median = 13%
Std Dev = 13%
Natural Gas-Fired Plants
On-Shore Wind Plants
10
Mean = 16%
Median = 12%
Std Dev = 14%
8
Frequency
Count
2
6
4
1
2
0
-5% to
-1%
0
-15% to -10% to - -5% to -9%
4%
1%
0% to
4%
5% to 10% to 15% to 20% to 25% to
9%
14%
19%
24%
29%
Learning Rate
Dependent Variable = $/kW (n=9)
Dependent Variable = $/kWh (n=2)
30% to
34%
0% to 4% 5% to 9%
10% to 15% to
14%
19%
Learning Rate
20% to
24%
24% to
29%
30% to
34%
Dependent Variable for electric powerDependent Variable = $/kW (n=23)
Table ES-1: Summary of studies characterizing historical learning rates = $/kWh (n=12)
generation technologies.
Technology
Number
of studies
reviewed
Number
of studies
with one
factor
Number
of studies
with two
factors
Range of
learning rates
for “learning by
doing” (LBD)
2
1
8
4
35
24
2
1
6
4
29
22
0
0
2
0
6
2
5.6% to 12%
2.5% to 7.6%
-11% to 34%
<0 to 6%
-3% to 32%
10% to 53%
4
7
3
3
4
7
0
0
0
0
0
2
Range of rates
for “learning by
researching”
(LBR)
12% to 45%
0% to 24%
Years
covered
across all
studies
Coal*
PC
IGCC
Natural Gas*
Nuclear
Wind (on-shore)
Solar PV
BioPower
Biomass production
Power generation**
Geothermal power
Hydropower
0.48% to 11.4%
10% to 26.8%
10% to 18%
1902-2006
Projections
1980-1998
1975-1993
1980-2010
1959-2001
2.63% to 20.6%
1971-2006
1976-2005
1980-2005
1980-2001
2.38% to 17.7%
*Does not include plants with CCS. **Includes combined heat and power (CHP) and biodigesters.
E.S. Rubin, Carnegie Mellon
Source: Azevedo et al,2013
19. Effect of Learning Curve Uncertainties
for Power Plants w/ CCS
% COE REDUCTION AFTER
100 GW EXPERIENCE
•
Projected reduction in cost of
electricity (COE) for each
plant type varies by a factor
of ~2 to 4
Nonetheless, the judicious
use of learning curves can
suggest a pathway from
FOAK to NOAK costs for
novel technologies
Source: Rubin et al., 2007
Percent Reduction in COE
•
30
25
20
15
10
5
0
NGCC PC IGCC Oxyfuel
E.S. Rubin, Carnegie Mellon
21. Common Applications of
Expert Judgment
•
Estimating the future value of a parameter
used in a model or cost calculation
•
Direct estimates of the future value of a
cost or other characteristic of interest
E.S. Rubin, Carnegie Mellon
22. Example of Expert Elicitations (1)
One of four amine
system parameters
estimated by experts
and used to calculate
expected cost reductions
from advanced solvents
(below)
E.S. Rubin, Carnegie Mellon
23. Example of Expert Elicitations (2)
Elicitations by Nemet et al. of future energy penalty and avoidance cost
for seven capture technologies as a function of three policy scenarios
Scen.3: Minimum cost of CO2 avoided ($/tCO2) across 7 technologies
0.08
Absorption
0.07
Abs=0.20
Ads=0.03
Mem=0.02
0.06
Oth=0.04
Pre=0.30
density
0.05
Oxy=0.24
Precombustion
CLC=0.17
0.04
2025
0.03
Oxyfuel
0.02
0.01
Chemical Looping
0
30
40
50
60
70
80
90
100
110
$/tCO2
Source: Jenni, K.E., Baker, E.D., Nemet, G.F. (2013).
IJGGC, 12, 136-145.
E.S. Rubin, Carnegie Mellon
Source: Nemet, G.F., Baker, E., Jenni, K.E. (2013).
Energy, 56, 218–228.
120
24. How does expert judgment compare to costs
projected using learning rates ?
Nuclear
NGCC w/o CCS
5,400
1,200
5,300
5,200
Original EPRI
EXPERTS
1,100
Base Case
ALT 1 - CES
1,050
1,000
950
ALT 2 - CTAX20
Cost ($/kW)
Cost ($/kW)
1,150
Original EPRI
EXPERTS
5,100
Base Case
5,000
ALT 1 - CES
ALT 2 - CTAX20
4,800
ALT Base Case
4,900
ALT Base Case
Conclusion:
4,700
4,600
• 2050
Expert judgment is2015 2020 2025 2030 2035 2040 2045 2050
often
2015 2020 2025 2030 2035 2040 2045
inconsistent with learning
On Shore Wind curve projections
Solar PV (Rooftop)
2,500
4,000
3,500
3,000
EXPERTS
Original EPRI
1,500
Base Case
ALT 1 - CES
1,000
ALT 2 - CTAX20
ALT Base Case
500
Cost ($/kW)
Cost ($/kW)
2,000
Original EPRI
EXPERTS
2,500
Base Case
2,000
ALT 1 - CES
1,500
ALT 2 - CTAX20
1,000
ALT Base Case
500
0
0
2015 2020 2025 2030 2035 2040 2045 2050
E.S. Rubin, Carnegie Mellon
2015 2020 2025 2030 2035 2040 2045 2050
Source: Azevedo et al,2013
26. Avoid Cost Estimates for Processes
at Early Stages of Development
•
This approach favors the use of performance metrics,
rather that cost estimates, to evaluate and screen novel
components or process designs in the early stages of
development (e.g., TRL levels 1-3 or higher)
•
For novel CO2 capture processes, projected reductions
in the power plant energy penalty is a favored metric
•
No guarantee, however, that improvements in key
performance metrics will necessarily result in lower
overall cost (e.g., COE) for the novel system
E.S. Rubin, Carnegie Mellon
27. So where does this leave us ?
E.S. Rubin, Carnegie Mellon
28. Conclusions and
Recommendations
Conclusions
•
A variety of methods are available to estimate the future
cost of new, pre-commercial capture systems; but …
•
Some methods are not commonly used; others are often
used inappropriately; no clear consistency across method
Recommendation
•
Establish a task force to develop improved guidelines,
emphasizing needs for transparency and incorporation
of uncertainties in cost estimates for novel processes
E.S. Rubin, Carnegie Mellon
29. A Final Word of Wisdom
“It’s tough to make predictions,
especially about the future”
- Yogi Berra
Future
CCS
Costs
E.S. Rubin, Carnegie Mellon