Life Cycle Assessment of Fuel Cell Vehicles
– Dealing with Uncertainties
BY
JOSÉ FERNANDO CONTADINI
B. S. ( Federal University of São Carlos – Brazil) 1987
M. S. ( Federal University of Minas Gerais – Brazil) 1997
DISSERTATION
Submitted in partial satisfaction of the requirements for the degree of
DOCTOR OF PHILOSOPHY
in
ENVIRONMENTAL ENGINEERING
in the
OFFICE OF GRADUATE STUDIES
of the
UNIVERSITY OF CALIFORNIA
Davis
COMMITTEE IN CHARGE:
Prof. Daniel Sperling ( Chair)
Prof. Patricia L. Mokhtarian
Dr. Robert M. Moore
May 2002 Copyright by
José Fernando Contadini
2002
ii
Life Cycle Assessment of Fuel Cell Vehicles – Dealing with Uncertainties
Abstract
Life cycle assessment ( LCA), or “ well to wheels” in transportation terms, involves some subjectivity and uncertainty, especially with new technologies and future scenarios. To analyze lifecycle impacts of future fuel cell vehicles and fuels, I developed the Fuel Upstream Energy and Emission Model ( FUEEM). The FUEEM project pioneered two specific new ways to incorporate and propagate uncertainty within an LCA analysis. First, the model uses probabilistic curves generated by experts as inputs and then employs Monte Carlo simulation techniques to propagate these uncertainties throughout the full chain of fuel production and use. Second, the FUEEM process explicitly involves the interested parties in the entire analysis process, not only in the critical final review phase.
To demonstrate the FUEEM process, an analysis has been made for the use of three different fuel cell vehicle technologies ( direct hydrogen, indirect methanol, and indirect hydrocarbon) in 2010 within the South Coast Air Basin ( SCAB) of California ( Los Angeles). The analysis covered topics such as the requirement of non- renewable energy sources, emissions of CO2 and other greenhouse gases, and emissions of several criteria pollutants generated within SCAB and within other regions. The results obtained from this example show that the hydrogen option has the potential to have the most efficient energy life cycle for the SCAB, followed by the methanol and finally by the Fisher- Tropsch naphtha option. A similar pattern is observed for the greenhouse gas emissions. The results showing criteria pollutants emitted within SCAB highlight the importance of having a flexible model that is responsive to local considerations. This
iii
dissertation demonstrates that explicit recognition and quantitative analysis of the inherent uncertainty in the LCA process generates richer information, explains many of the discrepancies between results of previous studies, and enhances the robustness and credibility of LCA analyses.
iv
Acknowledgments
Several professors and researchers from academia, government agencies, and laboratories helped me a lot on this journey. I want to thank all of them in the figure of Dr. Robert M. Moore, who was always willing to share his knowledge, experience, advice, and especially his friendship with me.
In the same way, dozens of experts and managers from different industries helped me, and trust me by answering my technical questions, providing me with data ( sometimes confidential), and sharing with me their thoughts about the future. I would like to say thanks to all of them in the figures of three people who not only participated on our panel of international experts, but also made personal efforts to seek within their industries specific experts to cooperate with me in more detailed technological investigations. Thanks to Nitin M. Patel from Air Products and Chemical, Inc.; Mark Allard from Methanex Corporation, and J. Steve Welstand from former Chevron Products Company ( now ChevronTexaco).
Finally, but not less important, thanks to Claudia V. Diniz and Yanê D. Contadini for their love, patience, and encouragement.
This study was funded by the CNPq ( Brazilian Council of Science and Technology) and by the Fuel Cell Vehicle Modeling Program at the Institute of Transportation Studies at UCDavis.
v
Table of Contents
1 INTRODUCTION AND PROBLEM CONTEXT................................. 1
1.1 Background and context definition..................................................................... 1
1.2 Problem definition................................................................................................ 4
1.3 Research approach and contributions................................................................ 6
2 LITERATURE REVIEW......................................................................... 10
2.1 Life Cycle Assessment ( LCA) - General overview........................................... 10
2.2 LCA in the fuel/ transportation industry........................................................... 15
2.3 Qualitative analysis of existing fuel/ transportation LCIs............................... 17
2.3.1 Scope............................................................................................................. 18
2.3.2 Boundaries.................................................................................................... 21
2.3.3 Time frame.................................................................................................... 23
2.3.4 Data............................................................................................................... 25
2.4 Methodology of calculus of existing fuel/ transportation LCIs....................... 28
2.4.1 Calculus for the stage ( activity) level........................................................... 29
2.4.2 Co- products allocation.................................................................................. 35
2.4.3 “ Average emissions” versus “ marginal emissions” calculation................... 38
2.4.4 Pathway level calculation............................................................................. 40
2.5 Quantitative analysis of existing fuel/ transportation LCIs............................. 43
3 FUEEM METHODOLOGY.................................................................... 51
3.1 Input data treatment for future technologies................................................... 52
3.1.1 The General Process..................................................................................... 54
3.1.1.1 The Expert Network.................................................................................. 56
3.1.1.2 Data Search............................................................................................... 58
3.1.1.3 Industry Survey......................................................................................... 59
3.1.2 Expert network activity details..................................................................... 62
3.1.2.1 Scenario Construction............................................................................... 63
3.1.2.2 Workshop Discussion............................................................................... 66
3.1.2.3 Group Discussions.................................................................................... 68
3.2 FUEEM uncertainty calculation........................................................................ 76
3.2.1 Monte Carlo simulation technique................................................................ 81
3.2.1.1 Short history.............................................................................................. 82 vi
3.2.1.2 The probability theory basis...................................................................... 83
3.2.1.3 Monte Carlo Sampling.............................................................................. 87
3.2.1.4 Output data analysis.................................................................................. 89
3.2.1.5 Latin Hypercube Sampling....................................................................... 90
3.2.1.6 Dependencies among variables................................................................. 92
3.3 FUEEM Characteristics..................................................................................... 97
3.3.1 Scope, boundaries and time frame................................................................ 97
3.3.2 The software................................................................................................ 101
3.3.3 FUEEM Calculation Example.................................................................... 104
3.3.3.1 Fuel upstream pathway composition...................................................... 105
3.3.3.2 Stage energy requirement....................................................................... 110
3.3.3.3 Stage emissions....................................................................................... 113
4 FUEEM COMPONENT MODELS.............................................................. 116
4.1 Hydrogen marketing activities......................................................................... 116
4.1.1 Energy Requirement and Pipeline design................................................... 118
4.1.2 Pipeline Pressure......................................................................................... 120
4.1.3 Bulk Storage................................................................................................ 121
4.1.4 Flow Rates x Pipeline Diameters................................................................ 123
4.1.5 Pipeline Length........................................................................................... 126
4.1.6 Hydrogen Compression.............................................................................. 128
4.1.7 Turbo- compressor at the bulk storage......................................................... 130
4.1.8 Hydrogen Refueling Station....................................................................... 130
4.1.9 Emissions.................................................................................................... 132
4.2 Hydrogen production........................................................................................ 133
4.2.1 The plant design.......................................................................................... 135
4.2.2 Data search.................................................................................................. 140
4.2.3 Industry survey............................................................................................ 144
4.2.4 FUEEM inputs............................................................................................ 150
4.2.5 Results......................................................................................................... 150
4.3 Liquid Fuels Marketing Activities................................................................... 156
4.3.1 Retail Activities.......................................................................................... 156
4.3.1.1 Vehicle refueling:.................................................................................... 156
4.3.1.2 Storage at the Fuel Station:..................................................................... 166
4.3.1.3 Fuel Distribution..................................................................................... 171
4.3.1.4 Fuel Terminal Activities:........................................................................ 176
4.3.2 Marine activities.......................................................................................... 180
4.3.2.1 In Port Operations................................................................................... 180
4.3.2.2 Fuel Transportation................................................................................. 181
4.4 Gas- to- Liquids production............................................................................... 184
4.4.1 Syngas production....................................................................................... 184
4.4.2 Methanol production................................................................................... 189 vii
4.4.2.1 Methanol synthesis.................................................................................. 190
4.4.2.2 Methanol distillation............................................................................... 192
4.4.2.3 FUEEM assumptions.............................................................................. 194
4.4.3 Fisher- Tropsch Naphtha production........................................................... 200
4.4.3.1 Fisher- Tropsch synthesis........................................................................ 205
4.4.3.2 FUEEM assumptions.............................................................................. 209
4.5 Fuel Characteristics.......................................................................................... 215
4.5.1 Natural Gas................................................................................................. 215
4.5.2 Hydrogen..................................................................................................... 216
4.5.3 Methanol..................................................................................................... 217
4.5.4 Fisher- Tropsch Naphtha.............................................................................. 217
4.5.5 Secondary fuels........................................................................................... 220
4.5.6 Life cycle values for secondary fuels.......................................................... 221
4.6 Equipment characteristics................................................................................ 223
4.6.1 Marine tankers............................................................................................ 223
4.6.2 Trucks......................................................................................................... 225
4.6.3 Stationary diesel engines............................................................................. 228
4.6.4 Residual Oil Boiler..................................................................................... 228
4.6.5 Natural gas engines..................................................................................... 230
4.6.6 Natural gas turbines.................................................................................... 231
4.6.7 Natural gas boilers...................................................................................... 231
4.7 Correlations among variables.......................................................................... 237
4.8 Greenhouse gases assessment........................................................................... 240
5 FUEEM DEMONSTRATION EXAMPLE......................................... 243
5.1 Analysis details.................................................................................................. 243
5.2 Fuel cell vehicles assumptions.......................................................................... 245
5.3 Fuel upstream pathway scenarios................................................................... 252
5.3.1 Gaseous Hydrogen...................................................................................... 253
5.3.1.1 Pathway 1: Centralized production......................................................... 254
5.3.1.2 Pathway 2: Decentralized production..................................................... 255
5.3.1.3 Combined Scenario................................................................................. 256
5.3.2 Liquid Fuels Marketing............................................................................... 257
5.3.3 Methanol..................................................................................................... 258
5.3.3.1 Pathway 1: Typical size plant within uncontrolled situation.................. 259
5.3.3.2 Pathway 2: Mega size plant within uncontrolled situation..................... 259
5.3.3.3 Pathway 3: Typical size plant within controlled situation...................... 260
5.3.3.4 Pathway 4: Mega size plant within controlled situation......................... 260
5.3.3.5 Combined Scenario................................................................................. 260
5.3.4 Fisher- Tropsch Naphtha ( FTN).................................................................. 262 viii
5.3.4.1 Pathway 1: Slurry and pure O2 within an uncontrolled situation............ 263
5.3.4.2 Pathway 2: Tubular and air within an uncontrolled situation................. 264
5.3.4.3 Pathway 3: Slurry and pure O2 within a controlled situation.................. 264
5.3.4.4 Pathway 4: Tubular and air within a controlled situation....................... 264
5.3.4.5 Combined scenario.................................................................................. 265
5.3.5 Natural Gas Feedstock................................................................................ 266
5.3.5.1 For Gas- to- Liquids Production............................................................... 266
5.3.5.2 For Hydrogen Production....................................................................... 268
5.3.6 Electricity.................................................................................................... 269
5.4 Analysis Results................................................................................................. 270
5.4.1 Total energy requirement ( TEreq)................................................................ 271
5.4.1.1 Bounding scenarios................................................................................. 272
5.4.1.2 Fuel upstream analysis............................................................................ 273
5.4.1.3 Secondary fuel calculation...................................................................... 274
5.4.1.4 Non- renewable fuels and petroleum dependency................................... 275
5.4.2 Assessment of global warming................................................................... 276
5.4.3 Criteria Pollutants....................................................................................... 279
5.4.3.1 Global Emissions.................................................................................... 279
5.4.3.2 NOx emissions within SCAB.................................................................. 280
5.4.3.3 NMOG emissions within SCAB............................................................. 281
5.4.3.4 CO emissions within SCAB.................................................................... 284
5.4.3.5 PM10 emissions within SCAB................................................................. 285
5.4.3.6 SOx emissions within SCAB................................................................... 285
5.5 Regression Sensitivity Analysis........................................................................ 286
5.6 Analysis of the use of dependency among variables...................................... 290
6 CONCLUSIONS AND SUGGESTIONS.......................................... 293
6.1 Input Data Treatment Methodology............................................................... 293
6.2 Fuel cell vehicle life cycle assessment.............................................................. 296
6.3 Suggestions for future improvements and studies......................................... 298
7 REFERENCES...................................................................................... 300
ix
List of Tables
Table 2- 1: Equipment share of US natural gas processing (%)........................................ 47
Table 2- 2: Analysis of existing data for methanol production plant................................ 48
Table 2- 3: Hydrogen physical parameters used by existing studies................................. 49
Table 3- 1: Summary of the procedures adopted in the FUEEM group discussion according to the type of information involved.......................................................... 74
Table 3- 2: Examples of probabilistic distribution functions............................................. 86
Table 3- 3: Energy data output format............................................................................. 102
Table 3- 4: Emissions data output format ( per pollutant considered).............................. 102
Table 4- 1: FUEEM input assumptions for the hydrogen marketing activities............... 127
Table 4- 2: FUEEM input assumptions for the hydrogen marketing activities............... 132
Table 4- 3: Emission rates for small stationary NG reciprocating engines..................... 133
Table 4- 4: Existing values for centralized hydrogen plants using SMR and no extra- steam produced.................................................................................................................. 141
Table 4- 5: Existing values for centralized hydrogen plants using SMR and producing extra steam for exportation..................................................................................... 142
Table 4- 6: Existing values for decentralized hydrogen plants ( small units for fuel stations) ............................................................................................................................... . 142
Table 4- 7: Reformer Emission rates assumed by existing models to calculate the emissions of hydrogen production plants................................................................ 144
Table 4- 8: Origin of the lines used for the efficiency calculation.................................. 147
Table 4- 9 : Other major inputs for the energy requirement in hydrogen plants............. 151
Table 4- 10: Fuel requirements of hydrogen plants using typical NG composition ( GJfuel / GJH2- produced - HHV)................................................................................................. 152
Table 4- 11: Calculated emissions for a 27 MTPD hydrogen plant with extra steam exportation and typical natural gas used ( grams / GJ- H2- produced - HHV)................ 153
Table 4- 12: Calculated emissions for a 1 MTPD decentralized hydrogen plant with catalytic burner, SCR and typical natural gas used ( grams / GJ- H2- produced - HHV) 154
Table 4- 13: Emission rate curves for in port activities................................................... 181
x
Table 4- 14: H2/ CO ratio summary for natural gas feed reforming ( source: Tindall et al., 1995)....................................................................................................................... 187
Table 4- 15: Heterogeneous processes for methanol production ( source: Lee, 1990)..... 192
Table 4- 16: Purge gas composition of a typical- size methanol plant with SMR............ 195
Table 4- 17: FUEEM emission rates for a typical- size methanol plant ( HHV)............... 196
Table 4- 18: Air emission control efficiency for a typical- size methanol plant with SMR. ............................................................................................................................... . 196
Table 4- 19: Emission rates for a typical- size methanol plant from current studies........ 197
Table 4- 20: Energy requirement rates for a mega- size methanol plant from current studies...................................................................................................................... 197
Table 4- 21: Emission rates for a mega- size methanol plant from current studies.......... 197
Table 4- 22: Air emission control efficiency for a typical- size methanol plant with SMR. ............................................................................................................................... . 199
Table 4- 23: FUEEM emission rates for a mega- size methanol plant ( HHV)................. 199
Table 4- 24: FUEEM emission rates for a uncontrolled FT plant using slurry reactor, O2 syngas plant and no steam exportation ( HHV)....................................................... 213
Table 4- 25: FUEEM emission rates for a uncontrolled FT plant using tubular reactor, air syngas plant and no steam exportation ( HHV)....................................................... 214
Table 4- 26: FUEEM natural gas composition assumed................................................. 215
Table 4- 27: FUEEM assumptions for hydrogen fuel...................................................... 216
Table 4- 28: Methanol physical parameters used by existing studies.............................. 219
Table 4- 29: Fisher- Tropsch characteristics from the literature....................................... 219
Table 4- 30: FUEEM assumed curves for low- temperature Fisher- Tropsch fuels.......... 219
Table 4- 31: FUEEM assumptions for the life cycle emissions of residual oil............... 222
Table 4- 32: FUEEM assumptions for life cycle emissions of diesel.............................. 222
Table 4- 33: Emissions from the engines of a marine tanker using residual oil.............. 224
Table 4- 34: Fuel economy of diesel trucks assumed by existing studies....................... 226
Table 4- 35: Diesel truck emission rates assumed by existing studies............................ 227
Table 4- 36: Emission rates for stationary diesel engines from the literature................. 234
Table 4- 37: Emission rates for residual oil boilers from the literature........................... 234
Table 4- 38: Emission rates for natural gas engines from the literature.......................... 235
xi
Table 4- 39: Emission rates for large natural gas turbines from the literature................ 235
Table 4- 40: Emission rate for natural gas boilers from the literature............................. 236
Table 4- 41: Deterministic Global Warming Potential values assumed ( based on the International Panel of Climate Change - 100 year horizon)................................... 240
Table 5- 1: Fuel economy of fuel cell passenger cars – data from the literature............. 249
Table 5- 2: Emissions of indirect methanol fuel cell passenger cars – data from the literature.................................................................................................................. 250
Table 5- 3: Emissions of the indirect hydrocarbon fuel cell passenger cars – data from literature.................................................................................................................. 250
xii
List of Figures
Figure 2- 1: Graphical representation of LCA according to ISO 14.040 ( 1997)............... 11
Figure 2- 2: Boundaries concept for a life cycle inventory ( source: Humphreys et al., 1996)......................................................................................................................... 16
Figure 2- 3: General idea of the fuel upstream calculation................................................ 29
Figure 2- 4: Input/ Output idea at the stage level................................................................ 32
Figure 2- 5: Establishment of the equipment activity or load process............................... 33
Figure 2- 6: Graphical representation of the emission calculation at the stage level........ 34
Figure 2- 7: Complexity of the life cycle calculation showing the interdependence among fuels ( LC = life cycle results).................................................................................... 42
Figure 2- 8: Mark ( 1998) comparison of the total upstream emissions for fuel cell vehicles in existing models..................................................................................................... 44
Figure 2- 9: Mark ( 1998) comparison of the hydrocarbon emissions calculated by existing studies, for gaseous hydrogen fuel produced in centralized plants ( SMR process).. 45
Figure 2- 10: Total emissions associated with natural gas recovering and processing from existing models......................................................................................................... 46
Figure 2- 11: Life cycle energy consumption for methanol fuel from existing studies..... 46
Figure 3- 1: FUEEM Participatory Scheme for Future Technology Assessment.............. 58
Figure 3- 2: Graphical representation of the concept that group opinion can produce better information than the individual opinion.................................................................... 63
Figure 3- 3: Expert opinion combination procedure for the FUEEM Delphi method....... 73
Figure 3- 4: Relation between F( x) and f( x)...................................................................... 85
Figure 3- 5: Monte Carlo sampling idea............................................................................ 88
Figure 3- 6: Latin Hypercube sampling idea..................................................................... 92
Figure 3- 7: An example of the Envelope Method application.......................................... 94
Figure 3- 8: Graphical format of two different rank order correlation degrees of normal distributions. ( source: Iman and Davenport, 1982)................................................... 96
Figure 3- 9: An example of the fuel own- use calculation................................................ 108
Figure 4- 1: Schematic of a hydrogen fuel pipeline design............................................. 119
Figure 4- 2: Two possible approaches for the flow rate assumption............................... 127
Figure 4- 3: Simplified scheme of a typical new SMR hydrogen plant........................... 136
xiii
Figure 4- 4: Correlation study between hydrogen plant efficiency and utilized NG energy content ( HHV)......................................................................................................... 146
Figure 4- 5: Rank order correlation study between the extra- steam produced by a hydrogen plant and its total thermal efficiency....................................................... 148
Figure 4- 6: CO and NOx emission rates from hydrogen production plant ( HHV)........ 149
Figure 4- 7: First set of assumptions for the vehicle refueling calculation..................... 158
Figure 4- 8: Second set of assumptions for the vehicle refueling calculation................. 159
Figure 4- 9: Third set of assumptions for the vehicle refueling calculation.................... 161
Figure 4- 10: Fuel Reid Vapor Pressures used for the evaporative emission calculations. ............................................................................................................................... . 164
Figure 4- 11: Evaporative control technologies for liquid fuel marketing ( source: AP- 42, 1995 and CEC, 1996).............................................................................................. 165
Figure 4- 12: Input assumptions for emission control at the fuel station......................... 167
Figure 4- 13: Input assumptions for the emission control of marketing activities.......... 168
Figure 4- 14: More input assumptions for the emission control of marketing activities. 170
Figure 4- 15: Other set of input assumptions for the emission control of marketing activities.................................................................................................................. 172
Figure 4- 16: Input assumptions for the emission control of the fuel terminal and truck. 174
Figure 4- 17: Methanol true vapor pressure ( source: EPA- AP42, 1995)........................ 175
Figure 4- 18: Liquid fuel bulk storage technologies ( source: EPA- AP- 42).................... 177
Figure 4- 19: FUEEM assumptions for the liquid fuel terminal activities...................... 178
Figure 4- 20: FUEEM assumptions for the liquid fuel terminal activities...................... 179
Figure 4- 21: FUEEM assumptions................................................................................. 180
Figure 4- 22: Emission rates assumed for the marine transportation activities............... 182
Figure 4- 23: Syngas production technologies ( source: Lange, 2001)............................ 188
Figure 4- 24: Simplified scheme of a new methanol plant.............................................. 193
Figure 4- 25: Energy requirement assumptions for a typical size methanol plant ( HHV). ............................................................................................................................... . 194
Figure 4- 26: Energy requirement assumptions for a mega- size methanol plant ( HHV). 198
Figure 4- 27: Sasol Fisher- Tropsch reactors ( sources: Steynberg et al., 1999 and Espinoza et al., 1999)............................................................................................................. 202
xiv
Figure 4- 28: Sasol new generation of FT reactors ( source: Jager, 1998)....................... 204
Figure 4- 29: General Fisher- Tropsch synthesis reaction scheme ( source: Dry, 1981)... 206
Figure 4- 30: Generic Fisher- Tropsch yields associated with the catalyst used ( source: Tijm et al, 1995)...................................................................................................... 207
Figure 4- 31: Selective hydrocracking process performance with FT product ( source: Sie et al, 1988).............................................................................................................. 209
Figure 4- 32: FUEEM assumptions for a FT plant with a slurry reactor and no SCR..... 211
Figure 4- 33: FUEEM assumptions for a FT plant with a tubular reactor and no SCR... 211
Figure 4- 34: FT production cuts for the slurry plant configuration ( a and b) and for the tubular plant configuration ( c and d)....................................................................... 212
Figure 4- 35: FUEEM assumptions for the methanol characteristics.............................. 218
Figure 4- 36: Emission rates for the marine tanker engines using bunker fuel............... 225
Figure 4- 37: FUEEM assumptions for diesel trucks ( class 8b)...................................... 226
Figure 4- 38: FUEEM emission rates for uncontrolled stationary diesel engines ( HHV). ............................................................................................................................... . 229
Figure 4- 39: FUEEM emission rates for uncontrolled residual oil boilers ( HHV)........ 230
Figure 4- 40: FUEEM emission rates for uncontrolled natural gas engines ( HHV)....... 231
Figure 4- 41: FUEEM emission rates for uncontrolled natural gas turbines ( HHV)....... 232
Figure 4- 42: FUEEM emission rates for uncontrolled natural gas boilers ( HHV)......... 233
Figure 4- 43: Delucchi's Economic Damage Index search values ( source: Delucchi, 1997). ............................................................................................................................... . 241
Figure 4- 44: FUEEM probabilistic curves assumed for EDI factors ( grams of CO2- equivalent / grams of pollutant)................................................................................................ 242
Figure 5- 1: Vehicle fuel efficiency assumed.................................................................. 251
Figure 5- 2: Emissions assumed for the indirect methanol fuel cell vehicle ( IMFCV)... 251
Figure 5- 3: Emissions assumed for the indirect hydrocarbon fuel cell vehicle ( IHFCV) ............................................................................................................................... . 252
Figure 5- 4: Boundaries definition for the hydrogen pathway 1...................................... 255
Figure 5- 5: Boundaries definition for the hydrogen pathway 2...................................... 255
Figure 5- 6: Share of the pathway 2 in the combination scenario................................... 256
Figure 5- 7: Liquid fuels pathway boundaries................................................................. 258
xv
Figure 5- 8: Methanol pathways representation............................................................... 259
Figure 5- 9: Possibilities of occurrences assumed for the methanol pathways................ 261
Figure 5- 10: Fisher Tropsch Naphtha pathways representation..................................... 263
Figure 5- 11: Possibilities of occurrences assumed for the Fisher Tropsch naphtha pathways.................................................................................................................. 266
Figure 5- 12: Life cycle result of the total energy requirement....................................... 272
Figure 5- 13: Example of " bounding scenarios" analysis for the total energy requirement of DHFCV and IMFCV cycles............................................................................... 273
Figure 5- 14: Example of fuel upstream analysis for the energy requirement................. 274
Figure 5- 15: Example of the secondary fuel calculation share....................................... 275
Figure 5- 16: Example of non- renewable fuel and petroleum dependency..................... 276
Figure 5- 17: Global warming potential versus economic damage index - comparison for the DHFCV case..................................................................................................... 277
Figure 5- 18: Greenhouse gases emissions ( means)........................................................ 277
Figure 5- 19: Assessment of global warming using EDI factors..................................... 278
Figure 5- 20: Global emission share of criteria pollutants............................................... 279
Figure 5- 21: Life cycle result of the NOx emissions within SCAB................................ 280
Figure 5- 22: Life cycle result of NMOG emissions within SCAB................................. 282
Figure 5- 23: Details of the NMOG emissions within SCAB ( IMFCV example)........... 283
Figure 5- 24: Sensitivity analysis over the " zero- evap" vehicle assumption ( 90 % confidence interval)...................................................................................... 284
Figure 5- 25: Life cycle result of CO emissions within SCAB....................................... 285
Figure 5- 26: Life cycle result of the PM10 emissions within SCAB............................... 286
Figure 5- 27: Life cycle result of the SOx emissions within SCAB................................ 287
Figure 5- 28: Regression sensitivity for the DHFCV - Total energy requirement.......... 288
Figure 5- 29: Regression sensitivity for the IMFCV - Total energy requirement........... 289
Figure 5- 30: Regression sensitivity for the IHFCV - Total energy requirement............ 289
Figure 5- 31: Regression sensitivity for the H2 pathway 1 - NOx emissions within SCAB........................................................................................................................... ..... 290
Figure 5- 32: Regression sensitivity for the MeOH pathway 1 - NMOG emissions within SCAB...................................................................................................................... 290
xvi
Figure 5- 33: Comparison example of the models assuming independent and dependent variables – total NOx emissions for the hydrogen pathway 1................................ 291
Figure 5- 34: Comparison example of the models assuming independent and dependent variables – NOx emissions within SCAB for the hydrogen pathway 1.................. 292
Figure 5- 35: Comparison example of the models assuming independent and dependent variables – process NOx emissions for the methanol pathway 3............................ 292
xvii
Abbreviations and Acronyms
atm
Atmosphere
bbl
Barrels
Btu
British thermal unit
CARB
California Air Resource Board
CEMS
Continuous Emission Monitoring System
CH4
Methane
CO
Carbon monoxide
CO2
Carbon dioxide
DH
Direct hydrogen
DHFCV
Direct hydrogen fuel cell vehicle
EPA
US Environmental Protection Agency
FT
Fisher- Tropsch
FTN
Fisher- Tropsch Naphtha
FUEEM
Fuel Upstream Energy and Emission Model
gal
Gallon
GJ
Giga Joule
HHV
High heating value
IH
Indirect hydrocarbon
IHFCV
Indirect hydrocarbon fuel cell vehicle
IM
Indirect methanol
IMFCV
Indirect methanol fuel cell vehicle
km
Kilometer
L
Litter
lb
Pounds
LHV
Low heating value
MBtu
Mega or million Btu
MeOH
Methanol
MJ
Mega Joule
mg
milligram x viii
MPa
Mega Pascal
MTPD
Metric tons per day
NG
Natural gas
NOx
Nitrogen Oxides
N2O
Nitrous Oxide
PM10
Particulate matter smaller than 10 microns
PSA
Pressurized Swing Adsorption
psi
Pounds per square inch
SCAB
South California Air Basin ( Los Angeles area)
scf
Standard Cubic Feet
SCR
Selective Catalytic Reduction unit
SMR
Steam methane reforming
SNCR
Selective Non- Catalytic Reduction unit
SOx
Sulfur oxides
xix
1
1 INTRODUCTION AND PROBLEM CONTEXT
1.1 Background and context definition
Transportation is an important contributor to world energy consumption ( Greene, 1996). According to the US Department of Energy ( TEDB, 2000), in 1997, the United States, the most automobilized country in the world, consumed 18.6 million barrels of oil per day, equivalent to 25.5 % of the world oil consumption. From that total, 67 % was used directly in transportation. In 1999, according to the same report, the transportation sector was 97.4 % dependent on petroleum energy. Similar values were presented in older studies. The USA was responsible for 30 % of the world energy use in the early 90’ s ( Ackerson et al., 1993). According to Gordon ( 1991), 41 % of that energy was spent directly or indirectly on transportation, and 97 % of the 22.66 quads directly used by the USA transportation sector were produced from petroleum.
Transportation- related air emissions can be also associated with greenhouse gas ( carbon dioxide- CO2, methane- CH4, nitrous oxide- N2O, carbon monoxide- CO, chlorofluorocarbons - CFCs, etc.). The concentration of these gases in the stratosphere may cause a global warming, and climatologists are expecting a global climate change to be associated with a lot of environmental impacts ( Beckmann et al., 1991; Walsh, 1993 and IPCC, 2000). CO2 produced in the combustion of fossil fuels, such as petroleum, is the major contributor to the global warming. In 1998, the US emitted 6,514 million metric tons of CO2- equivalent per year and from that 32.6 % was attributed to transportation ( DOE, 1999). In the early 90’ s the USA transportation sector was responsible for 30 % of the total 5.600 million tons of CO2 emitted per year ( EPA, 1992).
2
Transportation- related air emissions can be associated with urban air quality in terms of ozone formation, criteria pollutants ( non- methane organic gases- NMOG, carbon monoxide- CO, nitrogen oxides- NOx, sulfur oxides- SOx and particulate matter- PM), and toxic pollutants ( benzene, lead, etc.). Several health problems are associated with human exposure to these pollutants. In 1998, according to Davis ( 2000), 63.8 million tons of CO were emitted within the US by all transportation modes. From this total, 71.7 % is attributed to vehicles. A similar situation occurs with NOx when, in 1998, the transportation sector emitted 11.8 millions tons of the pollutant with 59.5 % being accounted for by vehicles. For volatile organic compounds ( VOC), 68.4 % of the total of 7.1 million tons emitted is attributed to vehicles. In the early 90’ s the transportation sector was responsible for 78 % of the USA’s emissions of CO, 30 % of the NMOG, 5 % of the SOx and 23 % of the PM10 ( EPA, 1995).
These concerns are present in all developed countries but also in several developing ones. Within the same development and technologies pattern the situation tends to get worse especially with the increasing vehicle mileages traveled ( VMT) in developed countries and with the rapid motorization occurring in developing countries ( UNDP, 2000).
Solving these problems is the principal motivation for introducing new vehicle technologies and alternative fuels. However, since transportation is a very complex system, a change in practice to alleviate one problem could well exacerbate others. A comprehensive evaluation of the environmental aspects ( air criteria pollutants, greenhouse gases emissions, non- renewable energy consumption, etc.) and the trade- off on the environmental impacts ( human health, biodiversity, sustainability of the future 3
generation, etc.) should include all life cycle activities from the vehicle operation to the feedstock ( oil, natural gas, coal, etc.) extraction.
Urban air quality improvement, climate change concerns, and a reluctance to depend on non- renewable sources have been as well the main motivations for the development of fuel cell technologies and their applications in fuel cells vehicles ( FCVs). Fuel cells are electrochemical devices that directly generate electricity using a fuel ( hydrogen, in general) as required and an oxidant ( oxygen) to complete the process. It emits only water vapor as a by- product and it is also more efficient than internal combustion engines ( ICE) due to the possibility of controlling the electrochemical reaction.
The rapid development of these new vehicle technologies may also require the establishment of a new fuel infrastructure soon. Hydrogen can be used directly as the fuel cell fuel, as can other alternative fuels, such as methanol, or, alternatively, some special kinds of hydrocarbon fuels can be used indirectly as hydrogen carriers. Again, a technology change of this magnitude may require a good understanding of the major risks of environmental impacts in the entire cycle of activities. This understanding may be necessary in order to prevent “ second order” problems and/ or to help in the selection of the best social strategy to establish policy, allocate subsidies, and drive R& D programs.
The Life Cycle Assessment ( LCA) methodology has the potential to be an important management tool in assisting decision- makers to achieve a holistic understanding of the entire system associated with a single product/ service to be introduced. In spite of being a scientific management tool in development, LCA has been used more and more frequently, even presenting some necessity for improvements as 4
discussed on the next sections. The amended ZEV rule ( Zero Emission Vehicles), approved by the California Air Resource Board ( CARB) in November 1998, highlighted the importance of vehicle life cycle analysis comparisons when it established partial credits for vehicles with low tailpipe emissions that use a cleaner fuel process than gasoline. Which alternative is more environmentally positive and by how much? This is the kind of question that LCA tries to answer.
1.2 Problem definition
Basically the methodology of Life Cycle Assessment has an inventory phase where the environmental aspects should be measured and a second phase where all the environmental impacts related with the aspects inventoried are assessed for a final comparison. As detailed section 2.1, the history of the methodology development has been marked by result manipulation attempts in order to push organization agendas or product benefits. Because of that, lack of credibility is a problem that LCA must reverse and the methodology improvements should prevent or minimize.
A critical element in the methodology is the subjectivity of the assessment phase in order to prioritize the importance of different environmental impacts. On the other hand, the inventory phase ( sometimes called LCI – Life Cycle Inventory), that deals with the system input data compilations and also with the calculations of the system environmental aspects outputs, is in some sense considered a more mature methodology with more than 20 years of development. Apparently, it has all the ingredients to be an objective tool; however, analysis done over LCA result discrepancies have shown that the inventory phase is still presenting serious problems too and improvements should be
5
interesting. The major problems for the inventory of the environmental aspects of an existing product are the lack of data, quality of existing data, lack of a single methodology to fill up the gaps, and also the decisions to deal with boundaries and co- products.
More complicated than that is a common characteristic in the fuel cell vehicle kind of situation where the “ cleaner technology” will always occur in the future and, therefore, there will always be some subjectivity in the analysis, even in the inventory phase.
Transportation life cycle studies suffer from similar problems pointed out by studies done in other sectors according to my initial study done in 1998, when a comparison of the existing " cradle- to- grave" or " well- to- wheels" studies related to fuels for transportation and vehicle technologies was done. In general, these kinds of studies focus on the inventory of air emissions ( grams) and energy requirements ( Joules or BTUs) over the entire range of fuel upstream activities ( life cycle) associated with the vehicle operation ( per km or mile). Some of the studies also do an assessment analysis for the climate change effected by the greenhouse gas emissions by using global warming potential factors ( GWP). As a general statement, it can be said that the existing studies do not agree in their results and, depending on the case, they disagree to the extent of several orders of magnitude. More details of this comparison are presented in section 2.5.
Basically, I identified three levels of disagreement:
Geographical differences ( US national average, South Coast California Air Basin, Canada, UK, etc.). Geographical differences are related to the initial study objective and, in general, are clearly delineated in the reports. Problems arise only if attempts are made
6
to generalize the result. Such an attempt is very common in conference presentations, study comparisons, and study press releases.
Technology scenario composition ( for example, natural gas pipelines propelled by turbines, reciprocating engines or electric motors, pressure of the gas pipe, electricity production mix per region). Within the same area and under the same technology umbrella ( for example, natural gas feedstock), the assumptions can be very different and generate different results. The use of a single situation to represent all the feasible and viable technologies possible in the real world is very common. There are few studies that perform sensitivity analysis at this level.
Technology data ( efficiencies and emission factors of different equipment). A lack of data for some equipment, as well as the use of deterministic values to represent a complex system ( the average of the USA methanol production plant efficiencies, for example), generate part of the disagreement in results. A robust study should be very clear about the technology considered and kind of data used. Several studies do only a kind of bookkeeping process, with generic assumptions about generic technologies and do not go to the level of calculation involving equipment design, level of equipment activity, and physical parameters. Even for the studies that do go to this level of detail in calculation, a lack of reported information about the details and assumptions used is unfortunately frequent.
1.3 Research approach and contributions
To deal with these uncertainties in the fuel cell vehicle life cycle assessment, I decided to develop a new model called FUEEM ( Fuel Upstream Energy and Emissions
7
Model). The FUEEM operational unit is kilometer driven and the time frame is 2010, due to the development characteristics of the fuel cell vehicles and fuel development level. The boundaries are from the natural gas extraction to the vehicle operation, since the initial comparison is among three special fuels that use natural gas as feedstock ( Hydrogen, Methanol and Fisher- Tropsch Naphtha). The model uses the global warming potential ( GWP) and Economic Damage Index ( EDI) to calculate greenhouse gas emissions ( CO2, CH4 and N2O) in terms of CO2- equivalent and it also calculates the total energy required disaggregated in terms of petroleum and fossil fuel use. For five of the criteria pollutants ( NOx, CO, NMOG, PM10 and SOx) which are considered in the study, the effort was to quantify how much is released in urban areas.
The model used two major approaches to explicitly recognize and quantitatively include the inherent uncertainties in LCAs:
1. For the technology data problems in the inventory, FUEEM works with specified equipment and system design performing a quantitative uncertainty analysis. This approach is suggested in the ISO 14041 ( 1998). To my knowledge, this project was the first to put it into practice. To use the approach, FUEEM establishes probabilistic curves as inputs and propagates the uncertainties over the calculation by using Latin- Hypercube sampling, Monte Carlo simulation, and rank order correlations. This approach is similar to performing thousands of sensitivity analyses at once, with the advantage of establishing the importance of each scenario ( expressed in the occurrence probabilities) at the end. More details about it are presented in section 3.2.
8
2. The other uncertainties are related to subjective and necessary decisions, such as the future technology compositions ( scenarios), the modeling approach that affects the results ( allocation of co- product credits, for example), the filling process for missing data, etc. I made all these major decisions with the participation of the interested parties. 1 This participation occurred during the entire process and not only in the critical review process. This procedure takes item 7.3.3 of the international standard for Life Cycle Assessment ( ISO 14040, 1997) a step further and is designed to enhance the credibility of the study results. This step is not a simple one since, in general, what differs among the parties are their different, and in most cases, conflicting interests. The methodology adopted in FUEEM to take maximum advantage of this participation and the explanation of the rationale behind the decisions made are presented in section 3.1.
Finally, since the inventory results are geographically specific it brings into question the advantage of having a flexible model to perform the analysis for different areas and situations. FUEEM performs most of it calculations at the level of detail where some physical parameters and scenarios ( distances, temperatures, gas composition, level of control enforcement, etc.) can be manipulated to better represent the local situations. To demonstrate the FUEEM process, I conducted an analysis for three Fuel Cell Vehicle Technologies concepts hypothetically running in the South Coast Air Basin ( SCAB) of California in 2010. The analyzed vehicle concepts were Direct Hydrogen Fuel Cell Vehicle, Indirect Methanol Fuel Cell Vehicle, and Indirect Hydrocarbon Fuel Cell
1 The definition of interested parties according to the ISO 14.040 ( 1997), is an “ individual or group
concerned with or affected by the environmental performance of a product system, or by the results of the
life cycle assessment”.
9
Vehicle. The analysis investigates the operational upstream activities of three zero- sulfur fuels ( hydrogen, methanol and Fisher- Tropsch naphtha) produced from the natural gas. Several fuel pathways and scenarios were explored. The experts and I chose SCAB because of its well- known air quality problems and its high probability of leading fuel cell vehicle introduction. The details of the analysis and the results are presented in section 5.
10
2 LITERATURE REVIEW
2.1 Life Cycle Assessment ( LCA) - General overview
Wouldn’t it be great if well- intentioned decision- makers had right in front of them a classification of the most environmentally friendly policy, process, product or technology? In fact, this is the dream of the scientific systemic management and the right way to go according to my view. But, how far are we from this dream?
The most important tool that has been developed for this purpose is called Life Cycle Assessment ( LCA). Therefore, Life Cycle Assessment is an environmental management tool that generates information about the environmental consequences of the existence of a product or service through all of its life activities. It is in general called “ from cradle to grave analysis” or in the transportation sector “ from well to wheel analysis.” The definition of the international standard is: “ Compilation and evaluation of the inputs, outputs and the potential environmental impacts of a product system throughout its life cycle” ( ISO 14.040, 1997). The international standard also presents the general methodology to conduct a LCA, which and it can be found in several other studies as well ( SETAC, 1993; Vigon et al., 1993; Graedel, 1998).
Basically, the methodology has three phases with a general interpretation step for each phase: First the definition of the project goal, time frame considered, the functional unit, scope, and, most important, the activities boundaries, assumptions, allocations procedures, etc. The second phase is the life cycle inventory analysis where the data is collected and analyzed, and the calculations of the energy and material flows occur. The idea is to quantify all inputs and outputs of the product system focusing on the released waste for the environment ( air, water and soil). Finally, the last phase is called life cycle 11
impact assessment where, based on the inventory results, the significance of the potential environmental impact is evaluated. The evaluation may focus on resource depletion, on human health impacts, on ecological impacts such as biological diversity and habitat alteration, and on economic impacts such as damage to infrastructures, land requirements ( food production), aesthetic values, etc. A graphical representation of these ideas is presented in Figure 2- 1. Depending on the author, the improvement suggestion and analysis are separated from the impact assessment into a new phase called improvement assessment ( Ayres, 1995).
1996: Standardization of Life- Cycle Assessment( LCA) ISO ( International Standard Organization) – ISO series 14.0001996: 14.000LCI• DefinitionsAspects InventoryImpact AssessmentInterpretation• Air• Water• Soil• Human Health• Resource Depletion• Biodiversity, etc.
LCI• LCILCI• • Scope, • Boundaries• Data, etc. AssessmentInterpretationDefinitionsDefinitionsAspects InventoryAspects AssessmentImpact AssessmentInterpretationInterpretation• Figure 2- 1: Graphical representation of LCA according to ISO 14.040 ( 1997)
Life Cycle Assessment ( LCA) has been designed and used in different arenas. Companies have been using it internally for product development and improvement of the environmental characteristics of their system, and as a baseline for environmental audits. The European Commission has been motivating industries to perform internal
12
LCAs, and according to Ecobilan ( 1996) European car companies have conducted several studies in the last decade. Most of them focus on the material use. To some extent the companies have been using LCA for strategic planning and marketing to make comparisons with concurrent products ( Lee et al., 1995). The idea of environmental labels for a product is based on this concept of product comparison. In the public policy making arena LCA could provide a framework for environmental taxes and incentives/ subsidies for technological development ( Lee et al., 1995). The amended ZEV rule ( Zero Emission Vehicles), approved by the California Air Resource Board ( CARB) in November 1998, highlighted the former idea when it established partial credits for vehicles with low tailpipe emissions that use a cleaner fuel process than gasoline. A life cycle study was used to support the amendment ( Acurex, 1996).
The Life Cycle Analysis concept is attributed to Harry Teasley from the Coca- Cola Company who, in 1969, sponsored a comparison of different beverage containers. The analysis was conduced by MRI ( Midwest Research Institute) and the concept became known as REPA ( Resource and Environmental Profile Analysis). It was the basis of the Life Cycle Inventory methodology development within the existing LCA idea ( Hunt et al., 1992). Several REPAs were conduced in the U. S. A. in the 70’ s and 80’ s initially focusing on the energy issue and later shifting to hazardous waste. A similar development pattern occurred in Europe inspired by the REPA studies. Christiansen ( 1993) comments on the 1984 Swiss model called BUS and the 1985 German qualitative model called PLA. Lee et al. ( 1995) complete the list with the Boustead model developed in the early 70’ s, and with the Sundström model in the mid 80’ s. Pedersen and Christiansen ( 1992) discovered that by that time 90 Life Cycle Assessments had been performed and
13
published and that 50 % of them were done on packaging materials and 10 % on energy production and building materials. Derenne ( 1995) three years later reported 274 studies, with 36.9 % on packaging, 8.8 % on energy, and 4 % on transportation.
With more than 20 years of development the quantitative inventory phase ( sometimes called Life Cycle Inventory - LCI) methodology is claimed by various authors ( Hunt et al., 1992; Boustead, 1992 and implicitly the international standard ISO 14.040) to be well established. On the other hand, existing problems in this phase are always unanimously attributed to the lack of comprehensive data and data quality. This study does not share the vision of the previous authors. The hypothesis here is that uncertainties in the data will always occur and therefore the LCI methodology should incorporate them in the calculation and data treatment. This point is discussed later. If we move to the impact assessment phase, the LCA problems become much worse and we can say that a long time will be necessary to mature some acceptable methodology for the assessment final result – to provide an environmental ranking of the compared products, services or policies.
It is important to point out that there is no such thing as a single environmental problem. Several problems caused by several causes with strong interdependency among them are the common figure. A change in practices to alleviate one problem could well exacerbate others. If on one hand this is the situation that generates the necessity for the LCA development it also requires that the impact assessment compare the losses and gains in each area and prioritize them. Monetary valuation of the impacts using the contingency valuation approach ( willingness to pay or willingness to accept payment surveys) appears to be one step ahead of other approaches such as single or
14
multidimensional non- monetary measures ( net- energy, material intensity per unit of service, etc.) or from other attempts using multi- objective decision- theoretic approaches ( Ayres, 1995). Depending on the pollutant/ impact in question, other complex calculations should be necessary such as external chemistry reactions, level of expositions, etc. These calculations are, in general, performed under the label environmental risk assessment. Each of these points is an entire study area and for logistic reasons the focus of this study covers none of them except the Life Cycle Inventory ( LCI) phase and its previous and necessary definitions. Global warming impacts, for potential warming or economical damage, expressed in terms of CO2- equivalent, are the only assessment performed in this study so far.
With all these uncertainties and potential economic interests on LCA results, it is easier to find comments in the literature about lack of credibility. Currently life cycle assessment ( LCA) methodology involves many decisions, choices and exclusions that may intentionally or unintentionally influence the outcome of the study. A classical example is presented by Christiansen ( 1991; in UETP, 1996) where five studies comparing milk containers generate five different answers with the characteristic that the results always favor the product of the company sponsoring the study. The explanation for the differences is related to different qualities of data, different boundaries of the life cycle, different types of technologies and different priorities in the evaluation stage. Ekvall ( 1992) also presents a comparison of two LCAs of similar cardboard. In this case the two studies use the same data profile but the results differ 30 % in the thermal energy requirement, 60 % in the electrical energy requirement, 30 % to 100 % on air emissions and 80 % on solid waste. Several topics were pointed as the main differences, among 15
them the content of the recycled fibers, share of waste going for incineration, energy recovered in the incineration process, mix of electricity generation, and the “ avoided emission” approach assumed.
2.2 LCA in the fuel/ transportation industry
A complete Life Cycle Assessment ( LCA) in the fuel/ transportation industry should be performed following these basic steps: For each stage in the life cycle ( vehicle operation, fuel distribution, fuel production, feedstock transportation and storage, and feedstock extraction and processing) the idea is to quantify the water, soil and air emissions for different phases of the project. These phases are Pre- operations ( R& D, Site Development and Construction), Operations and Post- operations ( Recycling, Decommissioning and Dismantling). Figure 2- 2 presents a graphical representation of these boundaries. The impact on the environment should be assessed and somehow compared after the inventory analysis. Photo- oxidant formation, acidification, eutrophication, global warming, stratospheric ozone depletion, ecotoxicological impacts, bio- diversity reduction, and habitat alterations are examples of environmental impacts.
For reliable results it is necessary to obtain data from different processes, which necessitates development of an ongoing data library and, as discussed before, the subjectivity involved in the evaluation phase of the LCA method is still critical. Christiansen ( 1993) reports the existence of several LCA done internally by the companies and never published but to the extent of my current knowledge only one “ complete” LCA study has been published in the transportation sector so far. Spirinckx and Ceuterick ( 1996) include a comparative impact assessment of air, water, and soil
16
emissions, and it is a comparative life- cycle assessment of fossil diesel and biodiesel. Unfortunately they did not publish their input assumptions for the inventory. For the evaluation, they used weighting factors from a Dutch report on eco- indicators ( Goedkoop, 1995). Their conclusion is that the environmental index of biodiesel is a factor of 2 higher than the one for diesel with the following statement “ However, weighting factors to a large extent have a subjective nature.”
Life Cycle StagesPrimary Resource Extract. & PreparationEnd- Use ServiceProduct Transpor- tation Storage & DistributionConversion& ProcessingTransport& StorageLife PreparationPrimary ServiceEnd- DistributionProduct ProcessingConversion& StorageTransport& Storage
Pre- operation: R& D, Site Development & ConstructionOperationPost- operation: Decommissioning & Dismantling Life Cycle PhasePre- Phase
Figure 2- 2: Boundaries concept for a life cycle inventory ( source: Humphreys et al., 1996)
All the other life- cycle studies in the fuel/ transportation sector perform the inventory phase of the methodology only. Some of them perform an assessment of the global warming potential in terms of amount of CO2- equivalent. A well- done life- cycle inventory ( LCI) is already an important management tool providing interesting outcomes. The inventory results can be associated with costs to perform a cost effectiveness analysis, or, in a more simple way, by assuming that “ less is better” for the energy requirement analysis and for the pollutant emissions analysis. More important, this kind of comparison for local situations can define where tradeoffs in the system may occur,
17
providing information where attention should be concentrated. It is essential to point out that the LCA, and especially the LCI, was created as a technical tool and, in spite of the necessity to consider some economic and social factors to discuss the technology used in the calculation, it does not automatically take these factors into account ( Derenne, 1995).
Examples of existing studies are: Unnash et al. ( 1996 and 2000), Delucchi ( 1991, 1993 and 1997), Greet ( 1998, 1999 and 2000), ETSU ( 1996, 1997 and 1998), GM ( 2001), MIT ( 2000), Pembina- Suzuki ( 2000), Methanex ( 2000), Adamson and Pearson ( 2000), Leveton ( 1999), Armstrong and Akhurst ( 1999), ANL ( 1998), Ogden et al. ( 1998 and 1995), DTI ( 1998), Ekdunge and Raberg ( 1998), Specht et al. ( 1998), ADL ( 1996), Berry ( 1996); Borroni- Bird ( 1996), Darrow ( 1994), Mark et al. ( 1994), Shelef and Kukkonen ( 1994), and Chang et al. ( 1991)
2.3 Qualitative analysis of existing fuel/ transportation LCIs.
A qualitative analysis is performed here to highlight some of the difficulties in conducting a quantitative comparison among the results of selected existing studies. Andress ( 1998) did a qualitative comparison between Greet and Delucchi’s Model for the ethanol fuel cycle, in which some general similarities and differences are addressed. No quantitative comparison was done in Andress’ study and the results can be summarized in terms of how they calculate greenhouse gas emissions and make parametric assumptions ( determined inside or outside of the models). A quantitative analysis however is possible at a more detailed level, some of which are discussed later.
18
2.3.1 Scope
Based on what was presented above, the principal motivation to evaluate the existing fuel use and eventual new alternative fuel use is to assess the potential to consume less petroleum and non- renewable fuels, so as to reduce air pollution and greenhouse gas emissions. According to Kordesch et al. ( 1995) spills, leaks, strip mining and other environmental aspects are also important points to consider; however, most of the existing LCI in the transportation sector focus on the energy requirement and air emissions ( criteria pollutants and greenhouse gases) only. This simplification was adopted as a strategy to reduce the cost and necessary effort in the projects as well.
An exception to that can be attributed to the NREL studies ( NREL 1991 and 1992) on bioethanol and reformulated gasoline, Mann and Spath ( 1997) on biomass gasification plants, and also to a similar study done by Spath and Mann ( 2000) on a hydrogen steam methane reforming plant. They included in their analyses the solid waste generation and the water emissions. The total amount of water pollutant was found to be small compared to other emissions ( 0.2 g / kg of H2 produced) and the waste generated is reported in an aggregated form ( 205.6 g / kg of H2 produced) attributed mainly to electricity consumption grid with coal generation. A similar conclusion was reached by the previous studies. The studies assess the criteria pollutants ( NO2, NMOG, SO2, CO and PM10), the toxic pollutant benzene ( C6H6), and the greenhouse gases ( CO2, N2O, CH4 and CO2- equiv.). All the emissions are calculated with a U. S. A. global perspective, i. e., without separating them into urban area emissions. The energy requirement is presented in terms of the total and feedstock content.
Other examples of the scope of some of the most robust and updated studies:
19
1. Delucchi ( 1991, 1993 and 1997): Calculated in a spreadsheet ( Lotus123), this study focuses on standard greenhouse gas emissions ( CO2- equiv., CO2, CH4, and N2O) and also includes some criteria pollutants ( CO, NO2, and NMOG). The criteria pollutants, including SOx and PM10, are calculated with a global perspective. The total energy is presented as well as at the activities’ phases ( feedstock recovery, feedstock production, fuel production and fuel distribution). The model includes the following U. S. A. pathways: reformulated gasoline, standard gasoline, and diesel from crude oil; LPG from crude oil and natural gas ( NG); compressed NG and Liquefied NG; methanol from NG, coal, and wood; ethanol from wood and corn; hydrogen from solar, hydrogen from nuclear, and several electricity generation technologies.
2. Greet ( 1998, 1999, 2000 and 2001): Calculated in a spreadsheet ( Microsoft Excel), this model independently focuses on standard greenhouse gas ( CO2- equiv., CO2, CH4, N2O) and criteria pollutant emissions ( NMOG, CO, NOx, PM10 and SOx). It creates a “ virtual” urban area for roughly local criteria pollutant analysis. The energy is presented in terms of petroleum consumption, fossil fuel consumption, and total. It has 26 fuel USA pathway calculations and 49 vehicle technologies. The result is a comparison of 77 fuel/ vehicle combinations.
3. Unnash et al. ( 1996 and 2000): Done in a relational data base environment ( Microsoft Access), this model focuses on the photochemical reactivity of NMOG for California's South Coast Air Basin but also assesses other emissions such as NOx, CO, CO2, and CH4 and their regional occurrence ( California, USA, and rest of the world). The initial study investigated the following fuels: gasoline and
20
reformulated gasoline, diesel, LPG from crude oil, methanol from NG and biomass, ethanol from corn, compressed and liquefied NG, hydrogen and electricity ( aggregated mix). The latest report evaluates diesel, reformulated diesel, and LPG from crude oil; synthetic diesel, methanol, and LPG from NG; methanol from landfill gas and biomass, and electricity from crude oil, NG, coal, biomass and hydroelectric.
4. ETSU ( 1996, 1997 and 1998): Calculated in a spreadsheet, the model focuses on the criteria pollutants ( NOx, NMOG, CO, SOx, and PM10) and on CO2 and CH4. All the pollutants are calculated in a global perspective for the UK cases. The initial study is done for the following fuels: gasoline, diesel, and LPG from crude oil; compressed NG, electricity, biomethanol, bioethanol, and biodiesel, and includes the generic passenger car, light- duty and heavy- duty vehicles, and buses. The following studies incorporate in the calculations some new and more detailed vehicle technologies: gasoline vehicle, diesel passenger car, methanol fuel cell vehicle, and NG fuel cell vehicle.
The other studies referenced before have a much more limited scope, or different goals than the ones selected here, for example, a cost analysis goal. Some of them used the data generated in one of the above selected robust studies; others were out- of- date. Whenever possible these studies were used for a more detail analysis or in data acquisition.
21
2.3.2 Boundaries
According to the definition of the international standard ( ISO 10.040, 1997) the system boundary is “ the interface between a product system and the environment or other product system.” Complex systems like industrial and fuel production systems have practically no final limit. One can trace back materials and energy indefinitely depending on the level of detail used. Therefore, every assessment must limit its analysis at some point. Different studies having different system boundaries may have different results and this detail must be taken into account when comparing them. In fact, several LCA result manipulations used this flexibility in the past. Lee et al. ( 1995) present the example of washing machine studies including or not the services ( heating, lighting, compressed air, etc.) of the manufacturing plant and having different conclusions. Ayres ( 1995) comments on the classical McDonalds’s study comparing groundwood ( papier- mache) and polystyrene hamburger shells.
The main sequence of operations in the product production and consumption is usually the easiest to identify. In the fuel/ transportation case, for example, the sequence should be the feedstock recovery ( crude oil, coal, NG, etc.), feedstock processing, feedstock transportation and storage, fuel production ( gasoline, methanol, etc.), fuel transportation and storage, fuel distribution, and vehicle operation. The idea is that the boundaries include all important activities that may change the final results. However, this definition is not so direct and in most cases a previous study must have been completed to make sure it was accurate ( ISO 14041, 1998). The solution presented by the ISO 14040 ( 1997) is that the system boundaries shall be identified and justified, but only these do not prevent situations found in Blinge and Lumsden ( 1995) where several
22
subjective justifications were presented not to include the raw material in the energy balance involving ethanol analysis. In general, when the activities get far from the main operational sequence, the probability of their significantly changing the final results decrease and, therefore, the importance of including them in the calculation also decreases. However, several studies ( Delucchi, 1993 and 1997; Greet, 1996 and 1999; ETSU, 1996 and NREL 1992 and 1997) investigating fuels from biomass showed the importance of including the fertilizers and other materials used in the agricultural activities. Similar problem can be found in Unnash et al. ( 1996 and 2000) that include the fuel consumption of the farm equipment but do not include the material to farm ( fertilizers, herbicides, etc.).
The objective of the study defines on the first hand the minimum necessary boundaries. Some studies, in spite of the name life cycle, truncate the analysis at some point because the study is only a piece of a bigger puzzle to be assembled over time. This is the case of Spath and Mann ( 2000) and most of the NREL studies where the objective is to analyze the hydrogen production only. The Unnash et al. ( 1996 and 2000) studies present the results in terms of pounds of pollutants per mile but they do not include the vehicle operation in the analysis. Vehicle fuel efficiencies are used to bring all the fuel results to the same operational unit but the studies are a fuel upstream analysis only and do not include the emissions of the vehicle operations, for example.
All the other analyzed fuel/ transportation studies ( Delucchi, Greet and ETSU) consider at least the energy requirement of the main operational activities “ from the well to wheels.” The energy requirement calculation includes all the primary energy consumption ( input in the main operational activities) and also the secondary energy 23
consumption ( input in the production activities of the fuels required in the primary activities). This secondary energy calculation is not performed in Unnash et al. ( 1996 and 2000). From the existing studies it is not possible to analyze the importance of Unnasch’s decision since the results are calculated in an aggregated form; however, from the pathways analyzed in this dissertation, it can be said that they are not significant. See section 5.4.1.3 for more details.
The emissions and energy requirement involved in the construction material of the plants ( concrete, steel, etc.) are calculated in the Delucchi and NREL studies. Therefore, the final ( or total) result incorporates these boundary differences and it must be considered for purpose of comparison. Greet’s model includes the emission associated with the vehicle material but not with the plant construction. According to Delucchi ( 1997), for light duty vehicles the energy requirement and CO2 emissions increase about 2.7 to 3.6 % when the plant and retailers location are considered and also they increase 9 to 12 % when the vehicle material is considered. For the special case of solar- hydrogen vehicles ( with Internal Combustion Engines) where the operational emissions are lower the increment is 19 and 72 % respectively.
2.3.3 Time frame
The time frame considered in the analysis is very important because it defines the technology to be considered in the study. It becomes more critical for the impact assessment phase, especially when the boundaries involve disposal, recycling, and decommissioning of plants. Material decomposition time, atmosphere reaction time, system regeneration time, and the life of the product/ components may play an important
24
role ( for example, consider the replacement of batteries for electric vehicles within the time frame of 5 years and 10 years).
Unnash et al ( 1996) calculate their scenario 1 based on the year 1990 and other three scenarios ( 2, 3 and 4) for the year 2010. Unnash et al. ( 2000) present the evaluation for one scenario in the year 1996 and two scenarios for the year 2010. Greet ( 1998, 1999 and 2000) is a model that has two levels of combustion technology: one called “ current” that was done in the early 90’ s before the 1990 Clean Air Act Amendment took effect, and one called “ future” that does not specify any precise time. Theoretically, changing the percentage of current and future combustion technology for different calendar years can be analyzed. However, the model default for near- term vehicle technology analysis is 20 % for current and 80 % for future combustion technologies set for the year 2006 according to Greet ( 1999). ETSU ( 1996, 1997 and 1998) reports do not state the time frame of their analysis but at the same time they use the UK power generation mix composition of the year 1996 and analyze future vehicle technologies ( i. e., fuel cell vehicles) that will not be on the market in the short term. NREL ( 1997 and 2000) studies give no specific time of consideration. NREL ( 1997) is done for a hypothetical plant that could be placed at any time and it considers that the life of the plant has been 30 years; however, for the material analysis it uses the TEAM – Tools for Environmental Analysis and Management data that is a software developed by Ecobalance, Inc. containing data for current processes. Finally, Delucchi ( 1991 and 1993) has the base case for the year 2000. On the other hand, according to Delucchi ( 1997) the model user can specify any year between 1995 and 2015 so that the model applies factors to scale up and down to the
25
base year. Unfortunately, his model was not available, and in the report, results are presented for the year 2000 and 2015, but somehow all the tables, and results are equal.
2.3.4 Data
According to the ISO 14041 ( 1998), Life Cycle Inventory is “ a collection and analysis of input/ output data” and the data treatment is the most important phase of the entire assessment that will be done based on the LCI results. On the other hand, the majority of the authors investigating the LCA methodology agree that there is a lack of comprehensive data available for these studies and also that the quality of the existing data is in most of the cases questionable ( Hendrickson et al., 1997; UETP, 1996; Ayres, 1995; Lee et al., 1995; Boustead, 1994; Denison, 1993; Franklin and Hoffsommer, 1992 and Hunt et al., 1992). Data collection and data analysis have been pointed out as important sources of LCI results discrepancies.
The common advice provided by the studies presented before is that a company that can work with their suppliers’ information should prefer primary data ( collected by the study). However, the cost of doing this is always a problem, for a very extended analysis it may not be possible, and finally, proprietary information cannot be checked or published. In addition, if the analysis involves a more generic product such as fuel, a single company’s data may not be sufficient to represent the possible mix of technology. According to the authors ( referenced above) secondary data ( from literature) can be out- of- date, especially for advanced technologies, to represent a large range of technology and in most of the cases gaps must be filled in. The solution suggested so far is that the steps used to fill the gaps must be identified in the report.
26
As a basic principle of any scientific study the data should be available for all researchers who want to reproduce the results. Today it is more and more common in LCI publications for only the results to appear and very few comments are made about generic assumptions in the model. Those studies are in most cases useless because they generate the situation of “ believe me or not.” The selection of the so- called “ most comprehensive” existing studies analyzed here was based mostly on the concern of the authors to publish their assumptions. Even with these selection criteria, one trying to reproduce the studies’ results may have no success due to the lack of necessary information. All the assumptions used in Greet ( 1999 and 2000) can be checked since the model is publicly available; however, several inputs are the author’s subjective assumptions with no explanation of the rationale for the decision. Some reports, like Unnash et al. ( 1996) and ETSU ( 1997), publish the spreadsheet table which helps somewhat more than the ones that do not publish them ( e. g., Delucchi’s report). When a subjective assumption is not the case, a common practice is the use of a single source of reference as input; could be cleaver sometimes it is not the case of a lack of other sources. A critical example is Unnash et al. ( 1996) using data from the early 70’ s for hydrogen plants. Similarly, ETSU ( 1996) uses U. S. EPA emission factors from 1985.
The data problem in the LCI methodology is so critical that Derenne ( 1995) suggests that all studies should establish an independent authority charged with supervising data collection and processing. Also, Ayres ( 1995) suggests that when more than three firms use the same process at the national level the data about that process should be available. The international standards ( ISO 14040, 1997 and ISO 14041, 1998) suggest several levels of critical review: from an internal expert, from an external expert,
27
or from a panel of experts representing the interested parties. Denilson ( 1993) goes further and suggests that the peer review should not be only a post- study activity but should also be integrated into the study design and execution phases.
According to Denison ( 1993) aggregation of data has been used to mask proprietary information. It is also used to preserve a standard routine in the model when external calculations are performed to generate a standard input format ( like the plant energy efficiency). An important difference pointed out by Andress ( 1998) in his qualitative comparison between Delucchi and Greet models for ethanol fuel is the higher amount of external calculation performed by the Greet model. The problem with this external calculation approach is that, in general, the input and methodology of the external calculation is not published and the situation “ believe me or not” appears again.
Denison ( 1993) comments about the difficulties of comparing different studies and figuring out the importance of some decision when the results are generated and/ or presented in an aggregated form in the fuel/ transportation studies. For example, Greet ( 1998, 1999 and 2000) include in all calculations the secondary emissions and energy requirement in such a way that one cannot check the importance of the secondary pathway in the calculus or compare the result with another study that does not include the secondary calculation in it. A similar problem occurs with Delucchi ( 1991 and 1997) in reference to the material for plant construction.
Finally, another common reporting problem in these analyzed studies is related to the technologies that they are considering. In general, they report well the combustion engine assumed ( turbines, reciprocating 2 strokes, etc.); however, for air emission controls a certain kind of control is assumed without specifying it. It is useful to point out
28
that the EPA/ AP- 42 ( 1995) reports several emission factors for equipment like boilers, reciprocating engines, etc. and that all of them are for uncontrolled situations. For some control technologies a factor is provided to reduce the uncontrolled emission factor, but not for all. The transparency in the assumed air control technology is also important to understand the potential for improvements in the future.
2.4 Methodology of calculus of existing fuel/ transportation LCIs
What the previous Life Cycle Inventories ( LCI) studies did well was to establish the calculus methodology to inventory the air emissions and energy requirement in the fuel/ transportation sector. Basically, for each fuel that is analyzed one can define two different aggregations: the fuel pathway, defining the process involved in specific upstream- connected activities ( or stages), and the system definition. For example, in the first aggregation, one pathway example is hydrogen fuel delivered as compressed gas at the fuel station, distributed by pipelines from bulk storage and produced from natural gas ( NG) in a centralized steam reformation plant inside the analyzed area. A similar specific pathway is extended for the NG ( feedstock) back to the extraction process.
The second aggregation is related to the system definition. For example, considering only the hydrogen pipeline pressure, some systems may assume the pressure of 200 psi ( Greet, 1998) and others 1000 psi ( ADL, 1996). Each new alternative considered should define a new pathway in a tree configuration; however, in practice, a single pathway may contain more than one system definition. The Figure 2- 3 presents this idea. It is essential to point out that a single change in the system aggregation or in the pathway aggregation will change the final result.
29
The calculus is performed initially at the stage ( or activity) level, and later a composition of the various stages defines the pathway result. This sequence idea is presented next.
FuelOptionHydrogenMethanolGasolineOthersCompressedLiquid….. nNG- SRH2 PipelineH2- prod….. nTruckStorageoutside areaNGCalifornia NGPathway AggregationSystem AggregationFuel StationFuel Storage and TransportationFuel ProductionFeedstock T& S and Production270 - no steamElectrol. H2- prod.. nComp. gas.. noutside areaStorageinside area….. ninside area270 - no steaminside area27 - no steaminside area27 - steaminside area270 - steamPipelineTexas NGCanada NGOthersFuelOptionHydrogenMethanolGasolineOthersCompressedLiquid….. NGOthersFuelOptionHydrogenMethanolGasolineOthersCompressedLiquidLiquid….. n….. NGOthers
Figure 2- 3: General idea of the fuel upstream calculation
2.4.1 Calculus for the stage ( activity) level
The existing emission factors are, in general, established at the equipment level and they are associated with the equipment load or activity level. For example, grams of pollutant emitted per fuel consumed by a boiler, or pounds of pollutant emitted per work produced by an engine. These factors should be the representative average value of a long- term process activity and, in general, they are reported by organizations such as EPA and CARB. The EPA/ AP- 42 ( 1995) is the typical example of an emission factors publication. One interesting point here is that the AP- 42 presents the emission factors for 30
uncontrolled equipment only, and for certain equipment it presents a factor to adjust the uncontrolled value to an air control device assumed. The percentage of uncontrolled equipment versus controlled ones, as well as the percentage per type of air control technology assumed for a region is, most of the time, a subjective assumption due to lack of specific data, especially for a broad national analysis like the ones performed by Delucchi ( 1997), Greet ( 2000) and ETSU ( 1996). What is more common in the existing studies is the assumption of an aggregated emission factor, theoretically a weight average of all technologies assumed, without too much explanation or the rationale behind the assumption. Environmental policies and policies enforcement level may help make the decision of the assumption. In complement of that, the police analysis can be easier for a more restricted area, like in SCAB performed by Unnash et al. ( 1996) where it does not present the state’s diversity of laws, enforcement strategies and success in their execution.
On the other hand, equipment of different sizes may also have different emission factors; therefore, by assuming one specific emission factor a scenario composition is created ( explicitly or not). For this dissertation, the explained system aggregation is called the technological scenario composition. All these necessary assumptions in the technological scenario composition lead to discrepancies among the existing studies and also to the discrepancies in their final results.
Other information commonly available is the thermal efficiency of fuel production plants, or other activities, used, in general, in cost analysis. The thermal efficiency is defined as the total usable energy output from the system divided by the total energy input into the system. As presented in Figure 2- 4, the thermal efficiency is the
31
energy content in the products divided by the energy content in the fuels and feedstocks. When the thermal efficiency value is available, a required connection with the energy consumed at the equipment level is necessary. This connection is achieved by understanding the plant design and translating it into the energy share and equipment share. The energy share ( Eshare) is defined as
Σ= nnnnshareFFE1/)(, Equation 2- 1
where F is the energy consumed from each different source ( natural gas, oil, electricity, etc.) and n is the number of energy sources used by the stage. Similarly, the equipment share ( Eqshare) is defined as
Σ= mmmmshareQQEq1/)(, Equation 2- 2
where Q is the energy consumed by each different equipment type ( boilers, engines, etc.) and m is the number of equipment type used by the stage. Figure 2- 5 shows the details of this idea where the “ fuel- 1” is the feedstock for the products production process. Eventually, depending on the available data, another pre- calculation is done to achieve the equipment load or the equipment energy requirement. These pre- calculations specify the detail level of the model. According to Andress ( 1998) a major difference between the Delucchi ( 1997) and Greet ( 1998) models for the ethanol calculation was that the detail level was much higher in Delucchi’s case. The majority of the existing studies use this
32
efficiency approach presented that, in fact, is a complex “ bookkeeping process.” In other words, until this level of calculation ( well represented by Greet models up to the version 1.5a), the analysis assumes the character of “ if- then.” For example, if the efficiency of the process “ X” is “ w,” then the result “ Y” is equal to “ z.” This situation reinforces the necessity for well- discussed input assumptions, in order to avoid a “ garbage in – garbage out” situation.
Stage orActivity Energy/ fuelsMaterials/ feedstockProductsWastes/ emissionsStage emissions
Figure 2- 4: Input/ Output idea at the stage level
For certain activities Unnash et al. ( 1996) uses physical parameters like work, volume, etc., to calculate the activity level of the system analyzed. This “ component- model” brings the analysis to the level where a worker- expert from a plant ( or system) similar to the one analyzed can provide accurate information and even some new data. Of course, because the ultimate target is the energy ( fuel, materials, electricity) consumed, it is necessary that a unit conversion involving some kind of efficiency concept ( vehicle fuel efficiency, compressor efficiency, etc.) be made.
33
• EquipmentShareTotal energy consumedFuel 1CombustionEquip. 1Fuel nCombustionEquip. n- 1CombustionEquip. nFuel 2... …. • Efficiency = energy out / energy in• Energy Share ( in) VARIABLES: • Products ( out)
Fuel ProcessProcess
Figure 2- 5: Establishment of the equipment activity or load process
A small difference should be considered for transportation stages where the final efficiency is associated with the distance transported. The lack of this distance association in the calculation was one of the main constraints for this project to use the available Greet ( 2000) model at that time for local analysis ( the other one was that all the equipment assumptions are supposed to reflect the U. S. national average data only). He and Wang ( 2001) solved the distance dependence problem in the Greet version 1.6.
Having established the equipment loads, the final result is the sum of the multiplication of every equipment load per its associated emission factors. In this dissertation, these emissions are called process emissions and, in most cases, they are associated with combustion activities and with the designed air control equipment. Another kind of emission is the fugitive emissions associated with maintenance, malfunctioning, spills, leaks and losses in junctions, purges, etc. For certain kinds of equipment or activities ( e. g., natural gas extraction or fuel storage tanks) there are similar emission factors, as explained before, and the way to calculate the emissions is the same.
34
However, in most cases no emission factors are available and a percentage of the fuel consumed by the equipment or activity is assumed to be lost. The amount of pollutants presented in the composition of the fuel lost is then calculated and added to the process emissions to give the total emission of the activity. Figure 2- 6 shows a graphical representation of this idea. By looking into the literature, one can note that the assumption of the average process design, translated into energy share and equipment share, as well as the amount of fuel lost, is not well documented. To some degree the input assumption becomes a subjective matter and a source of uncertainties and disagreement about the final results of existing studies.
Total EmissionsProcess EmissionsFugitive EmissionsEmissionFactorEnergy ConsumedPercentagelostFuel Lost• Pollutant on the fuelcomposition ( CH4, NMOG, etc.)
Fuel XX+
Figure 2- 6: Graphical representation of the emission calculation at the stage level
The emissions calculated are attributed to the geographical region where the activity is considered. For the life cycle approach it is also necessary to consider the
35
emissions associated with the production and distribution of the fuels consumed. The life cycle of these fuels, called here secondary emissions, may occur in different regions and should be kept separate if geographical occurrences are considered. None of the models but Unnash et al ( 1996) considers the geographical occurrences, and what they do is to sum, in most cases, the secondary emissions into the primary emissions calculation. These aggregated results also make comparative analysis difficult to understand whether the eventual differences are related with the secondary emissions, and to understand the importance of these emissions.
2.4.2 Co- products allocation
Since a single process can generate more than one product ( with market value) the energy requirement and the emissions generated by the process should be allocated among all these co- products. Some authors like Weidema ( 1993) call main- product the co- product which is used in the next step of the investigation, and by- products the co- products that are outside of the investigation’s scope. For simplification and following Vigon et al. ( 1993) denomination, only the term co- product will be used and it will be applied every time the activity generates a product different from the main product investigated.
Different approaches can be used to allocate the co- products credits ( or debits) of the environmental aspects calculated and, in most cases, the final result is very sensitive to the allocation procedure assumed. Currently, there is a search for an acceptable single allocation criterion to become standard to eliminate this source of disagreement and eventual manipulation. So far, all the proposed criteria suffer from several limitations,
36
and according to EETP- EEE ( 1996) none of the allocation criteria is universally applicable. Therefore the choice must depend on the type of product.
Allocation connected with physical properties such as weight, energy content, or chemical equivalents has been used and sometimes suggested as a general procedure ( Hunt et al., 1974; Consoli et al., 1993 and Vigon et al., 1993). According to Boustead ( 1992) the benefit of using physical properties is to keep the allocation stable under a given technology. However, one should not use weight allocation, which works well for metals, for energy services or use chemical equivalent for agricultural crop products and so on.
The economic or market value of the co- products has the obvious advantage of being universally applicable. According to Weidema ( 1993) Basler and Hofman had first used this approach in 1974, and Heijungs et al. ( 1991) suggested it as a general methodology. According to EETP- EEE ( 1996), the transient nature of economic values is the main problem adopting in this approach. Even when an averaged price over long periods of time is used, fluctuations are unavoidable and emissions will vary without any change in the technology itself.
The market displacement approach works with the rationale that most of the co- products can replace or substitute for other products, eliminating the environmental aspects associated with the ones replaced. In other words, the accumulated environmental aspect of the process minus the accumulated environmental aspects of all co- products will be the associated environmental aspect of the analyzed product. Vigon et al. ( 1993) used this approach to analyze waste incineration and Heintz and Baisnee ( 1991) suggested it as a general method. This approach involves the addition of a new life cycle
37
“ branch” to the process tree for every co- product, and it may be too complicated if several co- products are involved. Also, the decision of the replaced product can be subjective and it also may change over time ( Weidema, 1993).
The international standard ISO 14041 ( 1998) suggested three ranked steps for the allocation procedure: first, wherever possible avoid the allocation necessity by splitting the unit process or by expanding the product system. Second, where allocation can not be avoided, use some kind of physical relationship between the products, and, finally, where physical relationship is not possible, another kind of relationship like economic value shall be applied as last choice. Weidema ( 1993) presents an interesting comment that allocation by physical properties can be seen as a special case of the allocation by economic value. In the fuel analysis, for example, the market value of the fuel has a strong correlation with the energy content in the fuel and therefore the fuel energy content should be chosen as the allocation method. However, this idea does not apply for fuel productions that involve other kinds of co- product, like food in the corn ethanol case.
For fuel co- products, like natural gas liquids, Greet ( 1998, 1999 and 2000) and Delucchi ( 1991 and 1997) use the energy approach. For the ethanol production from corn, the Delucchi study uses the co- product displacement approach and Greet gives the option to alternate between the displacement approach and a mix of market value and energy content. Wang et al ( 1997) did a sensitivity analysis to test the importance of using this approach for ethanol calculation. According to his analysis the most significant factor in the study was the co- product credit allocation. Using different approaches the authors got results with differences up to 40 %.
38
It is not clear in the report how ETSU ( 1996) handle the allocation process. The only statement about the issue is this one: “ by- products are excluded from the analysis in the case of well- established processing operations. However, for new biofuels where by- products markets are weak and under- development, potential energy credits for by- products have been included in the range of possible outcomes.”
2.4.3 “ Average emissions” versus “ marginal emissions” calculation
Unnasch et al. ( 1996 and 2000) have been pushing the idea of using the life cycle approach to calculate the “ marginal emission” as opposed to the “ average emission” performed by all the other studies. Unnasch et al ( 1996) take no internal co- product credits into account and use the “ average emissions” in their study, but they introduced the idea of “ marginal emissions” for the electric generation inside South Coast Air Basin ( SCAB). The “ marginal emission” idea was inspired by the fact that SCAB has a law called RECLAIM that caps the amount of NOx emitted inside the basin, based on a fixed amount of emission credits that the companies must have to emit NOx. The companies claim that it does not matter which technology is used because the final emission must comply with the law and therefore the “ marginal emission” in this case will be always zero.
Unnasch et al. ( 2000) uses only the “ marginal emissions” approach to come up with their results. In order to do that, it was necessary to use the co- product replacement method, which presents the problems discussed by Weidema ( 1993). The life cycle “ branches” of the co- products are not included in the Unnasch et al. ( 2000) analysis introducing much more uncertainty.
39
It is also my personal opinion that the “ marginal emissions” idea goes against the general idea of life cycle analysis which has been developed to compare the environmental aspects of two different products or services. Some of the outcomes are interesting to analyze if the “ marginal emission” approach is considered in a study. For example, if a methanol car replaces a gasoline car, no upstream benefit is found since the oil refineries inside the area are going to produce gasoline for exportation. In fact, the methanol ship tankers, fuel terminals, etc., can introduce more upstream emissions. In a landfill gas case example, if one tries to analyze what the best way ( in terms of NOx emission) to use the gas would be, the answer will be “ it doesn’t matter” - the marginal emission will be zero whether producing electricity or methanol. However, in reality, methanol production may emit less and other mechanisms ( e. g., RECLAIM) will allow the pollutants to be generated later somewhere else.
It is important to point out that various important aspects of the “ marginal emissions” approach have been previously incorporated in the “ average emissions” approach calculation when the technologies designs are selected; i. e., to choose the technologies design it is necessary to analyze the local emission control enforcement, cost, existing fuel production capacity, etc. In fact, the “ average emissions” approach is an average calculation of the “ marginal technology” selected.
In summary, the decision about which is the best methodology for the calculation is based on what fundamental question the study wants to answer. If the question is “ Which fuel technology has the highest potential to emit less air pollutants?” or “ How much emissions will be released by a specific technology over its life cycle activities?”, then the best methodology, in my opinion, is to use the “ average emission” approach.
40
On the other hand if the goal- question is “ What amount of pollutants will the population breathe?” then I am afraid a very complex model will be necessary, accounting for the location of the emission sources, atmospheric conditions ( wind directions, temperature, etc.), population densities, and so on. What the “ marginal emission” approach tries to do is to figure out the “ net” emissions considering all the sources in an area. In some sense this approach is one step towards the solution for the proposed second goal- question, but, unfortunately, it is moving towards greater complexity and therefore a more complex model is needed. My suspicion is that making huge assumptions without modeling them, as it is the case in Unnasch ( 2000), only increases the uncertainties of the study without knowledge of these uncertainties. Specifically, I am talking about the assumptions for displacement of fuels ( without any economic or demand modeling being performed) and the emission credits based on the displaced technology ( without a complete life cycle analysis of that technology within the study).
2.4.4 Pathway level calculation
At the pathway level, the calculation is basically the total of the emissions and energy requirement calculated for every stage of the pathway – this is what I called, above, a “ bookkeeping calculation.” However, some other sections are important to point out here. The first one is the downstream own- use factor that takes care of the consumption of the analyzed fuel in the downstream activities. In other words, if part of the input fuel in a stage has been consumed there, for example, transporting diesel in a diesel truck or losing fuel in fugitive emissions, the previous stage must supply more fuel 41
to account for the delivered and consumed fuel in each stage. In general, an own- use factor is generated in each stage ( similar to efficiency) and a multiplication of them gives the downstream own- use factor. In spite of considering minor mistakes, some pathways of some studies present problems in this calculation when they try to aggregate activities in one single block, like considering storage and transportation together.
At this level may also occur the mixing of more than one pathway to generate different combined scenarios for analysis. To do that, generally a weight average of the pathways results is used. Also, if the study accounts for areas of the emission occurrence, as in Unnasch et al. ( 1996 and 2000), the separation of the areas occurs at this level too, allocating each stage’s results into different cells and totaling them later.
A second important point is the consistency in the values of energy content used to add up the results of each stage. Theoretically, for some stages where the fuel combustion water remains as a gas; if the sensible heat and latent heat of a water vaporization is not used by the process, then net calorific value or low heat value ( LHV) can be used in the stage calculation. Examples of these stages are truck transportation and pipelines. On the other hand if the stage utilizes the heat of the water condensation, like refineries and power plants, the use of the gross calorific values or high heat value ( HHV) is recommended. The most important point is to care about the use of both heat assumptions to add the stages’ results and get the final pathway results. This inconsistency was not found in any study, but all the studies do choose one single heating system to perform all the calculations and, therefore, to compare their results one should account for these possible differences. Delucchi ( 1991 and 1997), Unnash et al. ( 1996 and 2000) and ETSU ( 1996, 1997 and 1998) use HHV. Greet ( 1998, 1999 and 2000) use
42
LHV. Boustead et al. ( 1992) suggest the use of HHV for fuel calculation and so do I. Since the sensible and latent heat of water vaporization is there, in the process, and can even be measured, it is only a technological strategy to use them or not. Low heating values are only a subterfuge to show better efficiency in a process.
LC- NGNatural Gas RecoveringNatural Gas ProcessingNatural Gas Transportation and DistributionNatural Gas Life- cycleLC- OilLC- DieselLC- CoalLC- NGLC- NG
Power PlantsTransmissionElectricity Life- cycleLC- ElectPower Elect
HydrogenProduction PlantHydrogen Transportation and DistributionHydrogen Fuel StationHydrogen Life- cycleLC- H2HydrogenProduction H2
Figure 2- 7: Complexity of the life cycle calculation showing the interdependence among fuels ( LC = life cycle results).
The third section is the interdependence among fuel production processes. The Boustead model was the first study to solve the problem of interdependence of energies by performing simultaneous interactions in the model where the first results are used as inputs for the second interaction and so on, until some convergence is achieved ( Boustead, 1992). It accounts, for example, for the convergence of the energy consumption of electricity and natural gas use in Figure 2- 7. Since electricity can be used in the natural gas process that later can generate electricity, the circular calculation is 43
necessary. The importance of it will depend on the initial inputs assumed, and simplification of the system into linear sequences to avoid the problem may give rise to significant errors according to Boustead ( 1992). Delucchi ( 1991 and 1997) and Greet ( 1998, 1999 and 2000) use this approach. Unnasch et al. ( 1996 and 2000) and ETSU ( 1996, 1997 and 1998) do not.
A possible problem occurring in the existing studies that do use the circular calculation approach is that it should be done geographically, and there is no evidence that it was done. For example, the electricity produced in the US is not used to process natural gas ( NG) in Canada, even if some power plant in US uses Canadian NG. Another example can be a bunker fuel consumed and refueled by a crude oil ship- tanker in a US port and in a remote area port.
2.5 Quantitative analysis of existing fuel/ transportation LCIs
Based on all the sections presented above one can realize the difficulty of matching a similar scenario and pathway to compare the existing studies results in a fair way. Similar problems will be found trying to create a composite result from existing studies. Mark ( 1998) did a comparison among the upstream emission results of some models for compressed hydrogen fuel, produced in a centralized steam reformation process, from natural gas. Figure 2- 8 and Figure 2- 9 show examples of his findings. A strong need for better comparative evaluation studies was clear, since no agreement was found among the models for either the total emissions or their detailed origins. Similar analysis was done at the beginning of this project, agreeing with the Mark ( 1998) findings. Figure 2- 10 presents my result for natural gas recovering and processing for the gas used as feedstock 44
in hydrogen fuel productions. Figure 2- 11 shows another example of mine for methanol upstream activities in terms of total energy consumption. It is good to keep in mind that some adjustments are necessary in order to present all results in the same units, and that some differences in the scenarios are still present, but they serve well as examples for discussion.
Total Upstream Emissions00.050.10.150.20.250.30.350.40.450.5HCNOxCOg/ miWang 1998Acurex 1996Wang 1998DTI 1998Acurex 1996Mark 1996NREL 1994CompositeGV ( 27.5 mpg) HFCV ( 55 mpg)
Figure 2- 8: Mark ( 1998) comparison of the total upstream emissions for fuel cell vehicles in existing models
These result mismatches can be extrapolated to other fuels and pollutants as well, and some different examples were published in Contadini et al. ( 2000a). The paper also discusses the three main reasons for the result mismatches. The geographical differences are the first one. They are related to the initial study objective and, in general, can be found stated in the reports. Problems arise when they are put together for comparisons, such as the case of the Mark ( 1998) presentation, comparing a SCAB analysis with a US national analysis. Similar problems occur in the three figures presented in this section, and a good solution is to identify the geographical differences very clearly in the slides,
45
as shown in Figure 2- 11. Another situation where this problem arises is in the attempt to generalize the results. A classical example was the press release of the Pembina/ Suzuki ( 2000) study. In the press release, some results were presented and discussed as universal but nowhere was the scope of the calculation done for Canadian scenarios clarified.
HC05101520253035404550Acurex 1996DTI 1998Wang 1998g/ mmBtu ( CH2 delivered - LHV) extractionproductiondistribution & compressiondistribution & compression5.81.59.2production18.50.51.1extraction22.027.11.7Acurex 1996DTI 1998Wang 1998HC05101520253035404550Acurex 1998
Figure 2- 9: Mark ( 1998) comparison of the hydrocarbon emissions calculated by existing studies, for gaseous hydrogen fuel produced in centralized plants ( SMR process).
A second problem is related to the technology composition scenario. Using the methanol analysis as an example, one can check that Delucchi’s results are based on a combination of coal and natural gas to methanol. Greet’s results are 100 % natural gas to methanol, but they are also a combination of 20 % of current technologies with 80 % of future technologies, and the Acurex results are the combination of 50 % advanced steam reformation plants and 50 % of advanced combined partial oxidation plants.
46
NG feedstock emissions related with H2 fuel production0.0010.0020.0030.0040.0050.0060.0070.00NMOGNOx COPollutantsEmissions ( g/ MBtu- H2_ deliv.) LHVAcurex, 96 ( 2010 sc) DeLuchi, 97 ( 2000 sc) DTI, 98 ( 2000 sc) Wang, 96 ( 2006 sc)
Figure 2- 10: Total emissions associated with natural gas recovering and processing from existing models.
0.0000.1000.2000.3000.4000.5000.6000.7000.8000.900MBtu- cons / MBtu- delivGRI ( 1994) Greet 1.4a ( 1998) Greet 1.5a ( 2000) DeLucchi ( 1997) Acurex ( 1996) ETSU ( 1997) Life Cycle Energy Consumed ( HHV) FeedstockProductionMarketing( U. Kingdom) ( SCAB) ( USA average)
Figure 2- 11: Life cycle energy consumption for methanol fuel from existing studies.
The level of the technology scenario composition goes as deep as the calculation detail performs. For example, for a very specific process such as the natural gas processing within the same area, different studies assume different combinations of
47
equipment consuming the gas in the process ( equipment share). Table 2- 1 presents the values.
Table 2- 1: Equipment share of US natural gas processing (%).
Study
Darrow ( 94)
Greet 1.5
Harrison 99
ADL( 96)
NG Turbines
50
54
NG Recip. eng.
46
67
NG Boiler
100
50
Process Heat
33
The third level of disagreement, and also very important for this project, is the use of deterministic values to represent complex systems. For example, even considering a very specific technology, such as a typical- size ( 2,500 metric tons per day) production plant of methanol, using steam reformation process to produce the syngas, the efficiency numbers, without the consideration of extra steam for exportation ( second column of Table 2- 2), generate a lot of mismatches among the existing studies, going from 59 % to around 70 %. A reason for that is clear: the measurement over time of a single plant efficiency will vary according to the natural gas composition variation, operational adjustments, equipment malfunctions and maintenance, catalyst deactivation and so on. Trying to represent with a single number the average of several similar plants operated by different organizations and placed in different regions within the same country is not an easy task. It is hard to defend one study value as better than another one, especially because there is a lack of detailed information in the literature, such as operational pressures, catalyst load and life, etc. For example, EPA and CARB do not release emission factors of plants, such as methanol or hydrogen, and what the existing studies try to do is extrapolate them from boiler data and other equipment. In all these cases, the
48
uncertainties behind each single number are also not apparent, and may contribute to eventual manipulation of the technological data, also generating mismatches.
Table 2- 2: Analysis of existing data for methanol production plant.
Typical Size: 2,500 metric tons of MeOH per day – Steam reformation syngas
HHV
Efficiency (%)
Electric. used (%)
NG used as fuel (%)
Extra Steam/ Electricity
Without
With
Without
With
Without
With
Greet 1.5a ( 2000)
69.6
71.6
0.2
- 3.33
17
24
Acurex ( 1996)
-
68.3
-
- 0.02
24.1
-
Delucchi ( 97, 93)
65
-
0.2
-
-
-
Greet 1.4 ( 1998)
65.6
-
0.2
-
- 100
-
Darrow/ GRI ( 1994)
-
66.1
-
- 0.007
-
22.6
Ogden et al ( 1994)
67.4
-
1.8
-
-
-
DTI ( 1998)
64
-
-
-
-
-
Chem. Ecn. HB ( 96)
-
71.3
-
-
-
-
Dybkar ( in Wang)
66
71.6
-
-
-
-
Islan ( in Wang)
63
-
-
-
-
-
Borroni- Bird ( 96)
59
70
-
-
-
-
DOE ( 89)
61.1
70.4
Sweeney ( 98)
65
-
-
-
-
-
AMI ( 98)
60
70
Allard ( 2000)
64
LeBlanc ( in Cheng, 94)
69.4
0.81
Leveton ( 2000)
64.0
Pembina/ Suzuki ( 2000)
61.8
One way to represent these systems variation is by using distribution curves. This solution can be applied to all systems and even for values that everybody expects to be constant. As an example, Table 2- 3 presents some physical parameters of hydrogen ( H2) used by existing studies. Of course, the study results are more sensitive to some variables modification than others.
Sensitivity analysis is an interesting approach to focus attention on the aspects that are important for the overall results of the assessment. One idea is to develop the LCI study in an interactive process starting with a simplified version of the product life cycle
49
and after a sensitivity analysis concentrate the effort in the critical areas ( EETP- EEE, 1996). The ISO 14041 ( 1998) suggests performing sensitivity analysis on significant inputs, outputs and methodological choices of the Life Cycle Inventory ( LCI).
Table 2- 3: Hydrogen physical parameters used by existing studies. Analized FuelHydrogen - H2SourcesDTIDeLuchi Ogden Greet Acurex PembinaMITHeywood19981993199920001996200020001988Energy content ( Btu/ scf - HHV) 325338324324324325343Energy content ( Btu/ scf - LHV) 273.4274274274274290290Fuel density ( g/ scf) 2.522.402.402.5462.549Molecular Weight ( g/ mol) 2.015
What some of the existing studies do is to present different scenarios that are primarily variations of the system and/ or pathway aggregations only. In this kind of analysis several inputs are changed at the same time and the significance of the changes is never discussed by any of the authors. The possible variation among the input data at the equipment level can also be critical but, in spite of its importance, this variation is not considered very seriously yet. Unnasch et al. ( 1996) is one of the studies that devotes some space to this kind of sensitivity analysis. Their report shows a huge variability of the individual NMOG emissions for the reformulated gasoline case, though the data and the methodology related to the calculation are unclear.
Greet 1.6 ( 2001), following early recommendations of this project, implemented as well the concept of uncertainty analysis in the model, using Monte Carlo simulation and probabilistic curve as input. In spite of the right direction adopted it is still suffering from several misunderstandings of the methodology proposed ( section 3). Basically, Greet 1.6 uses triangular curves to represent bounding scenarios without realizing that by doing so it is accounting for zero probability of that scenario to occur. It also made no
50
attempt to correlate the variables, and no regression sensitivity analysis – to understand the importance of each curve – was reported. In complement of that, unfortunately, the major problem of the model was, perhaps, the calibration of the curves with a very biased pool of experts. All of the experts were from only three oil companies that had explicitly engaged in pushing the gasoline pathway as the best solution for the future fuel cell fuel infrastructure problem.
In conclusion of this section, it is good to reinforce the difficulty of relying on one single value as input for the model or as the result of it. Based on that, it is almost impossible to do a fair comparison on the final results of existing studies. However, comparisons done at the detailed level are possible and very informative, as the examples presented here show. In fact, the comparison at the level of equipment and single stages is easier to perform and guarantees that only similar technologies and assumptions are present. This dissertation completed several of these detailed comparisons. They are incorporated in the database of the created FUEEM model and some of them are presented in section 4.
51
3 FUEEM METHODOLOGY
As described in section 1.3, this project and the development of the Fuel Upstream Energy and Emissions Model ( FUEEM) was targeting to deal with uncertainties in life cycle assessment studies, in particular, in the analysis of fuel cell vehicles and their potential new fuels.
As explained in previous sections ( 2.3 to 2.5), some subjectivities and uncertainties will always be an inherent part of this kind of study and, therefore, the basic idea was to explicitly recognize the uncertainties and quantitatively include them in the model and analysis results. By doing that, my expectation was to generate richer information, minimize possibilities for future result mismatches, and create a higher level of credibility for a life cycle assessment study.
Three main necessities were recognized. First, the necessity to improve the analysis of data input trying to minimize situations described as “ garbage in, garbage out,” thereby increasing the level of credibility of the study. A second necessity was to choose a variable format that better represents uncertainties than a single deterministic value. Finally, a third necessity was to create a model that combines and propagates the uncertainties through the calculation. I established as a parallel contribution the creation of a model that allows more flexibility for local analysis.
Working on the solution of these three necessities, this study generated two major original approaches. The first was the development of a methodology for the input data treatment for future technologies based on the concept of interested parties. The details of this methodology are presented in section 3.1. The second was the adaptation of economical risk analysis techniques into FUEEM to represent and propagate uncertainties
52
into the calculation. The techniques involve the use of probabilistic curves, Monte Carlo simulation, and rank correlations. The details of these techniques are presented in section 3.2.1 and some details of the model are presented in section 3.3.
3.1 Input data treatment for future technologies
Most of the development of the FUEEM methodology to treat input data for the fuel cell vehicle analysis was based on the possibility of having an international panel of experts cooperating with the project goals. The objective of this section is to present the methodology adopted to take maximum advantage of the expert participation and to explain the rationale behind the decisions made. Several other methodologies exist to deal with the same necessity; therefore, this project solution is presented here as a case study. By doing th