Life cycle model of alternative fuel vehicles: emissions, energy, and cost trade-offs

Life cycle model of alternative fuel vehicles: emissions, energy, and cost trade-offs

Transportation Research Part A 35 (2001) 243±266 www.elsevier.com/locate/tra Life cycle model of alternative fuel vehicles: emissions, energy, and c...

401KB Sizes 0 Downloads 1 Views

Transportation Research Part A 35 (2001) 243±266

www.elsevier.com/locate/tra

Life cycle model of alternative fuel vehicles: emissions, energy, and cost trade-o€s Jeremy Hackney a,*, Richard de Neufville b a

b

University of Berne, Switzerland Technology and Policy Program, Massachusetts Institute of Technology, Cambridge, MA 02139, USA Received 10 August 1998; received in revised form 9 February 1999; accepted 28 September 1999

Abstract This paper describes a life cycle model for performing level-playing ®eld comparisons of the emissions, costs, and energy eciency trade-o€s of alternative fuel vehicles (AFV) through the fuel production chain and over a vehicle lifetime. The model is an improvement over previous models because it includes the full life cycle of the fuels and vehicles, free of the distorting e€ects of taxes or di€erential incentives. This spreadsheet model permits rapid analyses of scenarios in plots of trade-o€ curves or eciency frontiers, for a wide range of alternatives with current and future prices and levels of technology. The model is available on request. The analyses indicate that reformulated gasoline (RFG) currently has the best overall performance for its low cost, and should be the priority alternative fuel for polluted regions. Liquid fuels based on natural gas, M100 or M85, may be the next option by providing good overall performance at low cost and easy compatibility with mainstream fuel distribution systems. Longer term, electric drive vehicles using liquid hydrocarbons in fuel cells may o€er large emissions and energy savings at a competitive cost. Natural gas and battery electric vehicles may prove economically feasible at reducing emissions and petroleum consumption in niches determined by the unique characteristics of those systems. Ó 2001 Published by Elsevier Science Ltd. Keywords: Alternative fuels; Electric vehicles; Fuel cell vehicles; Energy eciency; Emissions; Environmental policy

*

Corresponding author. Fax: +1-617-253-7568. E-mail address: [email protected] (R. de Neufville).

0965-8564/01/$ - see front matter Ó 2001 Published by Elsevier Science Ltd. PII: S 0 9 6 5 - 8 5 6 4 ( 9 9 ) 0 0 0 5 7 - 9

244

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

1. Introduction 1.1. The issue Motivated by dependence on oil and by urban air pollution, environmental and energy policies in the United States and other countries include initiatives to promote transportation systems based on alternative fuel vehicles (AFV). A comparison of performance trade-o€s between combinations of vehicle and fuel alternatives would help support this e€ort by identifying the systems with superior characteristics, before too much is invested ine€ectively. Policies embedded in the Clean Air Act Amendments of 1990 (CAAA90), the Energy Policy Act of 1992 (EPACT92), and associated state and federal programs attempt to stimulate the growth of AFV markets in the US with a menagerie of incentives. The high ®nancial stakes in the potential market for AFV systems encourages stakeholders to become closely involved in policy making that a€ects their interests, and to steer rules in their own favor. These groups readily provide information that presents their products in the best light, using favorable assumptions and points of comparison. This results in a confusing lack of comparability between the several alternative fuel and vehicle systems, and diculty deciding on a policy which supports the aims of the CAAA90 and EPACT92. Decision-makers need a simple, comprehensive overview that compares alternatives in a common context. Decisions based on technical merit require quanti®ed comparisons between the trade-o€s of each alternative in the light of economy. These assessments are dicult to make because the relative performance of alternative systems depends both on the physical circumstances (e.g., the state of the art of producing each fuel) and on speculative future prices of alternative fuels. The problem requires a framework for analysis that allows the easy calculation and comparison of the implications of di€erent assumptions that a€ect each fuel/vehicle combination. There is no single comprehensive reference comparing transportation systems based on AFV. As Congress debated new clean air and energy legislation in the late 1980s, the subject of alternative fuels and vehicles attracted renewed interest of federal agencies and departments, and therefore the private and academic sectors. A large body of work has emerged over the last decade after passage of the CAAA90 (see e.g., Transportation Research Board, 1997), but its inconsistent coverage of di€erent types of fuels, vehicle technology, and systems-level concerns hinder simple comparisons. 1.2. The response This paper presents a model that provides a clear comparison of the emissions, energy eciency and cost performance of di€erent fuel and vehicle technologies on a level playing ®eld, over an identical life cycle free of tax incentives and subsidies. The program combines fuel chain and vehicle models to present side-by-side comparisons of the trade-o€s o€ered by various AFV systems. It presents these results graphically, for ease of analysis and presentation. This life cycle model provides technical and economic insight into the AFV systems, without the distorting in¯uence of marketing claims. To investigate policy e€ects, users can control di€erent aspects of the model.

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

245

Results show that a consistent handful of alternative fuel and vehicle combinations o€er economically ecient trade-o€s between energy eciency and emissions. The combinations with the most consistent ecient trade-o€s currently are those powered by internal combustion engines (ICE) and based on liquid petroleum, especially reformulated gasoline (RFG). Economically ecient alternatives in the future are methanol blends in ICEs, fuel cell vehicles, and natural gas and battery electric vehicles (NGV and BEV, respectively) in niche markets, due to special considerations regarding performance and refueling. Thorough sensitivity analyses show that these results are signi®cant over a wide range of assumptions concerning the prices of the resources, the technology of the fuel production chains, and the performance of the vehicles. Realistically different assumptions about each stage in the fuel chain or vehicle models result in minor e€ects to the life cycle cost trade-o€s, but do not alter the overall conclusion. 1.3. Contributions The advantages of this life cycle model are its: (a) life cycle comprehensiveness, accounting for operating costs and changes in vehicle performance over a 12-yr life of driving; (b) level playing-®eld approach, removing taxes and subsidies from the model as completely as possible to provide a picture of how each technology might compete without laws which favor certain alternatives; (c) broad view, considering 17 AFV technologies and 23 fuel chains simultaneously. The model can be used in two ways. Analysts can evaluate the output to compare the performance of di€erent AFVs under various scenarios of technology and resource prices. They can then use the results either to guide long-term policy decisions or to test the e€ects of a proposed policy on the life cycle trade-o€s of AFVs by adjusting the implied changes in input prices and technology.

2. Model structure 2.1. General structure The program is a multiple-page Excel spreadsheet with table-formatted inputs, calculations, and a graphical display for presenting results. The user chooses parameters that represent the level of sophistication of certain major technologies that are used in each vehicle and each fuel production industry, and the percentage of each fuel that derives from each resource. The user can also modify or replace fuel production or vehicle models. The programÕs strength lies in its comprehensive, simple treatment of a broad question for policy-oriented goals, rather than in sophisticated modeling of speci®c components. Complete details are in Hackney (1997). This life cycle model is based on the fuel chain model of the ®rm Arthur D. Little (ADL, 1996), which compiled the preliminary information about the fuel chain and provided the part of the spreadsheet model that calculates fuel chain emissions and energy eciency. This core fuel chain model traces the emissions and energy eciency of the production of vehicle fuel.

246

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

Hackney (1997) extended ADL's core work with recent research on the life cycle costs of the fuel chains, and the emissions and cost performance of a wide range of AFV (see Table 1). These extensions drew upon methodology outlined in DeLuchi (1989) and Wang et al. (1993). Speci®cally, the modi®cations add a cost model for each fuel chain, several new fuel chains, and the vehicle models that simulate 12 yr of driving in di€erent types of vehicles (US DOT/FHWA, 1992). Calculations of emissions and cost from the fuel chain are expressed per gigajoule (GJ) contained in the resulting fuel, and weighted according to the amount of fuel that came from each resource. The vehicle models calculate the amount of energy in the form of the fuel used by each Table 1 Elements of life cycle modela Program step

a

1.

Resource extraction

2.

Transport of resource

3.

Conversion of resource

4.

Fuel transport

5.

Refueling

6.

Vehicle

Model structure

Cost ($/GJ)

Eciency ratio

Emissions (Kg/GJ)

DOE price projections industry studies Avg. transport distance transport mode split industry sophisticationb Plant technology plant location resource type production volume learning industry sophisticationb Avg. transport distance transport mode split industry sophisticationb Final form of fuel losses new infrastructure refueling frequency industry sophisticationb Fuel chain used common chassis distance per year data maintenance cost data avg. eciency parameter avg. emissions parameter

Of extracted, harvested or collected fuel resource Above + (km  $/km GJ)

1 ± ratio of energy used to that before processing Above times same calculation

CO; CO2 ; SO2 ; NOx ; NMHC; CH4 ; PM10 Above + same calculation

Above + ($/GJ)

Same

Same

Above + (km  $/km GJ)

Same

Same

Above + $/GJ of fuel delivered to vehicles

Same

Same

Above + (GJ/ km  km) over 12 vehicle service years

Above times eciency at converting fuel to kinetic energy

Above + tailpipe (kg/km  km) CO; CO2; NOx ; NMHC; CH4 ; PM10

Calculations for each fuel chain follow steps 1±5. Cost, eciency and emissions are expressed per unit energy and used in di€erent vehicle models in step 6. Original work from ADL (1996) is italisized. b Industry sophistication refers to emissions and energy policy that would a€ect performance in all fuel chains (e.g., coal boiler controls, diesel emissions controls, etc.).

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

247

vehicle over its lifetime, and multiplies this value by the corresponding cost or emissions per unit energy of fuel from its production chain. Adding the cost and emissions of using the vehicle to this former result from the fuel chain gives the life cycle output. The fuel chains trace the stages of fuel production through the 5 steps in Table 1: resource extraction, its transport, re®ning into fuel, transport of fuel to a retailer, and its ®nal delivery as fuel into a vehicle. Computationally, the vehicle models act like a ®nal stage of fuel processing that consumes all the energy in the fuel. There is one fuel chain per resource, per fuel, though multiple types of vehicles may use the same fuel, and the same type of vehicle may use multiple fuels. The program presents an overview of AFV performance, and makes no attempt to model market competition. The fuel chains and vehicle models are mostly independent of each other in the spreadsheets. However, the fuel chains share parameters that represent industry standards for emissions controls in various technologies such as coal boilers or diesel emissions caused by trucking. Likewise, the vehicle models share a common ``gliding'' vehicle chassis, which is used as a base for further independent modi®cations that depend on each vehicleÕs power plant and fuel. 2.2. Building blocks The extensions built into the original ADL model derive from a range of these sources. The supplementary data on the costs and e€ects of the production, distribution and use of the various fuel chains come from: (a) biomass resources (Seymour, 1992): the International Energy Agency (IEA/OECD, 1994), and the US National Renewable Energy Laboratory (US DOE/NREL, 1995a); (b) hydrogen: DeLuchi (1989); (c) RFG: the US DOE/ODIP (1996), US DOE/EIA (1994) and a fuel manual for the gasoline retailers from Downstream Alternatives (1996); (d) methanol: the US DOE/ODIEP (1989) and (Webb et al., 1990). The models for the AFVs relied on emissions and performance information from: (a) reports on life cycle greenhouse gas emissions (DeLuchi, 1991, 1993); (b) works on emissions from AFVs ± models of emissions and broader costs (Wang et al.,1993), results on natural gas vehicle emissions (Alson et al., 1990), and in-use emissions of AFVs (Gabele,1995); (c) a report on international electric vehicle policies (US Government Accounting Oce, 1994); (d) a comprehensive summary of the state of the art in vehicle batteries (Kalhammer, 1996); (e) papers on the technology of fuel cells for vehicles (Iwase and Kawatsu, 1996), and their early market opportunities (Maceda, 1996). Information for methanol direct-conversion fuel cells was based on early experimental results (Seshan, 1997; Halpert et al., 1997). Data about the costs and more information about energy eciency and emissions from AFVs come from ®eld studies of ¯eets in the United States by the Department of Energy (US DOE/EIA/OIAF, 1992; US DOE/ODIEP, 1991, 1992; US DOE/ NREL, 1995b). Data on current and possible future AFV systems come from the US National Research Council (NRC/NAS, 1990), Sperling (1990) and the US DOE's analysis of the market potential of light duty AFVs in 2000 and 2010 (US DOE/ODIEP, 1996).

248

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

2.3. Availability of model The model is available on request for non-commercial use. It is continually being updated from the original version (Hackney, 1997). 3. Model details 3.1. Costs Cost calculations refer to a steady average scale of production over the life cycle. Fuel costs include the cost of constructing new fuel processing and retail facilities. According to the published models referred to above, fuel costs are expressed as average costs of capital and operation over the service life of the fuel re®nery in dollars per unit energy of fuel delivered. Vehicle costs include purchase price, fuel cost, and maintenance. Other vehicle costs are considered to be identical across vehicle types and are ignored. The costs of using the fuel and vehicles are discounted into the future at a rate entered by the user. The default discount rate is 10% for the results shown here. The user can choose ranges of the steady state production scale of each fuel chain, parameters that represent the technology of the fuel production facilities, and other parameters relevant to extracting resources and delivering fuel. In this way, the impact on the AFV life cycle of advances in technology or advantages of scale in one fuel industry can be compared in scenarios. Representative re®nery con®gurations simplify the analysis. More exact descriptions are in fact speculative since re®neries and reforming plants can produce di€erent ratios of fuels and other valuable by-products, at di€erent capacities, depending on the state of the market, on particular regulations or on the price and quality of the input resource. The methanol, biomass, and hydrogen cost models use a capital recovery factor derived over an output stream of 30 yr to account for production costs, the depreciation of the capital investment, the opportunity cost of the capital, and economic growth. Di€erent levels of sophistication and scales of production for these fuel chains were programmed from the literature to simulate ``learning'' and other improvements in the industries. In a simulation, the user can alter the ratio of ``advanced'' to ``standard'' and ``large scale'' to ``small scale'' production facilities that supply the fuel. Costs in the remaining mature petroleum fuel market for the current and future scenarios are linear functions of resource price (e.g., Hadder, 1992) and are considered optimal (e.g., no opportunity for unit cost reductions other than reduction in resource price). The spreadsheet adds the resulting average cost of fuel produced to the cost of the amount of resource consumed, to arrive at the total cost of the fuel per unit energy contained in the fuel. Transportation costs depend on costs per unit of fuel per average unit distance transported, a ®gure that represents the density of the re®ning and refueling network. The costs of the refueling stations are functions of sales volume and geographic distribution. The model uses representative values from the ranges reported in the literature. These are added to either the cost of fuel or the cost of a vehicle. Per vehicle costs of installing and using a refueling station for various fuels over the 12-yr vehicle lifetime vary from $58 for liquid propane to $420 for cryogenic gas or battery rechargers. The energy eciency of a refueling station refers to

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

249

spillage, leakage, evaporation, and the energy used to pump, compress or cool fuels. Emissions due to the evaporation of fuels can be signi®cant relative to tailpipe emissions from a clean vehicle. 3.2. Emissions The program uses a single representative value over the life cycle for each tailpipe pollutant, in units of mass per distance. The tailpipe values chosen re¯ect standardized measurements (e.g., EPA Federal Test Procedure) for di€erent types of engines and fuels, gathered in government and industry studies (Gabele, 1995). Tailpipe emissions are scaled from the values in the literature according to the relative ratio of vehicle power in the literature and in the model. Measurements of tailpipe emissions are imprecise because many variables a€ect engine performance. A sensitivity analysis of the model inputs shows that approximately 65% of the total emissions from using a combustion vehicle come from the tailpipe. This implies that a 10% error in assuming these single value tailpipe emissions rates would a€ect the total lifetime emissions by roughly 6.5%. The user must establish a credible range of lifetime tailpipe emissions independently, and then run the life cycle model in order to ®nd a range of outputs to compare. Section 5 presents the e€ect of uncertainty in tailpipe emissions on life cycle comparison. Although the spreadsheet presents separate results from the fuel chain and the vehicles, this paper combines these results to illustrate the insight this life cycle analysis can give to national level policy makers. This provides a useful view of the competing alternatives for the purpose of an overview comparative analysis. Major advantages or concerns identi®ed by this approach should help policy makers to focus more speci®c questions. The model reports locally active pollutants including small particulate matter (PM) and the hydrocarbon and nitrogen oxide (NOx ) precursors to ground-level ozone and the formation of further liquid aerosols. These chemicals react or settle to the ground relatively quickly and thus present a more signi®cant problem closer to their source. It might thus be useful to consider the fuel chain and vehicle sources of these emissions under separate ``stationary'' and ``mobile'' source guidelines. The model also reports carbon dioxide (CO2 ) and methane (CH4 ) emissions that continue to react in the atmosphere a relatively long time so that the life cycle combination of the fuel chain and the vehicle is a preferable method of evaluation. Carbon monoxide (CO) is tracked in the program, but it is not reported in the results in Section 6 because, since it is mostly emitted at the vehicle tailpipe, its consideration in a life cycle emission adds little insight. The model converts total hydrocarbon and NOx emissions to mass and adds them to represent their mutual contribution to the formation of ozone or liquid aerosols. The particular spectrum of hydrocarbon emissions from a vehicle is normalized to a representative amount of ``hydrocarbon'', based on the Maximum Incremental Reactivity Adjustment (MIR/RAF) scale as outlined by Wang et al. (1993), and used by the California Air Resources Board (CARB). The model also adds methane and CO2 and calls them ``greenhouse'' emissions. To represent the two pollutants as ``greenhouse gases'', the model multiplies the emitted mass of the methane by 9 to simulate its higher heating e€ect than CO2 in its longer lifetime in the atmosphere (DeLuchi, 1991). The model also reports PM of 10 lm or less (PM10).

250

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

3.3. Energy eciency At each stage the energy eciency is de®ned as the energy contained per unit mass of the fuel after it has been processed through the stage, divided by the sum of the energy in the fuel or resource before and that used during the processing. Lower heating value is used as per DeLuchi (1993). No step is 100% ecient because it either takes energy to transform or transport the fuel or there are thermodynamic or mechanical losses. 3.4. Fuel chains The program allows the user to analyze 23 fuel chains which might be likely contenders for providing alternative motor fuels, and 17 types of vehicles. Table 2 presents the chains that are now available for analysis in the life cycle model and the chains used in the results presented here.

Table 2 Fuel chains available in the life cycle model of alternative fuel vehicles (percentage of fuel made from each resource for results presented here)

a

Energy resource

Transportation fuel

Percentage

Petroleum

Gasoline, M85, E85 Reformulated gasoline (RFG) Diesel Lique®ed petroleum gas (LPG ˆ propane)

100a 100a 100 50

Natural gas

Compressed, liquid natural gas (CNG, LNG) Methanol, M85 Hydrogen Lique®ed petroleum gas (LPG ˆ propane)

100 100a 90 50

Coal

Methanol, M85 Hydrogen

Cellulosic biomass (wood)

Ethanol, E85 Methanol

Maize (corn)

Ethanol, E85

100a

Various

US electricity mix (1990) Northeast electricity mix NY/NJ electricity mix Southern California electricity mix

100 ± ± ±

US electricity mix (1990) Northeast electricity mix NY/NJ electricity mix Southern California electricity mix Nuclear power Renewable electric technologies

Hydrogen

0 10 0 0

± ± ± ± ±

0

M85, E85, RFG contain constituents made from multiple resources. Major source indicated here.

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

251

3.5. Vehicles Table 3 summarizes vehicle costs and energy eciency. The internal combustion vehicles in the program use gasoline; RFG; diesel; lique®ed petroleum gas (LPG or propane); ethanol or methanol; CNG or LNG; compressed or liquid hydrogen (CH2 , LH2 ) or hydrogen hydride cells. The direct conversion fuel cell vehicles can use methanol or CH2 . The hybrid electric vehicle uses RFG. The battery electric vehicle uses the US electricity generation mix, which is 48% coalpowered. The USA mix was chosen in this paper for consistency with other fuel chains in an overall evaluation. Regional analyses should use the characteristics of local electric power plants. The eciency of each vehicleÕs drive train is represented by a normalization factor relative to the eciency of a conventional gasoline vehicle (CGV), which can either be entered by the user or calculated by the model from the fuel eciency of the base gasoline vehicle. The uncertainty of the eciency value used for the base gasoline vehicle is carried over into all vehicles by this method, so that any systematic errors have the same e€ect across vehicles. The model assumes that all vehicles are fuel ecient subcompact passenger cars of 40 hp (30 kW), driven the national average annual vehicle miles traveled (VMT). To ensure a common basis of comparison, even the diesel, LPG, and CNG vehicles in the model are passenger automobiles, though in the US most of these vehicles are heavy trucks. These fuels are however more commonly used throughout Europe and Japan in passenger vehicles. The user enters the price of a CGV, its average eciency (mpg), and a range of low and high incremental costs for each AFV. The low and high prices represent aftermarket conversion, new vehicle prices from original equipment manufacturers (OEM), and prices of ¯ex-fueled versions of the vehicles (life cycle emissions not reported here). The results reported in this paper use the Table 3 Incremental costs of alternative fueled subcompacts in the base case compared to $15 000 conventional gasoline subcompact, including battery costs for the electric vehicles Fuel or vehicle Gasoline RFG Diesel LPG CNG Methanol ICE Methanol FC LNG Hydrogen FC Hydrogen hydride Hydrogen CH2 Hydrogen LH2 Ethanol Battery electric M85 E85 Hybrid RFG/EV

OEM costs Low

High

Avg.

Relative vehicle eciency

0 0 0 200 400 0 2800 400 3600 690 800 800 0 160 0 0 4000

0 0 0 800 1000 300 147 000 800 148 200 7390 1200 1200 300 6160 300 300 6220

0 0 0 500 700 150 74 400 600 75 900 4040 1000 1000 150 3160 150 150 5110

1.00 1.00 1.00 1.10 1.10 1.15 2.33 1.10 2.67 1.29 1.50 1.50 1.40 3.56 1.00 1.00 1.60

252

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

average life cycle cost that results from the low and high cost calculations. The range of new vehicle costs provides the largest cost uncertainty in the model. These ranges may be considered a ®rst order representation of the band of uncertainty bracketing the life cycle costs for each vehicle. The cost of combustion vehicles was obtained from OEMs and conversion companies. The cost range is well known from the industryÕs years of experience with alternative combustion fuels. Variations in the price of conversion kits or OEM products represent the complexity of the engine control systems, the materials of certain fuel systems, production volume, etc. The model assigns high and low costs to experimental technologies such as fuel cell vehicles, ICE/electric hybrids, and hydrogen-fueled vehicles. This represents the wide spread of uncertainty associated with the optimism of laboratory or entrepreneurial developers, and skepticism about these predictions. The costs for direct-conversion hydrogen fuel cells, e.g., was cited by the privately funded fuel cell development ®rm, H-Power Corporation, to be $5000/kW in 1996, but falling quickly toward $200/kW (Maceda, 1996). A recent article in Business Week (3/2/1998) cites an incremental cost of $2500/kW for an indirect conversion, gasoline fuel cell vehicle at Ford, with the companyÕs goal set at $250±350/kW for the system. The cost of battery electric and hydrogen hydride vehicles is calculated automatically as the vehicle model sizes the battery and hydride tank appropriately to the required range-per-charge and according to vehicle energy eciency (DeLuchi, 1989). Table 3 shows the price increments for the results reported here. The model calculates the required size of the BEV battery in kWh (energy) from the eciency of the glider chassis and drive train, the relative weights of battery and chassis, the batteryÕs maximum depth of discharge (DOD), and the range of the electric vehicle. There is a constraint of reduced eciency with increased battery weight (DeLuchi, 1989). The model assumes that the BEV battery discharges to 80% depth at a slow rate with each use. The ``low tech'' and ``high tech'' batteries correspond to those yielding 600 and 1000 charging cycles at 80% DOD (USABC near- and longterm, respectively). Similarly, the model ®nds the size of the battery for an HEV based on its range under electric power only. At 30% DOD, it will last 10 000 cycles (Ovshinsky et al., 1996). The BEV modeled here needs 5 21 kWh batteries in 12 yr, and the HEV needs 4 8.1 kWh batteries. Each type of vehicle travels the same number of miles each year as the other vehicles. Annual VMT generally falls as the vehicle ages (US DOT/FHWA, 1992). Using a uniform annual VMT for all vehicles is consistent with the level playing ®eld approach of this research. Because the program uses an average fuel economy over the lifetime of a vehicle, declining VMT implies declining fuel costs as the vehicles age. Annual maintenance costs increase with the age of the vehicle. Actual costs for 12 yr are taken from (US DOT/FHWA, 1992). The model determines costs for AFVs by multiplying the maintenance cost of a CGV by a coecient corresponding to the AFVs' relative need for maintenance. These coecients range from 0.60 for electric drivetrain vehicles to 1.0 for combustion vehicles. Maintenance costs are less than 10% of fuel costs for most vehicles. 3.6. Miscellaneous elements The model includes the emissions and energy eciency of mining, drilling, and other resource extraction processes. It does not include the energy used in construction of the mining infrastructure. It assumes that the cost of extraction included in the resource prices.

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

253

The model represents the shipment of resources and fuels by modes and over distances based on the actual distribution of extraction sites, processing plants, and retailers in the US for each fuel and resource (US DOT/BTS, 1996; Webb et al., 1990; ADL, 1996). It accounts for electricity transport by an eciency loss in the fuel chain.

4. Base case results for 1996 4.1. Basic initialization Table 4 is a sample of an input table from the life cycle model, showing the resource prices used for the base case and the future scenarios. The crude oil, electricity, coal, and gas prices are from US DOE/EIA/OIAF (1992) and US DOE/ODIEP (1992). The prices are similar to those in other sources (Methanex, 1996; PennWell, 1996), but the DOE values were used in order to have a consistent reporting year for all resources. Corn, animal feed by-products (a source of revenue that o€sets the price of corn), and wood costs are from IEA/OECD (1994). The MTBE price is from Hadder (1992). The model calculated the diesel and gasoline costs. The base case scenario simulates current fuel production technology in the United States, and the vehicles currently available or likely to emerge in the short term. The base case gasoline vehicle costs $15 000 and is 15% ecient at turning the potential energy in its fuel into kinetic energy of the vehicle (assumed 30 mpg). Table 3 lists the relative costs and eciency of other vehicles. The life cycle costs of internal combustion vehicles that use liquid fuel are lowest, and their costs di€er only slightly. The costs of methanol and natural gas combustion vehicles are next lowest. Higher cost vehicles include ethanol and hydrogen internal combustion, battery electric, and hybrid electric vehicles. The experimental fuel cell vehicles are very expensive in the base case. The model compares life cycle emissions and energy eciency to costs in trade-o€ plots that de®ne the set of alternatives that provide the best balance between the characteristics of each type of vehicle. On each plot, ``feasibility frontier'' de®nes the most economically ecient performance, Table 4 Input table showing the estimated resource prices for the 1996 base case scenario and possible future scenarios (1996 $) 1996 base case Crude Oil Diesel Gasoline MTBE Electricity Natural gas Coal Corn Maize products H Maize products L Wood

21.00 0.65 0.62 0.85 0.07 1720.00 22.18 108.00 (200.00) (83.00) 50.00

Possible range

Units

Low

Medium

High

22.60 0.70 0.65 0.23 0.07 4000 31.6 85 Same Same Same

33.40 0.99 0.94 Same Same 4460 Same Same Same Same Same

40.20 1.19 1.13 Same Same 4650 Same Same Same Same Same

$/bbl $/gal $/gal $/gal $/kWh $/MMSCF $/ST $/MT $/MT $/MT $/MT

254

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

that is the lowest available cost for a given quantity of the characteristic (emission, energy eciency). (By construction, points representing combinations of cost and the characteristic being compared will lie on or above this frontier.) The closer an alternative is to the frontier, the more economically ecient it is. This economic eciency should not be confused with the actual energy eciency of a fuel chain and vehicle life cycle. 4.2. Greenhouse emissions Fig. 1 presents the total greenhouse emissions from using each vehicle, from both the fuel chain and the vehicle. Life cycle emissions range from 20 to 45 metric tons. Two extrapolations of the feasibility frontier show the potential of fuel cells to lower the cost of emissions improvements. The economically ecient alternatives are liquid petroleum ICEs, natural gas in ICEs and, at an extremely high cost, direct conversion fuel cell vehicles. The least expensive alternatives are conventional petroleum and LPG. LPG has 20% lower greenhouse emissions. The alcohol fuels, both pure and in 85% mixtures with gasoline, do not reduce total greenhouse emissions over conventional petroleum fuels. Compressed and liquid

Fig. 1. Best performance feasibility frontier for life cycle cost vs. total greenhouse emissions for the 1996 base case. The lower curves represent possible performance if fuel cell vehicles become market realities at current costs.

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

255

natural gas (methane) systems have similar costs to each other, and reduce greenhouse emissions 25% over petroleum systems, despite their high fuel chain emissions of methane. The more expensive compressed and liquid hydrogen fueled vehicles o€er the same low greenhouse emissions of natural gas vehicles. For these vehicles the carbon dioxide from the fossil resource is emitted from the fuel chain, not the vehicle. Electric and hybrid electric vehicles can lower greenhouse emissions about 30% from CGVs, but cost about 30% more. Both kinds of electric vehicles reduce greenhouse emissions due to their superior eciency. Fuel cell vehicles using methanol or hydrogen show potential greenhouse gas reductions of about 50% from ICEs. 4.3. Ozone precursors Fig. 2 presents ozone precursor emissions from the vehicle and fuel chain, which mutually contribute to the formation of ground-level ozone. The life cycle emissions range from 50 to 300 kg for most vehicles. The economically ecient alternatives are liquid petroleum fuels, and ICE/EV hybrids using RFG. Hydrogen and methanol fuel cell vehicles would lie on the feasibility frontier, but are extremely expensive. The diesel fuel chain does not appear on the chart because the high NOx emissions of the particular vehicle used in the model cause the data point to be plotted to the right, o€ the scale at 700 kg (see particulate emissions for explanation). RFG is the least expensive, low ozone

Fig. 2. Best performance feasibility frontier for life cycle cost vs. life cycle total ozone precursor emissions (NOx and hydrocarbon emissions) for the 1996 base case.

256

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

precursor alternative, and its performance is unmatched except for the vehicles with electric powertrains. Alcohol fuels di€er from each other by the higher NOx emissions from the ethanol fuel chain, but both emit an equal or higher life cycle amount of ozone precursor gases than gasoline. Natural gas and hydrogen perform the same as methanol for ozone precursor emissions. Battery and hybrid electric vehicles can lower ozone precursor emissions about 40% from the rough average of all other conventional drivetrains, due primarily to improved vehicle eciency, which lowers NOx , and the use of low-methane content fuels. The experimental fuel cell vehicles may o€er an 80% reduction in ozone precursors, due to their low running temperature that cuts NOx , and to their eciency. 4.4. Lifecycle PM10 emissions Fig. 3 shows these varying from almost none to almost 30 kg for most alternatives. Except when coal is used in the fuel chain, the PM emissions come primarily from fuel combustion in the engine, rather than from emissions from the fuel chain. Vehicles which do not combust fuel or which burn fuels that have relatively more energy stored as bound hydrogen, have dramatically lower PM emissions from the vehicle. The alternatives on the economically ecient feasibility

Fig. 3. Best performance feasibility frontier for life cycle cost vs. life cycle total particulate emissions for the 1996 base case.

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

257

frontier are liquid petroleum fuels and hydrogen ICEs. Methanol and hydrogen fuel cell vehicles would lie on the feasibility frontier, as well, at an extremely high cost. Battery electric and ethanol vehicles, which use coal in their fuel chains, have higher life cycle PM emissions. The high particulate emissions of the ethanol fuel chain result from using uncontrolled diesel engines for the corn harvest, and coal boilers for breaking down starches for fermentation of the corn. Points representing the ethanol particulate emissions fall far to the right of the plot scale, at 270 and 300 kg PM for E85 and E100, respectively. Ethanol vehicles emit the same amount of PM as methanol vehicles. Emissions from BEVs derive entirely from the electricity generation. The results in this paper assume that the industry standard for controlling the particulate emissions from the coal boilers is fairly poor in the base case, and signi®cantly improved in the future scenarios. All petroleum fuels have high PM emissions except for LPG, which is very clean because of its higher hydrogen/carbon ratio. The diesel tailpipe emissions modeled for this thesis use a diesel that is running lean, which causes it to emit high NOx , but low PM. Natural gas and methanol PM emissions fall near the average of emissions from petroleum fuels. The PM result for natural gas vehicles derives mainly from the tailpipe emissions value used in the base case, which is higher than many reported natural gas vehicle emissions values. However, this higher value was used because it was measured in the same test in which the other NGV tailpipe emissions were measured (Gabele, 1995), and is therefore being used in its proper context with the other emissions levels to represent a real vehicle. Hydrogen PM emissions are very low in the base case, because most hydrogen comes from natural gas. If the source of hydrogen were to change to coal, the PM emissions would rise very quickly. Fuel cell vehicles would lower PM emissions to almost nothing, because they do not combust their fuels. All of the PM emissions for these vehicles result from the fuel chains. 4.5. Lifetime energy eciency The life cycle energy is described above. As it costs more to achieve more ecient fuel and vehicle systems, the shape of the plot is reversed from the preceding plots. This means that alternatives above and to the left of the feasibility frontier are less desirable, and the range of currently impossible alternatives is in the lower right. Fig. 4 thus shows the life cycle energy eciency of each fuel processing chain and vehicle. Lifecycle energy eciency ranges from 7% for ethanol to nearly 27% for a hydrogen fuel cell vehicle. The most energy ecient alternatives are hydrogen fuel vehicles, hybrid electrics, and fuel cell vehicles. The life cycle eciency is sensitive to the vehicle energy eciency assumed in the model, as the vehicle is the least ecient ``processing'' step for all fuel chains. This value is a source of uncertainty in the analysis. For example, most AFVs can be more ecient than CGVs, but can be less ecient if poorly maintained. The vehicles on the economic feasibility frontier for life cycle eciency are those powered by liquid petroleum, perhaps methanol fuels; hybrid electric vehicles powered by RFG; and hydrogen fuel cell vehicles. The ICE hybrid is more energy ecient than a CGV because the engine is assumed to run at its optimum rpm and torque. An ICE hybrid electric could extend petroleum reserves by o€ering 33% higher fuel eciency (in this model). The life cycle energy eciency of the fuel cell vehicles derives from the high eciency of the fuel cell and the electric motor in the vehicle. The hydrogen used in fuel cell vehicles comes from a much more ecient fuel chain than the methanol. The

258

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

Fig. 4. Best performance feasibility frontier for life cycle cost vs. energy eciency for the 1996 base case. Di€erent curves re¯ect di€erent life cycle eciencies of the two fuel cell vehicles.

energy eciency of a battery electric vehicle su€ers from inecient electricity transmission and high vehicle weight. The properly tuned LPG engine is more fuel ecient than a CGV, so its life cycle energy eciency is higher despite a similar fuel chain. Unfortunately, despite LPGÕs high eciency use of its petroleum resource, its low cost, and its low emissions, the component of petroleum which can be used for LPG is very low, and resources are too limited for LPGÕs widespread use in vehicles. Conventional petroleum, natural gas fuels, and ®nally, methanol and then ethanol, are the least energy ecient fuels, in that order. Oil reserves could be used more eciently by electric hybrids. Natural gas vehicles are 50% more ecient at using the natural gas resource than methanol vehicles.

5. Sensitivity of output Errors can result from incorrect input data, model assumptions, and imprecise knowledge. Uncertainty in the output about the mean reported values compares to that of input data. For mature industries (petroleum, ICEs) trade and government publications provide reasonable

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

259

forecasts of cost, emissions, and energy with corresponding indications. However, estimates for developing technologies and market scenarios are most uncertain. For these situations the calculations cannot be checked against reality but have to be bracketed with arguments of plausibility. Sensitivities in the program were checked by changing the component models within a range of plausible uncertainty to identify parameters that most strongly in¯uence the life cycle outcomes. Table 5 summarizes the largest responses for each of the 6 components of the life cycle analysis. The most sensitive changes in output are a fraction of the size of the change in the model parameter. This re¯ects the linearity of the models. The major exception is the choice of discount rate, which makes vehicles with higher variable costs appear less expensive in current dollars. The life cycle cost of BEV and HEV are most sensitive to this parameter. In most cases only large changes in models outside the justi®ed values in the literature can a€ect the emissions, cost or energy ranking of most vehicle/fuel pairs. However, for some pairs which are nearly equal in a given measure, even small changes in the models in the table a€ect which combination emerges as superior when individual characteristics are considered alone, e.g., lowest cost or lowest NOx emitter. This observation implies that these alternatives are indistinguishable in these circumstances. However, it is more signi®cant though to view the sensitivity of the life Table 5 Sensitivity of the life cycle model, as a ratio of the percent changes in a parameter relative to the di€erence in the life cycle output of a vehicle and a gasoline ICE

a b

Model step number

Sensitivity test

Component a€ected

Life cycle e€ect: change input (%); change life cycle output (%)

1 ± resource extraction

Resource prices

Coal price Petroleum price Natural gas price

1:1 ethanol vehicles 1:1 petroleum vehicles Very small: vehicle cost weighs more

2,4 ± transport

Pipeline liquid cost Barge cost Truck cost

Non alcohol liquids All Alcohol fuels

Very weak Very weak Weak, but a€ects relative methanol/ gasoline costa

3 ± fuel production

ng ! MeOH technology and production scale NG ! H2 technology and production scale Coal boiler pollution controls

Methanol, RFG

Ethanol PM10

Weak, but a€ects relative methanol/ gasoline costa 1:1 medium cost e€ect, fuel compression or liquefaction still costly 1:1 strong individual e€ect

5 ± resource mix

NG vs: Coal ! H2 NG vs: Coal ! MeOH

CO2, PM, $

>1:1 for H2 and MeOH

6 ± vehicle model

Discount rate Relative eciency Tailpipe emissions Price

BEV, HEV ICE ICE High vehicle cost High other costs

1:3 large (future costs)b E€ect according to eciency Medium 1:0.65 Medium 1:1 Weak

H2 fuels

Life cycle methanol and gasoline ICE vehicle costs indistinguishable. Costs of HEV, BEV depend heavily on battery replacement, which appears less expensive with higher discounting.

260

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

cycle analysis as a whole, in which even large uncertainty does not change the overall conclusion that a handful of alternatives share favorable trade-o€s, and others remain technically inecient. 6. Results for possible future conditions 6.1. General scenarios The authors used the model to compare AFVs under 8 possible scenarios. These include the three sets of possible future prices for fossil resources shown in Table 3, each involving low and high rates of technological advancement in fuel chains and vehicles. Two additional analyses investigated optimistic ``best case'' technology and cost developments for electric drive vehicles. The authors selected these analyses because of the uncertainty in the cost predictions for immature technologies, and the current interest in zero emission vehicles. These analyses impacted fuel cell vehicles, battery electric vehicles, and RFG hybrid electric vehicles. The low technology case, compared to the base case, assumes an increase in the penetration of emissions controls on basic building blocks of industry, but no improvement in the eciency, emissions or cost of vehicle technology and fuel chains. This case re¯ects an emphasis on tighter emissions restrictions on stationary sources over mobile sources. The high technology case further assumes that all fuel chains improve to the maximum eciency and lowest emissions and cost possible in the model, according to the model assumptions. ``High technology'' improvements include cost advantages from an increased scale of production and increased eciency of manufacturing (learning). The vehicle model assumes that the eciency of a CGV is 17% instead of 15%; that the higher eciency and lower emissions of CGVs cost $1000 more than the low technology alternative (i.e., $16 000); and that batteries achieve the characteristics of the USABC near-term battery. Conventional internal combustion vehicles become more expensive in each scenario over the life cycle because these vehicles are sensitive to resource prices. In the case of better technology, the mature technology of conventional vehicles requires added cost to improve its energy eciency or already lowered emissions. Alcohol fueled and natural gas vehicles experience a small rise in cost and a rise in energy eciency because of the high technology improvements to the fuel chain, but because these vehicles are slightly less sensitive than CGVs to changes in resource prices, the cost rise is smaller. Hydrogen powered vehicles experience cost and emissions reductions due to improvements in fuel processing. Hybrid and battery electric vehicles are less a€ected by the changes in resource prices because hybrid electric vehicles are more fuel-ecient than conventional combustion vehicles, and because the cost of electricity barely changes in the three scenarios (US DOE/EIA/ OIAF, 1992; US DOE/ODIEP, 1992). Emissions from the production of electricity drop slightly in the model as more natural gas powered plants are built, and emissions from coal plants are controlled. Fuel cell vehicles are still handicapped in the future scenarios by the high cost of fuel cells. 6.2. Special scenarios The two additional scenarios illustrate exceptional developments that could alter conclusions derived from simple projections of future costs and technologies. Advances in fuel cells and

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

261

electric vehicles could result in high cost savings for these vehicles. In the ``best case fuel cell'' scenario, the incremental vehicle costs drop from $75 000 ($2600/kW) more than a CGV, to only $2800 more (80% of cost of CGV to represent rolling chassis, plus $200/kW). Compressed hydrogen fuel systems still add some additional cost to the vehicle. A large cost reduction in fuel cells would lower the feasibility frontier for greenhouse emissions to a revolutionary extent (Fig. 5). Fuel cell vehicles could then be a dominant solution, o€ering the highest greenhouse gas emissions reductions at a medium-level cost. Even at higher costs, fuel cell vehicles may become a competitive alternative if controlling greenhouse gas emissions becomes a priority. In the ``best case electric drive'' scenario, battery and hybrid electric vehicles use batteries meeting the USABC long-term battery cost and performance. The case also assumes that manufacturing techniques for electric vehicles have lowered the production costs of vehicles to the lowest level in Table 3. A hybrid electric would cost $4000 more than a CGV in this scenario, $1000 less than the base case increment. The cost reductions correspond to high production volume, ground-up design, and integrated electronics. In this scenario, both hybrid and battery electric vehicles lie on the feasibility frontier for cost/greenhouse emissions trade-o€s (Fig. 6).

Fig. 5. Best performance feasibility frontier for life cycle cost vs. greenhouse equivalent emissions for the case of the best development of fuel cells.

262

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

Fig. 6. Best performance feasibility frontier for life cycle cost vs. greenhouse gas emissions, for the case of the best development of electric vehicles. Both improved hybrid and battery electric vehicles arrive at the feasibility frontier for the three future fuel price scenarios.

7. Model implications 7.1. Performance of alternative fuel vehicles AFVs have potential advantages over petroleum vehicles in two areas: reduced emissions from the tailpipe and fuel chain, and less reliance on imported oil for vehicle fuel. The cost trade-o€s for these bene®ts in the level-playing ®eld model results do not identify a clear dominant alternative that would compel a nationwide policy to change motor fuel in the near future. The important conclusions for each fuel are: (a) The best performing, all-round alternative for a low cost is RFG. (b) Natural gas vehicles may reduce greenhouse and PM emissions with an economically ecient trade-o€. (c) Slightly higher cost alcohol blends or natural gas vehicles may yield lower ozone precursor emissions than RFG, but these alternatives do not appear to be as economically ecient. (d) LPG is a high quality fuel that is used on a widespread basis in the United States already, but its reserves are too low for it to replace other crude oil derived fuels.

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

263

(e) Hydrogen fuels could substantially reduce particulate, greenhouse gas, and CO emissions at high cost, but would not change life cycle ozone precursor emissions from RFG levels. (f) The high cost battery electric vehicles could emit an equivalent amount of PM from stationary sources as conventional petroleum vehicles do from the tailpipe, but other life cycle emissions and energy consumption are lower. (g) Electric hybrid vehicles powered by RFG perform and cost about the same as BEVs, but emit as much tailpipe CO as cheaper alcohol or natural gas vehicles, and still use oil. (h) Research claims for fuel cell vehicles promise to reduce or eliminate all of the major pollutants at a competitive cost, though their cost is currently extremely high. Several fuel alternatives could provide economically ecient ways to reduce certain emissions. Comparing in future scenarios the mid-range of high and low life cycle costs of every alternative, the base case results hold in general through the three sets of resource prices and two cases of technology development. Changes in the future scenarios suggest that compressed hydrogen combustion vehicles could become economically ecient alternatives for reducing greenhouse gases and ozone precursor emissions, and ethanol fuels could bene®t from fuel chain improvements to become economically ecient at reducing greenhouse gas emissions. 7.2. Policy implications The nationÕs dependence on personal motor vehicles and freight trucks suggests that prudent long-term national policy should prepare strategies for two possible situations: the need to reduce vehicle emissions in the face of continued VMT growth, and an eventual constraints on oil supply. Based on the life cycle model, a coherent strategy would continue to use petroleum while emphasizing low tailpipe emissions, continue the enforcement of regional air quality standards, and take steps to prepare nationally for a gradual introduction of natural gas as the main alternative transportation fuel resource. Both natural gas and fuels containing liquid derivatives from natural gas (RFG, M85 or M100) should be the primary fuels, and used in internal combustion vehicles. Where it is possible to use vehicles in conjunction with centralized refueling locations, planners should look at natural gas as a fuel, and at battery electric and hydrogen vehicles as eventual alternatives if there is pressure for very low emissions. A mainstream ¯eet of fuel cell vehicles that use liquid fuels should be a realistic long-term goal. These suggestions recognize the superior trade-o€s o€ered by RFG and the availability of crude oil, and recommend continuing its use in high pollution regions as a low cost method of reducing emissions. However, they also recognize the desirability of avoiding multiple incompatible fuel systems or changes in fuel systems over time by emphasizing the continued use of liquid fuel for the mainstream consumer now and in the proximate future. According to this analysis, an early step would be to modify the existing infrastructure for gasoline storage and refueling to make it compatible with alcohol across the country. Methanol is the likely future ``alcohol'' fuel for the majority of the country. The nation is unlikely to subsidize the widespread use of ethanol even though it has regional appeal in the midwest. This recommendation is consistent with the commitment, in the Clean Cities Program of the CAAA90, to establish ``clean corridors'' between cities supported by refueling stations for clean burning alternative fuels. Whether and where alcohol fuels are actually adopted will depend on emissions

264

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

requirements, which is a local consideration after the CAAA90, and on relative fuel prices, which is partly a matter of national policy. Pursuant to the CAAA, the federal government has created opportunities for AFVs by regulating tailpipe and regional pollutants and by supporting, with EPACT, the conversion of certain ¯eets to AFVs, at least through 2004. Similar controls on emissions and fuel consumption will be necessary beyond 2004 for AFVs to enjoy a market in the US. The cooperative development of markets by inter-industry partnerships o€ers the best chance for introducing and sustaining new vehicle technologies. Some DOE projects, like the US Advanced Battery Consortium (USABC), the Partnership for a New Generation of Vehicles (PNGV) or Clean Cities Program, have taken steps in this direction. The federal government coordinates industry and municipal initiatives in these programs to encourage consistent e€orts. The clearest incentive to encourage industry to o€er AFVs to consumers may be an atmosphere of stable government commitment to uphold policies that favor the environmental and energy security advantages of these technologies. 8. Conclusion The life cycle model of AFV provides a consistent way to compare these technologies, based on a level playing ®eld without subsidies and assuming standards of use. It is based on Excel spreadsheets and can be easily used by almost anyone. The current version uses data based on the best available data, but is not tied to these parameters. Users can modify the inputs to exploit more recent data and explore any scenario they might desire. The model is available on request for non-commercial use. Using the model as outlined in the paper shows the range of possible alternatives for the future vehicle ¯eet in the US. This analysis did not test some signi®cant fuel/vehicle combinations, and it is possible that one of them may turn out to be better suited to replace current vehicles. However, vehicles based on petroleum fuels using some form of internal combustion engine are likely to dominate the automotive ¯eet for the conceivable future. The policy recommendations following from the life cycle analysis introduce change gradually. They recognize the uncertainty of both of the inputs and outputs of any model. In these circumstances a narrow policy that promotes a particular technology based on this or a similar study would be a mistake. Instead, modest steps taken now to prepare for a number of future alternatives, along with coordinated research by stakeholders in the markets, o€er less risk with more chance for a technically superior solution (de Neufville, 1997). Acknowledgements Several sources have provided important intellectual and ®nancial support. The major one has been the Ford/MIT China Project that focused on life cycle costs of automotive fuels in China (MIT, 1997). Dr. Charles Stokes, the US Fulbright Scholarship, the American Methanol Institute and the MIT Technology and Policy Program provided additional moral and tangible support. The authors gratefully acknowledge the reviewers' helpful suggestions.

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

265

References ADL (Arthur D. Little), 1996. Integrated Fuel Chain Analysis Model (IFCAM). User Instructions for Version 3.1, Filename IFCAMV31.xls. Report to Ford Motor Company, 27 February, Cambridge, MA, USA. Alson, J.A., Adler, J.M., Baines, T.M., 1990. Motor vehicle emission characteristics and air quality impacts of methanol and compressed natural gas. In: Sperling, D. (Ed.), Alternative Vehicle FUELS: An Environmental and Energy Solution. Quorum Books, CA, USA. DeLuchi, M.A., 1989. Hydrogen vehicles: an evaluation of fuel storage, performance, safety, environmental impacts, and cost. International Journal of Hydrogen Energy V14 (2), 81±130. DeLuchi, M.A., 1991. Emission of Greenhouse Gases from the Use of Transportation Fuels and Electricity. Argonne National Laboratory, Energy Systems Division, IL, USA. DeLuchi, M.A., 1993. Emission of Greenhouse Gases from the Use of Transportation Fuels and Electricity: Appendices. Argonne National Laboratory, Energy Systems Division, IL, USA. Downstream Alternatives, 1996. Changes in Gasoline III. Auto TechniciansÕ Quality Guide. Bremen, IN, USA, 1996 update. Gabele, P., 1995. US Environmental Protection Agency, exhaust emissions from in-use alternative fuel vehicles. Journal of Air and Waste Management Association 45 (October), 770±777. Hackney, J., 1997. Alternative fuel vehicle policy: A life cycle analysis tool for emissions, costs and energy eciency with an application to the United States. Master of Science thesis, Technology and Policy Program, MIT, Cambridge, MA. Hadder, G.R., 1992. Future re®ning impacts of the Clean Air Act Amendments of 1990. Energy V17 (9), 857±868. Halpert, G., Marsh, T., Giner, J., Kosek, J., 1997. The direct methanol liquid-feed fuel cell. Jet Propulsion Laboratory, Electric Power Section, January. Overhead Slide Copies Presented by R. Lewis at MIT in February, Cambridge, MA, USA. IEA/OECD (International Energy Agency/Organization of Economic Cooperation and Development), 1994. Biofuels, IEA/OECD Energy and Environment Policy Analysis Series, Paris, France. Iwase, M., Kawatsu, S., 1996. Electrocatalysts for Polymer Electrolyte Fuel Cells. In: Roller, D. (Ed.), 29th International Symposium on Automotive Technology and Automation (ISATA) Proceedings. Electric, Hybrid, and Alternative Fuel Vehicles. Toyota Motor Corporation, Japan, Automotive Automation, Ltd., Croydon, UK. Kalhammer, F.R., 1996. Batteries for CaliforniaÕs Zero Emission Electric Vehicle Program. In: Roller, D. (Ed.), 29th International Symposium on Automotive Technology and Automation (ISATA) Proceedings. Electric, Hybrid, and Alternative Fuel Vehicles, Electric Power Research Institute (EPRI), USA, Automotive Automation, Ltd., Croydon, UK. Maceda, J.P., 1996. Costs, performance, delivery and competitiveness of renewable and logistic fueled fuel cell hybrid, heavy vehicles. In: Roller, D. (Ed.), 29th International Symposium on Automotive Technology and Automation (ISATA) Proceedings. Electric, Hybrid, and Alternative Fuel Vehicles, H-Power Corporation, USA, Automotive Automation, Ltd., Croydon, UK. Methanex, 1996. Annual Report 1995. Vancouver, Canada. NRC/NAS (National Research Council/National Academy of Sciences), 1990. Fuels to Drive our Future. National Academy Press, Washington DC, USA. de Neufville, R., 1997. Alternative fuels for tomorrowÕs cars. In: Proceedings of the World Methanol Conference. Tampa, FL, 1997, pp. 199±214. Ovshinsky, S.R., Stempel, R.C., Dhar, S., Fetcenko, M.A., Gi€ord, P.R., Venkatesan, S., Corrigan, D.A., Young, R., 1996. Ovonic NiMH batteries technology ± advanced technology for electric vehicle and hybrid electric vehicle applications. In: Roller, D. (Ed.), 29th International Symposium on Automotive Technology and Automation (ISATA) Proceedings. Electric, Hybrid, and Alternative Fuel Vehicles, Ovonic Battery Company, Inc., USA, Automotive Automation, Ltd., Croydon, UK. PennWell Publishing, 1996. Oil and Gas Journal Databook 1996. Tulsa, OK, USA. Seshan, P.K., 1997. System engineering for direct methanol fuel cell systems. Jet Propulsion Laboratory, California Institute of Technology, March Version of Overhead Slide Copies, Pasadena, CA. Seymour, A., 1992. Re®ning and Reformulation: The Challenge of Green Motor Fuels. Oxford Institute for Energy Studies. Aldgate Press, Oxford, UK.

266

J. Hackney, R. de Neufville / Transportation Research Part A 35 (2001) 243±266

Sperling, D. (Ed.), 1990. Alternative Transportation FUELS: An Environmental and Energy Solution. Quorum Books, USA. Transportation Research Board, 1997. E€ects of transportation on energy and air quality. Transportation Research Record 1587, Washington DC. US DOE/EIA (US Department of Energy/Energy Information Administration), 1994. Assessment of RFG, vol. 2, US DOE Energy Information Administration, Oce of Oil and Gas, September 29, Washington DC, USA. US DOE/EIA/OIAF (US Department of Energy/Energy Information Administration/Oce of Integrated Analysis and Forecasting), 1992. Annual Energy Outlook 1992 with Projections to 2010, Washington DC, USA. US DOE/NREL (US Department of Energy/National Renewable Energy Laboratory), 1995a. Biofuels for transportation: the road from research to the marketplace. NREL Golden, CO, USA. US DOE/NREL (US Department of Energy/National Renewable Energy Laboratory), 1995b. Clean ¯eet ®nal report, vol. 7, Vehicle emissions. DOE, NREL, DOE/CH/10093-T25, Golden, CO, USA. US DOE/ODIEP (US Department of Energy/Oce of Domestic and International Energy Policy), 1989. Assessment of the costs and bene®ts of ¯exible and alternative fuel use in the US transportation sector: Report 3, Methanol production and transportation costs, November, Washington DC, USA. US DOE/ODIEP (US Department of Energy/Oce of Domestic and International Energy Policy), 1991. Assessment of the costs and bene®ts of ¯exible and alternative fuel use in the US transportation sector: Report 7, Environmental, health, and safety concerns, October, Washington DC, USA. US DOE/ODIEP (US Department of Energy/Oce of Domestic and International Energy Policy), 1992. Assessment of the costs and bene®ts of ¯exible and alternative fuel use in the US transportation sector: Report 10, Analysis of alternative-fuel ¯eet requirements, May, Washington DC, USA. US DOE/ODIEP (US Department of Energy/Oce of Domestic and International Energy Policy), 1996. Assessment of the costs and bene®ts of ¯exible and alternative fuel use in the US transportation sector: Report 14, Market potential of alternative fueled light duty vehicles in 2000 and 2010, January, Washington DC, USA. US DOT/BTS (US Department of Transportation/Bureau of Transportation Statistics), 1996. Transportation statistics. Annual report 1996, Washington DC, USA. US DOT/FHWA (US Department of Transportation/Federal Highway Administration), 1992. Cost of owning and operating automobiles, vans, and light trucks, 1991. FHWA, Oce of Highway Information Management, Washington DC, USA. Jack Faucett Associates, Bethesda, MD, USA. US Government Accounting Oce, 1994. Report on electric vehicles ± likely consequences of US and other nationsÕ programs and policies. GAO/PEMD-7-95, December, Washington DC, USA. Wang, Q., Sperling, D., Olmstead, J., 1993. Emission control cost-e€ectiveness of alternative vehicles. In: Future Transportation Technology Conference, SAE, June 14, ANL/ES/CP ± 80072. Webb, R.F., Moyer, C.B., Jackson, M.D., 1990. Distribution of natural gas and methanol: costs and opportunities. In: Sperling, D. (Ed.), Alternative Transportation FUELS: An Environmental and Energy Solution. Quorum Books, USA.