10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 1 FTF Coordinating Subcommittee Meeting Model Structure Discussion Deanne Short October 12,

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 2 Modeling Methodology Overview Inputs from Study Step 1 (a) Vehicle technology costs Vehicle attributes Fuel and infrastructure technology costs Fueling infrastructure attributes Biofuel supply curves Other inputs AEO2010 oil price scenarios GHG factors for fuels Fueling infrastructure cost analysis a)Inputs selected from ranges of values defined by the literature reviewed in Step 1 of the Study b)Determines vehicle attributes (mpg, vehicle cost) based on fuel and vehicle technology costs consistent with payback criteria in VCM c)Estimates market share of new light duty (LD) vehicle sales based on vehicle attributes and vehicle/fuel costs d)Tracks LD vehicle stocks, stock fuel efficiency, fuel consumption (by type), fuel expenditures, VMT and GHG emissions by year Outputs for Analysis Fuel Consumed by Type per Year GHG Emissions (tCO2/year) Fleet Share of Vehicles by Type Vehicle Cost of Fleet (cents/mile) Fuel Cost (cents/mile) Dispensing Infrastructure Investment Vehicle Miles Traveled Vehicle Choice Model (c) VISION Tool (d) Vehicle Attribute Model (b) Vehicle & Fuel attributes Mkt. share by year for new sales Fuels & Infra cost analysis

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 3 Choice Variables Selected Under Various Assumptions Vehicle costs and fuel efficiencies were posited under combinations of cost assumptions –Vehicle technology high and low cost functions Covers aerodynamics, tires, light weighting, etc. used by all Subgroups For liquid ICE vehicles, includes power train and hybridization –Specific technology high and low cost functions For electric, includes energy batteries For CNG, includes fuel storage, engine systems and production pathways For fuel cells (FCEV), includes fuel cell system, fuel storage and power battery –Specific dispensing high and low cost functions which were factored into the fueling or charging costs For electric, vehicle costs and fuel efficiencies were posited assuming residential charging alone and residential plus public charging For FCEV, vehicle costs and fuel efficiencies were posited assuming high and low dispensing cost estimates, which were incorporated into the gasoline gallon equivalent cost of fueling For CNG, vehicle costs and fuel efficiencies were posited assuming fuel prices that linked gas feedstock prices to oil prices, and a mix of dedicated new and upgraded fuel dispensing infrastructure 3

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 4 Diversity in Studies Led to a Set of 384 Possible Future States The diverse cost estimates in published data led the Subgroups to posit multiple sets of vehicle costs and fuel efficiencies under 3 fuel or charging cost scenarios derived from the AEO2010 oil price cases Evaluating the impact of the inputs required analyzing 64 combinations of the inputs under each of the 3 oil price scenarios and under the possibility of high or low biofuel availability To reflect uncertainty in outcomes, inputs were mapped to outputs using VCM and VISION Independent Subgroup inputs for each oil price case Possible combinations 4 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 5 Integration Tool: Vehicle Choice Model (1 of 3) Vehicle choice model (VCM) is a spreadsheet tool that applies a system of nested multinomial logit (NMNL) functions to calculate new vehicle market shares across fuel/vehicle options. The coefficients and functional forms are taken from AEO VCM estimates market share based on the following attributes –Price of vehicle –Fuel cost of a mile (price of gasoline gal equivalent/mpg, so input fuel price and mpg) –Range (distance between refueling) –Acceleration (seconds 0-60mph; function of hp & wt) –Luggage space (% of Liquid ICE) –Maintenance cost –Make/model Availability (variety of models w/tech) –Fuel availability (% fueling stations w/particular fuel) –Home refueling option for EVs Attributes are weighted by coefficients that reflect the marginal (negative) dollar value of an attribute. The sum of the weighted attributes is a generalized cost of the fuel/vehicle option 5

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 6 Integration Tool: Vehicle Choice Model (2 of 3) Only vehicle price and fuel cost of a mile are being varied across fuel/vehicle options. –Other attributes are set equal to the reference case conventional values. –The implicit assumption is that in the long run, at commercial volumes, the other attributes will equalize across technology pathways. –The analysis will identify whether an alternative technology is competitive on a pure vehicle and fuel cost basis in the long run Outputs of the model do not provide insights on the transition to alternative technologies. –New vehicle market shares in each year are mature commercial volumes. This follows from equalizing the fuel availability and make/model availability attributes, along with fuel cost assumptions built on fully burdened dispensing infrastructure. –The study’s discussion on transition may use insights from the model by altering these attributes. –VCM is also capable of imposing upper bound constraints on market shares by specifying s-curve parameters. S-curves are not active constraints in the most recent model runs.

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 7 Integration Tool: Vehicle Choice Model (3 of 3) For consistency with VCM, the vehicle attribute model minimizes the cost of driving over 47,500 miles. –The VCM coefficients on vehicle cost (C v ) and fuel cost of a mile (C f ) imply incremental vehicle cost increases (  vc) must be offset by lower fuel cost of a mile (  fc) over 47,500 miles for no change in the generalized cost of the fuel/vehicle option. f = C v * vc + C f * fc + … C v *  vc = C f *  fc for  f=0  vc = C f /C v *  fc As an example, the vehicle cost coefficient for small cars is and the fuel cost coefficient is The ratio, C f /C v, equals The coefficients for the other classes are different, but the ratios are all

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 8 VCM to VISION The market shares, fuel mpg, fuel cost per gge, and percent splits for multi-fuel vehicles from VCM are entered into VISION case by case –VCM aggregates the 5 size class market shares and fuel mpgs into autos and light truck VISION calculates the impact of the technology pathway mix on fuel consumption, GHG and VMT. –VISION contains fleet assumptions, VMT and carbon coefficients –A vehicle value calculation was added to VISION to get a fleet total cost of driving We loaded VISION with the high and low oil price cases from AEO 2010, thus there are 3 VISION and VCM files in our set of tools

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 9 Back-up

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 10 Modeling Flow Detail (Light Duty Fleet) Vehicle Attribute Model Infrastructure Spreadsheet Vehicle Choice Model (VCM) VISION Tool (1 st iteration) High/Low Retail Fuel Cost ($/GEG) High/Low Specific Vehicle Tech. Cost Common Vehicle Tech. Attributes All vehicle attrib., mpg, vehicle cost % fuel split for multi-fuel vehicles Mkt. shares of new autos & LT by tech; Fuel economy of new autos & LT Assumes “0” biofuel in 1 st iteration Biofuel Volumes Calc. Fuel volumes (BTU/year) VISION Tool (2 nd iteration) Gasoline / Diesel Price in Case Biofuel volume; “XX” in Bxx Blend GHG emissions by year VMT by Vehicle Technology BTU by fuel type Fleet vehicle & fuel cost of driving Fuel efficiency of fleet Infrastructure Accounting Infrastructure portion of retail cost ($/GEG) max fuel consumption (GEG) Total Infrastructure cost (2008 $ billions) Biofuel Scenario (High/Low) as of 10/3/11 Cost of new vehicles Fuel cost of driving of new vehicles

10/3/2011 DRAFT – DO NOT CITE OR QUOTE For NPC Study Discussion Only 11 S-curve specifications