H 2 NextSTEPS Sustainable Transportation Energy Pathways Incorporation of Consumer Demand in Energy Systems Models and their Implications for Climate Policy Analysis June 19-21, 2013 32 nd International Energy Workshop, Paris, France Kalai Ramea, Christopher Yang, Sonia Yeh, David Bunch, Joan Ogden, www.steps.ucdavis.edu
Outline Background Mo5va5on COCHIN- TIMES (COnsumer CHoice INtegra5on in TIMES) Scenarios and Preliminary results Future Work
Background 4E (Engineering- Economic- Environment- Energy) models like MARKAL/TIMES are widely used for transi5on scenarios for mul5disciplinary subjects Iden5fy cost- effec5ve paqerns of resource use and long- term technological deployments Powerful tool for policy analysis for the energy system: Policy scenarios If- then scenarios Sensi5vity analysis Very rich in represen5ng supply- side technological details But, typically behavioral elements are represented much more simplis5cally.
Motivation System- engineering models typically assume society is homogenous, i.e. there is only one decision- maker at the societal level Consumer behavior cannot be ignored in system- wide modeling! It is a cri5cal aspect of policy and decision making 4E Models 4E models typically are linear op2miza2on models Winner takes all phenomenon (least cost op2miza2on)
Motivation Typically, consumer choice has been modeled using non- linear simula5on approaches The objec5ve of this project is to develop a bridging approach to bring in consumer behavioral parameters, to the linear programming framework of TIMES 4E models (with consumer preferences)
Motivation Why transportation sector? United States is one of the leading emiqers with transporta5on sector cons5tu5ng about 28% of emissions in the year 2011 59% of energy use comes from light- duty vehicles in the US within the transporta5on sector Consumer choice is very important in light- duty vehicle adop5on The policy analysis framework therefore has to be closer to the real- world for effec5ve implementa5on Source: US Environmental Protection Agency, 2011
Current-generation bottom-up models experience several limitations with regard to the inclusion of consumer behavior, particularly in transportation. - Andreas Schäfer Source: Schäfer, Andreas. "Introducing Behavioral Change in Transportation into Energy/Economy/ Environment Models." Draft Report for Green Development." Knowledge Assessment of the World Bank, 2012.
Consumer Choice Models Consumer choice theory in microeconomics states that a ra5onal consumer would have preference over a product in such a way that it maximizes his overall u5lity Discrete Choice Modeling (DCM) approach is specifically designed to capture the discrete preferences of consumers It is a simula5on model (non- linear) Typically follows a logit or nested- logit modeling approach The output of a discrete choice model is the probability at which a consumer is willing to make a given choice
Desirable qualities in an integrated model Objective: How can we combine the strengths of choice and energy modeling approaches? Be able to capture qualita5ve parameters as part of decision- making process Be able to include consumer heterogeneity on the demand- side Be compa5ble with the linear- op5miza5on framework of TIMES
Proposed Integrated Approach Integra5on of a choice model MA 3 T (Market Alloca5on of Advanced Automo5ve Technologies) and TIMES model MA 3 T is a nested mul5nomial logit model developed by Oak Ridge Na5onal Laboratory Na5onwide model (9 regions in the US) Based on NHTS (Na5onal Household Survey) and historical vehicle sales data 20 light- duty car technologies, 20 light- duty truck technologies Has 1458 consumer groups na5onwide segregated based on region, agtude, driving behavior, and charging infrastructure Input parameters are vehicle technology parameters, fuel prices, policies and other structural aqributes such as fuel availability
Disutility Cost MA 3 T model generates a cost term called generalized cost that has the direct and indirect cost components Direct costs: Vehicle prices, fuel costs Indirect costs or disu%lity cost components: Range Anxiety cost Refueling sta5on availability Model availability New technology risk premium Towing capability
COCHIN-TIMES model COnsumer CHoice INtegra5on in TIMES Illustra5ve tool that represents only light- duty car technologies for the California region Does not predict the future, but a learning tool to simulate how different parameters might affect vehicle purchase choices 20 light- duty car technologies, 27 consumer groups Vehicle prices, fuel efficiency and fuel prices are obtained from AEO 2012 data
Consumer Demographics The vehicle miles traveled end- use demand is divided between these groups based on Census data and percentage alloca5ons of driving behavior and agtude profiles. Each group will further be divided into 25 instances or clones. A random disturbance addi5ve term that follows cumula5ve extreme value func5on will be added to each instance in the group. This is included to capture the varia5ons of preferences in each segment.
Vehicle Technologies Gasoline Diesel Natural Gas Conven5onal Hybrid Hydrogen ICE Fuel Cell Gasoline Fuel Cell Extended Range EV 10- mile range CD 20- mile range CD 40- mile range CD BaQery Electric Car 100- mile range 150- mile range 200- mile range *ICE = Internal Combus5on Engine *CD = Charge Deple5ng Range
Disutility Cost profiles of vehicle technologies
Schematic of a typical optimization Model Fuel Cost g e h E Vehicle Cost LDV Demand Tech 1 Tech 2 Tech 3 Op5mal Technology Tech 4 System
Schematic of COCHIN-TIMES
Scenarios Reference Case Scenario No constraints Subsidy Scenario $5000 per vehicle government subsidy for electric, fuel cell and natural gas vehicles Carbon Tax Scenario Carbon intensity profiles are given for the fuels Carbon tax starts from the year 2010 @ $10/ton and increases linearly un5l 2035 @ $100/ton
Results: Reference Case (Percentage of Vehicle Sales)!"#$"%&'(")*+),"-.$/")0'/"1) $!!"#,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"#!"# Diesel %!!)# %!!*# %!!+# %!!,# %!$!# TIMES Model (No Consumer Choice) %!$$# %!$%# %!$&# %!$'# %!$(# %!$)# %!$*# %!$+# Plug-in 10-mile %!$,# %!%!# %!%$# %!%%# %!%&# %!%'# %!%(# %!%)# %!%*# %!%+# %!%,# %!&!# %!&$# %!&%# H 2 ICE %!&&# %!&'# %!&(# -.-/#'!# -.-/#%!# -.-/#$!# 01#234567#'!# 01#234567#%!# 01#234567#$!# 8%#91-# 04:3#1:33# -3:;<=6;#%(!# -3:;<=6;#$(!# -3:;<=6;#$!!# >?#234567#'!# >?#234567#%!# >?#234567#$!# @>#8AB=6C# D6:E:3#8AB=6C# >FEG367:#8AB=6C# @F<4=F3#>FE# D6:E:3# >FEG367:#!"#$"%&'(")*+),"-.$/")0'/"1) $!!"#,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"#!"# %!!)# %!!*# %!!+# %!!,# %!$!# COCHIN- TIMES Model FCV H2 ICE G.Hybrid Gasoline %!$$# %!$%# %!$&# %!$'# %!$(# %!$)# %!$*# %!$+# %!$,# %!%!# %!%$# %!%%# %!%&# Plug-in %!%'# %!%(# Diesel TIMES model investments follow 'winner takes all' phenomenon (in this conceptual model, there are no market constraints). %!%)# %!%*# %!%+# %!%,# %!&!# %!&$# %!&%# %!&&# %!&'# %!&(# -.-/#'!# -.-/#%!# -.-/#$!# 01#234567#'!# 01#234567#%!# 01#234567#$!# 8%#91-# 04:3#1:33# -3:;<=6;#%(!# -3:;<=6;#$(!# -3:;<=6;#$!!# >?#234567#'!# >?#234567#%!# >?#234567#$!# @>#8AB=6C# D6:E:3#8AB=6C# >FEG367:#8AB=6C# @F<4=F3#>FE# D6:E:3# EREVs >FEG367:# COCHIN- TIMES investment decisions are far diverse, mainly dominated by gasoline cars, followed by gasoline hybrids and gasoline plug- in cars in the later years. No Supply restric5ons: Faster penetra5on of plug- in hybrids since currently there are no manufacturer supply limita5ons in the model.
Results: Subsidy Case (Percentage of Vehicle Sales)!"#$"%&'(")*+),"-.$/")0'/"1) $!!"#,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"#!"# Diesel %!!)# %!!*# %!!+# %!!,# %!$!# %!$$# %!$%# %!$&# %!$'# TIMES Model %!$(# %!$)# %!$*# %!$+# %!$,# %!%!# %!%$# %!%%# Plug-in 10-mile %!%&# %!%'# %!%(# %!%)# %!%*# %!%+# %!%,# %!&!# %!&$# H 2 ICE %!&%# %!&&# %!&'# %!&(# -.-/#'!# -.-/#%!# -.-/#$!# 01#234567#'!# 01#234567#%!# 01#234567#$!# 8%#91-# 04:3#1:33# -3:;<=6;#%(!# -3:;<=6;#$(!# -3:;<=6;#$!!# >?#234567#'!# >?#234567#%!# >?#234567#$!# @>#8AB=6C# D6:E:3#8AB=6C# >FEG367:#8AB=6C# @F<4=F3#>FE# D6:E:3# >FEG367:#!"#$"%&'(")*+),"-.$/")0'/"1) $!!"#,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"#!"# %!!)# %!!*# %!!+# %!!,# %!$!# COCHIN- TIMES Model H2 ICE G.Hybrid Gasoline %!$$# %!$%# %!$&# %!$'# %!$(# %!$)# %!$*# %!$+# %!$,# %!%!# %!%$# %!%%# %!%&# Plug-in NG %!%'# Diesel %!%(# %!%)# %!%*# %!%+# %!%,# %!&!# %!&$# %!&%# %!&&# %!&'# %!&(# EREVs -.-/#'!# -.-/#%!# -.-/#$!# 01#234567#'!# 01#234567#%!# 01#234567#$!# 8%#91-# 04:3#1:33# -3:;<=6;#%(!# -3:;<=6;#$(!# -3:;<=6;#$!!# >?#234567#'!# >?#234567#%!# >?#234567#$!# @>#8AB=6C# D6:E:3#8AB=6C# >FEG367:#8AB=6C# @F<4=F3#>FE# D6:E:3# >FEG367:# $5000 per vehicle subsidy is given for electric, natural gas and fuel cell vehicles. The subsidy starts in the year 2015 and lasts 5ll 2019. COCHIN- TIMES: there is penetra5on of the subsidized technologies in the order of 2% to 5% in the specified years. Lowering the investment costs of these technologies through subsidies puts them in a set of more probable technologies to be chosen in the model.
Results: Carbon Tax Case (Percentage of Vehicle Sales) TIMES Model COCHIN- TIMES Model D.Hybrid H2 ICE FCV EREVs!"#$"%&'(")*+),"-.$/")0'/"1) $!!"#,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"#!"# %!!)# Diesel %!!*# %!!+# %!!,# %!$!# %!$$# %!$%# %!$&# %!$'# %!$(# %!$)# %!$*# %!$+# %!$,# %!%!# %!%$# %!%%# %!%&# Plug-in 10-mile %!%'# %!%(# %!%)# %!%*# %!%+# %!%,# %!&!# %!&$# %!&%# %!&&# %!&'# %!&(# -.-/#'!# -.-/#%!# -.-/#$!# 01#234567#'!# 01#234567#%!# 01#234567#$!# 8%#91-# 04:3#1:33# -3:;<=6;#%(!# -3:;<=6;#$(!# -3:;<=6;#$!!# >?#234567#'!# >?#234567#%!# >?#234567#$!# @>#8AB=6C# D6:E:3#8AB=6C# >FEG367:#8AB=6C# @F<4=F3#>FE# D6:E:3# >FEG367:#!"#$"%&'(")*+),"-.$/")0'/"1) $!!"#,!"# +!"# *!"# )!"# (!"# '!"# &!"# %!"# $!"#!"# G.Hybrid Plug-in Gasoline Diesel %!!)# %!!*# %!!+# %!!,# %!$!# %!$$# %!$%# %!$&# %!$'# %!$(# %!$)# %!$*# %!$+# %!$,# %!%!# %!%$# %!%%# %!%&# %!%'# %!%(# %!%)# %!%*# %!%+# %!%,# %!&!# %!&$# %!&%# %!&&# %!&'# %!&(# -.-/#'!# -.-/#%!# -.-/#$!# 01#234567#'!# 01#234567#%!# 01#234567#$!# 8%#91-# 04:3#1:33# -3:;<=6;#%(!# -3:;<=6;#$(!# -3:;<=6;#$!!# >?#234567#'!# >?#234567#%!# >?#234567#$!# @>#8AB=6C# D6:E:3#8AB=6C# >FEG367:#8AB=6C# @F<4=F3#>FE# D6:E:3# >FEG367:# The new technology investments do not differ much from reference case scenario in TIMES model, except for the year 2035, where all new sales are diesel hybrid cars. COCHIN- TIMES: There is a reduc5on in investment for gasoline cars in the years arer 2020. In the year 2035 the investment in gasoline car reduc5on is about 10% from the reference case.
Summary COCHIN- TIMES model illustrates consumer demand response for vehicle choices for light- duty car sector. This is a significant improvement from the exis5ng methodologies of energy models, especially in TIMES model where decisions are made at a societal level. This methodology can act as a good learning tool to understand the 'costs' or disu5lity of the barriers in vehicle technology adop5on and how policies instruments can be designed accordingly. Scenario analysis contrasts COCHIN- TIMES with TIMES and the 'winner takes all' approach is avoided. There is a lot of room for improvement to accommodate deficiencies arising from MA 3 T model assump5ons.
Future Tasks Decompose the components of u5lity costs represented in the model. And beqer represent each through learning curves, calibra5on with other simula5on models, etc. Include more consumer demographic groups to accommodate more behavioral changes, develop consumer- informed recharging infrastructure (based on NEXTSTEPS hydrogen & EV studies) Modify MA 3 T model inputs accordingly and use it for calibra5on. Also, use other vehicle choice models such as TAFV, HYTRANS, Autonomie to calibrate the vehicle penetra5on. Include light- duty trucks in the mix.
THANK YOU! Contact: Kalai Ramea University of California, Davis kramea@ucdavis.edu
ADDITIONAL SLIDES
Existing Methodologies in Energy Models for Vehicle Adoption MARKAL / TIMES (IEA, ITS): Pure op5miza5on based on vehicle and fuel costs, along with growth constraints on vehicle technologies. MoMo (IEA): Expert judgment on growth rate of different vehicle technologies. GCAM (PNNL), CIMS (SFU): Hybrid models, Logit sub- modules based on vehicle and fuel costs for mode choice and vehicle choice. NEMS (EIA): Simula5on model, has an independent consumer vehicle choice model (CVCM) to feed into the main model inputs.
Intangible Cost Component Limited EV range Refueling station availability Model Availability Cost New Technology Risk Premium Towing Capability Description Cost of the consumer willing to spend on rental cars in a year based on their value of perceived anxiety due to range limitations of the owned vehicle. It is calculated based on the charge sustaining capability of the vehicle, how much or how long the consumer drives every day, and the attitude of consumer towards technology risk. This attribute monetizes the anxiety of the consumer when it comes to using limited range EVs. Cost associated with the ease of access to recharging and refueling infrastructure. This cost captures the fuel availability and the ease at which the consumer can have access to refuel his vehicle. It depends on the fuel infrastructure itself, as well as the driving behavior of the consumer; if the consumer is prone to drive more, he or she has the need to refuel often. For example, in the year 2010, gasoline cars have an easier access to fueling stations than hydrogen cars, hence the gasoline cars have a lower cost associated with this compared to hydrogen cars. Cost associated with the number of vehicle models available for a given vehicle technology. It is assumed that, when the vehicle technology is new to the market and has limited sales, the models available to sell are also limited. So, if the user prefers to have a different model car in the given new vehicle technology, it may not be readily available until there is a sizeable market demand for it. This disutility is captured in this cost attribute.. Cost calculated based on the willingness to accept the technology risk and the perceived riskiness of new vehicle technologies. The consumers in this model are divided into early adopters, early majority and late majority, based on their attitude towards technology risk. For example, when a certain vehicle technology is new to the market, early adopters are more willing explore them rather than the other two groups. They have a lesser risk premium cost compared to the other consumer groups. Cost calculated based on the towing capacity of the vehicle technology. This cost is technology specific, and not consumer group specific. A few vehicle technologies, such as gasoline cars or diesel cars have a better towing capability than electric vehicles, for example. If a consumer prefers to have a better towing capacity for his vehicle, this cost attribute captures it.
EV Range Anxiety Cost Calculated based on the perceived value of anxiety by the consumer groups Random varia5on of daily vehicle miles traveled follow the gamma distribu5on, varies based on weekday / weekend and driving behavior Daily VMT is checked with the range of EV. If it is not met, that day is added to the number of days the consumer needs to rent Perceived value of anxiety: Early adopter: $10/day Early majority: $20/day Late majority: $50/day Caveat: The model does not consider 2nd car ownership
Model Availability Cost in the MA 3 T model Vehicle Sales
Carbon Intensity Profiles '%!" '$!" '()*+,&-,".,/0"1&2304.&5156.&.70//0+,/89)(:.5"+)0./&!"#$%& '#!" '!!" &!" %!" $!" +,-./012" 302-2/" 4,567,/"+,-" 8/29570905:" ;:<7.=21" #!"!" #!!(" #!!)" #!'!" #!'#" #!'%" #!#!" #!#(" #!*!" #!*("