Вы находитесь на странице: 1из 9

Addressing Supply- and Demand-Side Heterogeneity and Uncertainty

Factors in Transportation Life-Cycle Assessment: The Case of Refuse


Truck Electrification
Dong-Yeon Lee Georgia Institute of Technology, dlee348@gatech.edu
Valerie M. Thomas Georgia Institute of Technology, valerie.thomas@isye.gatech.edu
Patrick S. McCarthy Georgia Institute of Technology, mccarthy@gatech.edu
Abstract. As a way of incorporating heterogeneity and/or uncertainty factors in life-cycle
assessment (LCA) in a systematic and predictive manner, here we identify key variability factors
and evaluate their relative significance in LCA results, using a case study of refuse truck
electrification. More specifically, we conduct a comparative LCA of conventional diesel, natural
gas, and electric refuse trucks in 10 largest metropolitan areas in the U.S., putting focus on the
potential of electricity and natural gas as bridge fuels. Unlike conventional LCA studies, we first
look at variability elements and analyze which of those variables is worth addressing or not. In
so doing, we take a parametric modeling methods including linear regression and discrete
choice method. Our goal is to develop a generic model that can predict life-cycle inventory over
the entire range of input variables with the heterogeneity factors included as operational
variables. And then we apply vehicle activity statistics data to our generic model for life-cycle
inventory analysis as well as life-cycle valuation.
In terms of the potential of electricity or natural gas as bridge fuels for next-generation refuse
trucks, our findings indicate that conventional natural gas is neither very cost-competitive to nor
less polluting than conventional diesel or electricity alternatives. This is largely due to severe
drive or duty cycles in which refuse trucks typically operate, offsetting the lower carbon intensity
of natural gas fuel in comparison to petroleum diesel. However, natural gas produced from
landfills shows a significant potential to be a bridge fuel, while conventional diesel is expected to
remain as one of the most common fuels of choice over the next 10 (or 20 in some regions)
years. With the bridge fuels such as natural gas and electricity, whether the transition will
ultimately lead to natural gas or electricity really depends on regions. Lastly, with regards to
variability factors, our study suggests that vehicle/fuel LCA studies would better specify or report
input values of some key parameters including average trip speed, type of feedstock, payload,
region, and cross-sectional and longitudinal changes in economic, driving pattern, technological
components.
Proceedings of the International Symposium on Sustainable Systems and Technologies (ISSN 2329-9169) is
published annually by the Sustainable Conoscente Network. Jun-Ki Choi and Annick Anctil, co-editors 2015.
ISSSTNetwork@gmail.com.
Copyright 2015 by Dong-Yeon Lee, Valerie M. Thomas, Patrick S. McCarthy Licensed under CC-BY 3.0.
Cite as:
Addressing Supply- and Demand-Side Heterogeneity and Uncertainty Factors in Transportation Life-Cycle
Assessment: The Case of Refuse Truck Electrification. Proc. ISSST, Lee, D.-Y., Thomas, V. M., and McCarthy, P. S.
Doi information v3 (2015)

Addressing Supply- and Demand-Side Heterogeneity and Uncertainty Factors in Transportation LifeCycle Assessment: The Case of Refuse Truck Electrification

Introduction. Life-cycle assessment (LCA) has become a common practice or tool for
evaluating comprehensive environmental and economic performances of various natural and
human-made systems. One of the limitations of typical LCA studies is a lack of generalizability.
That is, results found in one study are often not generalizable, because the analysis is based on
average or case-specific conditions (Lee et al. 2013). Uncertainty and sensitivity analysis can
help, but they dont always explain or predict exactly why and how much the results can
change under different circumstances.
When it comes to transportation LCA, it is well-known that the use phase is generally the
dominant contributor to energy consumption and environmental impact (Lave and Maclean
2000), except in cases of massive infrastructure requirements for transportation in general or for
specific technologies or systems. What we explore here is the extent to which the impact of use
phase components can be parameterized and thus incorporated as predictive variables, which
can help explain the variability and differences between transportation LCA studies and
assumed conditions. Other life-cycle components (e.g., fuel production) can also be
parameterized within an LCA. To demonstrate this, we use refuse truck electrification as a case
study. We conduct a comparative LCA of refuse truck technologies, adopting a parametricmicroscopic approach to systematically deal with an array of different and individual truck
operating conditions and fuel supply-chains. Vehicle use such as driving behavior and refueling
pattern are major demand-side heterogeneity and uncertainty elements. The supply-side
heterogeneity and uncertainty factors include local fuel-sourcing as opposed to centralized fuel
supply, spatio-temporal variations of electric power, and geographical differences in exogenous
factors (e.g., local climate, fuel price, etc.). Overarching uncertainty factors are future fuel price
evolution, the relative importance of current investment over future earnings (or discount rate),
the social benefit of air emissions reductions benefits, etc. Note that here we differentiate
variability and uncertainty. By and large, both variability and uncertainty are part of the
heterogeneity, but variability is deterministic or something that vehicle or fleet operators can
control, while uncertainty refers to those factors that are not very controllable. For example, fuel
price is determined once location and time are fixed. Also, trip characteristics of commercial
vehicles including refuse trucks are also manageable and predictable. However, future fuel price
or the specific feedstock of diesel fuel is not something that individual fleet operators or
managers can control or determine.
Research Objectives. We are interested in comparative environmental and economic benefits
of conventional diesel, compressed natural gas (CNG), and battery electric refuse trucks in the
10 largest metropolitan areas (e.g., New York, Los Angeles, Chicago, etc.) in the U.S. With
regards to natural gas, we consider not only conventional natural gas but also renewable natural
gas (or biogas) produced from landfills. All of the metropolitan areas being considered here
have landfills that collect methane for transportation or electricity generation with varying
methane resource availability. Currently, almost half of new refuse trucks sold in the U.S. are
CNG, and a battery electric model has recently become available. Given the increasing diversity
of options and the momentum for changes, it may be useful to know whether and how much
society benefits from a transition from the diesel trucks to alternative-fueled trucks such as
natural gas or electricity. So, answering these whether and how much questions is our major
research objective. Also, in doing the evaluation, we attempt to avoid basing our analysis on
fixed or average conditions. We rather look at overall possible ranges of key variables based on
a parametric and spectrum-based modeling approach, addressing or incorporating the supplyand demand-side heterogeneity factors.

D.-Y. Lee et al.

Model Year
Manufacturer
Curb Weight (ton)
Payload (ton)
Engine
Maximum Power (kW)
Li-ion Battery
Capacity (kWh)

Table 1. Vehicle Specifications and Key Input Variables


Diesel
Compressed Natural Gas
Refuse Truck
Refuse Truck
(DRT)
(CNGRT)
2015 - 2030
Mack
Peterbilt
18
18.5
12
11.5
Mack MP7
Cummins ISX12 G
265
261
-

4 Type-II Cylinders: 120


diesel gallons equiv.
Goodyear 315/80R22.5 (10 pieces per truck)

Battery Electric
Refuse Truck
(BERT)
Motiv Power Systems
18
9.2
280
200 (nominal)

Fuel Tank

135 diesel gallons

Tires
Lifetime Vehicle
Miles Traveled (VMT)
Purchase price ($)

500,000 miles (about 20 years of lifetime with annual VMT of 25,000 miles)

Refueling station
Data sources

0.15 0.21 million

0.18 0.25 million


0.67 0.7 million
Already in place or
Already in place
Level 3 (60 kW) charger
requires construction
Vehicle and parts manufacturers: www.macktrucks.com, www.peterbilt.com,
www.motivps.com, www.cummins.com

Methods. To compare the refuse truck technologies (see Table 1 for specifications), we adopt
a LCA framework for the system boundary which is shown in Figure 1. The functional unit is tonkm, which is a product of the weight (metric tons) of garbage that is collected and transported by
a refuse truck and the total distance traveled (or vehicle miles traveled, VMT). Life-cycle impact
assessment criteria are fresh water consumption (in million gallons) and social life-cycle cost (in
constant 2014 $) of the four target refuse truck technologies. In the absence (to our knowledge)
of a publicly-available LCA model that can answer our research questions or address the issues
aforementioned, here we develop a model called Hybrid PArametric-Microscopic Mobility LifeCycle Assessment (HPAM-LCA) which consists of several sub-modules as follows:
Vehicle and parts manufacture inventory. We take a process-based LCA approach for vehicle
manufacture inventory. We use heavy-duty truck materials composition (Gains et al. 1998) and
modified the data so as to reflect more recent model years (Davis et al. 2014). As for energy
and water requirement and emissions for the production of the materials as well as for vehicle
assembly and end-of-life, we use the GREET model (ANL 2014a). Since GREET is for light-duty
vehicles, we adjust the data based on the differences in vehicle specifications (e.g., engine
displacement, number of tires, frontal area, and etc.) between passenger cars and refuse trucks.
Vehicle use-phase energy consumption and emissions prediction. Based on hundreds of drive
cycle (speed vs. time profile) samples collected from publicly-available sources (FHWA 2004;
ANL 2014b; EPA 2014; ImagineMade 2014), we run vehicle dynamic simulations using
ADVISOR (ImagineMade 2014), augmented with tail-pipe emissions test data from Sandhu et
al.s work (2014). Using drive or duty cycle characterization parameters (e.g., average trip
speed, payload, etc.), we develop a linear regression-based energy and emissions prediction
model complemented with additional parameterized sub-models for predicting energy and
emissions impact for extreme climate (hot and cold) conditions, battery degradation, road grade,
etc. This parametric approach allows us to incorporate microscopic (or heterogeneous) vehicle

Addressing Supply- and Demand-Side Heterogeneity and Uncertainty Factors in Transportation LifeCycle Assessment: The Case of Refuse Truck Electrification

use conditions of individual vehicles or fleets into LCA in an integrated manner, avoiding the
potential bias of averages (Taptich and Horvath 2014) and providing predictive power to LCA.
Note that microscopic models sometimes refer to their functionality of providing per-second
output, which is not the case in our model. LCA is typically for an entire lifetime of a product or
service, so the results are to be aggregated over a long span of time.

Figure 1: System Boundary.


State-by-state and hour-by-hour electricity generation. We take a simplified load-filling approach
for electricity generation and consumption fuel mix estimation, based on power generation data
from EPA (2015) for 2014 as well as statistical hourly and monthly generation data from EIA
(2015). Fuel or energy input as well as net generation is checked to be the same as reported in
EPA and EIA data in the original data sets spatial and temporal scales. The amount of water
required for thermo-electric power plants varies widely not only across power generation fuel
types but also from region to region and by cooling system types (e.g., once-through,
recirculating, dry, hybrid, etc.). To provide spatially- and temporally-resolved analysis of water
use for power generation, we use EIAs water withdrawal and consumption survey data for 2013
(EIA 2015) and aggregate them from boilers to cooling systems, power plants by fuel type,
prime mover, water source type, and cooling system type for each state. We then estimate fresh
water withdrawal and consumption rates for generating electricity for each fuel type for each
hour of day in each state.
Life-cycle valuation. We calculate total cost of ownership and then add carbon emissions
damage cost as well as monetized marginal human health impact from non-GHG air pollutants
emissions (Muller et al. 2011). Both our parametric prediction model for life-cycle inventory and
our discrete choice model for comparative cost-effectiveness are generic so that they can be
applied to general or specific cases. For this study, we use refuse truck activity statistics data
from NRELs FleetDNA project (Walkowicz et al. 2014).
Results. Figure 1 shows part of our life-cycle inventory (LCI) results for greenhouse gas (GHG)
emissions and fresh water consumption for the four technologies being compared. Non-electric
refuse trucks LCI reveals an asymptotic pattern in the average trip speed spectrum, and typical
refuse truck operation condition is in the zone with the highest curvature. So, variation in trip
speed can lead to a significant variability or uncertainty in the LCI. As can be seen in Figure 1,

D.-Y. Lee et al.

changing average trip speed, say, from 4 mph (10th percentile) to 11 mph (90th) results in 30%
variability in the GHG emissions. Driving behavior (or skill) can also result in additional but
smaller variability, as does payload condition (empty to full). All these variability factors are
parametrized and incorporated as operational variables in our model, providing predictability
and explaining variabilities. Regional variations for non-electric trucks are relatively very small
compared to the electric truck.

Figure 2: Life-Cycle Inventory of GHG (CO2e ton). Emissions and Fresh Water Consumption (M gallons).
Overall, electric or landfill gas trucks are favorable over diesel and conventional natural gas
trucks in terms of GHG emissions. Although natural gas has a lower carbon-intensity than diesel
on a per-energy basis, the natural gas truck is less efficient than the diesel counterpart in low
speed operation conditions and has higher upstream GHG emissions, which tends to offset the
lower carbon-intensity advantage of the final fuel product. However, natural gas produced from
landfills provides significant GHG emissions reduction potential for natural gas trucks. As for
different types of feedstocks for petroleum and natural gas fuels (CRC 2013, Gordon et al.
2015), diesel shows the largest variability. With the fugitive methane emissions (1 4%) and
other variation factors, natural gas also shows large (but smaller than diesel) feedstock
variability. In terms of fresh water consumption, non-electric trucks consume virtually no water
during vehicle operation (driving) and thus are better than electric counterparts in general.
However, in some areas such as Boston, LA, and Miami, the electric trucks fresh water
consumption is lower than non-electric counterparts, owing to the higher level of reliance on

Addressing Supply- and Demand-Side Heterogeneity and Uncertainty Factors in Transportation LifeCycle Assessment: The Case of Refuse Truck Electrification

saline water for cooling in power plants in those areas.


Figure 3 shows probabilistic longitudinal variability of the social life-cycle cost performance of
the technologies, based on the monetized values of the LCI result above and discrete choice
model. Dotted lighter lines are baseline scenario, and thick solid lines are alternative scenarios.
Conventional natural gas is not cost-effective in any of the scenarios considered, which means
conventional natural gas is not a very attractive alternative fuel to incumbent diesel for refuse
truck application. Natural gas (blue lines) shown in Figure 3 is landfill gas. A few distinct
patterns emerge: First, incumbent diesel will still be an important fuel in the next decade.
Second, landfill gas can play a role as a bridge fuel. Third, in the long run, whether landfill gas
or electricity becomes next-generation fuel or not varies significantly from region to region,
mostly because of the regional/local fuel price differential variations, for instance, Dallas (one of
the areas with the cheapest electricity price in the nation) vs. Boston (having almost triply
expensive electricity rates than Dallas). Fourth, behavioral (e.g. driving skill) and/or
technological (e.g., internal combustion engine thermal efficiency improvement) changes will
have different implications and impact for different regions and fuels. Idle reduction will
decrease the attractiveness of the electric truck the largest in Dallas, whereas driving skill
improvement will increase the electric trucks performance the most in Boston. Also, it is
expected that the electric truck will eclipse landfill gas in 10 years in Dallas, regardless of the
aggressive performance improvement efforts for non-electric trucks, but this is not the case in
Boston where landfill gas will remain the best fuel option.

Figure 3: Probability that Each Technology Provides the Least Social Life-Cycle Cost.
Discussion. From the life-cycle inventory or conventional environmental life-cycle impact
assessment standpoints, it seems average trip speed and the type of feedstock can explain and
predict most of the variability. Non-average trip speed variability is not trivial but the significance
is relatively small, as does the payload. However, in different vocations (e.g., long-haul) or
different condition of the same refuse hauling vocation, the relative significance can change. In
the case of non-electric refuse trucks, regional variations are minimal. For example, local
climate will change over the course of day, month, and year, but the variations are cyclic, and
thus the impact of these variables are not significant in comparative LCA looking at entire
product or service lifetime. Depending on regions, driving behavior and payload impact turns out
to be very significant for electric trucks, which implies that these variables would better be

D.-Y. Lee et al.

specified or included in LCA of electric trucks to minimize confusion. In terms of life-cycle


valuation analysis, it seems both behavioral and technological changes expected in the future
should be included and addressed. Evaluating the potential of bridge fuels requires projections
and forecasts in which longitudinal variability can significantly change the results. Therefore,
when it comes to life-cycle valuation, along with cross-sectional variability components (e.g.,
fuel price projections) for the same model year, these longitudinal variability factors would also
better be addressed explicitly.
It is important to acknowledge the complexity of the problem, but it is equally crucial to
distinguish what to or not to bother with. For LCA practitioners/researchers not to be left with
confusion, we need better information about overall understanding of variabilities and individual
factors impact on LCA results in both absolute and relative manners. This in turn requires an
explanation of what causes the variability and how we can systematically predict or deal with
them in LCA studies. In coping with the heterogeneity and uncertainty issues in LCA, one might
attempt to throw all possible factors in analysis, but this needs caution, because higher
precision doesnt always result in higher accuracy. As we showed here, alternative and/or
foremost research needs is to identify and investigate the sources of variabilities and their
relative significance, which in turn will help understand data need and control data quality in
LCA. For example, based on the essential variables identified, LCA practitioners and scholars
can develop a guideline for LCA studies that recommends or requires important variables or
factors to be reported so that potential confusion stemming from the heterogeneity and/or
uncertainty can be minimized.
Acknowledgements. Part of this work was supported by the National Science Foundation
under Grant No. 1441208.

References
ANL (Argonne National Laboratory). (2014a). The Greenhouse Gases, Regulated Emissions,
and Energy Use in Transportation (GREET) Model. Retrieved from: https://greet.es.anl.gov
(Accessed April 5, 2015).
ANL (Argonne National Laboratory). (2014b). Autonomie. Retrieved from:
http://www.autonomie.net (Accessed June 20, 2014).
CRC (Coordinating Research Council). (2013). Transportation Fuel Life Cycle Assessment:
Validation and Uncertainty of Well-to-Wheel GHG Estimates.
EIA (Energy Information Administration). (2015). Electricity Data. Retrieved from:
http://www.eia.gov/electricity/data/detail-data.html (Accessed April 25, 2015).
EPA (Environmental Protection Agency). (2014). MOVES (Motor Vehicle Emission Simulator).
Retrieved from: http://www.epa.gov/otaq/models/moves/ (Accessed January 3, 2015).
EPA (Environmental Protection Agency). (2015). Continuous Emissions Monitoring. Retrieved
from: http://www.epa.gov/airmarkt/emissions/continuous-factsheet.html (Accessed April 11,
2015).
Davis, S. C., Diegel, S. W., & Boundy, R. G. (2014). Transportation Energy Data Book. Edition
33. Oak Ridge National Laboratory.
FHWA (Federal Highway Administration). (2004). The Next Generation Simulation (NGSIM)
Program. Department of Transportation. Retrieved from: http://ngsim-community.org/
(Accessed September 30, 2013).
Gaines, L., Stodolsky, F., & Cuenca, R. (1998). Life-Cycle Analysis for Heavy Vehicles. Argonne
National Laboratory.
Gordon, D., Brandt, A., Bergerson, J., & Koomey, J. (2015). Know Your Oil: Creating a Global
Oil-Climate Index. Carnegie Endowment for International Peace.
Lave, L., & Maclean, H. (2000). Life-Cycle Analysis of Alternative Automobile Fuel/Propulsion
Technologies. Environmental Science & Technology, 34(17), 3598.
Lee, D.-Y., Thomas, V. M., & Brown, M. A. (2013). Electric Urban Delivery Trucks: Energy Use,
Greenhouse Gas Emissions, and Cost-Effectiveness. Environmental Science & Technology,
47(14), 8022-8030. doi:10.1021/es400179w.
Muller, N. Z., Mendelsohn, R., & Nordhaus, W. (2011). "Environmental Accounting for Pollution
in the United States Economy." American Economic Review, 101(5): 1649-75.
Sandhu, G. S., Frey, H. C., Bartelt-Hunt, S., & Jones, E. (2014). In-use measurement of the
activity, fuel use, and emissions of front-loader refuse trucks. Atmospheric Environment,
92557-565. doi:10.1016/j.atmosenv.2014.04.036.
Taptich, M. N., & Horvath, A. (2014). Bias of averages in life-cycle footprinting of infrastructure:
truck and bus case studies. Environmental Science & Technology, 48(22), 13045-13052.
doi:10.1021/es503356c.

D.-Y. Lee et al.

Walkowicz, K., Kelly, K., Duran, A., and Burton, E. (2014). Fleet DNA Project Data. National
Renewable Energy Laboratory. Retrieved from: http://www.nrel.gov/fleetdna (Accessed
February 2, 2014).

Вам также может понравиться