ArticlePDF Available

Optimizing Hydrogen Fueling Infrastructure Plans on Freight Corridors for Heavy-Duty Fuel Cell Electric Vehicles

August 2023
SAE International Journal of Sustainable Transportation Energy Environment & Policy 5(1)

August 2023
5(1)

DOI:10.4271/13-05-01-0008

Authors:

Majbah Uddin

Oak Ridge National Laboratory

Show all 5 authorsHide

The development of a future hydrogen energy economy will require the development of several hydrogen market and industry segments including a hydrogen based commercial freight transportation ecosystem. For a sustainable freight transportation ecosystem, the supporting fueling infrastructure and the associated vehicle powertrains making use of hydrogen fuel will need to be co-established. This paper introduces the OR-AGENT (Optimal Regional Architecture Generation for Electrified National Transportation) tool developed at the Oak Ridge National Laboratory, which has been used to optimize the hydrogen refueling infrastructure requirements on the I-75 corridor for heavy duty (HD) fuel cell electric commercial vehicles (FCEV). This constraint-based optimization model considers existing fueling locations, regional specific vehicle fuel economy and weight, vehicle origin and destination (O-D), vehicle volume by class and infrastructure costs to characterize in-mission refueling requirements for a given freight corridor. The authors applied this framework to determine the ideal public access locations for hydrogen refueling (constrained by existing fueling stations), the minimal viable cost to deploy sufficient hydrogen fuel dispensers, and associated equipment, to accommodate a growing population of hydrogen fuel cell trucks. The framework discussed in this paper can be expanded and applied to a larger interstate system, expanded regional corridor, or other transportation network. This paper is the third in a series of papers that defined the model development to optimize a national hydrogen refueling infrastructure eco-system for heavy duty commercial vehicles.

Content uploaded by Majbah Uddin

Content may be subject to copyright.

 
Page 1 of 21 
Optimizing Hydrogen Fueling Infrastructure Plans on Freight Corridors for Heavy 
Duty Fuel Cell Electric Vehicles  
 
Adam Siekmann, Vivek Sujan, Majbah Uddin, Yuandong Liu, Fei Xie 
 
 
Abstract 
The development of a  future hydrogen energy economy will require 
the development of  several hydrogen market and industry  segments 
including  a  hydrogen  based  commercial  freight  transportation 
ecosystem.  For  a  sustainable  freight  transportation  ecosystem,  the 
supporting  fueling  infrastructure  and  the  associated  vehicle 
powertrains  making  use  of  hydrogen  fuel  will  need  to  be  co-
established. This paper introduces the OR-AGENT (Optimal Regional 
Architecture Generation for Electrified National Transportation) tool 
developed at the Oak Ridge National Laboratory, which has been used 
to optimize the hydrogen refueling infrastructure requirements on the 
I-75  corridor  for  heavy  duty  (HD)  fuel  cell  electric  commercial 
vehicles (FCEV). This constraint-based optimization model considers 
existing fueling locations, regional specific vehicle fuel economy and 
weight, vehicle origin and destination (O-D), vehicle volume by class 
and  infrastructure  costs  to  characterize  in-mission  refueling 
requirements for  a given  freight corridor.   The authors applied  this 
framework to determine the ideal public access locations for hydrogen 
refueling (constrained by existing fueling stations), the minimal viable 
cost  to  deploy  sufficient hydrogen  fuel  dispensers,  and  associated 
equipment, to accommodate  a growing population  of hydrogen  fuel 
cell trucks.  The framework discussed in this paper can be expanded 
and applied to a larger interstate system, expanded regional corridor, 
or other transportation network.  This paper is the third in a series of 
papers  that defined  the  model development  to  optimize a  national 
hydrogen  refueling  infrastructure  eco-system  for  heavy  duty 
commercial vehicles. 
Introduction 
The development of a  future hydrogen energy  economy will require 
the development of  several hydrogen market and industry  segments 
including  hydrogen  based  commercial  freight  transportation 
ecosystem [Error! Reference  source not found.,1,3,4,5].  As more 
countries  and  companies  commit  to  a  reduction  in  criteria  and 
greenhouse  gas  emissions  from  vehicles  to  reduce  pollution  and 
combat  climate change  concerns [6,7,8,9,10],  there needs  to be  an 
aligned strategy to provide the infrastructure to accelerate the adoption 
of  new  and  emerging  vehicle  powertrain  technologies.  This  is 
especially critical for the freight transportation industry where 24% of 
all transportation generated GHG emissions come from freight truck 
movement. At present, while R&D into  vehicle technologies for the 
use of electricity or hydrogen as a fuel is rapidly gaining momentum, 
the challenge in the near term towards sustainable large scale customer 
adoption remains [Error! Reference source not found.,11]. The rate 
and pace of  technology evolution  and how it will  affect the energy 
pathways  for  commercial  transportation  and  industrial  use  are 
dependent  on  multiple  variables  such  as  national  energy  and 
environmental  policies  and  public-private  partnerships  [Error! 
Reference source not found.]. As we migrate from a carbon intensive 
fossil fuel-based freight transport system to a substantially/completely 
decarbonized  freight  transport  system,  several  customer  centric 
challenges  need  to  be  addressed.  As  compared  to  BEV  or  H2 
powertrains, fossil fuel-based powertrains provide mission flexibility, 
and high  uptime at  a relatively  low total cost  of ownership  (TCO). 
While the incumbent carbon  intensive powertrains suffer  from poor 
efficiency and are not sustainable to support Global Climate Change 
initiatives  in  transportation  decarbonization,  techno-economic 
challenges  continue  to  create  complex  barriers  to  the  large-scale 
displacement of these with highly electrified powertrains architectures 
[13,14,15].  Migration  towards  sustainable  zero  emission  power  in 
commercial vehicles with steady long-term adoption rates is dependent 
on both vehicle and infrastructure solutions that are well aligned with 
commercial  vehicle  end-user  market  needs.  Their  priorities  are 
centered on:  Availability (i.e.  solutions are ready  when it  matters), 
Affordability  (i.e.  favorable  economics),  Efficiency  (i.e.  lower 
operational expenditure), Productivity (i.e. ability to get the job done), 
and  Sustainability (i.e.  emissions  or CO2  footprint/TCO/system-of-
system capabilities). With fuel cell vehicles being an attractive option 
for  the  replacement  of  diesel-powered  trucks,  the  refueling 
infrastructure along interstate highways demands a level of urgency to 
meet regulatory deadlines.    For a sustainable  freight  transportation 
ecosystem,  the supporting  fueling infrastructure  and  the associated 
vehicle  powertrains  making  use  of  hydrogen  fuel  will  need to  be 
optimized [16,17,Error! Reference source not found.]. 
Current refueling infrastructure (Figure 1) allows for diesel fuel to be 
transported, with relative ease, to various refueling locations including 
“behind  the  fence”  locations  such  as  distribution  centers  and  port 
authorities, public refueling  stations, and other shipping  origins and 
destinations.  The varying locations for a vehicle to refuel allows for 
optimal refueling strategies on the vehicular level, with flexibility for 
refueling  options  when  weather  and  traffic  adversely  affect  fuel 
economy. 
Current infrastructure requirements literature is based on the needs to 
reduce GHG emissions and meet regulatory targets between 2030 and 
2040 and the expected adoption rate of alternative fuel vehicles such 
as  compressed  and  liquid natural  gas,  hydrogen  FCEV,  BEV  and 
others  with the  main near-term  focus being  BEVs.   The European 
Automobile Manufacturer’s Association estimates there will need to 
be approximately 30,000 500kW public chargers required by 2030 and 
1,000 hydrogen refueling stations by 2030 [Error! Reference source 
not  found.].    The  International  Council  on  Clean  Transportation 
suggests that only 220 stations providing 4,800 kg of hydrogen a day 
will be required  in the  US, with that  number rising  to nearly  7,000 
stations by 2050 [0].  Of note it is suggested that only 12% of heavy 
trucks on the road will be FCEV by 2050 with the remaining being a 
combination of BEV and ICE. 
The first paper in this series explored the freight traffic demand and 
described methods of generating freight volumes and vehicle weights 
using  the  Freight  Analysis  Framework  (FAF),  Travel  Monitoring 
Analysis System (TMAS), and other data sources to determine demand 
for hydrogen along I-75 [18].  The second paper in this series explored 
the cost of installing hydrogen refueling infrastructure and the impact 
demand have on total system cost as well as the impact on refueling 
time for commercial trucks [22].  In this paper, the third in the series, 
introduces  the  OR-AGENT  (Optimal  Regional  Architecture 
Generation for Electrified National Transportation) tool developed at 
the Oak Ridge National Laboratory (ORNL), which has been used to 
optimize the hydrogen refueling infrastructure requirements on the I-
75 corridor for HD FCEVs. 

Page 2 of 21

Figure 1. Diagram showing typical diesel refueling infrastructure

Figure 2. Complete architecture for FCEV with highlighted section

indicating OR-AGENT focus for this paper

OR-AGENT is the evolution of the previous ORNL developed

infrastructure planning tool, REVISE-II [23] that explored the location

of electric chargers for battery electric vehicles. In the new OR-

AGENT framework, a systematic multi-layered approach to on-

highway freight transportation energy infrastructure is developed. The

long-term objective of the tool is to provide an optimal zero emission

vehicle powertrain and energy infrastructure architecture

recommendation, that is regional specific, and accurately represents

real world scenarios including freight movement, energy

infrastructure, operational characteristics, and local constraints. This

framework will assimilate the commercial vehicle first, middle, and

last mile operations data including vehicle O-D weight and volume,

the vehicle powertrain architecture options, supporting energy

infrastructure components (both behind the fence and public access

dispensing, storage, and DER), the electric grid energy production

assets, regional constraints, and operating environment factors. It then

uses advanced genetic algorithms to meet the objectives. Figure 2

shows the architecture considerations for OR-AGENT. The initial first

phase construct (Figure 3 and topic of this paper) will incorporate

various external data sets from powertrain simulations, existing

infrastructure, and other sources to generate a realistic representation

of the public access refueling requirements for HD FCEVs traveling

on the I-75 freight corridor network. This will focus on the public

refueling infrastructure typically found at travel centers and truck stops

and makes assumptions that a vehicle can only fill up at these locations,

as discussed later. Considerations into traffic, weather, grid impact

and hydrogen generation are not considering in this first phase model.

The OR-AGENT model in this study will provide decisions on:

• Where should a refueling station be located to meet

hydrogen demand?

• What is the hydrogen storage need at each station?

• How many dispensers should a refueling station have to

meet hydrogen demand?

• What type of dispenser technology provides the lowest cost

option while still meeting demand?

Previously studies have explored the optimization of refueling

locations for various vehicle types including both passenger and

commercial vehicles [24]. These studies have also explored the

requirements for fuel transportation to refueling locations and various

use cases for the specific refueling station using different mathematical

models to determine origin-destination pairs and distance traveled of

vehicles. The reoccurring assumption in these previous studies is that

a vehicle has a finite driving distance absent of road grade or vehicle

parameters. In our study, simulated vehicle dynamics will be

incorporated to provide a realistic representation of driving range and

refueling requirements.

Research on infrastructure needs revolve around

This paper will discuss the current data requirements for the model and

how the data is generated. Secondly, the methods and steps required

to incorporate the data and form a result. Lastly, a sample result will

be discussed with a discussion of the conclusions and future

opportunities.

Model data inputs

Information from multiple data sources including Federal databases,

simulations and other models have been combined to create an

infrastructure optimization strategy for electrified transportation

(Figure 3).

Figure 3. Data sources and structure of OR-AGENT for optimization

of hydrogen refueling stations for commercial vehicles

The following data flows will be discussed in this section:

• Freight Flow Information – How freight moves along a route

defined by an origin and destination

• Traffic and Weight Data – Distribution of heavy truck traffic

along major corridors and the vehicle class and weight

• Road grade and speed limit data – Elevation/grade and speed

limit along major road corridors

• Vehicle & Power Train Simulation – Simulation of vehicle and

powertrain dynamics to determine fuel economy for segments

along the given route

• Current Refueling Infrastructure – Location and demand

distribution of potential refueling locations

• New Infrastructure Economics – Capital cost of new

infrastructure deployment at a refueling station

• Infrastructure Optimization – Cost optimization for refueling

locations based on the previous data inputs

A. Freight Flow Information

FAF highway network was utilized for truck routing purposes [25]. To

speed up truck routing, a network simplification process was first

introduced to consolidate the network and to reduce the number of total

highway links without losing information. In the process, for any two

Inbound

Outbound

Sea / Inland

waterways

Rail

Diesel fuel

Public access D.C. access

Diesel powertrains

Fuel refinery

Port/term

authority

Freight

Operations

Inbound

Outbound

Electricity fuel

Port/term/ware

house authority Public access D.C. access

FCEV powertrains

Power plant

Vehicle arch

Refueling arch

Grid energy arch

Freight

Sea / Inland

waterways

Rail

New energy sources

Operations

RC/DC/FC

Freight flow

information (FAF)

Traffic count and

WIM data

OD, Truck type,

weight, volume

count

Vehicle /

powertrain

simulation

Road grade /

speed limit data

Current refueling

infrastructure

New fueling

infrastructure

economics

Infrastructure

optimizer

OR-AGENT first phase construct

 
Page 3 of 21 
adjacent links , , if (i) no additional truck links are connected 
to the junction of the two links, denoted by  node , and (ii) the two 
links have the same  speed limit, then the  two adjacent links  can be 
consolidated into one link connecting node  and : .  A simple 
network example is shown in Figure 4(a).  Green dots are the vertices 
of each road link before simplification while red circles represent the 
vertices after simplification.  The speed limit of each link is marked in 
the  figure.    In the  example,    and    cannot  be  consolidated 
because there is another link  that connect to node .  and  
cannot  be consolidated  because the  two links  have different  speed 
limits.   and  , on the other hand, can be merged into one link 
 given both conditions are satisfied.  Figure 4(b) shows a portion 
of the simplified FAF4 network.  Similarly, green dots are the vertices 
of  each road link  in the  FAF4 network.   Red  circles represent the 
vertices after simplification. 
 
Figure 4. (a) Network simplification example, (b) FAF4 network 
vertices vs simplified FAF4 network vertices 
Truck routing: It was assumed that all trucks will choose the shortest 
path between Origin (O) and Destination (D).  There is a total of 132 
origin FAF zones and 132 destination FAF zones in the FAF4 network.  
All truck movements were assumed to start from the centroid of the 
origin region and end at the  centroid of the destination region.   The 
same assumptions were adopted in previous studies [26].  The shortest 
path from 132 origin zones to 132 destination zones was obtained.  All 
paths that  traverse the  I-75 corridor were then  identified.  Figure 5 
shows  an example  of  the resulting  shortest path  from  the Detroit-
Warren-Ann Arbor  area to  the Birmingham-Hoover-Talladega area.  
The shortest path (green lines) uses the I-75 corridor (black lines) from 
node E1 to node E2. 
 
Figure 5.  Shortest path from Detroit-Warren-Ann Arbor area to 
Birmingham-Hoover-Talladega area 
Based on the routing results, each route from Origin to Destination is 
divided into three segments: (i) the segment from a FAF origin to the 
I-75 corridor entrance (e.g., O to E1 in Figure 5); (ii) The segment from 
the I-75 corridor entrance to the I-75 corridor exit (e.g., E1 to E2 in 
Figure  5);  and  (iii)  the  segment  from  the  I-75  corridor  exit  to  a 
destination (e.g., E2 to D in Figure 5).  Note that a truck may leave the 
I-75 corridor for a bypass and reenter the same corridor again before 
reaching the  destination.   In these  cases, there  are more  than three 
segments.  The resulting O-D, and routed distance for each segment 
are  the  primary  input  data  for  truck  hydrogen  fuel  consumption 
simulation. 
 
Figure 6.  Location of Origin and Destinations considered for Heavy 
Truck travel along I-75 
 The daily  average flow of  heavy truck  freight from  the Freight 
Analysis  Framework  (FAF)  was  determined  using  the  procedure 
described by Uddin [18].  Using the FAF network centroids for origin 
and destination, a routed distance to I-75 was calculated to determine 
how far a heavy truck would need to travel before reaching I-75.  For 
the purposes  of this  study, only origin  and destinations within  500 
miles of I-75 were considered to ensure that the vehicle would be able 
to reach a refueling station before running out of fuel.  The maximum 
routed distance traveled off  I-75 was 499.4 miles and the maximum 
distance traveled on I-75 was 717 miles.  Figure 6 shows the candidate 
origins and destinations used for the I-75 infrastructure optimization. 
In addition to the distance traveled from origin to I-75 and from I-75 
to destination, the distance traveled on I-75 was also calculated.  Figure 
7 shows the total daily distribution for vehicle trip length with many 
of the vehicles having a daily trip length of less than 600 miles, which 
aligns  well  with  Federal  Motor  Carrier  Safety  Administration 
(FMCSA) hours of service (HOS) regulations [27] allowing, in most 
cases, allowing a maximum of 11 hours of daily driving.  Trips over 
600 miles may be split over multiple days of travel. 

Page 4 of 21

Figure 7. Distribution of Total Trip Length from Origin to Destination

B. Traffic and Weight Data

Vehicle class data for this study is represented by the FHWA notation

which is based off vehicle configuration rather than weight (see Figure

8). Vehicle weight and class data for the previously generated O-D

pathways was obtained from various data sources, such as the FAF

Network tonnage, Travel Monitoring Analysis System (TMAS)

weigh-in-motion (WIM) data [26]. Using distributions of weight for

the vehicle class and location (example in Figure 9) a Gaussian

Mixture Model was used to define three weights for each vehicle class

(light, medium and heavy loaded). This info will be used to simulate

vehicle fuel efficiency for a range of weights.

Figure 8. FHWA Heavy Vehicle Classification. Data taken from Ref.

[28]

 
Page 5 of 21 
 
Figure 9.  Sample Monthly Distribution of Weight Data at a Florida WIM station for Class 12 trucks 
The  monthly  averaged  weight distribution for  Class 9-Class  13 
heavy trucks is shown below in Figure 10, with 90% of the population 
for Class 9,11 and 12 being less than 60,000 lbs. and less than 85,000 
lbs. for Class 10 and 11.   
 
Figure 10.  FHWA Vehicle Class Weight Distribution 
The volume of traffic along the route defined by any origin and 
destination pair, was calculated  from FAF Network data.  The daily 
average  truck  volumes  used  in  this  research  are  modified  by  an 
adoption rate described by the annual percentage of sale requirements 
of CARB [1,5] (Figure 11), with an upper limit of 45% of the total HD 
truck population.  CARB’s Advanced Clean Trucks (ACT) regulation 
requires manufacturers to sell increasing percentages of zero-emission 
trucks  and  is expected  to  further  reduce the  lifecycle  emission  of 
greenhouse gases  and eliminate  tailpipe emissions  of air  pollutants 
[1,5]. The ACT rule requires the sale of zero-emission or near zero-
emission HD  Trucks starting  with the  manufacturer-designated MY 
2024 (see Figure  11). Sales  requirements are  defined separately  for 
three  vehicle  groups:  Class 2b-3  trucks  and  vans,  Class  4-8  rigid 
trucks, and Class 7-8 tractor trucks [1,5]. The regulation is structured 
as  a credit  and  deficit accounting  system. A  manufacturer  accrues 
deficits based on the total volume of  on-road HD truck sales within 
California in a given  model year. These deficits must be offset with 
credits generated by the sale of zero- or near zero-emission vehicles 
(ZEVs/NZEVs) [1,5]. This adoption rate value was used to modify the 
daily truck volume, linearly, to determine the total amount of trucks 
that would be using hydrogen as percentage of the total population of 
trucks on I-75. 

Page 6 of 21

Figure 11. Adoption Rate of Vehicles by model year. Data taken from

Ref. [2]

C. Road grade and speed limit data

Road grade and vehicle road speed data are critical to assess the fuel

consumption, and consequently, the refueling requirements along the

I-75 corridor. Several data sources are combined to characterize this

roadway for vehicle operations. Specifically, the Freight Analysis

Framework 4 database is used to identify the GPS latitude and

longitude coordinates of the I-75 corridor. All road elevation (grade)

and speed limits are extracted from a higher accuracy, Nokia HERE

database [29] at each reference GPS latitude and longitude coordinate.

This process is used to leverage the preprocessing conducted by the

FAF4 tool on GPS latitude and longitude coordinates associated with

specific road names. In parallel, commercial tools such as the Nokia

HERE database provide high accuracy assessments of elevation and

road speed limits for a given GPS latitude and longitude coordinate.

For this phase of the OR-AGENT framework, road speed limits are

used as the vehicle speed targets (see Figure 12). Future manifestations

of this approach will also leverage the actual road traffic flow rates at

a given location, time, and date, to obtain a more representative real-

world operating state for the vehicles. This will provide a real-world

assessment of traffic and weather conditions.

(a) I-75 corridor (black) indicated among the entire U.S.

interstate road system (blue)

(b) I-75 elevation and road speed limits

Figure 12. Characterizing the road grade and speed limits along I-

75. Data taken from Ref. [25,29]

D. Vehicle & Powertrain Simulation

Fuel economy for heavy trucks was calculated using a 1-D FCEV

powertrain vehicle model as described by [16] with a range of fuel cell

sizes to represent different vehicle configurations (Figure 14). The

following equations from Sujan are the basis for the power calculations

used for fuel economy in this paper.

(a) Vehicle non-elastic 1-D longitudinal dynamics model

(b) Wheel-tire-surface 1-D longitudinal model

Figure 13. Modeling the vehicle and road dynamics. Data taken

from Ref. [16]

10%

20%

30%

40%

50%

60%

70%

80%

2022 2024 2026 2028 2030 2032 2034 2036

ANNUAL PERCENTAGE SALES REQS.

MODEL YEAR

CARB - Advanced Clean Truck Regulation

Class 2b-3: pickup trucks and vans

Class 4-8: non-tractor trucks

Class 7-8: tractor trucks

Paero

Paccel

Pwhl drag

Pgravity mg 20% Mg

40% Mg

Fz=mg

Ffr

Page 7 of 21

(1)

󰇗       (2)

     󰇛󰇜   󰆒  󰇛󰇜 (3)

  󰇛󰇜

 (4)

To simplify optimization calculations, an average fuel economy was

calculated for 34, approximately 50-mile, segments along I-75 (Figure

15) for Class 9-Class 12 vehicles with five weights each (Table 1).

Figure 14. 1-D FCEV powertrain vehicle model

Figure 15. I-75 Segments for Fuel Economy Averages

Table 1. Vehicle Weights for Fuel Economy Simulation

Figure 16 and Figure 17 capture the powertrain recommendations

motivated by the CapEx and FuelEx minimization while also

attempting to minimize product proliferation through inspection

(valuable to both developers and end-users). These assessments with

additional details were made previously [16]. In these tables energy

capacity options of both a generic Li-ion NMC (Nickel Manganese

Cobalt oxide – representing high energy battery chemistries) and LTO

(Lithium Titanate – representing high power battery chemistries)

batteries is provided [16,Error! Reference source not found.].

Characterizing the impact of battery chemistry is a complex problem

and impacts the architectures associated with electrified powertrains

including hybrid electric, battery electric, range extended electric, and

fuel cell electric vehicles. Several studies have been and will continue

to be conducted to explore battery characterization [16,Error!

Reference source not found.,30,31,32,33,34,35]. Our previous

studies have focused on a reduced order model on life, cost, power,

and packaging, with the key results being used here [16]. The TCO and

packaging studies conducted previously [16] marginally favor LTO

over NMC battery chemistry architectures. It is important to state that

significant development continues in the battery domain that will

necessitate closer examination of the wider range of chemistry options

that continue to be introduced (including gaps that currently exist in

real long term durability assessments over real world operating

scenarios). The range for both the fuel usage and the battery sizes is a

result of the two different powersplit control algorithms—load

following and energy minimization [16]. Figure 16 and Figure 17

summarize MY2020 and MY2030 Class 8 vehicle applications, but

data has been assessed for the MY2040 specified vehicles. The focus

of our work will use the architecture described by a vehicle architecture

that represents a mature technology market [16]. For the purposes of

this study, vehicles were assumed to have a 330-kW fuel cell (in

addition to the long haul parameters listed in Figure 16 and Figure 17),

and the resulting fuel economies, in kilograms of hydrogen consumed

per mile, are shown for model year 2020 and 2030 tractors in Figure

18 and Figure 19, respectively.

Figure 16. 2020 Architecture parameters

+ + + +

+ + +

5th Percentile 25th Percentile 50th Pe rcentile 75th Percentile 90th Percentile 95th Percentile

9 27,558 36,817 50,045 68,784 77,162 80,028

10 32,628 44,533 68,343 91,051 116,183 130,293

11 33,069 47,620 57,541 65,698 72,312 75,839

12 35,274 48,061 58,863 67,461 74,075 78,044

13 40,124 55,997 106,042 133,600 150,796 158,292

FHWA Class

Vehicle Weight (lbs)

Page 8 of 21

Figure 17. 2030 Architecture parameters

Figure 18. MY 2020 direction averaged fuel economies for 330 kW FC Vehicle (kg/mile)

Figure 19. MY 2030 direction averaged fuel economies for 330 kW FC Vehicle (kg/mile)

Fuel economy was calculated for both the north-to-south and

south-to-north directions on I-75 using real world road grade and road speed limit for each segment (as described previously). Figure 20

shows the averaged directional fuel economy of the 90th percentile

27,558 lb 36,817 lb 50,045 lb 68,784 l b 77,162 lb 80,028 lb 32,628 lb 44,533 lb 68,343 lb 91,051 lb 116,183 lb130,293 lb 33,069 lb 47,620 lb 57,541 lb 65,698 lb 72,312 lb 75,839 lb 35,274 lb 48,061 lb 58,863 lb 67,461 lb 74,075 lb 78,044 lb 40,124 lb 55,997 lb 106,042 lb133,600 lb150,796 l b 158,292 lb

1 10.39 9. 38 8.52 7.28 6.71 6.65 9.80 8.80 7.24 6.21 5.25 4.94 8.67 7.63 7.02 6.69 6.45 6.22 8.05 7.30 6.78 6.46 6.10 6.01 7.51 6.70 4.95 4.36 4.06 3.97

2 8.79 8.15 7.28 6.50 6.16 6.05 8.31 7.63 6.51 5.67 5.04 4.65 7.20 6.57 6.15 5.84 5.64 5.59 6.88 6.28 5.95 5.64 5.45 5.29 6.32 5.79 4.52 4.06 3.75 3.67

3 8.69 8.06 7.30 6.41 6.11 5.96 8.26 7.61 6.38 5.65 5.00 4.70 7.18 6.51 6.12 5.82 5.66 5.56 6.79 6.26 5.89 5.59 5.45 5.36 6.32 5.71 4.59 4.09 3.93 3.81

4 8.77 8.16 7.33 6.48 6.12 6.01 8.44 7.63 6.47 5.66 5.02 4.78 7.29 6.61 6.12 5.87 5.71 5.58 6.82 6.27 5.97 5.63 5.42 5.32 6.39 5.80 4.57 4.05 3.87 3.71

5 9.18 8.59 7.70 6.79 6.39 6.33 8.87 8.07 6.78 5.85 5.20 4.77 7.61 6.85 6.44 6.16 5.90 5.78 7.22 6.59 6.14 5.87 5.62 5.54 6.68 6.06 4.61 3.96 3.70 3.56

6 8.78 8.16 7.45 6.49 6.16 6.06 8.40 7.69 6.53 5.69 4.96 4.63 7.28 6.55 6.22 5.89 5.70 5.58 6.85 6.31 5.94 5.64 5.44 5.26 6.38 5.75 4.44 3.88 3.56 3.51

7 9.00 8.39 7.48 6.55 6.29 6.10 8.56 7.80 6.53 5.69 5.05 4.67 7.40 6.79 6.38 6.00 5.77 5.72 7.05 6.50 6.11 5.74 5.52 5.41 6.53 5.94 4.61 4.06 3.78 3.67

8 9.10 8.40 7.51 6.60 6.26 6.07 8.63 7.76 6.59 5.70 4.94 4.64 7.52 6.72 6.26 6.00 5.69 5.59 7.06 6.46 6.03 5.70 5.52 5.40 6.60 5.89 4.48 3.89 3.59 3.52

9 9.60 8.85 7.96 6.95 6.59 6.45 9.22 8.35 6.92 6.02 5.22 4.86 8.06 7.16 6.72 6.36 6.07 5.98 7.52 6.87 6.43 6.10 5.85 5.69 7.04 6.29 4.77 4.14 3.80 3.68

10 9.67 8.92 8.19 7.04 6.68 6.56 9.32 8.45 7.08 6.14 5.30 4.93 7.99 7.23 6.79 6.42 6.12 6.08 7.60 6.93 6.46 6.10 5.90 5.81 7.09 6.34 4.82 4.21 3.88 3.76

11 9.88 9.17 8.22 7.15 6.70 6.52 9.52 8.54 7.07 6.11 5.33 4.92 8.18 7.37 6.84 6.46 6.23 6.11 7.78 7.07 6.61 6.23 5.98 5.82 7.28 6.45 4.93 4.28 3.96 3.81

12 9.73 8.96 7.95 6.88 6.47 6.33 9.27 8.41 6.89 5.91 5.17 4.84 8.16 7.14 6.67 6.35 6.11 5.96 7.69 6.95 6.44 6.13 5.90 5.69 7.15 6.38 4.89 4.39 4.11 4.02

13 8.44 7.92 7.07 6.25 5.93 5.83 8.05 7.35 6.28 5.52 4.91 4.57 6.93 6.34 5.93 5.69 5.47 5.37 6.55 6.05 5.74 5.45 5.25 5.18 6.11 5.62 4.44 3.99 3.74 3.71

14 8.27 7.74 7.00 6.15 5.81 5.75 8.06 7.24 6.16 5.44 4.86 4.55 6.86 6.25 5.87 5.63 5.40 5.32 6.50 5.99 5.68 5.41 5.21 5.10 6.04 5.57 4.41 3.97 3.72 3.62

15 8.35 7.70 7.01 6.33 6.11 6.07 7.89 7.31 6.33 5.74 5.12 4.83 6.87 6.33 6.05 5.82 5.66 5.59 6.50 6.14 5.84 5.60 5.44 5.42 6.18 5.73 4.72 4.33 4.17 4.01

16 8.69 8.02 7.21 6.66 6.47 6.47 8.34 7.48 6.64 6.10 5.59 5.27 7.16 6.60 6.32 6.17 6.01 6.00 6.86 6.41 6.16 5.95 5.81 5.86 6.47 6.01 5.27 4.79 4.60 4.52

17 9.60 8.75 7.73 6.67 6.41 6.31 9.01 8.07 6.67 5.96 5.35 5.04 8.02 7.01 6.56 6.26 6.01 5.90 7.46 6.82 6.29 6.02 5.85 5.73 6.98 6.27 4.99 4.54 4.33 4.25

18 9.51 8.81 7.84 6.70 6.41 6.27 9.17 8.15 6.79 5.83 5.12 4.73 7.89 7.05 6.55 6.23 5.94 5.84 7.44 6.76 6.24 5.94 5.75 5.60 6.98 6.20 4.69 4.15 3.89 3.79

19 10.62 9. 63 8.52 7.38 7.02 6.84 10.02 8.98 7.37 6.40 5.61 5.26 8.75 7.82 7.24 6.95 6.59 6.51 8.29 7.55 6.94 6.65 6.40 6.29 7.77 6.90 5.37 4.82 4.47 4.38

20 9.34 8.55 7.63 6.71 6.37 6.25 8.93 7.97 6.70 5.85 5.24 4.95 7.80 6.89 6.47 6.16 5.94 5.90 7.27 6.65 6.26 5.89 5.68 5.59 6.82 6.16 4.87 4.41 4.14 4.06

21 8.69 8.02 7.14 6.20 5.93 5.73 8.39 7.46 6.20 5.42 4.77 4.44 7.25 6.46 6.05 5.78 5.57 5.44 6.84 6.24 5.80 5.52 5.32 5.24 6.38 5.77 4.44 3.96 3.74 3.57

22 11.82 10.69 9.30 7.96 7.46 7.30 11.16 9. 87 7.94 6.81 5.94 5.46 9. 81 8.58 7.85 7.42 7.04 6.88 9.27 8.30 7.53 7.16 6.76 6.73 8.66 7.58 5.59 4.97 4.61 4.52

23 12.55 11.28 9.79 8.32 7.90 7.66 11.74 10.49 8.42 7.10 6.06 5.65 10.56 9.08 8.35 7.87 7.46 7.33 9.93 8.74 8.04 7.49 7.19 6.92 9.17 7.97 5.79 5.07 4.71 4.54

24 11.93 10.96 9.70 8.37 7.80 7.61 11.42 10.14 8.30 7.07 6.06 5.63 9.90 8.89 8.11 7.74 7.32 7.14 9.36 8.55 7.85 7.34 7.07 6.87 8.75 7.82 5.72 4.92 4.55 4.41

25 8.75 8.09 7.29 6.48 6.09 6.03 8.30 7.57 6.48 5.63 4.90 4.57 7.21 6.60 6.18 5.86 5.62 5.47 6.85 6.23 5.86 5.55 5.37 5.22 6.30 5.69 4.38 3.85 3.49 3.39

26 10.20 9. 36 8.34 7.24 6.86 6.61 9.75 8.77 7.28 6.26 5.36 4.98 8.44 7.54 7.05 6.67 6.41 6.20 8.05 7.21 6.81 6.33 6.14 5.97 7.37 6.68 5.01 4.35 4.02 3.86

27 11.43 10.34 9.22 7.81 7.31 7.21 10.84 9. 61 7.73 6.59 5.72 5.32 9. 49 8.40 7.64 7.19 6.90 6.75 8.92 7.99 7.40 6.93 6.67 6.42 8.29 7.41 5.38 4.72 4.38 4.26

28 12.65 11.67 10.23 8.68 8.22 7.99 12.13 10.74 8.78 7.29 6.29 5.79 10.70 9.40 8.59 8.14 7.76 7.63 9.96 9.02 8.27 7.79 7.40 7.16 9. 31 8.21 5.97 5.15 4.73 4.64

29 11.13 10.16 8.96 7.82 7.36 7.17 10.65 9. 53 7.81 6.64 5.75 5.36 9. 31 8.30 7.69 7.19 6.90 6.68 8.81 7.97 7.32 6.85 6.63 6.46 8.16 7.21 5.36 4.69 4.34 4.21

30 9.35 8.61 7.72 6.70 6.37 6.17 8.91 7.97 6.74 5.83 5.00 4.63 7.71 6.91 6.46 6.10 5.89 5.73 7.27 6.65 6.17 5.89 5.56 5.54 6.74 6.09 4.56 3.95 3.64 3.48

31 8.91 8.22 7.52 6.56 6.16 6.07 8.56 7.68 6.61 5.71 4.92 4.58 7.36 6.67 6.22 5.93 5.71 5.56 6.91 6.32 5.98 5.66 5.45 5.30 6.42 5.86 4.44 3.88 3.54 3.42

32 8.56 7.84 7.14 6.42 5.96 5.90 8.19 7.50 6.35 5.54 4.84 4.48 7.09 6.41 6.02 5.70 5.49 5.37 6.69 6.15 5.77 5.45 5.24 5.21 6.22 5.66 4.26 3.70 3.40 3.25

33 8.53 7.89 7.17 6.38 6.05 5.95 8.12 7.48 6.44 5.59 4.86 4.51 7.03 6.44 6.04 5.79 5.53 5.45 6.65 6.17 5.75 5.48 5.27 5.13 6.21 5.69 4.28 3.72 3.39 3.22

34 8.91 8.30 7.55 6.60 6.24 6.14 8.71 7.78 6.57 5.69 5.01 4.67 7.45 6.73 6.32 6.01 5.79 5.63 7.05 6.50 6.00 5.68 5.50 5.37 6.57 5.90 4.47 3.84 3.55 3.43

Class 9

Class 10

Class 11

Class 12

Class 13

1 12.51 11.42 10.13 8.68 8.17 7.95 11.77 10.45 8.74 7.39 6.29 5.85 10.43 9.14 8.41 7.81 7.57 7.46 9.71 8.76 7.93 7.62 7.22 7.09 9. 04 7.94 5.81 5.13 4.77 4.62

2 10.71 9. 98 8.93 7.78 7.45 7.40 10.37 9.38 7.93 6.88 6.06 5.64 8.83 7.82 7.37 7.01 6.84 6.70 8.13 7.54 7.11 6.75 6.49 6.40 7.62 6.83 5.40 4.83 4.52 4.35

3 10.82 9. 81 8.83 7.70 7.35 7.15 10.16 9.28 7.82 6.75 5.96 5.60 8.67 7.84 7.34 6.97 6.68 6.56 8.14 7.47 7.03 6.62 6.37 6.28 7.51 6.85 5.35 4.87 4.55 4.45

4 10.86 10.13 9.02 7.90 7.50 7.41 10.45 9. 49 7.93 6.88 6.04 5.70 8. 76 7.89 7.43 7.12 6.81 6.71 8.22 7.49 7.06 6.77 6.54 6.41 7.70 6.92 5.42 4.88 4.62 4.52

5 11.41 10.60 9.42 8.35 7.93 7.84 10.85 10.00 8.41 7.28 6.35 5.92 9.23 8.40 7.89 7.46 7.20 7.01 8.62 7.97 7.46 7.13 6.78 6.72 8.05 7.37 5.65 4.98 4.65 4.45

6 10.89 10.07 9.11 7.93 7.58 7.42 10.52 9. 53 7.96 6.94 6.07 5.63 8. 82 7.94 7.51 7.12 6.83 6.69 8.26 7.68 7.18 6.77 6.48 6.38 7.66 6.98 5.42 4.79 4.46 4.27

7 11.16 10.19 9.20 7.95 7.54 7.40 10.61 9. 60 8.06 6.97 6.09 5.62 9. 03 8.18 7.53 7.18 6.94 6.85 8.50 7.79 7.24 6.93 6.65 6.49 7.77 7.07 5.47 4.88 4.56 4.40

8 11.13 10.28 9.30 8.03 7.59 7.37 10.61 9. 63 8.00 6.88 6.06 5.61 9. 09 8.10 7.57 7.16 6.91 6.70 8.48 7.81 7.27 6.82 6.63 6.43 7.91 7.11 5.37 4.73 4.37 4.24

9 11.81 10.85 9.78 8.43 8.03 7.88 11.39 10.23 8.54 7.33 6.36 5.92 9.67 8.65 8.05 7.72 7.39 7.17 9.02 8.23 7.64 7.31 7.05 6.83 8.41 7.50 5.74 5.03 4.63 4.52

10 12.06 11.14 9.95 8.74 8.16 8.08 11.55 10.51 8.75 7.53 6.55 6.07 9.78 8.73 8.22 7.80 7.49 7.41 9.22 8.43 7.85 7.40 7.09 6.98 8.49 7.68 5.84 5.12 4.74 4.61

11 12.11 11.24 10.06 8.67 8.14 8.06 11.62 10.53 8.74 7.47 6.38 5.99 9.86 8.85 8.26 7.84 7.55 7.36 9.23 8.59 7.86 7.46 7.15 6.93 8.67 7.80 5.85 5.18 4.79 4.62

12 11.86 10.72 9.44 8.10 7.57 7.45 11.21 9. 86 8.09 6.89 6.00 5.60 9. 70 8.52 7.88 7.52 7.13 6.96 9.00 8.21 7.56 7.14 6.86 6.67 8.44 7.49 5.60 4.99 4.70 4.62

13 10.46 9. 63 8.68 7.47 7.12 7.00 9.97 8.96 7.54 6.59 5.78 5.45 8.42 7.58 7.09 6.80 6.48 6.38 7.86 7.24 6.82 6.49 6.26 6.07 7.36 6.64 5.27 4.71 4.44 4.31

14 10.29 9. 47 8.35 7.43 6.93 6.87 9.88 8.85 7.42 6.51 5.66 5.34 8.35 7.46 7.01 6.66 6.44 6.26 7.85 7.09 6.73 6.46 6.16 6.04 7.23 6.57 5.15 4.74 4.40 4.24

15 10.19 9. 27 8.41 7.49 7.20 7.11 9.60 8.70 7.49 6.69 5.94 5.55 8.21 7.44 7.15 6.81 6.67 6.58 7.80 7.19 6.81 6.56 6.32 6.30 7.19 6.72 5.49 4.92 4.68 4.57

16 10.65 9. 53 8.51 7.65 7.51 7.35 9.98 8.92 7.66 7.06 6.21 5.90 8.65 7.67 7.31 7.13 6.85 6.82 8.16 7.42 7.08 6.80 6.71 6.68 7.54 6.89 5.87 5.38 5.15 4.99

17 11.69 10.58 9.26 7.88 7.44 7.35 11.05 9. 62 7.90 6.85 6.08 5.69 9. 52 8.29 7.69 7.30 7.04 6.85 8.92 8.05 7.39 6.98 6.70 6.65 8.15 7.31 5.68 5.12 4.84 4.75

18 11.71 10.57 9.53 8.08 7.53 7.40 11.15 9. 89 8.11 6.95 6.02 5.60 9. 49 8.45 7.88 7.37 7.12 6.93 8.89 8.18 7.46 7.10 6.76 6.67 8.24 7.29 5.53 4.87 4.55 4.50

19 12.76 11.57 10.25 8.75 8.22 8.02 12.07 10.67 8.79 7.47 6.48 6.06 10.44 9.22 8.52 8.02 7.72 7.53 9.90 8.94 8.22 7.72 7.44 7.25 9. 14 8.17 6.14 5.52 5.16 5.00

20 11.46 10.44 9.22 7.96 7.53 7.38 10.73 9. 74 8.01 6.88 6.12 5.78 9. 28 8.23 7.70 7.23 7.03 6.81 8.70 7.91 7.31 6.99 6.71 6.53 8.11 7.21 5.67 5.10 4.79 4.69

21 10.63 9. 72 8.51 7.34 6.91 6.84 10.18 9.01 7.35 6.35 5.52 5.13 8.79 7.67 7.17 6.71 6.47 6.30 8.16 7.33 6.78 6.41 6.21 6.08 7.52 6.73 5.14 4.55 4.26 4.13

22 14.16 12.79 11.21 9.43 8.80 8.64 13.39 11.82 9.35 7.97 6.93 6.40 11.59 10. 12 9.36 8.79 8.32 8.18 10.98 9.79 8.92 8.40 8.03 7.79 10.11 8.88 6.48 5.73 5.34 5.16

23 15.00 13.56 11.84 9.91 9.22 9.18 14.07 12.45 10.10 8.36 7.18 6. 55 12.48 10.78 9.79 9.25 8.81 8.59 11.72 10.40 9.52 8.94 8. 38 8.16 10.76 9.42 6.81 5.82 5.41 5.26

24 14.66 13.38 11.74 10.03 9.40 9.12 13.82 12.19 9.93 8.61 7.32 6.70 12.00 10.55 9.81 9.28 8.76 8.55 11.28 10.06 9.43 8.80 8.35 8.24 10.49 9.24 6.78 5.92 5.46 5. 26

25 10.82 9.96 9.06 7.99 7.53 7.39 10.38 9.51 7.91 6.90 6.02 5.63 8.78 7.88 7.45 7.09 6.79 6.73 8.24 7.58 7.03 6.75 6.48 6.36 7.60 6.88 5.33 4.72 4.37 4.22

26 12.44 11.61 10.22 8.95 8.39 8.16 11.95 10.69 8.93 7.58 6.50 5.98 10.17 9.11 8.47 8.04 7.69 7.47 9.49 8.71 8.06 7.64 7.35 7.12 8.96 7.97 6.00 5.16 4.83 4.68

27 13.83 12.54 11.11 9.34 8.70 8.56 13.04 11.69 9.35 8.01 6.82 6.32 11.32 9.91 9.22 8.68 8.19 8.02 10.59 9.54 8.83 8.31 7.95 7.71 9.87 8.64 6.38 5.56 5.15 5. 00

28 15.29 14.05 12.33 10.63 9.84 9.77 14.68 13.05 10.62 9.07 7.68 7.05 12.71 11.22 10.36 9.76 9.31 9.09 11.87 10.69 9.87 9.35 8.89 8.71 10.97 9.78 7.12 6.13 5. 77 5.54

29 13.54 12.34 10.97 9.47 8.84 8.58 12.90 11.64 9.51 8.06 6.96 6.43 11.17 9.77 9.17 8.68 8.31 8.07 10.39 9.44 8.80 8.23 7.89 7.70 9.71 8.70 6.37 5.58 5.18 5. 04

30 11.51 10.41 9.36 8.24 7.70 7.61 10.99 9.82 8.30 7.07 6.10 5.67 9.36 8.45 7.76 7.42 7.10 6.92 8.74 8.01 7.44 7.04 6.67 6.63 8.08 7.31 5.54 4.87 4.44 4.36

31 11.09 10.11 9.24 7.98 7.66 7.46 10.37 9.55 8.02 6.93 6.09 5.67 8.86 8.03 7.55 7.10 6.90 6.73 8.39 7.67 7.14 6.85 6.58 6.50 7.70 7.03 5.42 4.71 4.41 4.32

32 10.47 9.84 8.83 7.76 7.40 7.23 10.13 9.26 7.77 6.80 5.96 5.49 8.53 7.79 7.34 6.99 6.66 6.55 8.10 7.44 7.05 6.57 6.40 6.27 7.44 6.77 5.19 4.62 4.30 4.12

33 10.54 9.93 8.97 7.90 7.54 7.37 10.19 9.33 7.98 6.90 6.05 5.70 8.48 7.83 7.37 7.05 6.78 6.64 8.01 7.44 7.03 6.65 6.47 6.26 7.42 6.88 5.29 4.62 4.32 4.20

34 11.05 10.23 9.25 8.05 7.55 7.47 10.55 9.67 8.13 6.99 6.04 5.65 9.08 8.14 7.60 7.28 6.92 6.84 8.45 7.73 7.29 6.84 6.61 6.51 7.81 7.04 5.45 4.75 4.43 4.31

Class 9

Class 10

Class 11

Class 12

Class 13

Page 9 of 21

weight for each FHWA vehicle class using a 330-kW fuel cell. As one

would expect, the fuel economy decreases as the vehicle becomes

larger (more trailers and axles) and heavier. Due to the amount of

different origin to destination paths for the current research, fuel

economies for the trip length from an origin to I-75 and I-75 to the

origin were not explicitly calculated. To get a representative fuel

economy for each vehicle class and weight range, the average of the

34 segments on I-75 were used as the fuel economy for those segments

of road not on I-75. This is an area for future research and development

as it increases simulation and calculation time.

Figure 20. 90th percentile GCVW fuel economy for segments along

I-75 for MY2020 vehicles with 330-kW FC

Figure 21. 90th percentile GCVW fuel economy for segments along

I-75 for MY2030 vehicles with 330-kW FC

E. Current Refueling Infrastructure

The current refueling infrastructure is dominated by diesel stations

along major interstate corridors and distribution centers. Freight

typically originates from either a terminal (sea port, inland waterway

port, rail terminal or air freight terminal) and is transported by truck to

a distribution center, warehouse, factory, or other waypoint for further

processing and transport. Much of the freight movement is performed

by diesel trucks until a last mile portion of the system. Refueling can

happen at various locations depending on the carrier operations: 1) at

a private or access-controlled facility such as a port authority or

distribution center, or 2) at public access diesel stations. The fuel for

these stations is typically transported by truck to each location from a

central location, such as a pipeline. Figure 1 shows the typical

refueling infrastructure for diesel trucks.

The goal of the current model framework is to identify, characterize,

and use existing locations as candidate stations for refueling to

represent where vehicles would typically refuel. While a new station

may be readily identified and characterized by the process developed

in this paper, for the purposes of this study it is assumed that the

emerging hydrogen infrastructure will likely begin by phasing in

hydrogen fuel storage/dispensing capabilities at established

(brownfield) refueling sites as opposed to new (greenfield) sites. The

location of current diesel refueling infrastructure was obtained using

publicly available data from the Geotab Ignition [36] dashboard which

uses anonymized customer data to provide data trends. Using this

dashboard, the authors were able to filter the location data to determine

where HD trucks refuel and at what time of day to determine potential

refueling locations for hydrogen trucks. For the purposes of this study,

only refueling stations located in a State along the I-75 corridor, with

a minimum of 75% heavy truck traffic were considered (Figure 22).

This filtering ensured that the refueling stations exported were major

truck stops, where a heavy truck would likely refuel, rather than

refueling stations that may have at least one diesel dispenser.

Figure 22. Location of Refueling Locations near the I-75 corridor

from Geotab Ignition

Data elements exported from the Geotab Ignition dashboard include

the latitude and longitude of the refueling station, the percentage of

heavy truck traffic, and the hourly heavy truck demand. The hourly

demand (Figure 23) was normalized to the total heavy truck population

for the refueling station. While the Geotab data does not represent the

entire population of tractors on the roadways, it is assumed that the

hourly demand obtained is like that of the entire population for this

study.

Page 10 of 21

Figure 23. Distribution of station hourly demand for candidate stations. Each figure represents the demand for a single station along the I-75

corridor. While there are many more fueling stations along I-75, these 68 stations capture the commercial truck refueling stations.

F. New Infrastructure Economics

The capital cost of new infrastructure is critical to optimizing the

location of potential refueling stations for heavy trucks. Using the

methods defined by Liu [22] using the Heavy-Duty Refueling Station

Analysis Model (HDRSAM) [37], capital cost was calculated as a

function of daily demand (number of trucks), dispenser fill rate

(technology), number of required dispensers, and the hourly demand

for a given station. Figure 24 shows example costs calculated using

the HDRSAM model.

Figure 24. Total investment cost and Levelized cost by demand

To incorporate these results into a computational model, absent of the

HDRSAM algorithms, and without calculating every single

permutation vehicle demand, look up tables were generated for each

of dispenser technology (e.g. 1.8, 3.6 and 7.2 kg/min dispensers).

Individual values of demand, dispenser count, or truck volume that

were not explicitly in the table were interpolated between known

values to determine a reasonable cost for the station. Future work on

the model will aim to incorporate a more direct cost function to

accurately calculate the cost of a particular station.

G. Infrastructure Optimization

The optimization workflow for refueling location strategy is a

computational model developed using the previously discussed inputs

using the following steps:

• Step 1: Candidate station generation – Determine valid

refueling locations based on distance from I-75

• Step 2: Route segment and volume generation – consolidate

routes that overlap to reduce computational time

• Step 3: Feasibility calculation – calculate the locations of

refueling and determine how many trucks will stop daily and

how much fuel will be required

• Step 4: Station Cost optimization – calculate the cost of a

station meeting the minimum fueling and demand requirements

prescribed in previous steps

Step 1 – Candidate Station Generation

To determine the location of stations along I-75, the distance each of

the refueling stations obtained from the Geotab Ignition dashboard

from I-75 was determined using a geohash representation. A geohash

is a convenient way of representing latitude and longitude alpha

numerically using a grid-based system [38]. A Geohash is a unique

identifier of a specific region on the Earth. The basic idea is that the

Earth is divided into regions of user-defined size and each region is

assigned a unique identifier based on its latitude and longitude, which

is called its Geohash—an alpha numeric string. This string or Geohash,

will determine which of the predefined regions the point belongs to.

Thus, points within close geographical proximity will have the same

Geohash. The length of the geohash determines the accuracy to

specific location (e.g. 7 characters creates an area of approximately 0.2

square kilometers and 10 characters is approximately an area 0.7

square meters). For this study 7 characters were used for the precision

of accuracy and were considered to be a candidate location if the first

5 characters of the geohash matched a location on I-75 (24 square

kilometer area or within about 6 miles of I-75). This precision was

used because it is assumed that vehicles will be dispatched full of fuel

and not require a refuel for many miles. These candidate stations

(shown in Figure 25) represent the chromosomes in the genetic

algorithm previously discussed.

Page 11 of 21

Figure 25. 68 Candidate station locations along I-75

For some scenarios (Figure 26), multiple refueling stations could be

present at the same location (major junctions could have multiple truck

stops). In this scenario, the stations were grouped into a single location

with the demand averaged for the refueling stations and the maximum

number of dispensers allowed to increase to 20 (maximum parameter)

per station (i.e. 40 dispensers for a location with 2 stations).

Figure 26. Diagram showing grouping of common candidate

locations along I-75

The calculation of candidate stations is deterministic on the route

chosen. For this study, since the focus was on I-75, the candidate

stations were static for all iterations of the model unless parameters are

changed to allow for stations to exist farther from I-75. The genetic

algorithm will be discussed in later sections of this paper.

Step 2 – Route Segment and Volume Generation

Identifying the segments along a route that vehicle will travel is

required to calculate total distance traveled on a particular segment and

the amount of fuel used on that segment. To accomplish this, every O-

D pair is assigned an “entry” and “exit” point on I-75 to calculate both

the distance traveled off I-75 and the distance traveled on I-75. For the

distance traveled off I-75, the distance is simply a calculated routed

distance. Similarly for distance traveled on I-75, the distance is a

calculated routed distance, but additional node information is

available. Thus, a list of locations that a vehicle will travel through on

its trip is calculated.

To efficiently iterate through large lists of trips, and because it is

known that many of these route’s overlap, common routes are grouped

to calculate a total volume for the given segment of I-75.

Step 3 – Feasibility Calculation

To determine the daily total trucks and the amount of fuel required,

each unique route previously calculated is iterated through to

determine the total number of each vehicle class that will stop at each

refueling station and how many kilograms of hydrogen will be

required.

To calculate fuel level during a trip, and the total fuel required by a

particular refueling station, each segment of road in an O-D pair is

iterated over with the total amount of fuel required to traverse that

segment being subtracted from the previous tank level. While

iterating, if the total amount of fuel drops below a parameterized fuel

level, it will fill up at the next location. Since it is assumed as part of

this study that every vehicle starts the trip full, it is known exactly how

much fuel will be required to top off the tank.

After each O-D pair is iterated through, the result is a total number of

vehicles that will stop daily at a given station, and the total amount of

fuel required daily. If a vehicle could not make the trip because it ran

out of fuel due to a refueling station not existing on their route or

simply not having a large enough tank to make the trip, it is considered

infeasible. Each infeasible vehicle would essentially be a waste of

capital for a company and another vehicle capable of making the trip

would have to be used in its place, so a variable infeasibility cost is

applied for each truck that cannot complete the trip.

Figure 27. Refueling strategy for each O-D Pair

Step 4 – Station Cost Optimization

There are stochastic operations in hourly demand share  (%,  

󰇝 󰇞) between stations in one area. Given the total daily fuel

demand  (kg/day) for a station where the capacity is needed to be

designed, the actual demand share could be any of the possible

scenarios. With design refueling speed  (kg/hour) for single fuel

dispenser, the problem is to determine the number of dispensers, , to

be installed at the station to support the total fuel demand 

considering potential delays to truck drivers. As installing dispensers

is capital intense, one objective is to minimize  to support the fueling

demand. On the other hand, with fewer dispensers, delays are possible

to occur when the demand temporally exceeds the supply. Another

objective is the minimize the total delay which could be costly to the

truck business. Therefore, there is a tradeoff between the infrastructure

cost and the operational cost.

There are two simple, albeit limited, methods to determine . One is

based on the average fuel demand. Given the number of operational

hours  (  ) in a day, the average hourly fuel demand is .

Page 12 of 21

Then based on the average fuel demand, it can be determined:

  󰇳

󰇴 (5)

However, such design ignores the stochasticity in operations with non-

uniform distribution on . A station with this design cannot meet the

peak  and could cause significantly delay to drivers.

Another way is the design the capacity based on the maximum hourly

rate, namely:

  󰇳󰇛󰇜

󰇴 (6)

A station with this design will not have any delay to trucks in all

periods but could incur significant capital cost to the station

infrastructure.

A better way is to consider the hourly patterns in all scenarios and

determine  considering the cost tradeoff. A stochastic programming

model is formulated and described here with the following objective

function:

   

󰇝󰇞 (7)

Subject to the following constraints:

  (8)

       (9)

Table 2. Formulation Notation

Set



Indexed by , set of refueling demand pattern

scenarios

Parameters



Total daily fuel demand (kg)



Total daily number of trucks to refuel at the station



Design refueling speed (kg/hour) of the dispenser



Demand share (%) in time  in scenario 



Daily capacity cost ($/day) for a dispenser



Penalty cost on waiting and delay for refueling

($/hour/truck)



Probability share (%) of the scenario 

Variable



Non-negative integer, the number of dispensers to

be installed



Actual fulfilled demand (kg) in time  in scenario 



Carryover remet fuel demand for next hour

󰇛  󰇜

1. Genetic algorithm (GA) overview

To solve the model optimally for the large number of permutations, a

genetic algorithm was adapted from previous research [21] to

determine which refueling stations need to be open and what type of

refueling technology is present at that location. The total number of

possible refueling scenarios is calculated using the one of following

equations:

    



  󰇛  󰇜 (10)

   (11)

Where N is the number of candidate stations, T is the number of

refueling technologies (e.g. 1.8 kg/min @ 350bar, 3.6 kg/min @

350bar, 7.2 kg/min @ 350bar), and S is the number of station states

(i.e., equal to T+1, where the additional state is the null state

representing the lack of hydrogen infrastructure). For the current

problem in this research, the total number of permutations would be

468 which would take a significant amount of calculation time for an

exhaustive search for the optimal configuration.

Even though station capacity can be formulated as piecewise linear

relationship, the mixed integer program is still hard to solve optimally.

We developed a genetic algorithm based heuristic method [39] to solve

the model. Generally, the genetic algorithm involves a pool of

chromosomes. Each chromosome represents one candidate solution to

the problem, and the fitness of each candidate solution is measured by

the cost objective value. Between chromosomes, the one with a lower

objective value indicates better fitness. The algorithm has an iterative

process, during which chromosomes with bad fitness are more likely

to be removed from the pool while chromosomes with good fitness

have more chances to survive and give birth to a new child. The whole

process mimics natural selection. It is generally observed that through

multiple iterations the genetic algorithm can help find near optimal or

relatively satisfactory solutions.

While the genetic algorithm served as a great estimation of the

locations of fueling infrastructure for hydrogen FCEV, future

iterations of the OR-AGENT tool will adapt dynamic programming to

find a more global solution, at the cost of additional compute time and

computational resources.

2. Implementation details

Two features are important to efficiently apply a genetic algorithm.

One is the simplicity in coding the solution with chromosomes, and the

other is to be able to easily determine fitness given the genetic

information of one chromosome. However, both features are hard to

maintain for the proposed model. First, as in Li et al.’s study [40], only

two-dimension infrastructure related decision variables, i.e., where and

when refueling stations are opened, are needed to be encoded in the

chromosome. However, this task is more challenging for this study

because the model requires another dimension on capacity-related

decisions, i.e., how many dispensers per station. That creates

challenges in coding all decision information in each chromosome.

Second, as capacity is considered for each fueling, the heuristic

approach to check path feasibility with infinity capacity [38] is not

applicable to this study. Instead, mixed integer programming sub-

models need to be solved, which makes it difficult to efficiently

determine fitness for each chromosome generated in the solution

process.

To tackle the two challenges in this complex model we made the

following assumption to simplify the FCEV fueling behavior: If an O-

D path at one stage is feasible and selected given fueling infrastructure

deployment, refueling stations along the path will have sufficient

dispensers to satisfy all O-D traffic demands along the path in the same

time stage. This assumption is already inherently built into the

proposed model, which considers only the feasibility of each O-D pair

instead of each O-D trip. This assumption is reasonable to provide

equity for all travel demand along each O-D pair.

This method simplifies the solution process for the model in the

following ways. First, it will significantly simplify the determination

of fitness of each chromosome with heuristic methods. Given the

setting of refueling stations at each time stage encoded in each

chromosome (where and when to open stations), the heuristic approach

[38,39] can be used to efficiently check path feasibility of all O-D

pairs. Following this assumption, we can determine the number of

dispensers needed for each station to satisfy all refueling demand.

Page 13 of 21

Finally, the fitness of the chromosome can be calculated using the

objective function. Second, as suggested in the first benefit, the

number of dispensers per station can be post-calculated. Therefore,

only decisions on where and when refueling stations are opened need

to be encoded in each chromosome. That simplifies the representation

of each chromosome.

3. Settings of genetic algorithms

Based on the above properties and assumptions of the problem, we

made the following settings to apply the genetic algorithm to solve the

proposed model.

3.1 Encoding of chromosome

As previously described, only decisions on where and when fueling

stations are opened need to be encoded into each chromosome. Since

each opened refueling station will remain open once it is in operation,

only the time when the refueling station is first opened needs to be

recorded. Therefore, we used a single dimension integer valued string

to represent such information. Each digit entry along the chromosome

string represents one specific candidate location. Then, the total length

of the string is |N| (the number of candidate locations). Each digit can

take non-negative integer values. When the digit takes a value of 0, it

indicates no refueling station is opened at the location throughout the

time horizon. When the digit takes a positive value of i, it indicates the

type of dispenser technology at the opened station.

Consider the simple example chromosome string “01020” to

demonstrate its meaning on refueling infrastructure planning decisions

as follows:

• Five digits in the string indicate five candidate locations (|N|= 5)

for fueling stations

• No fueling station is opened at candidate locations 1, 3, and 5

• One fueling station is opened at candidate location 2 with 1.8

kg/min dispenser technology

• Another fueling station is opened at candidate location 4 with 3.6

kg/min dispenser technology.

3.2 Fitness of chromosome string solution

The fitness of each chromosome string solution involves several sub-

steps. This is summarized as follows:

• Sub-step 1: initialization that translates genetic information of the

chromosome into corresponding refueling location decisions

• Sub-step 2: check feasibility and select a path for each O-D pair

in all stages

• Sub-step 3: for all feasible O-D pairs determine refueling

activities based on the fuel level

• Sub-step 4: for each opened fueling station determine the total

number of refueling activities and the corresponding required

number of dispensers based on the design level of

• Sub-step 5: calculate the fitness of the chromosome with the

objective function described previously

3.3 Population pool

At beginning of the algorithm, we initialized a pool of chromosome

strings with a population size of N (e.g., 500). The pool is defined as

the population pool where crossover, mutation, and replacement are

applied through an iteration process.

3.4 Parent selection

For each iteration, four candidate chromosomes are randomly

selected from the population pool. The four chromosomes are

partitioned equally into two groups, and one chromosome will be

selected from each group. The chromosome with better fitness (lower

objective value) has a better chance to be selected, and the selection

probability is set to be inverse to its objective value. The final selected

two parents are defined as P1 and P2

3.5 Crossover

The crossover is then applied to the two selected parents (P1 and

P2) to give birth to a new child chromosome C. Let fP1 and fP2 be the

objective values of the parents P1 and P2, respectively, and let i

indexes digits of each chromosome, I = 1,…,|N|. Then the chromosome

C is created as follows:

(1) if P1i = P2i , then set Ci:=P1i or P2i;

(2) otherwise, then set Ci: = P1i with probability p = fP2/(fP1+fP2), and

Ci: = P2i with probability 1 − p.

3.6 Mutation

Once a child chromosome C is created, each element Ci,i = 1,…,|N|,

has a probability (e.g., 5%) to mutate. Let t′ be a random integer value

selected among 1, …, |T| with equal probability. If the element Ci is

selected to mutate, then it has two possible ways to change depending

on the original value of Ci:

(1) if Ci = 0, then set Ci: = t′;

(2) otherwise, then set Ci: = 0 with probability of 50%, and Ci: = t′ with

probability of 50%.

3.7 Replacement

After a new individual is added to the population (e.g., a child

created with both the crossover and mutation processes), it will replace

an existing individual in the population pool. Specifically, n (e.g., n =

3) candidate individuals are randomly selected from the pool, and the

one with the highest objective cost is removed from the pool. The

entire genetic algorithm for the problem can be demonstrated using the

flow logic and simple example as seen in Figure 28:

(a) GA flow logic and convergence selection

Generate Population Pool with Size N

with Chromosome for each Candidate

Station

Conduct Parent Selection Among

Population Pool Based on Individual

Objective Cost

Apply Crossover with Parents and

Yield New Child

Apply Mutation to Child

Add Child to Population

Stop Condition

Met?

Print Optimal Result

Remove Random Individual From

Population Pool

Calculate Child Objective Cost

Lowest

Cost?

New Optimal Result

Yes

Page 14 of 21

(b) Simple example of chromosome strings and their

modifications through various steps of the GA

Figure 28. Flow logic and behavior of the GA

The final solution is the best chromosome achieved with the lowest

objective value from the population. Figure 29 shows the number of

iterations required to converge on an optimal solution using both the

genetic algorithm and iteration process. As shown, many of the

solutions converge to near their optimal values within 10,000 iterations

which depending on computational power requires about 2-3 minutes

per adoption rate.

Figure 29. Generations to converge to optimal objective cost

The number of generations for this solution is similar across adoption

rates mainly due to the unconstrained locations for implementation of

hydrogen refueling stations. In an unconstrained paper, we would

expect fewer iterations as the number of potential station locations is

reduced by the number of stations in the previous generation.

Discussion of Results

For the current study, assumptions were made to reduce the

complexity of the current model framework and to demonstrate the

feasibility of such a model for infrastructure optimization. The

assumptions for the results in this section are as follows:

Table 3. Current Assumptions for Result Generation

Current Assumption

Future Adaptation in Model

All vehicles have a full tank

level prior to dispatch. This

also assumes that a vehicle

can arrive at its destination

with less than full tank

because it will be refueled

prior to being dispatched

again

Additional fueling stations and

infrastructure can be added along

an O-D pair to accommodate

refueling off I-75 (or other

highway system)

All vehicles were assumed to

have a 330 kW FC for this

study. Future iterations of

OR-AGENT will allow for

variable FC sizing depending

on vehicle classification.

Allow for a percentage of

vehicles in the population to have

varying size of FC based on O-D

and vehicle class

Tank sizes were assumed to

be 100 kg for Class 9-11 and

150 kg for Class 12-13.

Realistically, a manufacturer

would not create a different

tank size for each vehicle

unless manufactured at scale.

However, without a current

co-optimization for distance

and fuel tank size this

assumption is being made.

Refine the tank size based on real

world requirements. This can

also be co-optimized based on O-

D and vehicle class

Stations can only have a

single dispenser technology

and all hardware is assumed

to be new construction (not

upgraded)

Allow for a new cost model to

account for multiple dispensing

technologies at a station and

available upgrade paths for

dispensers and other refueling

infrastructure

Fuel economy off I-75 is

assumed to be the average of

the fuel economy on I-75.

This assumption allows for

the study to not calculate FE

on every single route.

Expand the model network to

include fuel economy simulations

for routes and O-Ds off I-75

Data from Geotab is

assumed to be representative

of the entire population of

heavy vehicles

Add additional data sources to

better realize the refueling

patterns of vehicles, (i.e. how

many actually fill up “outside the

fence”) and refueling station

demand

A vehicle is assumed to cost

approximately $400,000 and

is a loss if not able to make

the trip. This is based on the

cost of a typical diesel

vehicle with appropriate

additions to be a FCEV

Allow for varying vehicle costs

based on specific architecture

Gross vehicle weight is

assumed to be constant

throughout the trip, thus fuel

economy won’t drastically

change based on weight

Allow for some variance, but

understanding that the weight will

not change significantly

throughout the trip unless a

vehicle is offloading cargo prior

to destination

Page 15 of 21

No traffic or weather data is

incorporated currently into

the fuel economy

calculations

Allow for the additional of

weather and traffic data which

will likely reduce the overall fuel

economy for the vehicles, hence

requiring more refueling

Hydrogen is plentiful at each

station to accommodate

demand. This is

incorporated into the cost

Better understand or incorporate

methods of transporting or

generating hydrogen for refueling

to calculate total cost.

For this study, an initial result was generated for a scenario where

there are no constraints on what stations can be open or what

technologies can exist at any given station, other than the technology

must be uniform (no mix of 1.8 kg/min or 3.6 kg/min dispensers at a

single location). While this result does not consider all practical facets,

it illustrates what the refueling infrastructure requirements would be if,

overnight, a certain population of vehicles were to be hydrogen. Input

parameters for this result are listed below in Table 4.

Table 4. Parameters Used for Results

Parameter

Description

Default

Value

Number of dispenser

technologies

Integer value of different

dispenser types possible

Dispenser flow rate

Array of flow rate in

kg/min for each dispenser

type

{1.8 3.6 7.2}

Minimum dispensers

per Station

Minimum number of

dispensers a refueling

station must have

Maximum dispensers

per Station

Maximum number of

dispensers a refueling

station can have

Remaining Fuel Limit

Fuel tank level prior to

refilling

20%

Minimum Remaining

Fuel Limit

Minimum fuel tank level

prior to refilling

Fuel Tank Size

Tank size in kg

Class 9-11:

100 kg

Class 12-13:

150 kg

Infeasible Truck

Penalty

Cost of a truck not being

able to make the trip from

origin to destination

because fuel level was not

sufficient

$400,000

Technology Adoption

Rate

Percent of HD Truck

population that is using

hydrogen

User

Defined

The current primary objective of OR-AGENT is to optimize the

location of each refueling station such that the vehicle demand can be

met with the lowest possible cost. Optimization of the location of the

refueling stations based on refueling requirements is critical to reduce

costs, as opposed to building refueling stations at a set uniform location

along I-75. Figure 30 shows the optimal result generated by OR-

AGENT compared to the unoptimized scenario of every refueling

station having a minimum number of dispensers to handle demand. At

low adoption rates this illustrates infrastructure cost increases of >67%

of the unoptimized solution as compared with the optimized solution.

At high adoption rates we note cost increases of >350% of the

unoptimized solution as compared with the optimized solution. Table

5 quantitively provides a comparison of the total cost between the

optimized and non-optimized solutions. Additionally, allowing for not

only a dynamic amount of dispensers in parallel with varying dispenser

flow rates to manage demand has a large impact on total station cost.

Figure 30. Station Capital Cost

Table 5. Station Cost Optimization

The amount of required fuel for all stations, number of stations

required, and the total number of dispensers required for each

generation of adoption rate are shown in Figure 31,

Figure 32, and Figure 33, respectively. Of note, since this is an

unconstrained simulation, it was not required that a refueling station

remain open, or have hydrogen dispensers, as the adoption rate of

hydrogen vehicles increased. Due to this lack of constraint, there are

Optimized

Solution Cost

1.8 kg/min

Dispensers

Increase from

Optimized

3.8 kg/min

Dispensers

Increase from

Optimized

7.2 kg/min

Dispensers

Increase from

Optimized

5% 16,833,268 49,707,165 195.29% 28,230,278 67.71% 39,955,762 137. 36%

10% 18,806,984 57,553,235 206.02% 31,435,091 67.15% 44,185,944 134. 94%

15% 24,093,198 72,636,582 201.48% 42,781,186 77.57% 62,851,917 160. 87%

25% 29,519,634 94,448,890 219.95% 60,028,580 103.35% 82, 199,322 178.46%

35% 41,275,598 133,480,630 223.39% 92,560,844 124.25% 134,898,772 226.82%

45% 48,092,022 156,392,669 225.19% 116,846,795 142.97% 159,661,705 231.99%

100% 55,750,646 251,725,003 351.52% 266,673,109 378. 33% 372,836, 921 568.76%

Page 16 of 21

scenarios where opening a single station in a location can require

fewer dispensers at other locations or cause them to close all together.

This can be seen in the small increase in requirements for fuel and

station cost between the 5% and 10% adoption rates.

Another key finding is that small changes in the technology used

in parallel with specific refueling locations can reduce the number of

stations required and even reduce the number of dispensers required.

For example, the 45% adoption rate requires the most amount of fuel

but has the fewest number of total stations. However, this simulation

also has the largest amount of 3.6 kg/min dispensers and the only

result with a 7.2 kg/min dispenser to handle increased volume.

Figure 31. Daily average fuel requirements for each adoption rate

Figure 32. Number of stations required for each adoption rate

Figure 33. Total dispensers required for each adoption rate

Truck volume, as well as hourly demand, play a large role in where a

refueling station is located and how many dispensers are located at that

location. In the Figure 34 and

Figure 35, the locations of the refueling stations for each adoption rate

are shown with the number of dispensers at each location and the daily

truck volume at each location. There are major junctions that have

higher volume of trucks, especially near the Tennessee and Georgia

border. In part this occurs due to the confluence of several key freight

transport corridors, shipping ports, and distribution centers in this

region. This is well reflected in the need for additional refueling

infrastructure. We expect similar findings on other critical transport

corridors where high volume confluences will result in the need for

additional refueling infrastructure.

Page 17 of 21

Figure 36 shows the results of each generation of adoption rate for

number of stations, daily truck volume per station, and total daily fuel

required per station as a function of distance along I-75. Changes

between generations are highlighted. This is an unconstrained

simulation, where future studies will refine these results with critical

deployment constraints, which would better reflect how a station

grows over time when required to remain open.

Figure 34. Maps of station sizes at each location scaled to number of dispensers at each candidate site

Page 18 of 21

Figure 35. Map of truck volume at each station location

Figure 36. Results of Station size, truck demand and daily fuel requirements for multiple adoption rates (blue – optimum solution, red – change from

previous figure generation)

The resulting data from Figure 36 is shown in tabular form in Figure 37. Interestingly, 34 of the 68 candidate stations were never required

Page 19 of 21

to be open during the simulation, which as previously discussed is a

factor of vehicles have sufficient fuel to make their trip without refueling. This number of stations would increase as more O-D pairs

(further than 500 miles from I-75) were considered.

Figure 37. Resulting output for each adoption rate in the simulation

Diving into the results further, it became evident that the restriction of

O-D Pairs to 500 miles of travel off I-75 reduced the vehicle volume

significantly. In addition, there were many OD pairs that trip distance

was short enough that a vehicle would not be required to refuel at all

during the trip. If the assumption of a full tank prior to dispatch were

not in place, the result would be different, and could potentially require

additional stations off I-75 to be present to accommodate more

vehicles.

Conclusions

The development of a future hydrogen energy economy will require

the development of several hydrogen market and industry segments

including a hydrogen based commercial freight transportation

ecosystem. For a sustainable freight transportation ecosystem, the

supporting fueling infrastructure and the associated vehicle

powertrains making use of hydrogen fuel will need to be co-

established. This paper introduces the OR-AGENT (Optimal Regional

Architecture Generation for Electrified National Transportation) tool

developed at the Oak Ridge National Laboratory, which has been used

to optimize the hydrogen refueling infrastructure requirements on the

I-75 corridor for heavy duty fuel cell electric commercial vehicles.

This constraint-based optimization model considers existing fueling

locations, regional specific vehicle fuel economy and weight, vehicle

origin and destination, vehicle volume by class and infrastructure costs

to characterize in-mission refueling requirements for a given freight

corridor.

This paper has addressed the complexity of modeling the architecture

for the electrification of vehicles and has presented an initial unique

solution in OR-AGENT to optimize the infrastructure location and

cost. This research also demonstrates the ability to integrate different

classes of vehicles, vehicle technologies, and refueling technologies,

as well as other infrastructure requirements, into the OR-AGENT

optimization framework. The specific solution depends on many

different inputs from freight movement information, powertrain

simulations and cost models and illustrates the gaps in data currently

to create a true infrastructure strategy. While the initial results were

promising, there are still many questions on how hydrogen will be

generated at these locations (especially behind the fence at a dispatch

location), how will the components be powered, and other important

questions.

The development of OR-AGENT also makes possible the ability to co-

optimize vehicle components and infrastructure costs. Requiring a

smaller fuel cell on a vehicle, potentially allowing for more freight to

Station

Geohash

Distance

on I-75

(mi)

Pumps

Pump

Tech

(kg/min)

Average

Daily

Volume

Average

Daily

Fuel (kg)

Pumps

Pump

Tech

(kg/min)

Average

Daily

Truck

Volume

Average

Daily

Fuel (kg)

Pumps

Pump

Tech

(kg/min)

Average

Daily

Truck

Volume

Average

Daily Fuel

(kg)

Pumps

Pump

Tech

(kg/min)

Average

Daily

Truck

Volume

Average

Daily

Fuel (kg)

Pumps

Pump

Tech

(kg/min)

Average

Daily

Truck

Volume

Average

Daily

Fuel (kg)

Pumps

Pump

Tech

(kg/min)

Average

Daily

Truck

Volume

Average

Daily

Fuel (kg)

Pumps

Pump

Tech

(kg/min)

Average

Daily

Truck

Volume

Average

Daily

Fuel (kg)

dpkwrpf 206.0437 2 1.8 2 190 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dpkwxnx 210.1037 2 1.8 3 276 2 1.8 4 368 2 1.8 5 459 2 7.2 8 732 2 1.8 11 1005 2 1.8 13 1188 2 1.8 30 2738

dpkt7f7 222.7287 2 1.8 1 95 2 3.6 4 374 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0. 0 0 0

dpsr5j0 240.0427 2 1.8 10 755 0 0.0 0 0 2 3.6 16 1057 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 3 3.6 45 3277

dpsqeeh 247.2377 2 3.6 5 344 3 3.6 16 1179 0 0.0 0 0 4 3.6 24 1749 5 3.6 33 2407 6 3.6 40 2911 5 3.6 29 2157

dpkz9z8 273.8257 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dpksd0z 402.9946 0 0.0 0 0 0 0.0 0 0 2 1.8 6 553 2 1.8 8 737 2 1.8 10 920 2 1.8 13 1196 2 1.8 27 2484

dpkd0w2 432.2494 2 1.8 8 629 2 1.8 9 715 2 1.8 13 991 2 1.8 12 992 2 1.8 15 1269 2 1.8 17 1454 3 1.8 30 2594

dpkd023 434.7037 2 7.2 6 276 2 7.2 6 276 2 7.2 6 276 2 1.8 7 318 2 1.8 8 362 2 1.8 9 407 2 1.8 14 638

dpk232g 462.9283 2 1.8 1 93 2 1.8 1 93 2 1.8 1 93 2 1.8 1 93 0 0.0 0 0 0 0.0 0 0 2 1.8 2 185

dphpw2v 475.534 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 2 1.8 1 69

dphnsxm 486.82 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dphjh3g 507.4949 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 2 1.8 3 235 0 0.0 0 0 2 1.8 2 189

dph5k4p 529.9024 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 2 1.8 1 93

dph4uqb 534.1572 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dph45cg 545.6257 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dph0ctz 560.911 2 1.8 2 186 2 1.8 2 186 2 1.8 2 186 0 0.0 0 0 2 1.8 4 369 0 0.0 0 0 2 1.8 6 559

dph02g4 568.5231 2 1.8 2 189 2 1.8 2 189 2 1.8 3 284 2 7.2 6 561 2 3.6 4 378 0 0.0 0 0 2 1.8 9 851

dngv4mc 615.3501 2 1.8 31 1832 2 1.8 37 2145 3 1.8 42 2454 3 1.8 53 3027 2 3.6 72 4282 3 3.6 96 5807 4 3.6 158 8947

dngv43e 616.9798 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dngf7d1 651.7133 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dngcezh 659.0351 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dns5q18 746.5458 2 1.8 19 1512 2 1.8 19 1512 0 0.0 0 0 3 1.8 27 2071 0 0.0 0 0 0 0.0 0 0 9 3.6 205 13902

dns4mfz 758.3053 2 1.8 30 2029 3 1.8 36 2423 2 1.8 22 1594 4 1.8 51 3393 0 0.0 0 0 0 0. 0 0 0 0 0.0 0 0

dn7f5w9 762.6952 2 1.8 5 471 2 1.8 5 471 2 1.8 5 471 0 0.0 0 0 2 7.2 6 566 2 7.2 6 566 0 0.0 0 0

dnknfyp 806.7887 0 0.0 0 0 0 0.0 0 0 2 1.8 11 1024 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dnkh49j 807.6362 2 1.8 24 1679 2 1.8 27 1807 2 1.8 17 805 0 0.0 0 0 2 1. 8 32 1485 2 1.8 39 1795 6 3.6 188 11595

dnkm2h8 830.468 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 6 3.6 115 8596 7 3.6 132 9849 0 0.0 0 0

dnkkc6j 839.1506 0 0.0 0 0 0 0.0 0 0 4 3.6 42 3294 0 0. 0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dn78tzq 910.2876 2 1.8 16 1226 2 1.8 17 1303 2 1.8 21 1615 0 0.0 0 0 3 1.8 34 2553 4 1.8 40 2995 0 0.0 0 0

dn5xfyv 922.1542 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dn5rj1j 941.7203 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 9 3.6 81 5284 0 0.0 0 0 0 0.0 0 0 0 0. 0 0 0

dn5qvt0 943.4148 2 1.8 15 1128 2 1.8 15 1128 2 1.8 17 1268 0 0.0 0 0 2 1.8 22 1636 2 1.8 24 1775 0 0.0 0 0

dn5t3f0 968.0014 2 1.8 8 737 2 1.8 8 737 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dn5t45v 970.7255 0 0.0 0 0 2 3.6 3 281 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dn5se68 977.8101 2 1.8 2 190 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dn5eecs 990.7216 2 1.8 6 564 2 1.8 6 564 2 1.8 14 1270 2 1.8 14 1270 2 1.8 18 1630 2 1.8 19 1719 0 0.0 0 0

dn5dynm 999.8799 2 1.8 10 926 2 1.8 12 1111 2 1.8 14 1296 2 1.8 17 1572 2 1.8 24 2127 2 1.8 27 2496 5 3.6 169 13937

dn5dpf6 1012.729 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dn5cenu 1022.947 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dn5bqye 1041.102 2 1.8 5 468 2 1.8 5 468 2 1.8 5 468 2 1.8 5 468 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djgzxr4 1051.622 2 1.8 9 780 2 1.8 9 780 2 1.8 9 782 2 1.8 12 1055 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djgzr3v 1056.707 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djup0b8 1062.083 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djujnqm 1089.208 0 0.0 0 0 0 0.0 0 0 2 3.6 2 189 2 3.6 2 189 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djuhzht 1093.398 2 1.8 12 1108 2 1.8 12 1108 2 1.8 15 1388 2 1.8 21 1940 3 1.8 34 3106 3 1.8 39 3565 0 0.0 0 0

dju6qbk 1137.829 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dju3w2s 1147.075 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 8 3.6 207 13754

dju2wqd 1156.915 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djsqtbt 1183.938 2 3.6 4 379 2 3.6 4 379 2 3.6 4 379 2 3.6 4 379 2 1.8 7 663 2 1.8 7 663 0 0.0 0 0

djsmyhk 1191.626 2 1.8 9 820 2 1.8 14 1231 2 1.8 16 1501 2 1.8 18 1607 2 1.8 23 2133 2 1.8 25 2206 0 0.0 0 0

djsmy1q 1192.694 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0. 0 0 0 0 0.0 0 0 0 0.0 0 0

djsdg95 1232.871 2 1.8 9 816 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djsdey6 1234.078 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djkycje 1282.092 0 0.0 0 0 2 7.2 6 561 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 2 7.2 6 569 0 0.0 0 0

djky741 1290.962 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 2 1.8 2 283 0 0.0 0 0 2 1.8 2 283 0 0.0 0 0

djkvw51 1302.727 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 2 1.8 4 379 2 3.6 5 474 2 1.8 6 569 0 0.0 0 0

djmj0yf 1311.328 2 1.8 16 1459 2 1.8 16 1459 2 1.8 18 1644 2 1.8 20 1830 2 1.8 24 2201 2 7.2 101 8199 0 0.0 0 0

djm68sf 1348. 713 2 1.8 8 693 2 1.8 8 693 2 1.8 9 774 2 1.8 11 957 2 1.8 13 1140 0 0.0 0 0 0 0.0 0 0

djjw3xj 1405.313 2 1.8 14 1298 2 1.8 15 1390 3 1.8 33 2830 2 1.8 19 1757 3 1.8 28 2592 0 0.0 0 0 0 0.0 0 0

djjt47p 1422.402 2 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djjs5zq 1433.75 2 1.8 16 1329 2 1.8 17 1423 0 0.0 0 0 2 3.6 24 2009 2 3.6 32 2688 0 0.0 0 0 0 0.0 0 0

djjee9v 1442.21 2 3.6 2 188 0 0.0 0 0 2 3.6 2 188 2 3.6 2 188 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

djje4dr 1448.529 2 0.0 0 0 2 3.6 3 277 0 0.0 0 0 0 0.0 0 0 2 7.2 5 460 0 0.0 0 0 0 0.0 0 0

djj3neu 1475.968 2 1.8 1 94 0 0.0 0 0 2 1.8 2 188 2 1.8 2 188 0 0.0 0 0 0 0.0 0 0 2 1.8 8 753

dhv8ypd 1599.943 2 0.0 0 0 0 0.0 0 0 0 0. 0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dhtzdeg 1621.545 2 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0

dhwg7eg 1755.722 2 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0.0 0 0 0 0. 0 0 0 0 0.0 0 0

Total 76 311 24759 62 338 26631 62 372 29319 67 455 35028 60 582 45277 53 661 50209 63 1131 78721

100% (nS=18)

5% (nS=33)

10% (nS=30)

15% (nS=29)

25% (nS=27)

35% (nS=25)

45% (nS=20)

 
Page 20 of 21 
be moved, in conjunction with more stations being open for refueling 
could have many impacts on freight efficiency and how vehicles are 
produced.  Additionally, there are opportunities to explore adding new 
refueling  stations  to  remote  locations  to  accommodate  existing 
infrastructure, such as  the power  grid, rather  than retrofitting  diesel 
stations to accommodate hydrogen vehicles.  More so, safety impact 
of potential queuing lines at refueling stations, fuel handling, and other 
areas must be explored. 
The  results  presented  only  show  an  example  of  unconstrained 
parameters, future papers will focus on a more realistic approach to 
infrastructure  expansion  and  use  of  real-world  parameters  and 
operational scenarios. The addition of traffic modeling, realistic driver 
behavior (stopping for lunch etc.) and sources of data show promise in 
creating  an  effective  tool  at  designing  the  future  electrified 
transportation network 
References 
1. Eckerle, W., Sujan, V., and Salemme, G., "Future Challenges 
for Engine Manufacturers in View of Future Emissions 
Legislation," SAE Technical Paper 2017-01-1923, 2017, 
https://doi.org/10.4271/2017-01-1923.  
2. CARB (2020). Accessed December 20, 2021. 
https://ww2.arb.ca.gov/our-work/programs/heavy -duty-low-
nox/heavy-duty-low-nox-updates.  
3. ICCT (2020). Accessed December 20, 2021. 
https://theicct.org/publications/california-hdv-ev-update-
jul2020.  
4. EPA (2022). Accessed March 31, 2022.  
https://www.govinfo.gov/content/pkg/FR-2022-03-28/pdf/2022-
04934.pdf.  
5. CARB (2019).“Advanced Clean Trucks Regulation,” Accessed 
December 20, 2021. 
https://ww2.arb.ca.gov/rulemaking/2019/advancedcleantrucks.. 
6. Butler, D. (2021). Accessed December 20, 2021. 
https://www.washingtonpost.com/climate-
environment/2021/03/05/more-than-50-companies-have-vowed-
be-carbon-neutral-by-2040.  
7. Dean G. (2021). Accessed December 20, 2021. 
https://www.businessinsider.com/fedex-delivery-fleet-all-
electric-carbon-neutral-2040-sustainability-2021-3.  
8. Mathers (2020). Accessed December 20, 2021. 
https://business.edf.org/insights/walmart-commits-to-100-zero-
emission-trucks-by-2040-signaling-electric-is-the-future.  
9. Wikipedia (2021). Accessed December 20, 
2021.https://en.wikipedia.org/wiki/2021_United_Nations_Clima
te_Change_Conference.  
10. DOE (2021). Accessed May 31, 2022. 
https://www.energy.gov/articles/us-launches-net-zero-world-
initiative-accelerate-global-energy-system-decarbonization.  
11. Roeth, M., "Transformational Technologies Reshaping 
Transportation - An Industry Perspective," SAE Int. J. Adv. & 
Curr. Prac. in Mobility 3(1):5-48, 2021, 
https://doi.org/10.4271/2020-01-1945.  
12. NACFE (2020). “Guidance Report: Making Sense of Heavy-
Duty Hydrogen Fuel Cell Tractors”. North American Council 
for Freight Efficiency. Report published December 2020.  
13. Greene, D. L. and Plotkin, S. “Reducing greenhouse gas 
emission from U.S. transportation”. Technical Report; Center 
for Climate and Energy Solutions: Arlington, VA, 2011.  
14. Bui A., Slowik P., Lutsey, N. “Power play: Evaluating the U.S. 
position in the global electric vehicle transition”. International 
Council on Clean Transportation (ICCT), June 29, 2021 
Briefing. 
15. NACAD (2021). Ntnl. Acad. of Sciences, Engineering and 
Medicine. “Assessment of Technologies for Improving Light-
Duty Vehicle Fuel Economy—2025-2035”. Natl. Acad. Press: 
DC, 2021. 
16. Sujan, V.A. “Developing Fuel Cell Electric Powertrain 
Architectures for Commercial Vehicles”. Submitted to the SAE 
International Journal of Sustainable Transportation, Energy, 
Environment, & Policy, June 2022. 
17. Guerrero, A., Davendralingam, N., Raz, A., DeLaurentis, D., 
Shaver, G., Sujan, V., and Jain, N. “Projecting adoption of truck 
powertrain technologies and CO2 emissions in line-haul 
networks”. Transportation Research Part D: Transport and 
Environment, Volume 84, 2020. 
18. Sujan, V., Xie, F., and Smith, D., "Achieving Diesel Powertrain 
Ownership Parity in Battery Electric Heavy Duty Commercial 
Vehicles Using a Rapid Recurrent Recharging Architecture," 
SAE Int. J. Adv. & Curr. Prac. in Mobility 4(4):1166-1180, 
2022, https://doi.org/10.4271/2022-01-0751.  
 
19. van Grinsven, A.H. et al. 2021, Research for TRAN Committee 
– Alternative fuel infrastructuresfor heavyduty vehicles, 
European Parliament, Policy Department for Structural and 
Cohesion Policies, Brussels 
 
20. Minjares, Ray et all. 2021, Infrastructure to support a 100% 
zero-emission tractor-trailer fleet in the United States by 2040.  
 
21. Uddin, Majbah, Liu, Yuandong, Xie, Fei, Sujan, Vivek, & 
Siekmann, Adam. Modeling Freight Traffic Demand and 
Highway Networks for Hydrogen Fueling Station Planning: A 
Case Study of U.S. Interstate 75 Corridor. United States. 
 
22. Liu, Yuandong, Siekmann, Adam, Sujan, Vivek, Uddin, 
Majbah, Xie, Fei, & Shiqi Ou. Providing Levelized Cost and 
Waiting Time Inputs for Hydrogen Refueling Station Planning: 
A Case Study of U.S. I-75 Corridor. United States. 
 
23. Xie, F., Lin, Z. (2021). Integrated U.S. nationwide corridor 
charging infrastructure planning for mass electrification of inter-
city trips, Applied Energy, Volume 298, 2021, 117142. 
24. Hwang, S. W., Kweon, S. J., & Ventura, J. A. (2015). 
Infrastructure development for alternative fuel vehicles on a 
highway road system. Transportation Research Part E: Logistics 
and Transportation Review, 77, 170-183.  
25. ORNL (2021). http://faf.ornl.gov/faf5/. Accessed July 12, 2021. 
26. Hwang, H. L., Lim, H., Chin, S. M., Wang, R., & Wilson, B. 
(2019). Exploring the Use of FHWA Truck Traffic Volume and 
Weight Data to Support National Truck Freight Mobility Study 
(No. ORNL/TM-2019/1385). Oak Ridge National Lab, Oak 
Ridge, TN. 
 
27. FMCSA (2022). Summary of Hours of Service Regulations. 
Accessed June 30, 2022. 
https://www.fmcsa.dot.gov/regulations/hours-service/summary-
hours-service-regulations .   
 
28. FHWA. (2013). Traffic Monitoring Guide. 
 
29. Nokia (2022). Accessed June 30, 2021. https://www.here.com/.  
30. Santhanagopalan, S., Smith, K., Neubauer, J., Kim, G-H., 
Pesaran, A., and Keyser, M. (2014). “Design and Analysis of 
Large Lithium-Ion Battery Systems”.  2014, ISBN: 978-1-
60807-713-7   
31. He, H., Liu, Y., Liu, Q., Li, Z., Xu, F., Dun, C., Ren, Y., Wang, 
M-X., and Xiea, J. (2013). “Failure Investigation of LiFePO4 
Cells in Over-Discharge Conditions”. Journal of The 
Electrochemical Society, 160 (6) A793-A804 (2013) 
32. Xu, F., He, H., Dun, C., Liu, Y., Wang, M-X, Liu, Q., Ren, Y., 
and Xiea, J. (2012). “Failure Investigation of LiFePO4 Cells 

Page 21 of 21

under Overcharge Conditions”. ECS Transactions, 41 (39) 1-12

(2012)

33. FPRF (2011). “Lithium-Ion Batteries Hazard and Use

Assessment”. Exponent Failure Analysis Associates, Inc.,

published by the Fire Protection Research Foundation (2011)

34. Golubkov, A., Fuchs, D., Wagner, J., Wiltsche, H., Stangl, C.,

Fauler, G., Voitic, G., Thalera, A., and Hackere, V.

(2014).“Thermal-runaway experiments on consumer Li-ion

batteries with metal-oxide and olivine-type cathodes”. Royal

Society of Chemistry, RSC Adv., 2014, 4, 3633

35. Zhang, R., Xia, B., Li, B., Cao, L., Lai, Y., Zheng, W., Wang,

H., and Wang, W. (2018). “State of the Art of Lithium-Ion

Battery SOC Estimation for Electrical Vehicles”. Energies 2018,

11, 1820.

36. Geotab (2022). Acceessed October 15, 2021.

https://www.geotab.com

37. Mintz, M., Gillette, J., Elgowainy, A., Paster, M., Ringer, M.,

Brown, D., &Li, J. “Hydrogen delivery scenario analysis model

for hydrogen distribution options”. Transportation Research

Record: Journal of the Transportation Research Board, 1983

(2006), pp.114–120.

38. Geohash. Accessed July 27, 2022.

https://en.wikipedia.org/wiki/Geohash.

39. Vose, M.D., 1999. The Simple Genetic Algorithm: Foundations

and Theory. MIT press.

40. Li, S., Huang, Y., Mason, S.J., 2016. A multi-period

optimization model for the deployment of public electric vehicle

charging stations on network. Transport. Res. Part C: Emerg.

Technol. 65, 128–143.

41. Li, S., Huang, Y., 2014. Heuristic approaches for the flow-based

set covering problem with deviation paths. Transport. Res. Part

E: Log. Transport. Rev. 72, 144–158.

Definitions/Abbreviations

ACT

Advanced Clean Truck

BEV

Battery Electric Vehicle

CARB

California Air Resources Board

CO2

Carbon Dioxide

DER

Distributed Energy Resources

FAF

Freight Analysis Framework

Fuel Cell

FCEV

Fuel Cell Electric Vehicle

FHWA

Federal Highway Administration

FMCSA

Federal Motor Carrier Safety

Administration

Genetic Algorithm

GHG

Greenhouse Gas

Hydrogen

HOS

Hours of Service

Heavy Duty

HDRSAM

Heavy-Duty Refueling Station

Analysis Model

Kilowatt

Li-ion

Lithium Ion

LTO

Lithium Titanate

Model Year

NMC

Nickel Manganese Cobalt oxide

NZEV

Near-Zero Emissions Vehicle

O-D

Origin-Destination

OR-AGENT

Optimum Regional Architecture

Generation for Electrified National

Transport

ORNL

Oak Ridge National Laboratory

REVISE

Regional Electric Vehicle

Integrated System Evolution

TCO

Total Cost of Ownership

TMAS

Travel Monitoring Analysis System

WIM

Weight In Motion

ZEV

Zero Emissions Vehicle

Page 22 of 21

ResearchGate has not been able to resolve any citations for this publication.

Modeling Freight Traffic Demand and Highway Networks for Hydrogen Fueling Station Planning: A Case Study of U.S. Interstate 75 Corridor

Conference Paper

Full-text available

Jul 2022

The use of hydrogen as an alternative transportation fuel has gained much interest in recent years. Hydrogen can be utilized in electric vehicles equipped with hydrogen powertrains (including hydrogen internal combustion engines or fuel cells). Given that most of the freight in the U.S. is transported via diesel trucks, transition to hydrogen fuel would help achieve significant environmental benefits as well as accelerate the decarbonization of the freight transportation sector. This paper presents the methodology and results of a case study on modeling freight traffic demand and highway networks based on publicly available data for the Interstate 75 freight corridor. The purpose of this study is to prepare input traffic and network data that can support the planning of a hydrogen fueling station infrastructure. In particular, the data can be used for siting and characterizing an optimized framework of hydrogen fueling stations from candidate diesel stations along the Interstate 75 corridor. The methodologies developed and presented in this paper may be readily expanded and applied to any transport corridor given the data availability. This paper is the first in a series that will build out a comprehensive model to optimize a consolidated national hydrogen refueling infrastructure eco-system targeted at commercial vehicles.

Providing Levelized Cost and Waiting Time Inputs For Heavy-duty Vehicle Hydrogen Refueling Station Planning: A Case Study of U.S. I-75 Corridor

Conference Paper

Full-text available

Jul 2022

Widespread use of diesel fuels in freight transportation leads to greenhouse gas (GHG) emissions that play an important role in air pollution. The air pollution, as well as the energy crisis, drives the transition from diesel fuel to cleaner and more energy-efficient fuels such as hydrogen fuels. With such a transition, the strategic planning and optimization of supporting refueling infrastructure along the national highway for heavy-duty vehicles (HDVs) become a necessity. Cost and service level are two essential factors to be considered in hydrogen refueling station location and capacity optimization. The service level in this study is assessed by the delay (waiting time) at fueling stations as well as the fueling demand fulfilled. To this end, this study presents a methodology to provide waiting time (delay) and levelized hydrogen fuel cost inputs under different station configurations (number of dispensers and fill rates) for hydrogen refueling station planning. Determining the station configurations based on the trade-off between the two inputs is further discussed. Particularly, this study points out that the levelized cost and waiting time are affected by hourly demand patterns and estimates the two inputs under different peak hour fueling demand scenarios. The results suggest that, with the same daily demand, the cost, as well as the waiting time, increases obviously with the peak hour demand. In the case study, the cost grows at least 30% when the peak hour visit frequency of the daily total visits increases from 5% (evenly distributed pattern) to 10% (the most common pattern for existing diesel fueling stations along I-75). Accounting for this peak hour effect in hydrogen refueling station planning is recommended. Overall, the waiting time model and cost analysis provide key inputs for optimizing hydrogen refueling station location and configurations based on anticipated refueling demand patterns. The paper is the second in a series that aims to build a comprehensive modeling plan and optimize hydrogen refueling infrastructure along the Interstate 75 (I-75) corridor for HDVs.

Power play: Evaluating the U.S. position in the global electric vehicle transition

Technical Report

Full-text available

Jun 2021

The United States is the third largest electric vehicle manufacturer, behind China and Europe. From 2017 to 2020, the U.S. share of cumulative global electric vehicle production since 2010 decreased from 20% to 18%, while manufacturing increased in the other two regions. The electric vehicle production dynamics reflect where there has been the most electric vehicle market growth and where policies are encouraging the transition to electric vehicles. About 15% of the approximately $345 billion in global automaker electric vehicle investments appear to be destined for the United States. Based on the company announcements through 2020, about 5% of this global total is actively being invested in specific U.S. assembly plants to increase electric vehicle production. Similarly, of the automakers’ announcements that sum to 22 million annual electric vehicle sales by 2025, about 2.3 million (or about 10%) are slated to be manufactured in the United States. Seven of the 44 major U.S. assembly plants, representing about 16% of U.S. vehicle production and capacity, will manufacture only electric vehicles in 2025. The U.S. electric plants include three owned by General Motors, two by Tesla, and one each by emerging electric vehicle companies Rivian and Lucid Motors. Five automakers (Ford, Fiat Chrysler, Toyota, Honda, and Nissan) that each produce about 900,000 to 1.5 million vehicles annually have not announced plans for electric vehicle-only assembly plants. Based on this assessment, more investigation is needed to identify and recommend targeted electric vehicle policy actions to bolster the U.S. auto industry. With 44 major manufacturing plants and production volume of more than 10 million light-duty vehicles, the United States has both a risk of continuing to lag developments in China and Europe and also an opportunity to increase its share of global electric vehicle production.

State of the Art of Lithium-Ion Battery SOC Estimation for Electrical Vehicles

Article

Full-text available

Jul 2018

Sate of charge (SOC) accurate estimation is one of the most important functions in a battery management system for battery packs used in electrical vehicles. This paper focuses on battery SOC estimation and its issues and challenges by exploring different existing estimation methodologies. The key technologies of lithium-ion battery state estimation methodologies of the electrical vehicles categorized under five groups, such as the conventional method, adaptive filter algorithm, learning algorithm, nonlinear observer, and the hybrid method, are explored in an in-depth analysis. Lithium-ion battery characteristic, battery model, estimation algorithm, and cell unbalancing are the most important factors that affect the accuracy and robustness of SOC estimation. Finally, this paper concludes with the challenges of SOC estimation and suggests other directions for possible research efforts.

Developing Fuel Cell Electric Powertrain Architectures for Commercial Vehicles

Article

Dec 2022

Vivek Anand Sujan

The Simple Genetic Algorithm: Foundations and Theory

Book

Aug 1999

Michael D. Vose

The Simple Genetic Algorithm (SGA) is a classical form of genetic search. Viewing the SGA as a mathematical object, Michael D. Vose provides an introduction to what is known (i.e., proven) about the theory of the SGA. He also makes available algorithms for the computation of mathematical objects related to the SGA. Although he describes the SGA in terms of heuristic search, the book is not about search or optimization per se. Rather, the focus is on the SGA as an evolutionary system. The author intends the book also to serve as an outline for exploring topics in mathematics and computer science in a goal-oriented way. Bradford Books imprint

Achieving Diesel Powertrain Ownership Parity in Battery Electric Heavy Duty Commercial Vehicles Using a Rapid Recurrent Recharging Architecture

Conference Paper

Mar 2022

div class="section abstract"> Battery electric vehicles (BEV) in heavy duty (HD) commercial freight transport face challenging technoeconomic barriers to adoption. Specifically, beyond safety and compliance, fleet and operational logistics require both high up-time and parity with diesel system productivity/Total Cost of Ownership (TCO) to enable strong adoption of electrified powertrains. At present, relatively high energy storage prices coupled with the increased weight of BEV systems limit the practicality of HD commercial freight transport to shorter range applications, where smaller batteries will suffice for the mission energy requirements (single operational shift). This paper presents an approach to extend the feasibility of BEV HD trucking for a broad range of applications. The concept is based on the leveraging rapid and recurrent recharging of a BEV HD truck that may either already make frequent stops due to shipment drop-offs/reloading or be required to make frequent stops along with longer missions for recharging. While the challenges of the latter are well appreciated, the concept proposed explores making minimal impact to overall mission time by targeting high C-rate charging while optimizing the frequency (miles) through which these events must occur. The concept optimizes battery size and chemistry, such that the expected life (years and total energy throughput) is balanced with the depth of discharge between recharging events, thus making complete use of the energy available through the life of the battery system. The solution is constrained to minimize the impact on payload capacity. The paper analyzes critical levers in achieving diesel price parity (based on a simplified vehicle TCO), achieved through different purchase options (including lease versus buy) and operational models (half-life swap out). Finally, the paper identifies the application design domain where these solutions are viable with limited impact on fleet operations. </div

Integrated U.S. nationwide corridor charging infrastructure planning for mass electrification of inter-city trips

Article

Sep 2021
APPL ENERG

This study introduces an integrated modeling framework to evaluate long-term national corridor charging infrastructure requirements in the United States to support the growing inter-city charging demand with the rapid growth in the battery electric vehicle (BEV) market. The core model is an optimization model that considers spatial and temporal dimensions and models heterogeneous behaviors between travelers. The model also introduces the travelers’ inconvenience cost function by linking travelers’ acceptance of the charging infrastructure with exogenous technology and social factors. The inconvenience cost function simulates mode choice between BEVs and alternative modes by heterogenous travelers. We applied the framework to assess the inter-regional charging infrastructure requirements for the entire U.S. mainland interstate highway network. We evaluated impacts on the infrastructure design and its public acceptance with changes in policy, technology, and demographic characteristics, and we also quantified the importance of modeling full-scale inter-regional charging infrastructure requirements compared to the conventional regional level analyses.

Transformational Technologies Reshaping Transportation - An Industry Perspective

Conference Paper

Sep 2020

Michael Roeth

div class="section abstract"> Freight trucking has always embraced continuous improvement although the general public may be largely unaware of the tremendous changes that have occurred to date in vehicle safety, efficiencies, emissions mitigation and asset utilization. Today’s freight industry is in the early stages of dramatic technological advances in areas such as connectivity, automation and electrification. Fleet and manufacturer perspective on these transformational technologies are universally supportive, driven both by economic and regulatory factors. The early stages of these technological revolutions are characterized by a multitude of competing concepts confusing choices, complicated by a significant number of competing vested interests obscuring objectivity, and clouded by billions of dollars of venture capital. The objective of this report is to provide unbiased insight into the freight industry’s thoughts on these transformational technologies. </div

Projecting adoption of truck powertrain technologies and CO2 emissions in line-haul networks

Article

Jul 2020
TRANSPORT RES D-TR E

In this paper we present a mixed-integer linear program to represent the decision-making process for heterogeneous fleets selecting vehicles and allocating them on freight delivery routes to minimize total cost of ownership. This formulation is implemented to project alternative powertrain technology adoption and utilization trends for a set of line-haul fleets operating on a regional network. Alternative powertrain technologies include compressed (CNG) and liquefied natural gas (LNG) engines, hybrid electric diesel, battery electric (BE), and hydrogen fuel cell (HFC). Future policies, economic factors, and availability of fueling and charging infrastructure are input assumptions to the proposed modeling framework. Powertrain technology adoption, vehicle utilization, and resulting CO2 emissions predictions for a hypothetical, representative regional highway network are illustrated. A design of experiments (DOE) is used to quantify sensitivity of adoption outcomes to variation in vehicle performance parameters, fuel costs, economic incentives, and fueling and charging infrastructure considerations. Three mixed-adoption scenarios, including BE, HFC, and CNG vehicle market penetration, are identified by the DOE study that demonstrate the potential to reduce cumulative CO2 emissions by more than 25% throughout the period of study.

Optimizing Hydrogen Fueling Infrastructure Plans on Freight Corridors for Heavy-Duty Fuel Cell Electric Vehicles

Abstract

Recommended publications

Optimizing Long Term Hydrogen Fueling Infrastructure Plans on Freight Corridors for Heavy Duty Fuel...

Assessing Powertrain Technology Performance and Cost Signposts for Electrified Heavy Duty Commercial...

Modeling Freight Traffic Demand and Highway Networks for Hydrogen Fueling Station Planning: A Case S...

Achieving Diesel Powertrain Ownership Parity in Battery Electric Heavy Duty Commercial Vehicles Usin...