Geographical Modeling of Charging Infrastructure Requirements for Heavy-Duty Electric Autonomous Truck Operations

Abstract

This study presents an analysis of the charging infrastructure requirements for autonomous electric trucks (AETs) in a specified geographical region, focusing on the state of Texas as a case study. A discrete-time, agent-based model is used to simulate the AET fleet and consider various model parameters such as trip distance/duration, the number of trips, and charging speeds. The framework incorporates unique properties of the Texas road network to assess the sensitivity of charging infrastructure needs. By synergizing electrification and automation, AETs offer benefits such as reduced carbon emissions, enhanced transportation safety, decreased congestion, and improved operational costs for fleets. By simulating daily trips and energy consumption patterns, an analysis of the charging infrastructure needs for cities along the Texas highway triangle formed by I-35, I-45 and I-10 revealed that the total charging energy and average charging power for these major cities ranges between 443~533 MWh/day and 18.5~22 MW, with costs in the range of USD $7.74~$15.93 million for each city, depending on charging infrastructure design and exclusive of any enhancements to the distribution grid infrastructure needed to support the charging infrastructure. This data-driven approach may be replicated for other regions by adapting the simulation parameters to allow policymakers and stakeholders to assess the charging infrastructure requirements and related investments needed to support the transition to electric and autonomous heavy-duty trucking.

1. Introduction

The transportation sector, particularly heavy-duty trucks, plays a vital role in the global economy. However, the predominance of diesel-powered heavy-duty trucks has raised significant concerns due to their substantial contribution to carbon and greenhouse gas (GHG) emissions. In the United States, medium and heavy-duty trucks account for approximately 26% of the total GHG emissions from the transportation sector, while light-duty vehicles make up 57% of them [1]. While heavy-duty trucks make up a smaller proportion compared to passenger vehicles in terms of volume, their emissions have a significant impact due to their high mileage and relatively lower energy efficiency measured in miles per gallon (mpg). As an example, heavy-duty diesel trucks generally have a fuel efficiency of 6–7 mpg (6 mpg ≈ 40 L/100 km, 7 mpg ≈ 33 L/100 km), whereas passenger vehicles typically average between 25–30 mpg (25 mpg ≈ 9.4 L/100 km, 30 mpg ≈ 7.8 L/100 km) [2]. According to a report by the U.S. Department of Transportation, Federal Motor Carrier Safety Administration, in 2017, among the 272,480,899 total registered vehicles in the United States, 9,336,998 (3.42% of total) were single-unit trucks (straight trucks), and 2,892,218 (≈1% of total vehicles) were combination trucks (tractor-trailers) [3]. In addition, in 2017, there were 3212.3 billion VMT by all motor vehicles, but large trucks traveled 297.6 billion of those miles (9.3 percent of the total) [3]. This underscores the significant role that heavy trucks play in terms of total miles traveled and the potential impact on emissions reduction through electrification.

Diesel heavy-duty trucks also incur high maintenance and operational costs. These costs include fuel expenses, engine maintenance, and exhaust after-treatment systems required to comply with strict emission standards. For example, diesel particulate filters (DPFs) and selective catalytic reduction (SCR) systems require regular maintenance and can be costly to replace [4]. Transitioning to autonomous electric heavy-duty trucks (AETs) can help alleviate these challenges by reducing maintenance and operational costs.

AETs also offer a promising solution to address the safety challenges associated with diesel trucks. Currently, heavy-duty trucks are involved in a significant number of accidents and fatalities; in 2020, approximately 4965 fatalities occurred in crashes involving large trucks in the US [5]. The primary causes for these accidents include driver fatigue, distraction, and human error. Moreover, the trucking industry faces a driver shortage, with an estimated deficit of 60,800 drivers in 2020 [6], which exacerbates the problem. Because the majority of driven miles for heavy-duty trucks occur on highways, where traffic patterns are more predictable and pedestrian interactions are minimal, heavy-duty truck highway driving automation is a simpler technological target when compared to urban driving scenarios.

The development of an AET fleet holds immense potential to reduce GHG emissions, enhance safety, and address the challenges faced by the trucking industry. However, the successful implementation of AETs requires careful consideration of the charging infrastructure. This study aims to investigate the key factors influencing the implementation of a charging infrastructure for AETs and proposes strategies to accelerate their adoption.

Building upon previous studies, this research endeavors to provide valuable insights into the development and analysis of the charging infrastructure for AETs. By addressing the identified gaps, this study contributes to the overall understanding and progress in the field.

The existing body of research broadly investigates charging infrastructure requirements for electric vehicles and their distribution grid impacts. Comprehensive reviews such as those in [7,8] have examined various aspects of this domain, including transport networks, charging technologies, spatial localization methodologies, and challenges in gathering empirical data.

Several studies have delved into energy/power requirements and the distribution grid impacts of heavy-duty electric vehicle charging, including micro-level and macro-level analyses. At a micro level, a study carried out by the National Renewable Energy Lab (NREL) sought to conduct time-series simulations on various connection points of a distribution feeder to understand charging impacts, given charging load profiles and charging locations [9]. On a more macro scale, an NREL study developed synthetic depot charging load profiles for heavy-duty trucks from real-world operating schedules and applied depot charging load profiles to 36 distribution real-world substations within the Energy Reliability Council of Texas (ERCOT) network, concluding that most substations can accommodate high levels of heavy-duty EV charging without upgrades by optimizing charging times in order to take advantage of the off-shift dwell hours (average of 14.1 off-shift dwell hours) of the real world fleet data modeled [10].

Another study by NREL leveraged scale vehicle telematics data (>205 million miles of driving) of diesel-based semi-trailer fleets to estimate charging behaviors and infrastructure requirements for the U.S. battery-electric semi-trailer trucks within three main operating segments: local, regional, and long-haul. Their finding suggests that existing charging standards (≤350 kW) meet the charging requirements for 35 to 77% of the total energy demand for local trips with an operating radius (OR) ≤ 100 miles and regional trips with 100 miles ≤ OR ≤ 300 miles [11]. This assessment, again, was based on charging availability time windows due to the long dwell times of semi-trailers in these segments, which averaged 16.9 h/veh/day and 15.2 h/veh/day for local and regional truck segments, respectively.

Although these previous studies investigated the impact of added electrical loads on distribution systems, they did not explore the impacts of autonomous semi-trailer operations, which are analogous to electric semi-trailers. Existing research primarily focused on modeling electric semi-trailer duty cycles similar to their diesel counterparts, heavily relying on downtime for opportunities to meet charging requirements. However, the commercial trucking landscape is rapidly changing due to increased online purchases and same-day delivery. With 65% of U.S. consumable goods transported to market by trucks, between 12 and 15% of all U.S. purchases are now made from home. Amazon’s same-day delivery service, which is only a few years old, already accounts for 5% of all deliveries, and this figure could reach 15% by 2025 [12]. This demand for goods and services delivery is fueling the accelerated innovation in the heavy-duty trucking industry to reduce operational costs. McKinsey research anticipates a 45% decline in operating costs for fully autonomous trucks, saving the U.S. for-hire trucking industry between $85 billion and $125 billion per year [12]. A major contributor to these cost reductions is the industry trend known as “asset sharing,” which refers to businesses sharing capital-intensive assets such as trucks and warehouses to unlock unused asset capacity. This suggests that the industry will shift toward autonomous electric trucks (AETs), with them becoming shared assets among multiple shippers/businesses, significantly changing the logistical landscape. Fleet managers will aim to minimize truck downtime to maximize daily AET delivery numbers. In addition to driver shortage, regulatory authorities have put in place hours-of-service regulations and electronic-logging device mandates that limited the number of hours that a driver can be on the road [12]. Fully autonomous trucks, however, are not likely to be subject to the same constraints and conditions, further strengthening the case for minimal down time.

Several studies have evaluated the adoption of shared autonomous electric vehicles (SAEV) [13] and the charging infrastructure requirements in the context of providing ride sharing capabilities for commuter trips [14]. The study in [15] explored the operations of shared autonomous electric vehicles (SAEV) and their charging infrastructure requirements in a regional, discrete-time, agent-based model. The work in [15] builds on the framework outlined in [16] to include charging infrastructure estimations that are modeled based on charging characteristics, vehicle range, and fleet size. The work simulates the operation of SAEVs under various vehicle ranges and charging infrastructure scenarios in a gridded city model, with the optimization goal of maximizing the number of trip requests served by a given fleet size and charging station location distribution. These previous studies do not, however, fill the research gap for understanding the operations and charging infrastructure impacts of AETs.

This paper, therefore, aims to address this research gap by estimating the charging infrastructure required to accommodate a full or partial fleet of AETs at a macro scale within a geographical region. Leveraging a discrete-time model [15,16] and focusing on the state of Texas, we simulate daily trips of AETs and estimate the resulting charging infrastructure needs. This contribution extends the previous research by considering the minimal downtime and continuous operations that are characteristic of AETs, thereby offering a more contemporary model for the rapidly evolving heavy-duty trucking industry.

2. Methodology

2.1. Model Data Inputs and Exploratory Data Analysis

To simulate the discrete time series operations of the AETs, several data inputs were considered and explored. This section goes into details on the fleet operational schedules modeled, the heavy-duty vehicle modeled, and geographical location characteristics.

2.1.1. Fleet Operational Characteristics

The NREL fleet DNA project [17] offers access to vehicle fleet data summaries and visualizations similar to real-world telematics tools for medium and heavy-duty commercial fleet vehicles operating within a variety of locations. Figure 1 illustrates a breakdown of the diversity of the dataset used in this work to infer the modeling parameters to be discussed in subsequent sections. Figure 1 represents telemetry data collected for a fleet of Class 8 vehicles, i.e., large trucks that have a gross vehicle weight rating exceeding 33,000 lbs. The deployment ID here represents the group/fleet ID that the vehicles belong to.

The sample size of the NREL fleet DNA data included 70 vehicles and 1150 days of operation, providing a diversity of data which we deemed representative of the fleet operations in North America. Further exploration of fleet DNA data described in Figure 2 and Figure 3 were used to provide insights into the operational logistics of the fleet such as trip distance distribution, trip duration distribution (including and not including stops), daily trip counts, and trip start times and trip operating durations.

From the data analysis, we can see that there is a strong correlation between trip distances, trip durations, and the off-shift periods between the fleet DNA and the estimates detailed in [10,11]. Most trips are <300 miles in the distance, and most vehicles are in operation for <10 h in the day, indicating that the dataset explored is diverse and truly representative of fleet operations in the U.S.

2.1.2. Vehicle Characteristics

While several original equipment manufacturers (OEMs) have developed electric heavy-duty truck platforms [18] and delivered fully electric heavy-duty trucks [19,20,21], and many tech upstarts and incumbents have developed autonomous driver software [22,23,24,25], the combination of both technologies has not gained wide adoption as yet. In this study, we have modeled the AETs operated in a geographical location using the published specifications for an all-electric semi-truck that is the closest electric vehicle to achieving significant market penetration, and that includes significant autonomous operation plans [26].

Table 1. Vehicle specifications for a Class 8 tractor trailer “semi” used in this work [26].

Range (Average Load)	500 miles
Energy efficiency (Average Load)	2 kWh/mile
Battery capacity	1 MWh
Charging Speed	500 kW

shows the important specifications used in our model.

2.1.3. Geographical Location Trip Demand

In this study, the geographical location considered is the state of Texas. Texas is the second largest state in the U.S., with a population of about 27 million people, with nearly three-quarters of the population living within the so-called “Texas Triangle”, bordered by three major highways: I-35, I-45 and I-10, and linking the major cities of Austin, Dallas, Fort Worth, Houston, and San Antonio [27]. Figure 4 is a representation of the Texas Triangle showing major cities along each highway.

In 2014, more than 2.6 billion tons of freight moved in Texas, and this is anticipated to increase to nearly 3.8 billion tons by 2040, which is a 46% growth between 2014 and 2040 with trucks moving two-thirds of this tonnage [27]. In 2013, Texas contributed over $1 billion in congestion costs to the trucking industry, ranking only behind California. Dallas ranked fourth with over $406 million, and Houston ranked sixth with over $373 million in congestion costs to the trucking industry. The I-35 through Austin is the most congested roadway segment in Texas based on trucking delays, with over 116,000 h of delays in 2013 [27]. These staggering statistics make Texas and the Texas Triangle freight corridors excellent case studies for the development of the charging infrastructure for AETs.

The Statewide Traffic Analysis and Reporting System (STARS II) of the Texas Department of Transportation (TxDOT) is a database that provides detailed traffic data and statistics on Texas roads [28]. One of the data sets it provides is Average Annual Daily Traffic (AADT) data, which measures the average volume of traffic on a roadway segment over the course of a year. The total number of trucks on the road (which is estimated to be 5% of all vehicles) compared to passenger cars [29] can be used to estimate the number of trucks on the road at specific segments of major cities along highway corridors; however, this data may not truly depict the percentage of AADT that represents the heavy-truck segment.

To better estimate the percentage of AADT attributed to heavy-duty trucks, we can utilize the average vehicle miles traveled (AVMT) per vehicle provided by the U.S. Department of Transportation’s Federal Highway Administration [30], which encompasses truck AVMT, along with the daily vehicle miles traveled (DVMT) on highways, as reported by the Texas Department of Transportation [31]. This report also includes the total miles traveled by heavy-duty trucks (Truck DVMT). By incorporating these sources, we can more accurately approximate the percentage of AADT for heavy-duty vehicles.

Table 2. Summary of heavy-truck highway and total daily mileage used to estimate percentage of heavy-truck traffic on highway segments [30,31].

	All Motor Vehicles	Heavy-Duty Trucks
DMVT (All Roads in Texas)	711,966,641 miles/day	79,624,836 miles/day
DMVT (Highway only)	176,636,332 miles/day	33,636,598 miles/day
AVMT per vehicle (Yearly)	10,523 miles/year	60,356 miles/year
AVMT per vehicle (Daily)	28.83 miles/day	165 miles/day
AVMT per vehicle (Daily, Highway only)	7 miles/day	70 miles/day
Daily Trips (Highway only)	24,704,282 trips/day	481,928 trips/day

summarizes the data used in this work to estimate the percentage of AADT for heavy-duty vehicles. From the data, we can estimate that daily heavy-duty trips (heavy-duty truck AADT) are ≈2% of total trips on highway only roads [30,31].

3. Model Development

A modeling simulation framework shown in this work was developed using Python based on the work presented in [15,16]. At a high level, the simulation gridded area was first defined. Following this, an inter-city trip generation algorithm was developed that sought to match trip distances generated from the probability density curve from the NREL fleet DNA data to the corresponding cities, while biased by the estimated heavy truck AADT. Finally, a discrete time simulation with simulation steps of 1 h were run against the trips generated for an entire day, assigning heavy duty trucks to each trip while calculating and aggregating the charging requirements for each truck at each location using the conceptual Semi specs defined in Table 1.

3.1. Gridded City

In this research, the selection of cities within Texas, which covers an area of 268,597 square miles, was based on a grid representation. A 500-mile by 500-mile square was created to represent the state, which was further divided into one mile by one mile grids for more precise geographical locations. The aim of the model was to illustrate inter-city trips along the highway corridors of the “Texas Triangle.” To achieve this, prominent cities along the highway corridors were chosen at approximately even intervals of 96 miles, corresponding to the average trip distance from the NREL fleet DNA data analyzed in Section 2 [14]. We estimated the approximate positions of each city along the borders of the “Texas Triangle” using major highway corridor intersections as anchors.

Table 3. Average annual daily traffic and estimated heavy truck traffic for major cities along Texas Triangle highways [28,30,31].

City	AADT	Heavy Truck AADT
Houston	202,293	4046
Dallas	175,750	3515
Round Rock	164,505	3290
San Antonio	160,584	3212
Austin	142,491	2850
New Braunfels	126,365	2527
San Marcos	114,429	2289
Conroe	115,363	2307
Katy	84,397	1688
Waco	75,170	1503
Temple	73,640	1473
Seguin	53,460	1069
Huntsville	43,828	876
Centerville	39,011	780
Buffalo	36,262	725
Columbus	36,286	726
Corsicana	34,165	683
Fairfield	34,117	682
Madisonville	32,704	654

presents the AADT data and the estimated truck AADT for prominent cities situated along the “Texas Triangle” highway corridors. It should be noted, as described in Section 2, that the heavy truck AADT column is determined by multiplying the AADT by 2% (or 0.02) to estimate the number of heavy trucks on each road segment [28]. While this is a rough estimate and may not accurately represent actual heavy truck traffic volumes on these highways, it serves as a suitable representation of anonymized trip data for truck operations on these roadways due to the sensitivity and privacy concerns surrounding real-world fleet GPS trip data.

Figure 5 is a representation of the gridded state of Texas with the cities along the highway corridors. The size of each marker indicates the AADT for the cities.

3.2. Trip Generation Algorithm

To generate an algorithm that statistically simulates the daily inter-city trips for heavy-duty trucks along the I-10, I-35, and I-45 highway corridors in Texas, we used Monte Carlo simulations informed by the probability density functions (random samples of the trip distances based on the data distribution) shown in Figure 2, and the weighted selection based on AADT data shown in Table 3 (to match the estimated trip counts by city). The high-level outline of the algorithm included:

1.

Gathering the input data:

• AADT for all major/relevant cities along the corridors.
• Estimated daily miles travelled of heavy-duty trucks.
• The probability density curve of distances traveled by trucks from the NREL fleet DNA data (PDF_daily_distances).

2.

Preprocess data:

• An estimate of the heavy-duty truck traffic by multiplying the AADT at the road segments in each prominent city by 2%, which represents the number of the heavy-duty truck trips [28,30,31]. NOTE: We estimate that not all trucks operate in this local/regional segment [10], so a multiplier of 0.7 of the Heavy Truck AADT is used for estimating the trip counts per city.

3.

Initialize variables:

• Create an empty list to store the daily inter-city trips for heavy-duty trucks (daily_trips).

4.

Determine the total number of daily trips:

• Calculate the sum of all the truck traffic values for the cities.
• Divide this sum by two to account for the fact that each trip has an origin and a destination.
• Round the resulting number to the nearest integer to obtain the total number of daily trips (num_trips).

5.

Monte Carlo simulation:

• For each trip in the range of num_trips, perform the following steps:
  • Select the origin city randomly, biased by truck_traffic_normalized.
  • Select the destination city randomly, biased by truck_traffic_normalized, and ensure it is not the same as the origin city.
  • Sample a daily distance from the PDF_daily_distances.
  • Calculate the distance between the origin and destination cities using the Haversine formula or a similar method.
  • If the sampled daily distance is greater than or equal to the distance between the origin and destination cities, consider the trip to be feasible.
  • Calculate the trip duration based on the correlation of trip distance and the duration PDF from the fleet DNA data.
  • If the trip is feasible, add the trip details (origin, destination, and distance) to the daily_trips list.

6.

Repeat step 5 until the total number of generated trips matches the expected number of daily trips (e.g., based on AADT data).

7.

Using the total number of trip start counts for each city, divide the trips into 24 chunks so that an even number of trips originate from each city across the 24-h period of the day. Add the trip start times to the daily_trips list.

8.

Output the daily_trips list, representing the daily inter-city trips for heavy-duty trucks along the specified highway corridors.

This algorithm provides a statistical simulation of daily inter-city trips for heavy-duty trucks along the I-10, I-35, and I-45 highway corridors in Texas. By leveraging AADT data, the percentage of heavy-duty trucks, and anonymized truck driving patterns, the algorithm can generate realistic and accurate trip data for a single day.

Table 4. Aggregated trips across all cities for each hour of a day.

Hour	Trip Count
0	1062
1	1061
2	1070
3	1058
4	1061
5	1064
6	1059
7	1067
8	1059
9	1067
10	1054
11	1058
12	1074
13	1073
14	1069
15	1048
16	1065
17	1061
18	1056
19	1066
20	1054
21	1028
22	814
23	232

below shows the number of trips aggregated across the 24-h period of a day. Figure 6 is a visualization of trip connections that resulted from the algorithm presented above.

Figure 7 is a distribution of trip distances showing a strong correlation between the trip distances generated and the NREL fleet DNA data analyzed in Section 2 and described in Figure 2.

Table 5. Trip origin and destination counts for inter-city trips for one day.

Rank	City	Origin	Destination
1	Houston	2913	837
2	Dallas	2555	997
3	San Antonio	2225	1012
4	Round Rock	2326	1563
5	Austin	1969	1779
6	New Braunfels	1726	1558
7	San Marcos	1647	1895
8	Conroe	1613	886
9	Temple	990	1828
10	Katy	1155	1310
11	Waco	1045	1062
12	Seguin	741	1135
13	Huntsville	587	1187
14	Centerville	515	1310
15	Buffalo	484	1195
16	Corsicana	492	917
17	Fairfield	504	859
18	Columbus	491	1475
19	Madisonville	452	859

illustrates the trip origin and destination data for one day, generated by the algorithm defined. Combining the origin and destination traffic for each location results in approximately 70% of the heavy truck AADT for that city based on the factor we set modeling the portion of trips that fall into local/regional trucking segments that is being modeled in this work. An interesting phenomenon observed is that larger cities have more origin trips than destination trips, and smaller cities have more destination trips than origin trips. This is due to the way the trip generation algorithm is designed to work. The algorithm selects the origin city based on the AADT of each city. This means that larger cities with higher AADT values have a higher probability of being selected as the origin city for a trip. This results in more trips originating from larger cities.

For destination cities, the algorithm chooses a city based on two factors: the city’s AADT and the difference between the selected trip distance (based on the PDF of trip distances [17]) and the actual distance between the origin and destination cities. The algorithm calculates the probability of selecting a destination city by multiplying the AADT by the exponential of the negative absolute difference between the trip distance and the actual distance. This means that for any given origin city, nearby cities with similar distances to the randomly selected trip distance will have a higher probability of being selected as the destination city.

Since smaller cities generally have lower AADT values, they are less likely to be selected as origin cities. However, when the algorithm generates the destination city based on distance and AADT, smaller cities that are closer to the trip distance have a relatively higher chance of being selected as destination cities. This causes an imbalance in the number of origin and destination trips for smaller cities.

For medium-sized cities, their AADT values are relatively balanced with the distance factor, so their probabilities of being selected as origin and destination cities are closer to each other, resulting in a more balanced distribution of trips.

3.3. Trip Simulation

For each day of simulation, divide the 24-h period into 1 h time steps and follow the algorithm below.

The algorithm is divided into several steps:

• Initialize an empty DataFrame to store the hourly charging data for electric vehicles.
• Initialize an empty DataFame to store the aggregated daily energy consumption grouped by location.
• Loops through each day of the simulation.
• For each day, the algorithm iterates through all 24 h, filtering the trips for that specific hour.
• It then calculates the energy consumption and charging time for each trip and updates the hourly charging data.
• The energy calculation is done by first using the trip distance to calculate the kWh required for the trip based on the semi specs of 2 kWh/mi.
• The algorithm aggregates the energy consumption for each location and hour, storing the results in the hourly_charging_data DataFrame.
• The algorithm then does a grouping for each location and calculates the average power for each day by location based on the 24-h period.

4. Results

The main goal of this simulation is to understand the charging power and energy requirements of the charging stations at each city so that we can inform charging infrastructure decisions at a regional level. Instead of modeling the charging power for each vehicle and then scaling by the number of vehicles, the model presented here uses a probabilistic generation of trip origination/destination pairs to yield trip distances, from which a kWh/day load can then be estimated. From this number we can understand the power and energy required to complete all trips that originate from each city. Charging start times and trip start times for each city can then be optimized based on these estimates over the course of the entire day so that that the charging peaks are leveled off and there are no unnecessary spikes in charging power at charging stations in any of the cities. The average daily power and total daily energy per location in our map for a single day run the simulation for the base case of ≥70% of heavy truck AADT as represented by Figure 8 (100% of Local/Regional Segment Trips daily).

Table 6. Charging infrastructure cost estimates [10].

Equipment	Unit Cost Range	Installation Cost Range
50 kW DCFC EVSE	$20,000–$36,000	$10,000–$46,000
150 kW DCFC EVSE	$75,000–$100,000	$19,000–$48,000
350 kW DCFC EVSE	$128,000–$150,000	$26,000–$66,000
Distribution Transformer	$12,000–$175,000	N/A
Feeder Circuit (5+ MW Load)	$2 million–$12 million	N/A
Feeder Breaker (5+ MW load)	$400,000	N/A
Substation Installation (3–10+ MW)	$4 million–$35 million	N/A

below shows the equipment cost ranges for the charging infrastructure requirements based on NREL charging infrastructure estimates presented in [10].

In order to estimate the charging infrastructure equipment costs at specific locations, we need to make certain assumptions:

• The power demand will be evenly distributed among the direct current fast charging electric vehicle supply equipment (DCFC EVSE) units (50 kW, 150 kW, and 350 kW).
• The energy demand will be distributed evenly throughout the day.
• The required number of units is based on the power demand, divided by the power capacity of the equipment, and then rounded up.
• The total cost is calculated as the unit cost plus installation cost, multiplied by the number of units.

Based on these assumptions,

Table 7. Charging Infrastructure cost estimates for major cities based on simulation results for 100% penetration of regional trips (<300 miles).

City	Energy Demand (MWh)	Power Demand (MW)	50 kW Units	150 kW Units	350 kW Units	Total Cost Range
Houston	533	22	440	147	63	$9,680,000–$15,930,000
Dallas	459	19	380	127	55	$8,060,000–$13,240,000
San Antonio	443	18.5	370	124	53	$7,740,000–$12,740,000

presents the cost estimates for the cities of San Antonio, Austin, Dallas, and Houston.

5. Conclusions

This study presented a data-driven approach to estimate charging infrastructure requirements for heavy duty electric trucks using a conceptual autonomous electric heavy duty trailer (semi) within the major cities across the state of Texas. By simulating daily trips and energy consumption patterns, we have provided a comprehensive analysis of the charging infrastructure needs for cities along the Texas highway corridors I-35, I-45, and I-10. In this study, we have analyzed the charging infrastructure requirements for heavy duty electric trucks in 19 cities, with summaries focused on the three major Texas cities that anchor the intersections between the major highway corridors: San Antonio, Houston, and Dallas. Our analysis considers various factors such as daily trip patterns, energy consumption, and equipment costs for charging stations with different capacities (50 kW, 150 kW, and 350 kW DCFC EVSE). The results show that the total charging energy and average charging power for these major cities ranges from 443 MWh/day to 533 MWh/day and 18.5 MW to 22 MW, respectively, while the total cost for setting up the necessary charging infrastructure ranges from $7.74 M to $15.93 M for each city, depending on the charging capacity and installation costs.

For San Antonio, we estimate that 370 50 kW DCFC EVSE units, 124 150 kW units, or 53 350 kW units are needed, with a total cost range of USD $7.74 million to $12.74 million. In Houston, 440 50 kW units, 147 150 kW units, or 63 350 kW units are necessary, with costs varying between USD $9.68 million to $15.93 million. Dallas requires 380 50 kW units, 127 150 kW units, or 55 350 kW units, with total costs ranging from USD $8.06 million to $13.24 million. San Antonio requires 370 50 kW units, 124 150 kW units, or 53 350 kW units, with total costs ranging from USD $7.7 million to $12.7 million.

Note that these estimates do not include the distribution grid infrastructure such as distribution transformers, feeder circuits, feeder breakers, or substation installation, as the requirements for these would have to be analyzed on a case-by-case basis in order to determine the need. In a future study, we plan to investigate the approaches and guidelines for developing and prescribing these types of distribution system upgrades.

Our data-driven approach can be replicated for other areas or regions by adapting the simulation parameters such as trip patterns, energy consumption, and equipment costs to the specific context. This will allow policymakers and stakeholders to assess the charging infrastructure requirements and related investments needed to support the transition to electric and autonomous heavy-duty trucking.

It is important to recognize that our estimates are subject to uncertainties and assumptions related to factors such as trip patterns, power demand distribution, energy consumption patterns, and equipment costs. To refine these estimates, future research should integrate real-world data, such as actual charging station usage, and examine the impact of different charging strategies on infrastructure requirements.

Overall, this study offers valuable insights into the charging infrastructure needs for AET-based operations and duty cycles for major cities in Texas along the major highway traffic corridors. The study takes into account the road network and the trip distance distributions, laying the groundwork for further research and policy development in the area of autonomous electric heavy-duty trucking. By utilizing data-driven methodologies, policymakers and stakeholders can make informed decisions to support the transition to electric mobility, ensuring a more sustainable and environmentally friendly future.