Optimizing Electric Vehicle Charging Stations and Charging

Ports for Highways and Expressways

Aadesh Pochhi Akshat Lohia Nilotpal Chakraborty
Indian Institute of Information Indian Institute of Information Indian Institute of Information
Technology Guwahati Technology Guwahati Technology Guwahati
Guwahati, Assam, India Assam, India Assam, India
ABSTRACT if all chargers are currently in use and a new EV arrives, it does
In this work, we attempt to address the problem of strategically not have to wait longer than a predefined time before it gets its
placing electric vehicle charging stations and obtaining the optimal charging port. We assume that EVs arrive at charging stations at a
number of charging ports at each charging station, specifically for constant rate, given the continuous flow of traffic on highways. All
highways and expressways. We consider a setup where there is a these are achieved while minimizing the cost of building charging
deterministic inter-arrival time between the vehicles, and addition- stations and installing charging ports.
ally, there is a constraint on the waiting time for a vehicle at any
charging station. The overall objective is to minimize the cost of set- 2 PROBLEM FORMULATION
ting up the charging stations to support a given set of vehicles. We In this section, we model our problem where we represent the
first model the problem using Mixed Integer Linear Programming highway as Φ, divided into 𝑛 nodes (the candidate sites for CSs).
formulation and subsequently propose a greedy heuristic algorithm The selection of the node 𝑖 ∈ 1, 2, . . . 𝑛, where a CS can be built
to produce results efficiently. is represented by a binary variable 𝑥𝑖 , (𝑥𝑖 = 1 if the node 𝑖 is
selected, 0 otherwise). Furthermore, the distance (𝑑𝑖 𝑗 ) between any
CCS CONCEPTS two adjacent selected nodes 𝑖 (𝑥𝑖 = 1) and 𝑗 (𝑥 𝑗 = 1) should be less
• Applied computing → Transportation. than 𝐶 min , the minimum of the maximum driving range across all
the possible EVs that might travel on the highway.
KEYWORDS Once the locations of the CSs are selected, our next target is to
decide on the charging ports 𝑦𝑖 at each selected node 𝑖. We consider
Electric Vehicles, Charging Station Optimization, Highway
only fast chargers that would be installed and between the arrival of
ACM Reference Format: any two EVs, there is an inter-arrival time iat𝑖 , ∀𝑖 where 𝑥𝑖 = 1. We
Aadesh Pochhi, Akshat Lohia, and Nilotpal Chakraborty. 2024. Optimizing also ensure that the waiting time for car 𝑐 arriving at node 𝑖 (𝑊𝑐,𝑖 ) is
Electric Vehicle Charging Stations and Charging Ports for Highways and Ex-
less than a maximum value (𝑊𝑖 ). Furthermore, iat𝑖 should be more
pressways. In The 15th ACM International Conference on Future and Sustain-
able Energy Systems (E-Energy ’24), June 04–07, 2024, Singapore, Singapore.
than a minimum value 𝐼𝑖 because very low values of iat𝑖 mean
ACM, New York, NY, USA, 2 pages. https://doi.org/10.1145/3632775.3661981 that the waiting time is increased and it could lead to inefficient
or suboptimal solutions. Operating under these constraints, our
1 INTRODUCTION objective is to minimize the cost of installing (𝑐 ins,𝑖 ) and maintaining
Due to the global rise in pollution and carbon emissions caused by (𝑐 ops,𝑖 ) charging stations at nodes 𝑖 and installing of charging ports
the transportation sector, electric vehicles (EVs) are being promoted in these CSs on the highway. Representing the cost of installing one
worldwide. EVs are known to be environment friendly as they are charging port at node 𝑖 as 𝑐 p,𝑖 , mathematically we can represent
free from the burden of burning fossil fuels, making them a sustain- the optimization problem as follows.
able mode of travel. However, ensuring adequate infrastructure for ∑︁  
EVs is paramount to consider them reliable and robust, particularly 𝑀𝑖𝑛𝑖𝑚𝑖𝑧𝑒 𝑥𝑖 (𝑐 ins,𝑖 + 𝑐 ops,𝑖 ) + 𝑦𝑖 𝑐 p,𝑖 (1)
along highways for intercity travel. While most of the existing 𝑖 ∈1,2,...𝑛
research works have primarily focused on infrastructure develop- We aim to solve (1) subject to the following constraints.
ment in urban areas [1] [2], there is a need to focus on highway 𝑊𝑐,𝑖 ≤ 𝑊𝑖 , ∀𝑖 where 𝑥𝑖 = 1 (2)
infrastructure to support long-distance travel requirements.
In this study, we conceptualize a highway as a series of nodes 𝑖𝑎𝑡 𝑖 ≥ 𝐼𝑖 , ∀𝑖 where 𝑥𝑖 = 1 (3)
representing potential candidate sites for charging stations (CS).
𝑑𝑖 𝑗 ≤ 𝐶 min, ∀𝑖, 𝑗 where 𝑖 and 𝑗 are adjacent selected nodes (4)
The number of required charging ports is determined to ensure that
Permission to make digital or hard copies of all or part of this work for personal or 𝑦𝑖 ≤ 𝑃, ∀𝑖 where 𝑥𝑖 = 1 (5)
classroom use is granted without fee provided that copies are not made or distributed ∑︁
for profit or commercial advantage and that copies bear this notice and the full citation 𝑥𝑖 ≤ 𝑀, ∀𝑖 ∈ Φ (6)
on the first page. Copyrights for components of this work owned by others than the
𝑖 ∈Φ
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specific permission 𝑥𝑖 ∈ {0, 1}, ∀𝑖 ∈ Φ (7)
and/or a fee. Request permissions from [email protected].
+
E-Energy ’24, June 04–07, 2024, Singapore, Singapore 𝑦𝑖 ∈ Z , ∀𝑖 where 𝑥𝑖 = 1 (8)
© 2024 Copyright held by the owner/author(s). Publication rights licensed to ACM.
ACM ISBN 979-8-4007-0480-2/24/06 Eqns. (5) and (6) are prevention constraints, which ensure that
https://doi.org/10.1145/3632775.3661981 the number of ports at a node 𝑖 does not exceed the value 𝑃 and

492
E-Energy ’24, June 04–07, 2024, Singapore, Singapore Aadesh Pochhi, Akshat Lohia, and Nilotpal Chakraborty

Í
𝑖 𝑥𝑖 , which is the number of stations built on a highway, does not EV arrives after 20 minutes, its waiting time will be 40 minutes.
exceed a user-defined value 𝑀. Constraints (7) and (8) are defining As a result, we add another charging port for the third EV. We
constraints for our decision variables. 𝑥𝑖 is a binary variable with continue this process until the condition (𝑊𝑇𝑐,𝑖 ≥ 𝑊𝑖 ) becomes
value 1 or 0, and 𝑦𝑖 is a positive integer number. The above optimiza- false. In this case, after the 6th port is added, the condition becomes
tion is intractable and therefore, in the next section, we propose a false since the waiting time for the seventh EV will be zero (less
greedy strategy to obtain approximation solutions efficiently. than the threshold), and for every car arriving after the seventh EV,
a free port will be available to charge. Thus, a total of 6 charging
3 PROPOSED METHODOLOGY ports will be needed in this scenario.
The proposed methodology aims to determine the number of CS by
looking at the EV specifications such that the difference between 4 PRELIMINARY RESULTS
any two CSs is less than the minimum driving range for the EV We have evaluated the proposed approach considering the Delhi-
set. At each CS, the number of charging ports is obtained based on Mumbai expressway in India, spanning a length of 1,350 km, treat-
inputs such as𝑊𝑖 and 𝑖𝑎𝑡𝑖 for the EVs for each CS 𝑖. The pseudo-code ing each junction as a potential CS. We used EV range specifications
of the proposed approach is shown in Algorithm 1. to determine the maximum distance between adjacent selected
Algorithm 1 Proposed Heuristic Approach nodes or charging stations: Tata Nexon EV (325 km), Tata Tiago EV
(315 km), MG Comet EV (230 km), Mini Cooper SE (180 km), Smart
Input Waiting time threshold (𝑊𝑖 ), inter-arrival time of EV’s at Eq Fortwo Cabrio (95 km). The nodes are selected such that each
node 𝑖 (𝑖𝑎𝑡 𝑖 ), maximum number of charging ports 𝑃 CS will have a distance between them just less than the minimum
Output Number of ports at each CS (𝑦𝑖 ∀𝑖, where 𝑥𝑖 = 1). car driving range.
1: Set 𝑦𝑖 = 1 We experimented considering a total of 200 EVs chosen randomly
2: Set 𝑊𝑐,𝑖 = 0 from the above specifications, on three different scenarios to evalu-
3: for each selected node 𝑖 ∈ Φ do ate the performance of the proposed algorithm. Case 1: 𝑖𝑎𝑡 𝑖 < 𝑊𝑖 ,
4: Calculate 𝑅𝑇 𝑗 = 100−𝐶
CR
soc
for 𝑗 ∈ {1, . . . , 𝑦𝑖 } Case 2: 𝑖𝑎𝑡 𝑖 = 𝑊𝑖 , and Case 3: 𝑖𝑎𝑡 𝑖 > 𝑊𝑖 , ∀𝑖 ∈ Φ, results of which
5: Calculate 𝑊𝑐,𝑖 = min(𝑅𝑇 𝑗 ) is shown in Fig. 1. From the preliminary results obtained, it is ob-
6: while 𝑊𝑐,𝑖 ≥ 𝑊𝑖 do served that shorter inter-arrival times lead to a higher influx of EVs,
7: 𝑦𝑖 += 1 potentially causing congestion and longer waiting times at the CS.
8: Recalculate 𝑅𝑇 𝑗 and 𝑊𝑐+1,𝑖 for next arriving car On the other hand, lowering the waiting time threshold requires
9: end while adding more charging ports to prevent prolonged wait times if the
10: if 𝑦𝑖 ≤ 𝑃 then station can’t handle increased demand within the specified time.
11: Save the value of 𝑦𝑖 in a vector
12: Continue to next node
13: else
14: Invalid input (𝑖𝑎𝑡 𝑖 or 𝑊𝑖 too low)
15: end if
16: end for
17: return 𝑦𝑖 for each 𝑖 ∈ Φ

Initially, each CS has one port, and the waiting time for the first
car is 0. 𝑅𝑇 𝑗 in Step 4 represents the remaining time for the currently
placed chargers which is calculated using the State of Charge (𝐶𝑠𝑜𝑐 )
of the EV. The minimum of these 𝑅𝑇 𝑗 is the waiting time for the Figure 1: Total Cost for Different Scenarios
current EV (Step 5). The while loop from Step 6-9, depending on 5 CONCLUSION AND FUTURE WORK
the waiting time of the current EV, adds a new port and the new
We introduced a greedy approach for strategically placing and
waiting time for the next arriving car is calculated considering the
sizing EVCS and charging ports along highways and expressways.
inter-arrival time. Steps 10-15 are to ensure the upper bound on the
Our approach incorporates deterministic inter-arrival times and
number of ports. The final result is returned at Step 17.
waiting time constraints to optimize charging station deployment
Let us consider a simple working example to understand the
while minimizing costs. In the future, we will extend our model
functioning of the proposed approach. We will imagine a single CS
to accommodate dynamic adjustments based on varying demand
with the following assumptions: the time between Electric Vehicles
patterns and obtain extensive results in comparison to the existing
(EVs) arrivals is 10 minutes, and EV users should not wait for more
solutions.
than 10 minutes to get a charging port. Initially, the CS has 1 port
and we assume that all the arriving EVs would need 1 hour to get REFERENCES
charged fully. When the first car arrives, it is immediately allocated [1] Fayez Alanazi, Talal Obaid Alshammari, and Abdelhalim Azam. 2023. Optimal
to the charging port. After 10 minutes, the second EV arrives, and Charging Station Placement and Scheduling for Electric Vehicles in Smart Cities.
it will have to wait for 50 minutes. This waiting time exceeds the Sustainability 15, 22 (2023), 16030.
[2] Ying-Chao Hung and George Michailidis. 2022. A novel data-driven approach
waiting time threshold of 10 minutes. Therefore, a new charging for solving the electric vehicle charging station location-routing problem. IEEE
port is added, and the second EV is assigned to it. When the third Transactions on Intelligent Transportation Systems 23, 12 (2022), 23858–23868.

493