EV Charging Station

Placement Optimization
Dandu surath Preetham - BL.SC.U4CSE24212
S M Vedanth - BL.SC.U4CSE24242

2025/08/05
 Chall

EV Growth & Cha

Accelerating EV Adoption
Electric vehicle adoption is rapidly increasing, but public
charging infrastructure is struggling to keep pace with the
growing demand. This gap results in long queues and driver
range anxiety.
 Problem St
Objective
Problem Statement
To determine the optimal locations for installing EV
charging stations such that the overall weighted demand
coverage is maximized while keeping installation costs
minimal, considering both daytime and nighttime demand
variations and predefined geographic constraints

Challenges
Uneven charger usage → some stations overcrowded,
others underused.

Long wait times and user inconvenience in high-demand

zones.

Irregular demand patterns across time (day/night) and

locations.

Grid stability and load imbalance due to unpredictable

charging needs.
Why Optimization Is Needed
Rapid EV growth → long wait times & S
uneven charger usage.
Stations are overcrowded in some
areas and underused in others.
Causes resource wastage, grid load
imbalance, and poor accessibility.
Demand varies by time (day/night) and
location, requiring smarter planning.

Project Scope
The scope of this project includes analyzing real EV demand data, mapping existing chargers, applying
evolutionary, direct, and heuristic optimization methods, and evaluating coverage, cost, and day/night demand
distribution to determine optimal charger locations
 Related Work Review
Previous Optimization Models Limitations of Previous Models LRP Models

Earlier studies applied Particle Swarm Most previous models ignore Location-Routing Problem (LRP) models
Optimization (PSO), Genetic heterogeneous demand patterns, grid combine routing with facility location
Algorithms (GA), and Archimedes constraints, or cost-capacity but scale poorly for large datasets,
Optimization Algorithm (AOA) to trade-offs, limiting their practical making them less suitable for city-wide
facility location problems. applicability. optimization.

Motivation for Hybrid Approach

This study proposes a hybrid approach combining weighted
K-Means clustering and evolutionary search to address these
limitations and achieve better optimization results.
 1. Data Collection
Gather existing EV charging station data (lat/long, usage).
Collect demand points (malls, residential areas, public hotspots).
Clean and prepare the geographic dataset.

2. Baseline Coverage Estimation

Apply Weighted K-Means to identify demand clusters.
Map existing chargers and calculate baseline coverage radius.
Determine percentage of city currently covered.

3. Optimization Algorithm
Perform Evolutionary Search to evaluate candidate charger sites.
Score each candidate using coverage gain + cost trade-off.
Select high-impact chargers iteratively until target coverage is reached.
 (Continuation)
4. Cost Optimization
Assign cost scores to candidate chargers (Low / Medium / High).
Balance total installation cost with improvement in spatial coverage.

5. Day & Night Demand Modeling

Split usage into daytime (public zones) and nighttime (residential).
Compute weighted demand coverage separately for both time periods.

6. Results & Visualization

Display before/after coverage maps.
Show coverage improvement (46.85% → 87.41%).
Present cost table & day/night usage distribution.
Weighted K-Means Clustering Why weighted K-Means Clustering
Weighted K-Means clustering initializes candidate sites by Weighted K-Means is used to identify high-demand areas
grouping demand nodes into micro-regions for efficient by giving more importance to locations with higher EV
optimization. activity, helping place chargers closer to real demand
hotspots.

Why Evolutionary Search Process

Evolutionary Search Process
The evolutionary search encodes candidate solutions as The evolutionary search process is used because it
bit strings, using mutation, crossover, and elitism to explores many possible charger placements and
converge on optimal solutions. iteratively selects the ones that give the highest
coverage, making it effective for complex,
real-world optimization problems.
 Distance & Constraint
Modeling
Objective Function Overview
Haversine Distance Formula
The objective function selects charger
locations that give the highest total
demand coverage across day and night,
while subtracting a penalty for costly
installations. This ensures the model
prioritizes placements that serve more
users without exceeding budget
The Haversine distance formula is used to calculate the spherical distance
constraints.
between nodes, ensuring accurate geographic coverage modeling. By
accounting for the curvature of the Earth rather than using simple
Euclidean distance, it provides a more realistic measure of proximity
between chargers and demand points. This helps the model correctly
determine which locations fall within the service radius of each charger,
leading to more reliable coverage estimation and better placement
decisions.
 Purpose of Preparing
Data Sources Why Los Angeles? Final Dataset Contains
the Dataset
City chosen for good Two datasets were used for the Coordinates of existing To estimate baseline
EV adoption rates. analysis. Existing Charging Stations charging stations. coverage using only
Dense and diverse Dataset Contains latitude & Average number of users current chargers.
geography helps in longitude of real chargers in Los per charger. Coordinates To feed into Weighted
accurate mapping. Angeles. Includes average user
of all demand points. K-Means, Evolutionary
Easy to visualize and count for computing daytime &
Demand weights for day Search, and Heuristic
nighttime demand. Demand Points
compute spatial and night periods. Search.
Dataset (Points of Interest)
coverage. Candidate charger site list To compute:
Locations where a new EV charger
can potentially be placed. Includes for optimization. Coverage improvement
latitude & longitude of public Optimal charger
spaces, malls, residential regions, placement Installation
etc. cost category Day vs
night demand distribution
 Baseline

What is Baseline Coverage?

Measures how much area existing EV chargers currently cover. Used as the reference before applying optimization.

How It Was Computed

Applied Weighted K-Means to find high-demand clusters.
Mapped the nearest existing chargers to those cluster centers.
Plotted chargers with a fixed coverage radius.

Key Observations
Only 46.85% coverage with 5 existing chargers. Many demand
points fall outside the coverage radius. Shows clear service gaps
across Los Angeles.

Why It Matters
Helps compare improvement after optimization. Shows the
limitations of the current EVCS setup.
 Baseline
Goal
To place new chargers that maximize coverage along with existing ones.
How Optimization Works
Every candidate site is scored based on coverage gained. used Evolutionary Search + Heuristic
Search to pick the best locations. Highest-scoring chargers are added one by one (iteratively).
Process
Compute coverage gain for each potential site. Select chargers with positive, high scores. Add them
to the map and recalculate coverage. Repeat until coverage target or max chargers reached.
Outcome
New charger placements significantly improve city-wide coverage. Forms the optimized map shown in the
results section.
 Cost & Demand Calibration
Cost Model Day-Night Demand calibration

Each site is assigned a cost based on Demand is split into daytime public use and
installation difficulty & cable distance. nighttime residential use, with weights tuned by
Chargers are labeled as Low / Medium / evolutionary search to match real charging patterns.
High cost. Higher-cost sites get penalized
in the selection score. Cost is applied using
the below formula.

Integrated Optimization Impact

Cost and demand weights are combined to

score sites, ensuring the final placements
balance coverage, cost, and time-based
usage.
 Coverage & Cost Outcomes

Cost computation

Installation cost is estimated for each charger based on its

geographic location.

Chargers with cable distances and simpler setups fall

under low-cost installations.

Locations requiring longer cables or more complex

installation conditions show higher estimated costs.

The cost levels help compare and plan charger placement

more efficiently.
 Coverage & Cost Outcomes
Coverage Improvement

The optimized placement of chargers using Weighted K-Means

improved coverage from 46.85% to 87.41%.
 Coverage & Cost Outcomes

Understanding Daily Load Variation

Shows charger locations with corresponding

daytime and nighttime usage levels.

Helps identify which chargers experience high

demand during specific times of day.

Useful for understanding peak load periods and

optimizing charger placement or capacity.

Highlights overall usage distribution across the

entire day.
 Key Insights & Error Analysis

Error Analysis Key Insights

Simplified Demand weighting Demand weighting is too basic

Uses spatial data only Real demand varies by time and behavior, so spatial
Might mis-estimate demand and cause sub optimal placement density alone is not reliable.

Deteministic Model Deteministic Model

Assumes fixed demand and locations. EV usage changes hour to hour; deterministic
Cannot handle real time Ev usage fluctuations assumptions reduce real-world accuracy.

Fixed coverage radius Fixed coverage radius

Uses constant radius assumption Accessibility depends on roads and traffic, not a fixed
Real accessibility varies with roads, traffic and driver behavior. circular radius.
 Study Conclusions

Effective Optimization Tool Policy Implications Overall study insight

The hybrid evolutionary approach provides The findings offer actionable The model demonstrates that
a transparent, reproducible, and guidance for EV infrastructure combining evolutionary search with
data-driven method for identifying optimal development, supporting Bengaluru’s heuristic refinement can significantly
EV charger locations. It improves coverage, 2030 EV goals. High-demand zones enhance charger placement, raising
balances cost, and adapts to near metro feeders, commercial hubs, coverage from 46.85% to 87.41%. This
daytime–nighttime demand patterns, and logistics corridors emerge as establishes a scalable framework for
making it a practical planning tool for priority areas, helping planners future EV expansion in growing cities.
real-world deployment. allocate chargers efficiently and
sustainably.
 Future Research Directions

Real-Time Data Integration Multi-objective optimization

Future work involves integrating real-time traffic flow, Extending the framework with multi-objective optimizers
charger availability, and IoT-based telemetry to enable (e.g., PSO, GA) will allow simultaneous balancing of
adaptive and dynamic charger placement. coverage, cost, emissions, and grid stability.

Renewable Energy Integration Dynamic Demand Updating

Solar-powered micro-grid integration can Incorporating real-time demand updates from
reduce grid load, improve sustainability, and mobility data, events, and weather patterns will
lower long-term operational costs. make the model align even more closely with
real-world EV behavior.
THANK YOU

2025/08/05