Electric Vehicle Charging Station Problem

EVCSP is the problem of optimal allocation of charging station, across given terrain map. My implementation solves this problem by taking following parameters: budget, distances_map, demand_map, land_rental_cost_map and poi_map. Optimizer want's to minimize total cost, containing to users cost and operator cost.

Map Creator

For creating mentioned maps necessary for this optimization problem, following map Creator was created:

EVCSP Map Creator: Link

It allows user to create each type of map as he pleases and it posess option to download created maps in .json format managable, by this C++ application.

Downloaded jso file containing maps, must then be placed in ./data folder. This location is default directory, where model tries to find maps.json file.

Hyperparameters

Hyperparameters of EVCSP model are located in ./data/hyperparameters.json directory. They include:

• max_stations_per_cell (int): Maximal number of stations we can put in one cell.
• budget (double): Budget for the whole project.
• mip_gap (double): Approximated variation from optimal solution.
• stations_powers (double[2]): Powers of each station.
• initial_costs (double[2]): Cost of installation of each station.
• maintenance_costs (double[2]): Daily maintanance cost for each station.

Documentation

Formal paper for this problem, one can find in ./docs/EVCSP_2026.pdf directory. It contains mathematical formulation and comparison, between meta heuristic and linear programming approaches of solving this problem.

Commands

# Initiates project.
cmake -B build -S .

# Removes current license path.
unset GRB_LICENSE_FILE
# Creates system path to the license.
export GRB_LICENSE_FILE="example_path"

# Edit file for chaning license path.
nano ~/.zshrc

# Builds app.
cmake --build build

# Runs app.
./build/EVCSP_Application

# Generates documentation
doxygen -g

# Updates documentation
doxygen

# Updates documentation settings
doxygen Doxyfile

# Run documentation
./docs/html/index.html

Maps

Distances Map

Represents obstacles and costs from travelling from given node to another.

Point Of Intrest (POI) Map

Represents the locations, where we possibly can build our electric vehicle charging stations.

Demand Map

Represents energy consumption demand of electric vehicle users, at a given piece of terrain.

Land Rental Cost Map

Represents rental cost of specific point of intreset on the map.

BFOA Solution

Gurobi Solution

Comparison

We can see that linear programming model is slightly better, however we can tune meta heuristic model, to be much faster than the gurobi based solution.

Technologies