Mobile smart charging robot - Operating and charging strategy

Description

This project provides an environment for simulating a mobile smart charging robot. It can be used for simulating different scenarios and comparing different operating strategies. The mobile charging robot named GINI consists of a battery with CCS DC charging interfaces for charging Electric vehicles and recharging itself.

While an actual prototype is build within an research project founded by german ministry BMWK, this repository should help to

• Evaluate the potential of mobile charging robots in different scenarios with varying user requirements
• Test different smart charging and energy management strategies for a GINI or a fleet of GINIs
• Provide a base for testing reinforcement learning algorithms
• Offer a template for general charging management simulation and algorithm development

Getting Started

Dependencies

• Python
• Latex (if you want to generate plots with Latex font - as per default)
  • For Windows follow this
  • For Linux (Ubuntu): sudo apt-get install texlive texlive-latex-extra texlive-fonts-recommended dvipng
• Poetry for dependency management

Installing with poetry (recommended)

• Clone the repository
• Install Poetry if you haven't already.

pip install poetry

• Install the project dependencies using Poetry:

poetry install

• Activate poetry

poetry shell

• If you encouter problems refer to the Poetry Documentation

Installing with venv

pip install -r requirements.txt

Config Settings

Parameter	Default Value	Explanation
RENDER	true	Indicates whether the rendering/live visualitation should be executed.
SAVE_JSON	true	Specifies if the output data should be saved.
OUTPUT_GRAPHS	true	Controls whether graphs should be generated and displayed.
FOLDERPATH	"config\\Paper"	The path where files, from where the config files are loaded.
LOOP	false	If this parameter is set, it runs multiple configs (from FOLDERPATH) in a loop.

Executing program

In order to execute a simulation run, the following steps need to be executed:

• Create or modify a config file e.g. config/env_config/env_config gini_1day.json. For detailed information about the configuration parameters, please refer to the Configuration Parameters file.
• Open Main.py and change the path of the file config file run_main(fullfilename="config/env_config gini_1day.json"). Deactivate the Loop function LOOP: false in main_config.json.
• Ensure that you have installed and acitvated the poetry venv and execute the Main.py: python Main.py

Visualisizing results

If you have activated the options RENDER: true and OUTPUT_GRAPHS: true in the main_config.json the results should be saved into the OutputData/Plots or OutputData/videos folder at the end of the simulation, respectively. The plots can be generated again by running python helpers\Plot_json_logs.py or executing the file via your IDE. You can change the output file format in the line containing FILE_EXTENSION = ".svg" for example to PDF or PNG format. In the corresponding file the name of the json logs need to be provided e.g. run_2024-07-27_13-31-13_trace.json.

Comparing control strategies

Usually the environment provides some degree of stochastic behavior, especially for EV statistics (arrival and departure time, battery characteristics etc.). This might not be desired if different strategies are supposed to be compared. Therefore, a recording and replay functionality is provided. With helpers/generate_traffic_sim_recording.py the behavior can be saved in a json e.g. config\traffic_sim_config\arriving_evs_record_2024-11-08_16-30-30.json and be replayed by providing it as "recording_data_path" field within the env_config file you are using.

NOTE: The env_config will be checked if it fits with the recording. If e.g. the start dates differ an error will be thrown. Same if the assigner_mode is not set to "fixed" in the env_config.

MPC based control strategy

To use the MPC based control strategy, please see the explanation Controller_Agent/Model_Predictive_Controller/MPC_Explanation.md.

RL based control strategy

To use the RL based control strategy, please see the explanation Controller_Agent/Reinforcement_Learning/RL_Explanation.md.

Note: As the MPC and RL strategy have been integrated with the primary purpose of performing research and investigation in the scope of thesis, the usability might not always be on point. If issues or questions come up feel free to follow the instructions in help sections or contact the author.

Help

If you encounter any problems or have any questions about this project, please open an issue in this repository.

Authors

• Max Faßbender (max.fassbender@rwth-aachen.de)
• Milan Kopynske

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details