Repository meiner Masterarbeit "Entwicklung und Evaluierung eines Modells zur Prognose des State of Health von Traktionsbatterien auf Basis künstlicher neuronaler Netze"
Repository of my master’s thesis "Development and evaluation of a model for predicting the state of health of traction batteries based on artificial neural networks"
To use this project, you have to install a linux distribution. In this case you can use Ubuntu 20.04.3 LTS. It comes with the build automation tool make. It is used to run the commands in the terminal.
After you cloned this repository, you should use the following command to prepare your system.
make prepare
It installs all necessery libraries and Python packages.
After this you need to generate three directories.
mkdir models data
After the preperation you can start using the main steps to calculate the internal resistance of a specific battery.
- Place a single logfile into the directory
data
- Run
make signals
- Set the limits of the signal section in seconds in the makefile (
LO_LIM
andUP_LIM
) - Run
make resample
to resample the signal section - Run
make training
to train the MLP and SVM with the resampled signal section - Run
make simulation
to calculate the internal resistance.
If you want to predict the SoH, you have to repeat the steps shown in the example. Each time you calculate the interal resistance, save the values into the discharge_resistances.csv
file.
Remember to only place one logfile at a time in the directory data
.
After you calculated enough values you can run the command make resistance
to get a diagram with all the calculated internal resistances of the csv-file.
Now you can make a prediction about the SoH based on the course of the internal resistances.