EIS Data Analytics

EIS Data Analytics is a collection of code to analyze which fitting approach is suitable for your electrochemical impedance spectroscopy (EIS) data.
1

• EIS Data Analytics
• Overview
• Getting Started
  • Binder
  • Python
  • Microcontroller (µC)
• Example Usage
• Data
• Temperature Estimation
• Verification
• Colophon
  • Related Publications / Citation
  • License
  • Acknowledgement
• Further Information
  • Developer Info
  • Git and Jupyter Notebooks
• FAQ
• Sources

Overview

The main purpose of this tool is to process and analyze large amounts of EIS measurements. This implies that compromises have to be made in some respects. The following functions are already implemented: - Standardization of the frequency measuring points by interpolation - LinKK - ECM fitting - DRT fitting - Impedance extrema extraction - Various forms of visualization - SOC, SOH, Temperature, Phase and other parameter predictions - Linear Regression [SISO, MISO, MIMO] - Elastic Net Linear Regression [SISO, MISO, MIMO] - Support Vector Classification (SVC) [SISO, MISO] - Support Vector Regression (SVR) [SISO, MISO, MIMO] - Multi-layer Perceptron (MLP) Regression [MISO, MIMO] - Microcontroller (µC) code for SOC, SOH, Temperature, Phase and other parameter predictions - Linear Regression [SISO, MISO, MIMO] - Elastic Net Linear Regression [SISO, MISO, MIMO] - Support Vector Regression (SVR) [SISO, MISO] - Multi-layer Perceptron (MLP) Regression [MISO, MIMO]

Getting Started

Binder

Launch the project by clicking on the badge: .
If you want to process a lot of data or run it on the microcontroller, please run it on your hardware.

Python

1. Clone the project into your "pathtothisproject". Clone the submodules is important here: ```bash git clone --recurse-submodules https://git.rwth-aachen.de/isea/eisdata_analytics.git ```
2. Install Python 3.11.9 (https://www.python.org/downloads/)
3. Open a prompt and navigate to the path of this project bash cd path_to_this_project

4. Follow the instructions of the Python Read the Docs to create a virtual environment (venv) and activate it. E.g.:
   Windows bash python -m venv .venv .\.venv\Scripts\activate Linux / Mac bash python -m venv .venv . .venv/bin/activate

5. Install all necessary packages with:
   Windows bash pip install -r requirements_windows.txt Linux / Mac bash pip install -r requirements.txt

6. (Optional) Clean up your venv: bash python -m pip cache purge

7. Open this project with a Jupyter-Notebook editor of your choice, e.g. VS Code (needs to be installed separately) with: bash code

Microcontroller (µC)

The code is written for the NUCLEO-F429ZI by ST. The µC STM32F429 was also used by the foxBMS 1. Most of the custom code is optimized for the ARM Cortex-M4 core with a floating point unit (FPU) using the Common Microcontroller Software Interface Standard (CMSIS) library for vector operations. The following software versions are used during development: - STM32CubeIDE 1.16.1 - STM32CubeMX 6.12.1 - STM32Cube.AI 9.1.0 - STM32CubeFWF4_V1.28.1 - CMSIS 5.7.0 (including CMSIS-DSP 1.8.0)

See also the README.md in the microcontrollereisnetwork folder.

Example Usage

The Jupyter Notebooks contain details and step-by-step explanations. However, the data must first be analyzed. 1. adatapreparation.ipynb [Runtime with search_filters = ["LiFun_575166-01"] Intel i7-1265U: 4.8 hours; Intel W-2295: 1 hour] 2. bdatainspection.ipynb

Afterwards, Single-Input Single-Output (SISO), Multiple-Input Single-Output (MISO), and multiple-input multiple-output (MIMO) models can be fitted.

• dSISOLinear.ipynb
• eSISOElasticNet.ipynb
• fSISOSVC.ipynb
• gSISOSVR.ipynb
• hMISOLinear.ipynb
• iMISOElasticNet.ipynb
• jMISOSVC.ipynb
• kMISOSVR.ipynb
• lMISOMLPRegressor.ipynb
• mMIMOLinear.ipynb
• nMIMOMTElasticNet.ipynb
• oMIMOMultiOutputRegressor_SVR.ipynb
• pMIMOMLPRegressor.ipynb

Data

| Name Pattern | Temperature (°C) | SOC (%) | SOH (%) | Cite | | ------------------------------------- | ------------------------------ | -------------------------------------------------------- | ------------------- | ---------------------- | | LiFun_575166-01_* / "example_data" | -15, -5, 5, 15, 25, 35, 45, 55 | 0, 1, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 95, 99, 100 | ~117 to ~112 | 1 | | CATL_LEP71H3L7_01_* | 5, 15, 25, 35, 45, 55 | 0, 1, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 95, 99, 100 | 100 | 1 | | Rashid_Faraji-Niri_* | 15, 25, 35 | 5, 20, 50, 70, 95 | 100, 95, 90, 85, 80 | 2 | | Chan_Dickinson_Heins_* | 23 | 20, 35, 50, 65, 80, 100 | 100 to ~17 | 3 | | Mohtat_* | 25 | 0, 10, 20, 30, 40, 50, 60, 70, 80, 90 | 100 to 0 | 4 | | Zhang_Tang_* | 25, 35, 45 | 0, 50, 100 | 100 to 0 | 5 |

Not added yet:

| Name Pattern | Temperature (°C) | SOC (%) | SOH (%) | Cite | | ---------------- | -------------------- | ----------- | ----------- | ---------------------- | | Moy_Aadil_* | | | | 6 | | Luh_* | | | | 7 | | Mohtat_* | | | | 8 | | Naumann_* | | | | 9 |

Notes - LiFun_575166-01_* / "example_data": Cells are still aging, and data will be updated. - SOH (State of Health) values for LiFun_575166-01_*/example_data are nominal SOH values

See also the README.md in the exampleinputdata folder.

Temperature Estimation

A more detailed explanation can be found in the corresponding publication. 1
The cells “LiFun575166-01XXX.csv” in the exampleinputdata folder are used to develop an example SVR temperature estimation model. 1
The following results are obtained for three individual frequencies.
1
With the entire spectrum as input, the following errors can be obtained.
1

Verification

To verify the flexibility of the tool, the results of the publication "Accelerated state of health estimation of second life lithium-ion batteries via electrochemical impedance spectroscopy tests and machine learning techniques" by Mona Faraji-Niri et al. (https://doi.org/10.1016/j.est.2022.106295) on their dataset by Muhammad Rashid et al. (https://doi.org/10.1016/j.dib.2023.109157) is verified. Unlike their paper, we use the SVR to estimate the SOH. We used three different input combinations, 30.000 hyperparameter combinations were tested for each. Below, for each input combination, the best result.

SOH prediction based on Absolute Values of the Impedance

SOH prediction based on Absolute Values of the Impedance + Temperature + SOC

SOH prediction based on Absolute and Phase Values of the Impedance + Temperature + SOC

Colophon

Thanks to Spyridon, Xiaohan, Jacob, Anatolii, Niklas and Eren who supported us, Ole Kappelhoff , Sehriban Celik and Alexander Blömeke .

We use impedance.py for LinKK calculations and RC fitting: https://doi.org/10.21105/joss.02349. For DRT calculations, we use pyDRTtools: https://doi.org/10.1016/j.electacta.2019.135316.

Related Publications / Citation

Please cite our paper: https://doi.org/10.1016/j.jpowsour.2024.235049.

Overview of all archived versions of this git:
https://publications.rwth-aachen.de/record/997067.

If you use the EIS data from the exampleinputdata folder, please cite all relevant article.

License

This project is licensed according to the file LICENSE.

Acknowledgement

The authors acknowledge the financial support from the German Federal Ministry of Education and Research (BMBF) in the project OSLiB (project number 03XP0330C) within the competence cluster Battery Utilisation Concepts (BattNutzung).

Further Information

Developer Info

We use Git Large File Storage (LFS) and Git Submodules.

This project is written in Python 3.11.9 using Visual Studio Code and Jupyter Notebooks in a Python virtual environment on Windows. Ruff is used for formatting.

Git and Jupyter Notebooks

Consider to ignore the Jupyter Outputs in Git:

bash git config filter.strip-notebook-output.clean 'jupyter nbconvert --ClearOutputPreprocessor.enabled=True --to=notebook --stdin --stdout --log-level=ERROR'

FAQ

• pywin32 fails: https://github.com/microsoft/vscode-jupyter/wiki/Failure-to-start-kernel-due-to-failures-related-to-win32api-module
• generalfunv8 can't be found: Make sure you loaded the submodule. See step 1. at Getting Started
• Files are empty dummies: We use Git LFS. Normally, a normal pull should also trigger this. Additionally, you can try: git lfs pull.
• I want to run the hyperparameter search in parallel: It is recommended that you use Linux/WSL for this. Make sure you install Java.

Sources

[1] https://doi.org/10.1016/j.jpowsour.2024.235049
[2] https://doi.org/10.1016/j.est.2022.106295
[3] https://doi.org/10.1016/j.jpowsour.2022.231814
[4] https://doi.org/10.1149/1945-7111/ac2d3e
[5] https://doi.org/10.1038/s41467-020-15235-7
[6] https://doi.org/10.1016/j.dib.2024.111046
[7] https://doi.org/10.1038/s41597-024-03831-x
[8] https://doi.org/10.1149/1945-7111/ac2d3e
[9] https://doi.org/10.1016/j.jpowsour.2019.227666