PyDMA - Battery Degradation Mode Analysis

Python implementation of TUM-EES DegradationModeAnalysis framework

🔭 Overview

PyDMA is a Python package for performing degradation mode analysis of lithium-ion and sodium-ion batteries. Among others, both electrodes can be modeled as blends, and inhomogeneity is available for both electrodes. It reconstructs measured pseudo-OCV curves using half-cell electrode potential curves to quantify three degradation mechanisms:

• LLI: Loss of lithium inventory (charge carrier loss)
• LAM_an: Loss of active material at anode
• LAM_ca: Loss of active material at cathode

The core algorithm reconstructs full-cell OCV as:

OCV_cell(SOC) = U_cathode(α_ca · SOC + β_ca) - U_anode(α_an · SOC + β_an)

Where α scales capacity and β shifts the SOC window.

⚙️ Installation

Install from PyPI:

pip install pydma

Or install from source:

git clone https://github.com/tum-ees/pydma.git
cd pydma
pip install .

For development installation:

git clone https://github.com/tum-ees/pydma.git
cd pydma
pip install -e ".[dev,notebook]"

🎮 Quick Start

import pydma
from pydma import DMAAnalyzer, DMAConfig

# Load your electrode OCP data
anode = pydma.load_ocp("path/to/anode_ocp.csv", electrode_type="anode")
cathode = pydma.load_ocp("path/to/cathode_ocp.csv", electrode_type="cathode")

# Create analyzer with configuration
config = DMAConfig(
    direction="charge",
    weight_ocv=100,
    weight_dva=1,
    weight_ica=0,
)

analyzer = DMAAnalyzer(
    config=config,
    anode=anode,
    cathode=cathode,
)

# Run analysis on aging study data
# The analyzer uses config.direction when loading the study.
results = analyzer.analyze_aging_study("path/to/aging_data")

# Access degradation modes
print(f"LLI: {results['CU2'].degradation_modes.lli:.2%}")
print(f"LAM_an: {results['CU2'].degradation_modes.lam_anode:.2%}")
print(f"LAM_ca: {results['CU2'].degradation_modes.lam_cathode:.2%}")

# Plot results
results.plot_degradation_modes()

📖 Key Features

Blend Electrode Model

Supports blended electrodes (e.g., Silicon-Graphite anodes):

from pydma import BlendElectrode

config = DMAConfig(
    use_anode_blend=True,
    gamma_anode_blend2_upper=0.30,  # Max 30% silicon
)

si_gr_anode = BlendElectrode(blend1=graphite_ocp, blend2=silicon_ocp)

analyzer = DMAAnalyzer(
    config=config,
    anode=si_gr_anode,
    cathode=cathode,
)

Inhomogeneity Modeling

Models electrode inhomogeneity effects:

config = DMAConfig(
    allow_anode_inhomogeneity=True,
    allow_cathode_inhomogeneity=True,
    max_inhomogeneity=0.3,
    inhom_anode_offset=0.2,
    inhom_cathode_offset=0.0,
)

Multiple Fitting Objectives

Combine OCV, DVA, and ICA fitting with custom weights:

config = DMAConfig(
    weight_ocv=100,
    weight_dva=1,
    weight_ica=0,
    roi_dva_min=0.1,
    roi_dva_max=0.9,
)

Speed Presets

Choose optimization thoroughness:

config = DMAConfig(speed_preset="thorough")  # "fast", "medium", or "thorough"

🔧 Silicon OCP Generation

Generate silicon OCP from measured blend electrode data:

from pydma.silicon import generate_si_curve

result = generate_si_curve(
    blend_path="path/to/blend_ocp.mat",
    graphite_path="path/to/graphite_ocp.mat",
    gamma_si=0.245,
)

📊 Parameter Vector Layout

The optimizer uses an 8-element parameter vector internally:

Index	Parameter	Description
0	α_an	Anode scaling / capacity ratio
1	β_an	Anode offset / SOC shift
2	α_ca	Cathode scaling
3	β_ca	Cathode offset
4	γ_blend2_an	Anode blend2 fraction (0 if disabled)
5	γ_blend2_ca	Cathode blend2 fraction (0 if disabled)
6	σ_an	Anode inhomogeneity magnitude
7	σ_ca	Cathode inhomogeneity magnitude

📚 Documentation

See the Getting Started Notebook for detailed examples.

📝 Release Notes

See CHANGELOG.md for the full release history.

1.0.0 highlights:

• New: inhomogeneity offset (inhom_anode_offset, inhom_cathode_offset) lets a fraction of the maximum inhomogeneity spread be present already at SOC = 0, matching MATLAB's new inhomOffsetFraction argument.
• Numerical change: q0 is now the span of the normalized SOC axis (≈ 1.0), matching MATLAB. Previously, PyDMA multiplied DVA/ICA costs by the raw Ah capacity, so fits with weight_dva > 0 and/or weight_ica > 0 may produce small numerical differences compared with PyDMA ≤ 0.1.0. Fits are now cell-size independent and consistent with the MATLAB-tuned weight defaults. OCV-only fits are unaffected.

🎖️ Acknowledgments

This is a Python translation of the TUM-EES DegradationModeAnalysis MATLAB framework. We would like to thank Johannes Natterer for providing the aging data set of a cyclic aged P45B cell and for help in translating into Python.

📯 Developers

• Mathias Rehm, Chair of Electrical Energy Storage Technology, School of Engineering and Design, Technical University of Munich, 80333 Munich, Germany
• Josef Eizenhammer, Chair of Electrical Energy Storage Technology, School of Engineering and Design, Technical University of Munich, 80333 Munich, Germany
• Moritz Günthner (student research project)
• Can Korkmaz (student research project)

✒️ Citation

This framework is the Python implementation of the MATLAB DegradationModeAnalysis toolbox. If you use this repository in any publication, please cite:

M. Rehm et al., "How to determine the degradation modes of lithium-ion batteries with silicon–graphite blend electrodes," Journal of Power Sources, 2026, DOI: 10.1016/j.jpowsour.2026.239418

The framework is also applied and validated on commercial sodium-ion batteries in the following publication. We appreciate citing this work as well, and kindly ask you to do so if your work involves sodium-ion cells:

M. Rehm et al., "Aging of commercial sodium-ion batteries with layered oxides: how to measure and analyze it?," EES Batteries, 2026, DOI: 10.1039/D5EB00221D

📜 License

BSD 3-Clause "New" or "Revised" License - see LICENSE for details.