Per-Cell Parameter-Adaptive Charging Control for a Degrading Lithium-Ion Battery Pack
Developed by Aakash Nair (@Sky-Code06), Samuel John Saju, Krishnanunni U, and Dr. Rajesh R (Vellore Institute of Technology).
In multi-cell lithium-ion battery packs connected in series, cells degrade at non-uniform rates due to manufacturing variations and localized thermal gradients. Because series cells share an identical charging current, standard pack-level controllers (which regulate only average voltage or use static fresh-cell models) risk pushing degraded cells into severe overvoltage.
SmartCharge-Adapt addresses this with:
-
Parallel Online Estimation: Per-cell Recursive Least Squares (RLS) tracking internal Ohmic resistance (
$R_0$ ) and capacity ($Q$ ) using standard BMS terminal measurements ($V_i$ and$I$ ). -
Granular Voltage Constraints: A one-step constrained current selection law that enforces the 4.2 V ceiling on every individual cell:
$$\hat{V}_{i}(k+1 \mid I) \le 4.200,\text{V} \quad \forall i \in {1, \dots, M}$$ -
Dynamic Bottleneck Protection: Real-time identification of the most restrictive cell (
$\arg\min_i I_{\text{upper},i}$ ), safely throttling charge current to safeguard pack longevity.
| Evaluation Metric | Pack-Lumped Static | Pack-Lumped Adaptive | Per-Cell Adaptive (Ours) |
|---|---|---|---|
| Peak Voltage Overshoot (Cycle 500) | +1,955.3 mV | +304.0 mV | 0.0 mV (Safe clamp) |
| Peak Voltage Overshoot (Cycle 1000) | +3,986.3 mV (>8.19 V) | +393.3 mV | 0.0 mV (Safe clamp) |
| Lifetime Violation Duration | 2,415 s | 6,284 s | 0 s |
| Bottleneck Cell Identification | 0% (Blind) | 0% (Blind) | 100.0% Accuracy |
| Sensor Overhead | Standard BMS | Standard BMS | Standard BMS ( |
- Charging Time vs. Aging: Shows safe throttling and the natural cycle ~700 capacity fade transition (Figure 1).
- Mid-Life Voltage Profiles: Demonstrates exact 4.200 V clamping on Cell 4 vs. dangerous overshoot in baselines (Figure 2).
- Constraint Violations: Zero lifetime overshoots for the proposed controller (Figure 3).
- RLS Tracking: Online tracking of resistance growth and capacity fade (Figure 4).
- Bottleneck Detection: 100% identification of the accelerated outlier cell (Figure 5).
SmartCharge-Adapt/
├── simulation.py # Main multi-cycle simulation script
├── requirements.txt # Python package dependencies
├── PROJECT_REPORT.md # Technical report and mathematical formulation
├── results/ # High-resolution simulation result plots
│ ├── fig1_charging_time_vs_cycle.png
│ ├── fig2_voltage_trajectories_cycle500.png
│ ├── fig3_constraint_violations_vs_cycle.png
│ ├── fig4_rls_parameter_tracking.png
│ ├── fig5_bottleneck_cell_identification.png
│ └── simulation_summary_dashboard.png
└── presentation/ # HTML presentation deck and speaker notes
├── presentation.html
└── SPEAKER_NOTES.md
- Python 3.9+
- Packages listed in
requirements.txt:pip install -r requirements.txt
Execute the full 1,000-cycle aging evaluation and regenerate all figures:
python simulation.pyOutputs and plots are automatically written to results/.
- Aakash Nair — School of Electronics Engineering, Vellore Institute of Technology
- Samuel John Saju — School of Electronics Engineering, Vellore Institute of Technology
- Krishnanunni U — School of Electronics Engineering, Vellore Institute of Technology
- Dr. Rajesh R — Associate Professor, School of Electronics Engineering, Vellore Institute of Technology
