Home Pricing Archive About Privacy

Predictive Modeling of Li-ion Battery Health (SOH) and Lifetime Using the NASA PCoE Dataset

📑 5 slides 👁 22 views 📅 2/5/2026
0.0 (0 ratings)

Introduction & Overview

Project title: Predictive Modeling of Li-ion Battery Health (SOH) and Lifetime

Introduction & Overview
2

Motivation & Problem Statement

  • Critical importance of battery health for CubeSat power systems
  • Challenges in modeling battery degradation and predicting SOH/RUL
  • Need for accurate predictive models to ensure safety and efficiency
  • Focus on NASA PCoE dataset for reliable ground-truth data
Motivation & Problem Statement
3

Dataset Overview

  • NASA PCoE Li-ion battery dataset with multiple batteries
  • Thousands of charge-discharge cycles recorded
  • Data collected at different operating temperatures (4°C, 24°C, 43°C)
  • Comprehensive sensor measurements including voltage, current, temperature
Dataset Overview
4

Methodology Highlight

  • Leakage-proof validation using GroupKFold by battery ID
  • Three-phase pipeline: Exploration, Feature Engineering, Modeling
  • Feature selection via correlation analysis and RFE
  • Comparison of linear vs ensemble models for SOH/RUL prediction
Methodology Highlight
5

Key Insights & Conclusion

  • SOH prediction achieves 95% accuracy, RUL prediction ~74%
  • Simpler models (Lasso) generalize better than complex ones
  • Recall prioritized over accuracy for safety-critical applications
  • Validated approach provides strong foundation for real-world BMS
Key Insights & Conclusion
1 / 5