Project THIRD REVIEW_TEMPLATE_FINAL DEMO.pptx

important

1. DEPARTMENT OF ELECTRICAL AND ELECTRONICS ENGINEERING AI-driven optimization of battery materials for fast-charging applications Guide Name: Dr.L.UMASANKAR Associate Professor Department of EEE S.A. Engineering College Student Name/Reg.no : Pradheep G / 111922EE01051 Reegan K / 111922EE01058 Sathish R / 111922EE01063 Saran Kumar S /111922EE01308 Batch No: 03 Date: 06/10/2025 1 of 12 S.A. ENGINEERING COLLEGE, CHENNAI - 600077 (Autonomous – Institute Level Research Centre - Affiliated to Anna University) Accredited by NBA, NAAC with ‘A’ Grade & ISO 9001:20015 Academic Year : 2025 – 2026 ODD THIRD REVIEW

2. Introduction & Objectives Introduction : The rapid adoption of electric vehicles (EVs) and renewable energy storage systems demands the development of batteries that can charge faster, last longer, and operate more safely . Conventional lithium-ion batteries struggle with fast charging due to issues like slow ion transport , thermal instability , and lithium plating , which lead to capacity loss and safety hazards. Problem Statement : Current lithium-ion batteries struggle to balance fast charging with long-term stability and safety. Inefficient material selection leads to overheating, degradation, and reduced battery lifespan. Trial-and-error methods in material discovery are time-consuming and resource-intensive. 2 of 12

3. Introduction & Objectives Objective : To develop an AI-driven framework for identifying and optimizing novel battery materials with enhanced fast-charging performance, by integrating graph neural networks (GNNs), density functional theory (DFT), battery degradation modeling, and reinforcement learning-based charging protocol optimization. Relevance to Sustainable Development Goals (SDG) : SDG 13: Climate Action - Enables more efficient energy storage, supporting wider adoption of renewable energy and electric vehicles (EVs). SDG 9: Industry, Innovation and Infrastructure - Promotes cutting-edge AI technology integration in materials science, accelerating innovation in battery manufacturing. SDG 7: Affordable and Clean Energy - Reduces reliance on fossil fuels by making EVs charge faster and more viable, leading to lower greenhouse gas emissions.

4. Literature Review Summary 1. J.-Y. Hwang et al., "Hard carbon anodes for fast-charging Na-ion batteries," Nature Energy, vol. 8, no. 1, pp. 92–105, 2023 Contribution : Designed Na-ion anodes for fast charging Strength: Application-focused real-world testing Drawback / Gap : Doesn’t use ML or connect chemistry with performance modeling Solution: In our proposed methodology, this drawback is mitigated by coupling Na-ion chemistry with PyBaMM -based performance modeling to establish structure-to-performance relationships.

5. Literature Review Summary 2. C. Chen et al., “Graph networks as a universal machine learning framework for molecules and crystals,” Nat. Commun., vol. 13, no. 1, p. 2453, 2022. Contribution : Developed GNNs for materials prediction. Strength: High accuracy for crystalline materials. Drawback / Gap : Only single-target prediction (e.g., conductivity). Solution: In our proposed methodology, this drawback is mitigated by developing a multi-task GNN that simultaneously predicts both ionic conductivity and electrochemical stability.

6. System Architecture / Block Diagram

7. Simulation setup 1. Material Properties & Physics Parameters Materials Analyzed : LGPS, LLZO, Na₃SbS ₄ Key Input Parameters : Activation Energy (Eₐ): 0.20–0.28 eV Attempt Frequency ( ν): 10¹³ s⁻¹ Hop Distance (a): 3 Å Mobile Ion Density: 10²⁸ m⁻³ Reference Temperature: 298.15 K

8. Simulation setup 2. Ionic Conductivity Calculation Used  Arrhenius-based diffusion model :

9. Simulation setup 3. PyBaMM Battery Simulation Framework Model : DFN (Doyle-Fuller-Newman) for detailed electrolyte behavior Parameter Set : Chen2020 (Li-ion parameters adapted for Na-ion materials) Simulation Type : C-rate sweep (0.5C, 1.0C, 2.0C) Experiment Setup : Charge at specified C-rate for 1800/C seconds Rest for 300 seconds Discharge at C/2 until 3.0 V cutoff

10. Simulation setup 4. Numerical Stability Enhancements Applied  scale factors (10¹⁸)  for conductivity values Used  CasadiSolver  in "safe" mode for robust convergence Conservative C-rates to ensure simulation stability

11. Phase 4 – Simulation setup 5. RL Charging Optimization Setup Environment : Custom Gym environment with material-specific dynamics State Space : [SOC, Voltage, Temperature, Time, Conductivity] Action Space : 20 discrete C-rates (0.1C to 2.0C) Reward Function : Material-aware balancing of speed vs. safety Algorithm : Deep Q-Network (DQN) with experience replay

12. Simulation setup 6. Output Configuration Directory Structure : Organized by material (LGPS/, LLZO/, Na₃SbS ₄/) Data Outputs : Time-series CSV files (voltage, current, time) Performance summaries (capacity, energy, metrics) Parameter override logs (JSON) Comparative analysis plots

13. Simulation setup 7. Validation Approach Model 1 : Load & Test – Validates AI agents follow physics-based rules Model 2 : Comprehensive Analysis – Deep scientific benchmarking Model 3 : Quick Summary – Stakeholder-friendly insights

14. Simulation Results

15. Simulation Results

16. Analysis of Results Model 1 – Load & Test: Validates that the trained AI agents follow physics-based rules, ensuring reliable and safe operation. Model 2 – Comprehensive Analysis: Performs deep scientific analysis with graphs and benchmarking to interpret AI performance and logic. Model 3 – Quick Summary: Summarizes complex results into simple insights for stakeholders, including material rankings and recommendations. 2 of 12

17. Applications & Benefits For Fast-Charging Battery Applications: Primary choice: LGPS Alternative: LLZO (if cost/processing factors favor it) Avoid for fast-charging: Na₃SbS ₄ (better suited for standard charging applications) The results clearly demonstrate that  materials with lower activation energies and higher ionic conductivities enable better fast-charging performance  - with LGPS being the standout candidate among the three materials tested. 2 of 12

18. Limitations & Future Work LGPS - BEST for Fast Charging Conductivity : 6.24 × 10⁻³ S/cm  (Highest) Performance : Shows the most stable voltage profile at 1C Why it's best : Highest ionic conductivity = fastest ion transport 2. LLZO - MEDIUM Performance Conductivity : 8.91 × 10⁻⁴ S/cm Performance : Moderate voltage stability Limitation : ~7x lower conductivity than LGPS 3. Na₃SbS ₄ - SLOWEST Charging Conductivity : 2.78 × 10⁻⁴ S/cm  (Lowest) Performance : Poorest voltage stability Limitation : ~22x lower conductivity than LGPS

19. Conclusion This project is technically feasible, scientifically relevant, and will deliver meaningful outcomes by the end of the academic cycle. To integrate GNN-based material prediction, DFT validation, PyBaMM performance simulation, and RL-based charging optimization into a unified research framework. 2 of 12

20. References & Acknowledgements C. Chen et al., "Graph networks as a universal machine learning framework for molecules and crystals," Nature Commun., vol. 13, no. 1, p. 2453, 2022. A. D. Sendek et al., "Machine learning-assisted discovery of solid Li-ion conducting materials," Chem. Mater., vol. 31, no. 2, pp. 342–352, 2019. J.-Y. Hwang et al., "Hard carbon anodes for fast-charging Na-ion batteries," Nature Energy, vol. 8, no. 1, pp. 92–105, 2023. V. Sulzer et al., " PyBaMM : A flexible battery modeling framework," Joule, vol. 5, no. 4, pp. 871–881, 2021. P. M. Attia et al., "Closed-loop optimization of fast-charging protocols," Nature, vol. 578, no. 7795, pp. 397–402, 2020. R. P. Joshi et al., "Generative AI for battery material design," Adv. Mater., vol. 35, no. 12, p. 2204567, 2023. S. Chu et al., "AI for battery manufacturing," Nature Rev. Mater., vol. 9, no. 2, pp. 89–103, 2024. M. Raissi et al., "Physics-informed neural networks," J. Comput . Phys., vol. 378, pp. 686–707, 2019. 2 of 12

21. Acknowledgements The team sincerely acknowledges the guidance and support of our Project Guide, Dr.L.UMASANKAR throughout the course of this work. We also thank the Head of Department, Dr.S.SENDILKUMAR , the faculty members of the Department of EEE, and our Institution for providing necessary facilities and encouragement. Finally, we express our gratitude to our peers and family for their continuous support.