Table 2 Summary of recent studies on battery SoH Estimation using advanced deep learning models.
From: Cycle based state of health estimation of lithium ion cells using deep learning architectures
Study | Method | Dataset | Performance |
|---|---|---|---|
MLP-like memory model | NASA cells | MAE = 0.0075 | |
Neural networks for SoH | Custom dataset | RMSE = 0.0110 | |
MLP for SOC/SoH | Hardware-accelerated | MAE = 0.0082 | |
BERT + TimeGAN | NASA B0018 | R² = 0.995 | |
FPCA-SETCN framework | NASA B0005 | RMSE = 0.0094 | |
2D-CNN + Self-Attention | LFP, NMC, NCA datasets | RMSE = 0.0109 (LFP), 0.0026 (NMC) | |
LC-GDAT (Lossy Counting + Gated Dual-Attention Transformer) | NASA, Real-world EV | MAE = 0.0046 (Lab), 0.0223 (EV) | |
Vehicle-Cloud Collaborative Hybrid Model | Real-world BEV data | MAE < 0.025 | |
Hierarchical Feature Extraction + ML | Real-world EV data (300 EVs) | RMSE = 0.0105 | |
SSA-LSTM + Deep SHAP | NASA, CALCE, PolyU | RMSE < 0.05, MAE < 0.05 | |
Two-Stage Physics-Informed Neural Network (TSPINN) | NCA, NCM datasets | MAE = 0.00675 |
