Table 10 Overlapping vs. Non-Overlapping: SOH Error Metrics.
From: Explainable OptiCNN-SLSTM hybrid model for enhanced lithium-ion battery state of health prediction
Model / Metric | Overlapping RMSE | Non-Overlapping RMSE | Overlapping MAE | Non-Overlapping MAE | Overlapping MAPE | Non-Overlapping MAPE |
|---|---|---|---|---|---|---|
PINN | 3.7% | \(\sim\)5.2–6.5% | 3.0% | \(\sim\)4.5–6.1% | 1.8% | \(\sim\)2.2–3.2% |
KAN LSTM | 3.0% | \(\sim\)4.3–5.7% | 2.5% | \(\sim\)3.7–4.9% | 1.6% | \(\sim\)2.0–2.8% |
CNN+BiLSTM | 5.4% | \(\sim\)6.5–7.8% | 4.5% | \(\sim\)5.5–7.0% | 2.9% | \(\sim\)3.3–4.2% |
OptiCNN-SLSTM | 2.15% | \(\sim\)3.1–3.8% | 1.7% | \(\sim\)2.3–3.0% | 1.1% | \(\sim\)1.5–1.8% |
