Table 1 Comparison of LIB degradation prediction performance
From: Early prediction of lithium-ion battery degradation with a generative pre-trained transformer
Method | SOH Prediction RMSE (%) | SOH Prediction MAPE (%) | Knee Point Error | Knee Point MAPE (%) | EOL Error | EOL MAPE (%) | |
|---|---|---|---|---|---|---|---|
Regressive prediction | CNN | 6.33 | 5.88 | -380 | 67.26 | Out of range | Out of range |
LSTM | 2.61 | 2.04 | -149 | 26.37 | -24 | 2.84 | |
CNN-LSTM | 2.63 | 1.90 | -20 | 3.54 | -20 | 2.37 | |
Autoregressive prediction | LSTM | 6.56 | 5.93 | -491 | 86.90 | Out of range | Out of range |
Transformer | 6.49 | 5.73 | -226 | 40.00 | Out of range | Out of range | |
BatteryGPT (5) | 2.56 | 1.90 | -177 | 31.33 | -21 | 2.49 | |
BatteryGPT (30) | 0.21 | 0.14 | 13 | 2.30 | 10 | 1.18 |
