Table 3 The results for different training data volumes in the NASA dataset are compared according to different cycle stages.
From: Multi-expert fusion for state-of-health estimation of lithium-ion batteries
Dataset | Percentage of data used for training | Different stages of cycle | Mechanism | LSTM | Transformer | MEFNet | ||||
|---|---|---|---|---|---|---|---|---|---|---|
MAE | RMSE | MAE | RMSE | MAE | RMSE | MAE | RMSE | |||
NASA B - 2000mA | 25% | 0% – 30% | 0.186 | 0.047 | 0.159 | 0.030 | 0.150 | 0.025 | 0.150 | 0.025 |
30% – 60% | 0.082 | 0.010 | 0.181 | 0.037 | 0.164 | 0.030 | 0.163 | 0.029 | ||
0% – 60% | 0.144 | 0.034 | 0.181 | 0.037 | 0.157 | 0.027 | 0.157 | 0.027 | ||
60% – 100% | 0.153 | 0.029 | 0.201 | 0.043 | 0.242 | 0.060 | 0.228 | 0.053 | ||
0% – 100% | 0.148 | 0.032 | 0.267 | 0.073 | 0.196 | 0.044 | 0.189 | 0.040 | ||
50% | 0% – 30% | 0.186 | 0.047 | 0.138 | 0.030 | 0.151 | 0.029 | 0.151 | 0.029 | |
30% – 60% | 0.082 | 0.010 | 0.170 | 0.044 | 0.110 | 0.015 | 0.109 | 0.015 | ||
0% – 60% | 0.144 | 0.034 | 0.155 | 0.038 | 0.132 | 0.023 | 0.132 | 0.023 | ||
60% – 100% | 0.153 | 0.029 | 0.215 | 0.049 | 0.238 | 0.061 | 0.219 | 0.051 | ||
0% – 100% | 0.148 | 0.032 | 0.182 | 0.043 | 0.182 | 0.043 | 0.173 | 0.037 | ||
100% | 0% – 30% | 0.186 | 0.047 | 0.104 | 0.015 | 0.186 | 0.038 | 0.186 | 0.037 | |
30% – 60% | 0.082 | 0.010 | 0.153 | 0.027 | 0.244 | 0.060 | 0.236 | 0.057 | ||
0% – 60% | 0.144 | 0.034 | 0.131 | 0.022 | 0.217 | 0.050 | 0.213 | 0.048 | ||
60% – 100% | 0.153 | 0.029 | 0.270 | 0.076 | 0.232 | 0.054 | 0.203 | 0.042 | ||
0% – 100% | 0.148 | 0.032 | 0.199 | 0.051 | 0.223 | 0.052 | 0.209 | 0.046 | ||
