Table 3 Performance comparison under different data partitioning strategies.
From: Cycle based state of health estimation of lithium ion cells using deep learning architectures
Model | Split type | RMSE | MAE | R² |
|---|---|---|---|---|
MLP | Chronological (80:20) | 0.0069 | 0.0049 | 0.9955 |
Block-wise (60:20:20) | 0.0072 | 0.0051 | 0.9951 | |
Rolling-window CV | 0.0074 | 0.0052 | 0.9949 | |
TCN | Chronological (80:20) | 0.0071 | 0.0051 | 0.9951 |
Block-wise (60:20:20) | 0.0073 | 0.0052 | 0.9948 | |
Rolling-window CV | 0.0075 | 0.0053 | 0.9946 | |
LSTM | Chronological (80:20) | 0.0076 | 0.0055 | 0.9944 |
Block-wise (60:20:20) | 0.0079 | 0.0057 | 0.9941 | |
Rolling-window CV | 0.0080 | 0.0058 | 0.9939 | |
GRU | Chronological (80:20) | 0.0160 | 0.0111 | 0.9754 |
Block-wise (60:20:20) | 0.0163 | 0.0114 | 0.9749 | |
Rolling-window CV | 0.0166 | 0.0116 | 0.9745 |
