Table 5 Comparison with state-of-the-art ventilation prediction methods from recent literature.
From: Digital twin-driven deep learning prediction and adaptive control for coal mine ventilation systems
Study | Method | Dataset Duration | Prediction Horizon | MAPE | R² | Field Deployment |
|---|---|---|---|---|---|---|
Proposed | LSTM-Attention | 6 months | 12 steps (6Â min) | 2.87% | 0.9612 | 8 months |
Zhang et al57. | CNN-BiLSTM | 3 months | 10 steps | 3.12% | 0.9580 | Simulation only |
Wang et al58. | Transformer | 4 months | 15 steps | 2.45% | 0.9685 | 2 weeks pilot |
Liu et al59. | GRU-Attention | 2 months | 8 steps | 3.68% | 0.9420 | Not reported |
Chen et al60. | Ensemble DL | 5 months | 12 steps | 3.01% | 0.9605 | Laboratory test |
