Research Article
CNN-LSTM Model Optimized by Bayesian Optimization for Predicting Single-Well Production in Water Flooding Reservoir
Table 4
Performance of different models (84-3).
| Model | MAE | MAPE | | RMSE | Accurate | Training time/min |
| CNN-LSTM | 17.58 | 6.54 | 0.99 | 37.26 | 0.93 | 6.33 | LSTM | 20.33 | 7.02 | 0.99 | 44.36 | 0.9 | 5.18 | GRU | 21.45 | 7.15 | 0.99 | 47.14 | 0.9 | 3.04 | Bi-LSTM | 22.23 | 7.69 | 0.99 | 48.12 | 0.88 | 6.44 | CNN | 22.65 | 7.73 | 0.99 | 48.87 | 0.88 | 3.16 | ARIMA | 31.47 | 15.68 | 0.98 | 37.17 | 0.83 | 2.11 | Attention-LSTM | 40.26 | 18.26 | 0.97 | 60.68 | 0.82 | 8.26 | Self-attention | 60.15 | 25.36 | 0.96 | 80.67 | 0.66 | 5.01 |
|
|
84-3: using the data from the previous 84 months to predict oil production of the next three months.
|