Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

< Back to Article

Fig 1.

Network architecture of the ST-DAN model.

More »

Fig 1 Expand

Fig 2.

Structure of the Transformer-Encoder-based Temporal reasoning component.

More »

Fig 2 Expand

Fig 3.

Schematic diagram of the self-attention mechanism.

More »

Fig 3 Expand

Fig 4.

Structure of the Physics Aware GAT.

More »

Fig 4 Expand

Fig 5.

Graph structure of meteorological element correlation.

More »

Fig 5 Expand

Fig 6.

Schematic diagram of adaptive physical constraint matrix.

More »

Fig 6 Expand

Fig 7.

Architecture of the features fusion layer.

More »

Fig 7 Expand

Table 1.

Hyperparameter configuration of the ST-DAN for ERA5 dataset.

More »

Table 1 Expand

Table 2.

Hyperparameter configuration of the ST-DAN for buoy observed data.

More »

Table 2 Expand

Table 3.

Comparison of the performance of GAT, Transformer and ST-DAN in ablation experiments.

More »

Table 3 Expand

Table 4.

Comparison of the performance of eight models in comparative experiments on the ERA5 dataset.

More »

Table 4 Expand

Fig 8.

Comparison of MAE, MSE, and RMASE during the Temperature Prediction Training Stage of Eight Models.

More »

Fig 8 Expand

Fig 9.

Comparison of R2 and AR2 during the Temperature Prediction Training Stage of Eight Models.

More »

Fig 9 Expand

Fig 10.

Comparison of MAE, MSE, and RMASE during the Temperature Prediction Testing Stage of Eight Models.

More »

Fig 10 Expand

Fig 11.

Comparison of R2 and AR2 during the Temperature Prediction Testing Stage of Eight Models.

More »

Fig 11 Expand

Table 5.

Comparison of the performance of six models in comparative experiments on the buoy data.

More »

Table 5 Expand

Fig 12.

Comparison of MAE, MSE, and RMASE during the Wind Speed Prediction Training Stage of Six Models.

More »

Fig 12 Expand

Fig 13.

Comparison of R2 and AR2 during the Temperature Prediction Training Stage of Six Models.

More »

Fig 13 Expand

Fig 14.

Comparison of MAE, MSE, and RMASE during the Wind Speed Prediction Testing Stage of Six Models.

More »

Fig 14 Expand

Fig 15.

Comparison of R2 and AR2 during the Temperature Prediction Testing Stage of Six Models.

More »

Fig 15 Expand

Fig 16.

The effect of temperature data repair based on Transformer.

More »

Fig 16 Expand

Fig 17.

The effect of temperature data repair based on ST-DAN.

More »

Fig 17 Expand

Fig 18.

Comparison of temperature data repair effects in the green area of Figs 16 and 17 (left: Transformer, right: ST-DAN).

More »

Fig 18 Expand

Fig 19.

Comparison of temperature data repair effects in the red area of Figs 16 and 17 (left: Transformer, right: ST-DAN).

More »

Fig 19 Expand

Table 6.

Comparison of specific temperature repaired values and truth (°C).

More »

Table 6 Expand

Fig 20.

The effect of wind speed data repair based on the Transformer model.

More »

Fig 20 Expand

Fig 21.

The effect of wind speed data repair based on the ST-DAN model.

More »

Fig 21 Expand

Fig 22.

Comparison of repaired wind speed data in the green area of Figs 20 and 21 (left: Transformer, right: ST-DAN).

More »

Fig 22 Expand

Fig 23.

Comparison of repaired wind speed data in the red area of Figs 20 and 21 (left: Transformer, right: ST-DAN).

More »

Fig 23 Expand

Table 7.

Comparison of specific wind speed repaired values and truth.

More »

Table 7 Expand