Table 1.
Waterlogging depth data come from WMS.
Fig 1.
Flowchart of waterlogging depth prediction based on BiTCN-GRU model (Includes BiTCN and GRU model components).
Fig 2.
Model structure of dilated convolution.
Fig 3.
Forward and backward convolution.
Fig 4.
Model structure of bi-directional dilation convolution.
Fig 5.
BiTCN residual block.
Fig 6.
Structure of GRU unit.
Fig 7.
Schematic diagram of sliding process of sliding window.
Table 2.
Detailed information of the two datasets (Time Interval: 5 minutes).
Fig 8.
Different time steps in the sliding window on the Minshan Road dataset.
Fig 9.
Different time steps in the sliding window on the Huaihe Road dataset.
Table 3.
Comparison of different optimizers on two datasets.
Table 4.
Parameter settings on BiTCN-GRU for Minshan Road and Huaihe River Road.
Table 5.
Comparison of MAE, RMSE, and R2 for different models.
Fig 10.
Prediction error plots of different models on the Minshan Road dataset.
(a-f).
Fig 11.
Prediction error plots of different models on Huaihe Road dataset.
(a-f).
Table 6.
Comparison of BiTCN-GRU with other deep learning models.
Fig 12.
Comparison of prediction results of different models on Minshan Road dataset.
Fig 13.
Comparison of prediction results of different models on Huaihe Road dataset.
Table 7.
Multi-step prediction of BiTCN-GRU on Minshan Road and Huaihe Road datasets.