Research on Multi-Factor Coastal Waterway Depth Prediction and Application Based on Attention-Enhanced LSTM Model
ObjectiveAccurate prediction of water depth in coastal waterway is essential for ensuring the safety and efficiency of construction and transportation activities, particularly in environments characterized by complex and dynamic hydrological conditions. The Pinglu Canal, an inland waterway with tida...
Saved in:
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Editorial Department of Journal of Sichuan University (Engineering Science Edition)
2025-01-01
|
| Series: | 工程科学与技术 |
| Subjects: | |
| Online Access: | http://jsuese.scu.edu.cn/thesisDetails#10.12454/j.jsuese.202500084 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849329616757456896 |
|---|---|
| author | LING Ganzhan HAN Yu WANG Jiawei JIE Weiwei TANG Ruikai HU Jiakai LIU Xiang LIANG Guangyue CAO Lu LIANG Ming |
| author_facet | LING Ganzhan HAN Yu WANG Jiawei JIE Weiwei TANG Ruikai HU Jiakai LIU Xiang LIANG Guangyue CAO Lu LIANG Ming |
| author_sort | LING Ganzhan |
| collection | DOAJ |
| description | ObjectiveAccurate prediction of water depth in coastal waterway is essential for ensuring the safety and efficiency of construction and transportation activities, particularly in environments characterized by complex and dynamic hydrological conditions. The Pinglu Canal, an inland waterway with tidal influences, serves as an example where traditional prediction models fall short in accurately forecasting water depth under complex hydrological conditions. This study proposes a water depth prediction model based on an attention-enhanced Long Short-Term Memory (LSTM) network, which is integrated into a decision-support platform for real-time management of channel transportation.MethodsFirst, key hydrological factors, including upstream discharge, daily rainfall, tidal current velocity, and tidal level, are incorporated to construct the LSTM-based coastal waterway depth prediction model. The raw hydrological data is preprocessed to address issues such as missing values, noise, and irregular time intervals, ensuring that it is suitable for time-series modeling. Then, an LSTM model is used to capture the long-term temporal dependencies within this data, allowing the model to account for the complex interactions between different hydrological variables over time. To optimize the model's performance, an attention mechanism is introduced. This mechanism improves the model's architecture by enabling it to dynamically adjust the weight of each feature at every time step, prioritizing the most relevant factors based on the current data. The attention mechanism enhances both the accuracy and stability of the model, particularly for ultra-long-term water depth forecasting under dynamic and complex coastal hydrological conditions. Finally, the optimized model is embedded into a transportation decision-support platform, allowing for real-time water depth prediction, dynamic correction of predictions based on new data, and navigable time window evaluation. The model's effectiveness is validated through comparative analyses with existing prediction models and field measurements, demonstrating its superior accuracy and reliability in predicting waterway depth.Results and Discussions The results demonstrate at two monitoring points, 5 km from the coast (Point 1) and 30 km inland (Point 2), the traditional LSTM model exhibits larger prediction errors, especially in long-term forecasts. The MAE error ranges from 0.07m to 1.08m for short-term predictions, and from 0.12m to 1.74m for long-term forecasts. The model also tends to overestimate water depth. In contrast, the attention mechanism-based LSTM model consistently keeps the MAE below 0.15m, even under sudden rainfall or upstream discharge events, showing enhanced accuracy in both short-term fluctuations and long-term trends. The model's performance across seasonal variations further highlights its robustness. During the dry season, MAE is reduced by 64.67%, and in the wet season, it decreases by 72.37%. The RMSE is also reduced by 67.52% and 73.39% in the respective seasons, with the R² coefficient improving by 2.18% and 5.60%. This demonstrates the model's adaptability to both stable and volatile water conditions. The attention mechanism-based LSTM model significantly outperforms traditional LSTM models in predicting waterway depth. Compared to the traditional model, the MAE error is reduced by 65%–72%, and the R² coefficient increases by 2.2%–5.6%, demonstrating superior predictive accuracy and stability. This improvement is particularly evident under complex hydrological conditions, where the model effectively captures non-linear and dynamic relationships between key features such as tidal flow speed, daily rainfall, and tidal water levels. In addition, when compared to a single feature vector model, the three-feature vector combination (daily rainfall, tidal flow speed, and tidal water level) resulted in an MAE error of no more than 0.14m and an R² coefficient of no less than 0.99, substantially improving the model's accuracy and stability for predicting waterway depth under complex coastal hydrological conditions. Finally, when integrated into the waterway transportation decision-support platform, the model's capabilities—such as real-time water depth prediction, dynamic correction, and navigable time window evaluation—substantially enhance the platform's effectiveness. This integrated system provides reliable and accurate information for waterway transportation management, contributing to safer and more efficient navigation in complex coastal environments.ConclusionsThis study introduces an attention mechanism-based LSTM model for waterway depth prediction in complex coastal environments. The model significantly enhances prediction accuracy and stability, particularly in dynamic hydrological conditions. By integrating key hydrological features, the model adapts effectively to varying waterway conditions, offering improved precision in depth forecasting. When applied to a waterway transportation decision-support platform, the model facilitates real-time predictions, dynamic corrections, and navigable time window evaluations, advancing the intelligence and digital management of waterway transportation systems. This work provides reliable technical support for future engineering applications. |
| format | Article |
| id | doaj-art-4a7d1003bcd14e319a05a5ce2b934bfd |
| institution | Kabale University |
| issn | 2096-3246 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Editorial Department of Journal of Sichuan University (Engineering Science Edition) |
| record_format | Article |
| series | 工程科学与技术 |
| spelling | doaj-art-4a7d1003bcd14e319a05a5ce2b934bfd2025-08-20T03:47:13ZengEditorial Department of Journal of Sichuan University (Engineering Science Edition)工程科学与技术2096-32462025-01-01110110237712Research on Multi-Factor Coastal Waterway Depth Prediction and Application Based on Attention-Enhanced LSTM ModelLING GanzhanHAN YuWANG JiaweiJIE WeiweiTANG RuikaiHU JiakaiLIU XiangLIANG GuangyueCAO LuLIANG MingObjectiveAccurate prediction of water depth in coastal waterway is essential for ensuring the safety and efficiency of construction and transportation activities, particularly in environments characterized by complex and dynamic hydrological conditions. The Pinglu Canal, an inland waterway with tidal influences, serves as an example where traditional prediction models fall short in accurately forecasting water depth under complex hydrological conditions. This study proposes a water depth prediction model based on an attention-enhanced Long Short-Term Memory (LSTM) network, which is integrated into a decision-support platform for real-time management of channel transportation.MethodsFirst, key hydrological factors, including upstream discharge, daily rainfall, tidal current velocity, and tidal level, are incorporated to construct the LSTM-based coastal waterway depth prediction model. The raw hydrological data is preprocessed to address issues such as missing values, noise, and irregular time intervals, ensuring that it is suitable for time-series modeling. Then, an LSTM model is used to capture the long-term temporal dependencies within this data, allowing the model to account for the complex interactions between different hydrological variables over time. To optimize the model's performance, an attention mechanism is introduced. This mechanism improves the model's architecture by enabling it to dynamically adjust the weight of each feature at every time step, prioritizing the most relevant factors based on the current data. The attention mechanism enhances both the accuracy and stability of the model, particularly for ultra-long-term water depth forecasting under dynamic and complex coastal hydrological conditions. Finally, the optimized model is embedded into a transportation decision-support platform, allowing for real-time water depth prediction, dynamic correction of predictions based on new data, and navigable time window evaluation. The model's effectiveness is validated through comparative analyses with existing prediction models and field measurements, demonstrating its superior accuracy and reliability in predicting waterway depth.Results and Discussions The results demonstrate at two monitoring points, 5 km from the coast (Point 1) and 30 km inland (Point 2), the traditional LSTM model exhibits larger prediction errors, especially in long-term forecasts. The MAE error ranges from 0.07m to 1.08m for short-term predictions, and from 0.12m to 1.74m for long-term forecasts. The model also tends to overestimate water depth. In contrast, the attention mechanism-based LSTM model consistently keeps the MAE below 0.15m, even under sudden rainfall or upstream discharge events, showing enhanced accuracy in both short-term fluctuations and long-term trends. The model's performance across seasonal variations further highlights its robustness. During the dry season, MAE is reduced by 64.67%, and in the wet season, it decreases by 72.37%. The RMSE is also reduced by 67.52% and 73.39% in the respective seasons, with the R² coefficient improving by 2.18% and 5.60%. This demonstrates the model's adaptability to both stable and volatile water conditions. The attention mechanism-based LSTM model significantly outperforms traditional LSTM models in predicting waterway depth. Compared to the traditional model, the MAE error is reduced by 65%–72%, and the R² coefficient increases by 2.2%–5.6%, demonstrating superior predictive accuracy and stability. This improvement is particularly evident under complex hydrological conditions, where the model effectively captures non-linear and dynamic relationships between key features such as tidal flow speed, daily rainfall, and tidal water levels. In addition, when compared to a single feature vector model, the three-feature vector combination (daily rainfall, tidal flow speed, and tidal water level) resulted in an MAE error of no more than 0.14m and an R² coefficient of no less than 0.99, substantially improving the model's accuracy and stability for predicting waterway depth under complex coastal hydrological conditions. Finally, when integrated into the waterway transportation decision-support platform, the model's capabilities—such as real-time water depth prediction, dynamic correction, and navigable time window evaluation—substantially enhance the platform's effectiveness. This integrated system provides reliable and accurate information for waterway transportation management, contributing to safer and more efficient navigation in complex coastal environments.ConclusionsThis study introduces an attention mechanism-based LSTM model for waterway depth prediction in complex coastal environments. The model significantly enhances prediction accuracy and stability, particularly in dynamic hydrological conditions. By integrating key hydrological features, the model adapts effectively to varying waterway conditions, offering improved precision in depth forecasting. When applied to a waterway transportation decision-support platform, the model facilitates real-time predictions, dynamic corrections, and navigable time window evaluations, advancing the intelligence and digital management of waterway transportation systems. This work provides reliable technical support for future engineering applications.http://jsuese.scu.edu.cn/thesisDetails#10.12454/j.jsuese.202500084Pinglu CanalWaterway construction and transportationWaterway depth predictionattention mechanismLSTM modelWaterway transportation decision support platform |
| spellingShingle | LING Ganzhan HAN Yu WANG Jiawei JIE Weiwei TANG Ruikai HU Jiakai LIU Xiang LIANG Guangyue CAO Lu LIANG Ming Research on Multi-Factor Coastal Waterway Depth Prediction and Application Based on Attention-Enhanced LSTM Model 工程科学与技术 Pinglu Canal Waterway construction and transportation Waterway depth prediction attention mechanism LSTM model Waterway transportation decision support platform |
| title | Research on Multi-Factor Coastal Waterway Depth Prediction and Application Based on Attention-Enhanced LSTM Model |
| title_full | Research on Multi-Factor Coastal Waterway Depth Prediction and Application Based on Attention-Enhanced LSTM Model |
| title_fullStr | Research on Multi-Factor Coastal Waterway Depth Prediction and Application Based on Attention-Enhanced LSTM Model |
| title_full_unstemmed | Research on Multi-Factor Coastal Waterway Depth Prediction and Application Based on Attention-Enhanced LSTM Model |
| title_short | Research on Multi-Factor Coastal Waterway Depth Prediction and Application Based on Attention-Enhanced LSTM Model |
| title_sort | research on multi factor coastal waterway depth prediction and application based on attention enhanced lstm model |
| topic | Pinglu Canal Waterway construction and transportation Waterway depth prediction attention mechanism LSTM model Waterway transportation decision support platform |
| url | http://jsuese.scu.edu.cn/thesisDetails#10.12454/j.jsuese.202500084 |
| work_keys_str_mv | AT lingganzhan researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT hanyu researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT wangjiawei researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT jieweiwei researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT tangruikai researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT hujiakai researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT liuxiang researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT liangguangyue researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT caolu researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel AT liangming researchonmultifactorcoastalwaterwaydepthpredictionandapplicationbasedonattentionenhancedlstmmodel |