A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Study region: The Luohu hydrological station, located in southeastern China, which has unstable water level- discharge relationship caused by tides. Study focus: Real-time flow monitoring based on horizontal-acoustic Doppler current profiler (H-ADCP), which remains insufficient to deal with low monitoring accuracy, complex flow characteristics, and large data volumes caused by the construction and operation of hydraulic engineering, backwater, tides, siltation changes, and high-frequency monitoring. This study proposed a deep characteristic learning (DCL) model to identify and extract the nonlinear characteristics between flow velocity of H-ADCP cell and river cross section by incorporating multiple intelligent algorithms. New hydrological insights for the region: The DCL model performs efficiently with a determination coefficient (R2) of 0.93 between the simulated and observed discharge, which is obviously better than the single intelligent algorithm-based models. The DCL model allows for adaptive algorithm selection and parameter adjustment according to the characteristics of river cross section and H-ADCP data. It shows strong self-learning capability and good simulation accuracy even with few training samples. Additionally, the DCL model is demonstrated to be stable and applicable in terms of model structure and practical performance. This study can serve as a reference for real-time flow monitoring under complex hydrological conditions.
Study region: The Luohu hydrological station, located in southeastern China, which has unstable water level- discharge relationship caused by tides. Study focus: Real-time flow monitoring based on horizontal-acoustic Doppler current profiler (H-ADCP), which remains insufficient to deal with low monitoring accuracy, complex flow characteristics, and large data volumes caused by the construction and operation of hydraulic engineering, backwater, tides, siltation changes, and high-frequency monitoring. This study proposed a deep characteristic learning (DCL) model to identify and extract the nonlinear characteristics between flow velocity of H-ADCP cell and river cross section by incorporating multiple intelligent algorithms. New hydrological insights for the region: The DCL model performs efficiently with a determination coefficient (R2) of 0.93 between the simulated and observed discharge, which is obviously better than the single intelligent algorithm-based models. The DCL model allows for adaptive algorithm selection and parameter adjustment according to the characteristics of river cross section and H-ADCP data. It shows strong self-learning capability and good simulation accuracy even with few training samples. Additionally, the DCL model is demonstrated to be stable and applicable in terms of model structure and practical performance. This study can serve as a reference for real-time flow monitoring under complex hydrological conditions.
Deep characteristic learning model for real-time flow monitoring based on H-ADCP
2025
Article (Journal)
Electronic Resource
Unknown
Metadata by DOAJ is licensed under CC BY-SA 1.0
Deep characteristic learning model for real-time flow monitoring based on H-ADCP
| Elsevier | 2025
Using ADCP-Based Flow Monitoring to Improve Operations at Niagara
| British Library Online Contents | 2012
Real time quantification of relocated sediment using ADCP and OBS measurements
| British Library Conference Proceedings | 2004
Flow rate measurement uncertainty Relating to ADCP gauging
| Online Contents | 2009
Limitations in Extraction of Survey Data from Real-Time Kinematic GPS ADCP Systems
| British Library Online Contents | 2014