Author(s): Wentao Wei; Lixin He; Yan Long; Xiaohui Lei; Ping Xue; Kaige Chen

Linked Author(s): Xiaohui Lei

Keywords: Data cleaning; Water balance; Hydrodynamic model

Abstract: With the rapid development of long-distance open-channel water transfer projects, ensuring the reliability of flow monitoring data has become crucial for operational safety and efficiency. However, flow inversion anomalies — where downstream flow exceeds upstream values — seriously undermine data quality and hinder rational scheduling and hydrodynamic simulation. To address this issue, this study develops a data cleaning model integrating the dynamic water balance principle and the interval flow longest sequence method. The model quantifies the theoretical flow for each gate by accounting for channel water loss rates, identifies abnormal data beyond a confidence interval (relative deviation <1%), and performs corrections based on water balance principles. Furthermore, the causes of inversion anomalies are systematically analyzed, highlighting equipment errors and gate regulation dynamics. Validation via a 1D hydrodynamic model shows that using cleaned data reduces the mean absolute error and root mean square error of upstream water level simulations by 0.0757 m and 0.0895 m, respectively. The results confirm that the proposed model effectively enhances the spatiotemporal consistency of flow data and supports more accurate simulation of water transfer processes, demonstrating significant application potential in engineering practice.

DOI:

Year: 2026