To improve the accuracy of flow discharge at monitoring sections influenced by hydraulic structures and the performance of HADCP-based discharge monitoring system, the apparatus depth, drop height and other relevant factors were comprehensively considered. A multilinear regression model solved by least-square method was built to calculate the cross-sectional mean velocity. In order to improve the low accuracy of flow discharge in low-flow conditions, the relationship of the measured mean velocity and the velocity cell of the H-ADCP which has a high correlation with the measured-mean velocity was considered using the LASSO regression model in machine learning for parameter estimation in low-flow conditions. The approach was applied in the Baihe Hydrological Station and the results meet the requirements of specification. The research results provide references for the H-ADCP velocity derivation scheme design in the Baihe Hydrological Station and other hydrological stations affected by hydraulic projects.