Side-scanning radar flow online monitoring can improve the monitoring efficiency and quality, expand the monitoring range and density, but its application accuracy is challenged under complex water conditions such as those affected by water conservancy projects. In this paper, various flow influencing factors are considered comprehensively, and the flow calculation schemes of side-scanning radar online monitoring system based on multiple linear regression model, machine learning Least Absolute Shrinkage and Selection Operator(LASSO) model, and deep learning Long Short Term Memory(LSTM) model are constructed respectively, and compared and analyzed. The application in Yunjinghong Hydrological Station shows that: the three flow calculation schemes meet the specification requirements, and can provide reference for the side-scanning radar flow calculation scheme of Yunjinghong Hydrological Station and similar stations affected by water conservancy projects; the LASSO model is the best, and the accuracy is improved by 22.93% compared with the conventional method; the accuracy of the multivariate regression model is a little bit lower than that of the LASSO model, but the construction is simple, easy, and is suitable for the need to quickly and conveniently calculate the flow rate; the LSTM model is the highest in complexity but has the lowest accuracy. This study can provide new ideas and methods for the improvement and optimization of the side-scan radar flow calculation scheme.