Abstract: This paper describes a new BP neural network method that expresses turbine nonlinear characteristics with torque and flow neural networks in numerical simulation to solve the existing problems in processing the synthetic characteristic curve of hydraulic turbines. The overall design idea and the solving process are demonstrated, and followed by an analysis of the specific details and implementation steps, i.e. how to retrieve and extend sample data and how to select a neural network and its training process. A case analysis on a practical curve and the likely causes of its calculation errors are included. This work provides a new perspective for synthetic characteristic curve fitting.