Neural network model for predicting snow melting headwater inflow in irrigation areas

Neural network theory was applied to the simulation and forecast of the inflow process of the spring-fed stream, researching and identifying complex non-linear relationships between inflow process of the spring-fed stream and influence factors, providing one new method for forecast of irrigating inflow in the irrigation areas without storable and regulative installation. By contrast, the neural network model and the traditional multi-regression model were used in the inflow forecast of the Tashenkuergan River in Xinjiang region, the forecast results by the two models are consistent with the measured results, and the forecast results by the neural network model are better than those by the linear multi-regression model. On the aspect of choosing the flood forecast factors, the neural network model is more simple than multi-regression model, and it has the mature rationale. The research and analysis indicate the achievement can be applied to the project production for solving the problems of inflow forecast of the irrigation areas. The model proves to be widely applicable.