基于神经网络的灌区融雪型河源来水预报模型

该文将神经网络理论应用于冰雪融水补给为主的河流来水过程的模拟与预报，研究并识别冰雪融水补给为主的河流来水变化过程与其影响因子之间的复杂非线性关系，为无调蓄设施灌区灌溉来水预报提供一种新的方法和途径。在此基础上将神经网络模型与传统回归模型进行了对比分析，并用于新疆塔什库尔干河流来水量预报，两模型预报结果与实测结果一致，神经网络模型模拟精度更高；神经网络模型在预报因素选择方面较回归模型简单，有成熟的理论基础。研究分析表明其成果完全可以应用于工程生产，解决灌区来水的预报难题，有较好的应用前景。

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.