基于BP神经网络的冬小麦耗水预测

该文根据中国科学院禹城农业试验站2003－2006年冬小麦季的气象资料和大型称重式蒸渗仪观测资料，把实测作物系数作为作物因子指标，建立了以日最高温度、日净辐射、实测表层60 cm土壤含水率、日序数和作物系数为输入因子，蒸渗仪实测蒸散量为输出因子的BP神经网络预测模型，神经网络拓扑结构为5-9-1，训练函数为Trainbr。检验结果表明冬小麦耗水量模型预测平均相对误差为13.1%，预测值和实测值的均方根误差为0.88 mm，模型预测Nash-Sutcliffe效率指数为0.865，预测效果较好，可满足生产需要。

Prediction of winter wheat evapotranspiration based on BP neural networks

By adopting meteorological data and the data from 2003 to 2006 collected from large weighing lysimeter with the crop of winter wheat at Yucheng Comprehensive Experimental Station, Chinese Academy of Sciences, a predicted model for winter wheat evapotranspiration was developed. Based on BP neural network, the model performance was tested with inputs of daily maximum temperature, net radiation, soil water content of top 60 cm layer, date number and measured crop coefficient and output of observed evaportranspiration. The topology of the neural network was 5-9-1 and the training function was Trainbr. The results showed that the model was good in simulating water consumption of winter wheat with average relative error of 13.1%, standard error of 0.88 mm, and Nash-Sutcliffe efficiency coefficient of 0.865. And the model can meet the requiements of production.