基于人工神经网络的作物水-盐响应初步研究

利用人工神经网络中最成熟的BP网络对内蒙古河套灌区的主要经济作物油葵的水-盐响应进行初步研究，以油葵的6个生育阶段的土壤溶液相对浓度为输入样本，油葵的相对产量为输出样本，训练并确定了具有6个输入向量，3个隐含结点，1个输出向量的3层BP网络。经检验，该模型能正确表达油葵的相对产量和各个生育阶段土壤溶液相对浓度之间的关系，模型的预测产量与实际产量具有较高的相关系数，是对传统的作物水-盐响应模型的补充。

PRELIMINARY STUDY ON CROP-WATER-SALT RESPONSE WITH ARTIFICIAL NEURAL NETWORK

With the BP network, the most mature ANN, a preliminary study on crop-water-salt response of sunflower which is the most important economic crop in Hetao irrigation district had been carried out. The BP network's input sample is soil solution relative concentration in sunflower's six different stages. The output sample is sunflower's relative yield. Based on these data, A three lays BP network including six input vectors, three hidden units and one output vector had been trained. Simulating result showed that there was good correlation between this model's forcasting yield and actual yield. This model is reliable to describe the relationship between soil solution relative concentration and relative yield and is a good supplement to traditional crop- watersalt response model.