ET_0的人工神经网络模型与评估研究

Artificial neural network model of ET_0 and evaluation

在对内蒙古135个站点,30年气象资料及利用Penman—Monteith公式计算获得参考作物腾发量(ET0)分析的基础上,运用人工神经网络中最成熟的BP网络对内蒙古各区域在作物生育期(4～9月)建立了方便可行的ET0计算模型。并利用辐射项(ETr)和空气动力学项(ETa)对所建模型进行了评估与适用条件的分析。研究表明,大量样本训练而得的作物生育期模型具有很高的可信度,分区与分月建立的模型在一定条件下可满足生产的需要。

Based on meteorological data of 30 years of 135 sites in Inner Magnolia and ET_0 computed with Penman-Monteith formula, the ET_0 model of crop period(4～9 month) that is convenient and feasible has been established with BP network which is the most mature network of artificial neural network. Applying condition of the model has been evaluated from radiation item and air-dynamics item. The result shows that the model trained by lots of sample is high creditable, the subarea and time model can meet requirement in practice.