应用自适应神经模糊推理系统(ANFIS)的ET_0预测

Prediction of daily reference evapotranspiration using adaptive neuro-fuzzy inference system

参照作物腾发量是计算作物需水量和进行灌溉预报的基础要素。该文利用自适应神经模糊推理系统(ANFIS)所具有的直接通过模糊推理实现输入层与输出层之间非线性映射能力,和神经网络的信息存储和学习能力,将其应用于参照作物腾发量预测中。根据相关分析,输入变量选择日照时数和日最高气温;用5年共1827个数据组对系统进行训练,建立了参照作物腾发量预测系统。利用该系统对近年213个数据组进行了实际预测,与Penman-Monteith方法计算结果进行比较,结果相关性良好。

The estimation of evapotranspiration from vegetated surfaces is a basic tool to compute water balances and to estimate water availability and requirements. Reference evapotranspiration (ET_0) just reflects weather conditions. Adaptive Neuro-Fuzzy Inference System(ANFIS), has the capacity of non-linear mapping between input layer and output layer by fuzzy inference, and has storing and learning ability with the information of the neural network at the same time. In this paper, the computation of daily ET_0 by ANFIS is presented, comparing the results with the ET_0 calculated through FAO Penman-Monteith method in the same period. Sunlight hour and maximum air temperature are as input variables in ANFIS according to regression analysis between every weather factor. The ANFIS for ET_0 estimator is built from training data, whose array list includes 1827 data of five years. The result of testing data of 213 datum groups, to estimate ET_0 using the ANFIS, is acceptable comparing with the result of Penman-Monteith method.