ET0影响因素及ANFIS预测模型研究

Research on ET0 Influence Factors and ANFIS Forecast Model

作物参考蒸腾量(ET0)是作物生长过程中一个非常重要的数据,ET0反应的是大气蒸发能力与作物需水信息的关系。采用自适应神经模糊推理系统(ANFIS)研究易于获得的日最低气温、日最高气温、日平均气温、日平均相对湿度、实际日照时长及风速六项气象数据对作物参考蒸腾量(ET0)的相关程度,通过比较均方根误差找到相关程度最大的组合且结构简单的ANFIS模型来预测ET0值,并通过比较均方根误差来验证所建立的ANFIS模型预测的准确性。结果表明,综合考虑模型的预测精度及结构的复杂程度,日最高气温、日平均相对湿度和风速的三输入组合为最佳的,其平均绝对误差小且ANFIS结构简单。利用该输入组合训练的ANFIS模型预测ET0,其训练的均方根误差相比于用神经网络训练的预测模型小,通过比较可知ANFIS比BP神经网络训练的模型精度提高。

Crop reference evapotranspiration （ET0） is one of the most important data in the process of crop growth, and ETO reflects the relationship between atmospheric evaporation and crop water requirement. In this paper, the adaptive neuro fuzzy inference system （ANFIS） is used to study the correlation between the six meteorological data of daily minimum temperature, daily maximum temperature, daily mean temperature, daily average relative humidity, daily sunshine duration and wind speed to ETO. The ETO value is predicted by the combination with the least root mean square error which means the maximum correlation degree. The accuracy of the ANFIS model is verified by comparing the root mean square error. The results show that the training error of the four input combination of the Tmax,RHmean and Uh is best. The average absolute error is minimum and the ANFIS structure is simple. Using the ANFIS model to predict the ETO, the root mean square error of the training is small compared to the prediction model of neural network training. The accuracy of ANFIS neural network training is improved by comparing the BP neural network.