论文部分内容阅读
参考作物蒸发蒸腾量(ET_0)是计算作物需水量和进行灌溉预报的基本要素。本文利用天气预报可测因子和Penman Monteith(PM)公式ET_0计算值作为基础数据,分别建立BP神经网络模型和ANFIS自适应模糊神经推理系统模型,两种模型的估算值与PM公式的计算值没有明显差异,均表现出显著的相关性以及整体吻合度。本文对两种模型取相同的数据样本进行比较,BP-ET_0预测结果的MRE值为32.13%,RMSE为0.134 mm,而R2达到了0.971,说明模型预测精度高,稳定性良好。相较于ANFIS-ET_0的检验结果,BP-ET_0模型的均方根误差更小(0.134mm/d<0.188 mm/d),表明其预测精度更高;而ANFIS-ET_0模型估算值的平均相对误差明显小于BP-ET_0模型估算值(16.92%<32.13%),显示出ANFIS-ET_0模型更高的稳定性。两种预测模型的输入项完全可以从当前短期天气预报因子中取得而不需要专用测量设备,程序操作简单,具有实用价值,为实时灌溉预报提供了理论基础。
Reference crop evapotranspiration (ET_0) is an essential element for calculating crop water requirements and for forecasting irrigation. In this paper, BP neural network model and ANFIS adaptive fuzzy neuro-inference system model are established by using the predictable factor of weather forecast and the calculated value of ET_0 of Penman Monteith (PM) formula. The calculated values of the two models and the calculated values of PM formula are not Significant differences, both showed significant correlation and the overall agreement. In this paper, we compare the two data sets with the same data sample. The predicted MRE value of BP-ET_0 is 32.13%, RMSE is 0.134 mm, and R2 is 0.971, which shows that the model has high prediction accuracy and good stability. Compared with the test results of ANFIS-ET_0, the root-mean-square error of BP-ET_0 model is smaller (0.134mm / d <0.188mm / d), which indicates that the prediction accuracy of BP-ET_0 model is higher. The error was significantly less than the estimated BP-ET_0 model (16.92% <32.13%), indicating a higher stability of the ANFIS-ET_0 model. The input of the two prediction models can be obtained from the current short-term weather forecasting factor without the need of special measuring equipment. The program is simple and practical, which provides a theoretical basis for real-time irrigation forecasting.