摘要
实时、准确地对作物需水量的预测是实现智能节水灌溉的关键技术。预测模型的合理选择及精度提高是作物需水决策系统的核心。本文将陕西西安地区的气象数据环境信息引入自适应神经模糊推理(ANFIS)作物参考蒸腾量(ET_0)预测模型,应用卡尔曼滤波器对气象数据经ANFIS建模得到的ET_0预测值进行滤波去噪,以提高模型的预测精度,并通过仿真和实验验证,从理论和实践两个方面来验证模型的精度。仿真结果得到,反映模型预测值与真实值之间拟合程度的均等系数(EC)值校正前为0.93,校正后达到0.98。实验结果得到,ANFIS预测模型的平均绝对误差是28.94%,平均相对误差是26.37%,卡尔曼修正后的ANFIS预测模型的平均绝对误差是7.24%,平均相对误差是6.59%。仿真和实验结果表明,利用卡尔曼滤波对ANFIS预测模型进行修正,可以提高预测的精度,经卡尔曼修正后的ANFIS模型能更佳地反映ET_0的变化趋势。
Real time and accurate prediction for water demand by crop is the key technology to realize intelligent water-saving irrigation. The reasonable selection of forecasting model and the improvement of accuracy is the key to the decision making system on water demand. This article introduced the meteorological data on environmental information in Xi＇an of Shaanxi province to the forecast model of self adaptive neural fuzzy inference （ANFIS） reference crop transpi-ration （ETq） . The caiman filter was used to filter the noise of the value obtained by the ANFIS model to improve the forecasting accuracy of the model, thus improving the forecasting accuracy of model and verifying the accuracy of the model through simulation and experiment. The simulation results showed that the equal coefficient （E C） reflecting the fit-ting degree between the real value and the result of forecasting model was 0.93 and 0.98 after being adjusted. The re-sults from experiment showed that the ANFIS forecast model ＇ s mean absolute error was 28.94%, and the average relative error was 26.37%. After modification, the mean absolute error was 7.24%, and the average relative error was 6.59% . Simulation and experimental results indicated that the prediction model of ANFIS was modified by using caiman filter, which could improve the accuracy of prediction. The revised ANFIS model by the caiman had better reflection of the change trend of ET0.
出处
《干旱地区农业研究》
CSCD
北大核心
2017年第3期114-119,共6页
Agricultural Research in the Arid Areas
基金
新疆维吾尔自治区高技术研究发展项目"干旱区智能控制微灌技术与设备"(201413102)
