| 摘要: |
| 为实现大田作物灌溉的精细化管理，研究了基于气象因素的生育期ET0预测模型。采用灰色理论对ET0与日均、日最高、最低温度，日均风速，相对湿度以及日照时数进行灰色关联度分析，结果表明ET0与温度（包括日均、最高、最低温度）及相对湿度的灰色关联度较高。在分析ET0与上述气象因素间的相关系数基础上，采用日均温度、日均风速以及日照时数作为模型的输入，ET0作为输出，建立了BP神经网络（BPNN）预测模型；采用日均温度、日均风速、日照时数及灰色关联度作为输入，建立了模糊最小二乘支持向量机（FLSSVM）预测模型。研究结果表明，BPNN模型的训练值决定系数为0.8643，平均相对误差6.29%，预测值决定系数为0.8099，平均相对误差7.83%；FLSSVM模型的训练值决定系数为0.9684，平均相对误差2.89%；预测值决定系数为0.9663，平均相对误差3.43%。BP神经网络与FLSSVM模型的精度均较高，可以用来预测ET0日值，这为大田作物的精细化灌溉管理提供理论与技术支持。 |
| 关键词: ET0 预测 BP神经网络 灰色理论 模糊最小二乘支持向量机 |
| DOI： |
|
| 基金项目:水利部公益性行业科研专项资助(201301102)；新疆维吾尔自治区重大科技专项资助(201130103-3)；新疆水文学及水资源重点学科资助(XJSWSZYZDXK20101202) |
|
| Prediction of ET0 during growth period of crops in inner arid areas based on meteorological parameters |
| WEI Guang-hui, MA Liang, DONG Xin-guang, YANG Peng-nian |
| () |
| Abstract: |
| In order to realize fine irrigation management of field crops, the ET0 prediction model was studied based on meteorological parameters in growth period. The grey theory was applied to analyze the grey relational degree between meteorological factors (daily average temperature, maximu
m temperature, minimum temperature, wind speed, relative humidity and sunshine hours) and ET0. The results showed that the grey correlation degree between ET0 and temperature (including daily average, maximum and minimum temperature) as well as relative humidity was relatively high. Based on the correlation coefficients between ET0 and meteorological parameters, the BP neural network (BPNN) prediction model was established by using daily average temperature, wind speed and sunshine hours as input and ET0 as output of the model. The fuzzy least square support vector machine (FLSSVM) prediction model was also established by using daily average temperature, wind speed, sunshine
hours and grey correlation degree as input. The results showed that: in the BPNN model, the trained determination coefficient was 0.8643 with the average relative error of 6.29%, and the predicted determination coefficient was 0.8099 with the average relative error of 7.83%; in the FLSSVM model, the trained determination coefficient was 0.9684 with the average relative error of 2.89%, and the predicted determination coefficient was 0.9663 with the average relative error of 3.43%. Both BP neural network and FLSSVM model exhibited a high precision and could be used to predict daily value of ET0. The results could provide the theoretical and technical support for the fine irrigation management of field crops. |
| Key words: ET0 prediction BP neural network (BPNN) grey theory fuzzy least square support vector machine (FLSSVM) |
