基于最小二乘支持向量机的ET_0模拟计算  

Simulation of ET_0 based on LS-SVM method

作　　者：侯志强[1,2] 杨培岭[1] 苏艳平[1] 任树梅[1] 

机构地区：[1]中国农业大学水利与土木工程学院,北京100083 [2]中国水电顾问集团昆明勘测设计研究院,云南昆明650051

出　　处：《水利学报》2011年第6期743-749,共7页Journal of Hydraulic Engineering

基　　金：水利部公益性行业科研专项经费项目(200701025);国家自然科学基金项目(40971130);教育部新世纪优秀人才支持计划项目(NCET-07-0814)

摘　　要：参考作物蒸散量(ET_0)有多种计算方法。本文将ET_0看作是气象因素的复杂非线性回归,以日最高温度、最低温度、平均风速、太阳辐射以及相对湿度5个气象因子不同方式的组合作为输入数据,以FAO56 Penman-Monteith公式计算的结果作为预测校准值,建立了最小二乘支持向量机模型。选取河套地区临河气象站2005年逐日气象数据进行训练与验证,并将模拟结果同其他常用ET_0计算公式的计算结果进行对比研究。结果表明最小二乘支持向量机能够很好的反映ET_0同气象因素之间的非线性关系,模拟计算精度较高,但随着气象因素个数的减少模拟精度有所下降。当基于温度和辐射条件计算时,最小二乘支持向量机模拟计算结果较Priestley-Talor公式计算精度更高,当基于温度条件计算时,温度较低时最小二乘支持向量机的模拟精度高于Hargreaves公式。As a key factor in calculating the crop water requirement, the reference crop evapotranspiration （ET0） can be calculated with a variety of methods. In this paper, ET0 was considered as a non-linear regression of the meteorological factors. Different combinations of meteorological factors such as daily maximum temperature, minimum temperature, average wind speed, solar radiation and relative humidity were used as the input data. The results calculated by the FAO56 Penman-Monteith equation were used as the calibration value, and Least Square-Support Vector Machine （LS-SVM） regression model was established accordingly. The daily observation data in 2005 from the Linhe Meteorological Station in Hetao district were used to train and test the model, and the results calculated by LS-SVM and other commonly used ETo calculation formula were compared. It shows that the LS-SVM model can well reflect the non-linear relationships between ETo and the meteorological factors with high accuracy of simulation, but as the number of meteorological factor decreases, the accuracy of simulation will be lowered. The accuracy of the LS-SVM is higher than that of Priestley-Taylor, and the accuracy of the LS-SVM is higher than that of the Hargreaves, at lower temperature.

关 键 词：参考作物蒸散量 最小二乘支持向量机 气象因素组合 

分 类 号：S161.4[农业科学—农业气象学]