基于不同ELM的西北旱区参考作物蒸散量模拟模型  

Simulation Models of Reference Crop Evapotranspiration in Northwest Arid Region Based on Different ELM

作　　者：徐颖 张皓杰 崔宁博[1,2] 冯禹 胡笑涛[3] 龚道枝[4] XU Ying;ZHANG Hao-jie;CUI Ning-bo;FENG Yu;HU Xiaotao;GONG Dao-zhi(State Key Laboratory of Hydraulics and Mountain River Engineering ,College of Water Resourse and Hydropower, Sichuan University,Chengdu 610065,Sichuan Province,China;Provincial Key Laboratory of Water-saving Agriculture in Hilly Areas of Southern China,Chengdu 610066,Sichuan Province,China;Key Laboratory of the Ministry of Agriculture,Water and Soil Engineering, Dryland Agriculture and Forestry University of Northwest China,Yangling 712100,Shaanxi Province,China;State Engineering Laboratory for Efficient Water Use and Disaster Loss Reduction of Crops,Institute of Environment and Sustainable Development in Agriculture, Chinese Academy of Agriculture Science,Beijing 100081,China)

机构地区：[1]四川大学水力学与山区河流开发保护国家重点实验室水利水电学院,四川成都610065 [2]南方丘区节水农业研究四川省重点实验室,四川成都610066 [3]西北农林科技大学旱区农业水土工程教育部重点实验室,陕西杨凌712100 [4]中国农业科学院农业环境与可持续发展研究所作物高效用水与抗灾减损国家工程实验室,北京100081

出　　处：《中国农村水利水电》2019年第1期6-12,共7页China Rural Water and Hydropower

基　　金："十三五"国家重点研发计划课题(2016YFC0400206);国家自然科学基金项目(51779161);"十二五"国家科技支撑计划课题(2015BAD24B01);2017年中央高校基本科研业务费专项资金资助(2016CDDY-S04-SCU)

摘　　要：参考作物蒸散量(reference crop evapotranspiration,ET0)的精准模拟是智慧灌溉、农田水分高效利用及水资源优化调度的重要依据。为有效提高气象资料缺乏情况下对西北旱区ET0的模拟精度,在西北旱区选取4个代表性站点,并以FAO 56 Penman-Monteith(P-M)模型的ET0计算结果为标准值,基于"sin"、"radbas"和"hardlim"3种激活函数构建27种极限学习机模型(extreme learning machine,ELM,分别记为ELM-sini、ELM-radj、ELM-hardk),并将其模拟结果与Hargreaves-Samani(H-S)、Makkink(MK)、Irmark-Allen(I-A)模型进行比较。结果表明:ELM-sin7(输入u2、Tmax和Tmin)的R2和NSE均大于0.96,RMSE小于0.35 mm/d,其GPI排名第4,模型模拟精度较高; ELM-rad5(输入Tmax、Tmin和n)和ELM-sin8(输入Tmax和Tmin)的R2和NSE分别大于0.78和0.76,RMSE小于0.93 mm/d; H-S、MK和I-A模型的R2和NSE分别小于0.77和0.63,RMSE大于1.00 mm/d,可见ELM-rad5和ELM-sin8模型精度明显高于相同输入下的其他物理模型;基于ELM-sin7探究隐含层节点数对模型精度的影响发现隐含层节点数为60～100时模型精度最高;基于ELMsin7模型进行可移植性分析发现,ELM-sin7在西北旱区内各训练站点和模拟站点组合下模拟精度较高。因此,在相同气象因子组合输入下,ELM-sini和ELM-radj模型模拟精度明显高于ELM-hardk,其中ELM-sin7模拟精度较高适用于西北旱区气象因子较少时的ET0模拟;而较传统物理模型,仅有温度和日照时数时ELM-rad5模型在西北旱区适用性更好,仅有温度时ELM-sin8模型在西北旱区适用性更强。The accurate simulation of reference crop evapotranspiration( ET0) is an important basis for intelligent irrigation,efficient utilization of water resources in farmland and optimal scheduling of water resources. With the lack of meteorological data,to enhance the simulation accuracy of ET0 in the arid regions of Northwest China effectively,4 representative sites are selected,the ET0 calculations of the FAO 56 Penman Monteith( P-M) model are set as the values of standard,and then 27 kinds of extreme learning machine model( extreme learning machine,ELM,respectively for the ELM-sini,ELM-radj,ELM-hardk) are built based on 3 activation functions( "sin""radbas"and"hardlim"). Next,the results of the simulation are compared together with the models of Hargreaves-Samani( H-S),Makkink( MK)and Irmark-Allen( I-A). The results are as follows: in the case of ELM-sin7( input u2,Tmaxand Tmin) model,the value of R2 and NSE are more than 0.96,while the RMSE is less than 0.35 mm/d. This model gets the fourth greatest GPI among all the models,indicating its higher accuracy. For the models of ELM-rad5( input Tmax,Tmin and n) and ELM-sin8( input Tmaxand Tmin),the values of R2 and NSE are more than 0.78 and 0.76,while RMSE is less than 0.93 mm/d. As for H-S,MK and I-A models,the values of R2 and NSE are less than 0.77 and 0.63,while their RMSE is more than 1.00 mm/d. Thus,the ELM-rad5 and ELM-sin8 models are obviously more accurate compared with the other physical models under the same input. What’s more,it is discovered that with the ELM-sin7 model,when the number of hidden layer nodes are from 60 to 100,the model has the highest accuracy,and the results of ELM-sin7 model portability show that ELM-sin7 model in the arid north-western China has simulation accuracy based on different sites’ data. Therefore,under the same input of meteorological factor combination,the simulation accuracy of ELM-siniand ELM-radjmodels are remarkably higher than that of ELM-hardk. And ELMsin7 model has the highest accuracy among those models,which can

关 键 词：ET0 ELM 激活函数 隐含层节点数 中国西北旱区 

分 类 号：TV93[水利工程—水利水电工程] S27[农业科学—农业水土工程]