基于灰色关联与SVM的蒸发量季节性预测
|
摘要
为正确认识蒸发量与气象因子之间的关系,以山西太原气象站为例,应用灰色关联分析选择关联度较高的气象因子,利用Matlab软件构建以该站2010～2016年日气象数据作为训练样本、分5个时段建立26个气象因子组合下的130个蒸发量SVM预测模型,并以2017年日气象数据作为验证样本,对模型模拟结果进行验证。结果表明,季节性蒸发量预测模型模拟精度高于全年蒸发量预测模型,且气象因子组合对模型模拟效果具有重大影响,在太原地区最佳季节性模型春、夏、秋、冬四季所对应的因子组合及数量均不同,模型预测R2值分别为0.78、0.53、0.53、0.51,RMSE值分别为0.83、1.42、1.16、1.31,预测结果较好。
To understand the relationship between evaporation and meteorological factors,this article chose Taiyuan station of Shanxi as the research object.The grey relation analysis was used to choose high correlation factors,and Matlab software was used to establish 130 evaporation SVM prediction model under 26 meteorological factors combination with five time intervals.The site daily meteorological data from 2010 to 2016 was taken as the training sample.The daily meteorological data in 2017 was used as the validation samples to test the simulation results of the model.The results show that the prediction model accuracy of seasonal evaporation capacity is higher than the annual prediction model,and the meteorological factors combination has a significant impact on the simulation effect of the model.In Taiyuan region,factor combination and number for the best spring,summer,autumn,winter seasonal model are different.The R2 values of prediction models were 0.78,0.53,0.53 and 0.51,RMSEvalues were 0.83,1.42,1.16 and 1.31,respectively.The effect of prediction model is good.
To understand the relationship between evaporation and meteorological factors,this article chose Taiyuan station of Shanxi as the research object.The grey relation analysis was used to choose high correlation factors,and Matlab software was used to establish 130 evaporation SVM prediction model under 26 meteorological factors combination with five time intervals.The site daily meteorological data from 2010 to 2016 was taken as the training sample.The daily meteorological data in 2017 was used as the validation samples to test the simulation results of the model.The results show that the prediction model accuracy of seasonal evaporation capacity is higher than the annual prediction model,and the meteorological factors combination has a significant impact on the simulation effect of the model.In Taiyuan region,factor combination and number for the best spring,summer,autumn,winter seasonal model are different.The R2 values of prediction models were 0.78,0.53,0.53 and 0.51,RMSEvalues were 0.83,1.42,1.16 and 1.31,respectively.The effect of prediction model is good.
引文
[1] 崔亚强,朱元骏.陕西长武塬区蒸发皿蒸发量影响因子及预测[J].干旱气象,2017,35(4):575-580,597.
[2] 杨甫乐,荣艳淑,杨甫光.四川盆地蒸发皿蒸发量变化趋势及影响因子分析[J].水资源保护,2014,30(3):38-44,60.
[3] 李俊.基于改进的BP神经网络的水面蒸发预测[J].节水灌溉,2017(11):42-45.
[4] 刘思峰,党耀国,方志耕,等.灰色系统理论及其应用[M].北京:科学出版社,2004.
[5] 国家气象信息中心.中国国家级地面气象站基本气象要素日值数据集评估报告[R].北京:国家气象信息中心,2012.
[6] 李传刚,纪昌明,张验科,等.基于支持向量机的水资源短缺量预测模型及其应用[J].水电能源科学,2015,33(5):22-25.
[7] Si Wu,Shun-ichi Amari.Conformal Transformationof Kernel Functions:A Data-dependent Way to Im-prove Support Vector Machine Classifers[J].NeuralProcessing Letters,2002,15(1):59-67.
[2] 杨甫乐,荣艳淑,杨甫光.四川盆地蒸发皿蒸发量变化趋势及影响因子分析[J].水资源保护,2014,30(3):38-44,60.
[3] 李俊.基于改进的BP神经网络的水面蒸发预测[J].节水灌溉,2017(11):42-45.
[4] 刘思峰,党耀国,方志耕,等.灰色系统理论及其应用[M].北京:科学出版社,2004.
[5] 国家气象信息中心.中国国家级地面气象站基本气象要素日值数据集评估报告[R].北京:国家气象信息中心,2012.
[6] 李传刚,纪昌明,张验科,等.基于支持向量机的水资源短缺量预测模型及其应用[J].水电能源科学,2015,33(5):22-25.
[7] Si Wu,Shun-ichi Amari.Conformal Transformationof Kernel Functions:A Data-dependent Way to Im-prove Support Vector Machine Classifers[J].NeuralProcessing Letters,2002,15(1):59-67.