基于Prophet模型的湖北省月降水量预测  

Prediction of Monthly Precipitation in Hubei Province Based on Prophet Model

作　　者：许浩然 陈中举[1] 杨兆前 房梦婷 詹炜[1] XU Hao-ran;CHEN Zhong-ju;YANG Zhao-qian;FANG Meng-ting;ZHAN Wei(School of Computer Science,Yangtze University,Jingzhou 434023,Hubei Province,China;College of Earth and Environmental Sciences,Lanzhou University,Lanzhou 730000,China)

机构地区：[1]长江大学计算机科学学院,湖北荆州434023 [2]兰州大学资源环境学院,兰州730000

出　　处：《节水灌溉》2022年第2期7-12,20,共7页Water Saving Irrigation

基　　金：中国高校产学研创新基金-新一代信息技术创新项目2020(2020ITA03012)。

摘　　要：为提高降水量的预测精度,建立了基于Prophet算法的降水量时序预测模型。使用湖北省1960年1月-2006年5月的月降水量资料,对2006年6月-2017年12月共139个月的月降水量进行预测,并与传统的SVR、BP以及ARIMA预测模型进行对比,以探讨该模型在降水量预测中的可行性。实验结果表明:在湖北省各站点降水量预测中,Prophet模型的预测精度优于其他3种模型,其RMSE较其他模型的平均值低18.402,MAE低16.189,且对月降水量时间序列上波峰出现的波动预测更加准确。在空间分布方面,通过克里金插值法得到的Prophet模型的预测值与实际月均降水量值的空间分布相似,说明该模型能够较为精准的预测湖北省各区域的月降水量,在降水量预测方面具有普适性。Precipitation is an important climate variable.In order to improve the prediction accuracy of precipitation,a time series prediction model of precipitation based on Prophet algorithm was established in this paper.Based on the monthly precipitation data of Hubei province from January 1960 to May 2006,the monthly precipitation of 139 months from June 2006 to December 2017 was predicted,and compared with the traditional SVR,BP and ARIMA prediction models,in order to explore the feasibility of this model in precipitation prediction.The experimental results show that the Prophet model is better than the other three models in predicting precipitation in Hubei Province,and its RMSE and MAE are 18.402 and 16.189 lower than the average of other models.Moreover,the Prophet model is more accurate in predicting the fluctuation of monthly precipitation time series.In terms of spatial distribution,the predicted value of Prophet model obtained by Kriging interpolation method is similar to the spatial distribution of the actual monthly average precipitation value,indicating that the model can accurately predict the monthly precipitation of various regions in Hubei Province and has universal applicability in precipitation prediction.

关 键 词：Prophet模型 降水量 预测 湖北省 克里金插值法 

分 类 号：S27[农业科学—农业水土工程] P332[农业科学—农业工程]