摘 要:基于浙江省智能网格模式 2 m 温度预报产品,结合 NCEP 再分析资料、路桥区域站气象观测资料,对模式 2018—2020 年路桥地区最高气温预报能力进行检验分析,利用 BP 神经网络建立最高气温订正模型。结果表明,该模式对路桥区夏季和秋季 2 m 最高气温预报具有较好的指导作用,2 ℃以内平均准确率为 78%,平均绝对误差为 1.3 ℃,春季和冬季预报能力偏弱。订正后 2020 年各站最高气温≤ 2 ℃,平均准确率由 65% 提高到 90%,平均绝对误差由 1.8 ℃下降到 1.0 ℃,这说明 BP 神经网络对路桥地区 2 m 最高气温有很好的订正效果。
关键词:神经网络;最高气温;气温订正;智能网格预报
DOI:10.19850/j.cnki.2096-4706.2023.06.024
中图分类号:TP183;P423 文献标识码:A 文章编号:2096-4706(2023)06-0092-06
Study on the Correction of Maximum Air Temperature in Luqiao District Based on Neural Network
WANG Peng1, ZHANG Shaohua 1, QIN Caiwei 1, WANG Hongyu1, HUANG Xiaolong2
(1.Luqiao Meteorological Bureau, Taizhou 318050, China; 2.Taizhou Meteorological Bureau, Taizhou 318000, China)
Abstract: Based on the 2 m temperature forecast product of the intelligent grid mode of Zhejiang Province, combined with the NCEP reanalysis data and the meteorological observation data of the Luqiao regional station, the maximum temperature forecast ability of the model from 2018 to 2020 in the Luqiao district is tested and analyzed, and the maximum temperature correction model is established using the BP neural network. The results show that the model has a good guiding effect on the prediction of the maximum temperature of 2 m in summer and autumn in the Luqiao district. The average accuracy within 2 ℃ is 78%, the average absolute error is 1.3 ℃ , and the prediction ability in spring and winter is weak. After correction, the maximum temperature at each station in 2020 is less than or equal to 2 ℃, the average accuracy rate increases from 65% to 90%, and the average absolute error decreases from 1.8 ℃ to 1.0 ℃ , which shows that BP neural network has a good correction effect on the maximum temperature of 2 m in the Luqiao district.
Keywords: neural network; maximum air temperature; air temperature correction; intelligent grid forecasting
参考文献:
[1] 金荣花,代刊,赵瑞霞,等 . 我国无缝隙精细化网格天气预报技术进展与挑战 [J]. 气象,2019,45(4):445-457.
[2] 胡争光,薛峰,金荣花,等 . 智能网格预报应用分析平台设计与实现 [J]. 气象,2020,46(10):1340-1350.
[3] 吴君,裴洪芹,石莹,等 . 基于数值预报产品的地面气温BP-MOS 预报方法 [J]. 气象科学,2007(4):430-435.
[4] 任巧丽,马乾 .PCA-RBF 神经网络对兰州市日平均气温的预测 [J]. 宜宾学院学报,2018,18(12):111-115.
[5] 李倩,胡邦辉,王学忠,等 . 基于 BP 人工神经网络的区域温度多模式集成预报试验 [J]. 干旱气象,2011,29(2):231-235+250.
[6] 雷彦森,蔡晓军,王文,等 . 遗传算法优化的 BP 神经网
络在地面温度多模式集成预报的应用研究 [J]. 气象科学,2018,38(6):806-814.
[7] 李虎超,邵爱梅,何邓新,等 .BP 神经网络在估算模式非系统性预报误差中的应用 [J]. 高原气象,2015,34(6):1751-1757.
[8] 郭庆春,李力,张冉,等 . 我国气温变化的神经网络预测模型 [J]. 热带气象学报,2009,25(4):483-487.
作者简介:王鹏(1992.01—),男,汉族,浙江台州人,助教,本科,研究方向:气象预报。