基于卷积神经网络和双向长短期记忆网络的气温预测模型  

Temperature Prediction Model Based on Convolutional Neural Networks and Bidirectional Long Short-Term Memory

作　　者：叶剑 唐欢 殷华 高振翔 YE Jian;TANG Huan;YIN Hua;GAO Zhenxiang(Suqian Meteorological Bureau,Suqian 223800,China)

机构地区：[1]宿迁市气象局,江苏宿迁223800

出　　处：《现代信息科技》2024年第21期35-40,45,共7页Modern Information Technology

基　　金：江苏省气象局青年基金项目(KQ202420);“宿迁英才”群英计划培养资助项目;宿迁市级指导性科技计划项目共同资助。

摘　　要：气温与环境要素之间存在非线性关系,针对传统的预测方法难以捕捉数据的内在特征和时间相关性问题,提出一种基于卷积神经网络与双向长短期记忆网络相结合的气温预测模型。基于宿迁四个国家气象观测站的逐小时观测数据,首先通过一维卷积神经网络提取气象要素数据的空间特征,然后将这些特征引入双向长短期记忆网络中来全面学习并掌握气象要素的上下文信息,进而对气温进行有效预测。实验结果表明,与其他的预测方法相比,所提模型在空间特征提取和时序特征学习方面表现卓越,且其在气温预测的精度上有显著的优势。There is a nonlinear relationship between temperature and environmental factors.Aiming at the problems that traditional prediction methods are difficult to capture the inherent characteristics and temporal correlation of the data,a temperature prediction model based on a combination of Convolutional Neural Networks and Bidirectional Long Short-Term Memory is proposed.Based on hourly observation data from four national meteorological observation stations in Suqian,firstly,the spatial features of meteorological element data are extracted through the One-dimensional Convolutional Neural Networks,followed by these features are introduced into the Bidirectional Long Short-Term Memory to comprehensively learn and master the contextual information of meteorological elements,so as to effectively predict the temperature.The experimental results show that compared with other prediction methods,this proposed model performs excellently in spatial feature extraction and temporal feature learning,and it has significant advantages in the accuracy of temperature prediction.

关 键 词：深度学习 卷积神经网络 双向长短期记忆网络 气温预测 对比分析 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] P457.3[自动化与计算机技术—控制科学与工程]