基于GCN-BiLSTM-attention融合模型的气温预测研究

打开文本图片集

中图分类号：TP18 文献标识码：A 文章编号：2096-4706（2025）09-0053-04

Abstract： In view of the drastic increase in the scale and dimension of meteorological time series data， and the problems such as insufficient utilization of data information， relatively long training time， low prediction accuracy and weak generalization ability of existing models， a Bidirectional Long Short-Term Memory Network integrated with the Attention Mechanism （BiLSTM-attention） based on Graph Convolutional Neural Network （GCN） is proposed to predict meteorological data. When this model predicts the air temperature of Tongling meteorological station in Anhui province for the consecutive next week， the average RMSE error is 1.88 and the average MSE is 1.64. Compared with other models， it has the highest accuracy. At the same time， when predicting the air temperature of the four representative regions， the error differences among the various regions are not significant， which proves that this model has good generalization performance and high prediction accuracy.

Keywords： Bidirectional Long Short-Term Memory Neural Network; Graph Convolutional Neural Network; Attention Mechanism; air temperature prediction

0 引 言

近年来，天气变化对各行业的影响日益变大，严重影响到人类的生活。（剩余5965字）