地铁车站PM2.5浓度自注意力混合预测方法研究
Research on self-attention hybrid prediction method for PM2.5 concentration in subway stations
陈定宇 1高国飞 1袁泉2
作者信息
- 1. 北京城建设计发展集团股份有限公司,北京,100037
- 2. 广州地铁设计研究院有限公司,广东 广州 510010
- 折叠
摘要
建立可靠的空气质量预测模型对经济发展和污染治理至关重要,解决PM2.5浓度的预测问题成为当务之急.本文提出了一种基于自注意力机制的混合预测方法,旨在提高PM2.5浓度的预测精度.使用自注意力机制来捕捉序列中的关键信息;用GRU对序列进行预测;使用DBN对误差序列进行校正,以提高预测的准确性和稳定性,形成了最终的预测序列.为了验证模型的性能,以我国四个地铁车站的室外PM2.5数据为例进行数据处理和预测.结果表明,预测模型在准确性和稳定性方面优于其他参照模型,为决策者提供了科学依据,以更好地治理大气污染问题.
Abstract
It is of great significance to establish a reliable air quality prediction model for economic de-velopment and pollution control.Since PM2.5 is the main pollutant in most parts of China,it has be-come a top priority to solve the problem of predicting PM2.5 concentration.In this paper,we propose an error correction model based on the self-attention mechanism to improve the prediction accuracy of PM2.5 concentration.This paper uses a self-attention mechanism to capture key information in the se-quence.The GRU is used to predict the sequence.The DBN is used to correct the error series to im-prove the accuracy and stability of the prediction,and the final prediction sequence is formed.In order to verify the performance of the model,this paper takes the outdoor PM2.5 data from Beijing,Tian-jin,Shanghai,and Guangzhou in China for metro stations as examples for data processing and predic-tion.The results show that the prediction model in this paper is superior to other reference models in terms of accuracy and stability,and provides a scientific basis for decision-makers to better control the problem of air pollution.
关键词
PM2.5/预测/自注意力机制/门控循环单元(GRU)/深度信念网络(DBN)Key words
PM2.5/prediction/self-attention/GRU/DBN引用本文复制引用
基金项目
国家自然科学基金面上项目(52072412)
出版年
2024
NETL
NSTL
维普
万方数据