基于改进时空因果卷积网络的PM2.5浓度预测  

PM2.5 Concentration Prediction Based On Improved Spatiotemporal Causal Convolution Network

作　　者：王媛媛 沈俞 陈秀川 王超 张兴潮 WANG Yuan-yuan;SHEN Yu;CHEN Xiu-chuan;WANG Chao;ZHANG Xing-chao(School of Computer and Software Engineering,Huaian 223001 China;Jiangsu Internet of Things Mobile Interconnection Technology Engineering Laboratory,Huaian 223001 China)

机构地区：[1]淮阴工学院计算机与软件工程学院,江苏淮安223001 [2]江苏省物联网移动互联技术工程实验室,江苏淮安223001

出　　处：《自动化技术与应用》2025年第2期21-25,共5页Techniques of Automation and Applications

基　　金：江苏省自然科学基金面上项目(BK20211365);淮阴工学院—淮安经济技术开发区产学研合作项目(Z413H21522);大学生创新创业训练计划项目(202211049087Y,202211049268XJ)。

摘　　要：PM2.5的浓度受到多方面的影响,包括天气、空气中的其他物质成分以及周围地点的空气质量等。为提高深度学习模型对PM2.5浓度预测的准确率与泛化性能,提出了一种改进的时空因果卷积网络(STCN),使模型在预测时考虑到不同地点对预测地PM2.5浓度的不同影响。以北京市多个地点的空气质量数据进行实验并与原始的STCN网络进行对比,实验表明,改进后的时空因果卷积网络对PM2.5浓度预测的能力高于基线STCN。The concentration of PM2.5 is affected in many ways,including weather,other air components and the air quality of surrounding places.In order to improve the accuracy and generalization performance of the depth learning model for predicting PM2.5 concen-tration,an improved spatiotemporal causal convolution network(STCN)is proposed,which allows the model to take into account the different effects of different locations on PM2.5 concentration in the prediction site.The air quality data from multiple loca-tions in Beijing are used for the experiment and compared with the original STCN network.The experiment shows that the im-proved spatiotemporal causal convolution network has a higher ability to predict PM2.5 concentration than the baseline STCN.

关 键 词：空气质量预测 PM2.5 时空因果卷积网络 空间关联因素 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]