基于深度学习和支持向量机集成学习的PM_(2.5)浓度24 h预测  

PM_(2.5) 24 hours prediction based on deep learning and support vector machine stacking model

作　　者：韦惠红[1] 李剑[2] 张文言 雷建军[2] 陈璇[1] WEI Huihong;LI Jian;ZHANG Wenyan;LEI Jianjun;CHEN Xuan(Wuhan Central Meteorological Observatory, Wuhan 430074, China;School of Computer, Hubei University of Education, Wuhan 430205, China)

机构地区：[1]武汉中心气象台,武汉430074 [2]湖北第二师范学院计算机学院,武汉430205

出　　处：《华中师范大学学报（自然科学版）》2022年第2期262-269,共8页Journal of Central China Normal University：Natural Sciences

基　　金：国家重点研发项目(2017YFC0212604).

摘　　要：提前24 h准确预测PM_(2.5)浓度可以有效的避免严重污染天气对人体带来的不利影响.为了提高深度学习模型PM_(2.5)浓度24 h预测的性能和泛化能力,在传统循环神经网络(RNN)模型上添加支持向量回归(SVR)作为下采样层提取非线性特征并降维;然后添加多核卷积神经网络(CNN)提升特征表达能力;最后利用门控循环网络(GRU)可记忆时间序列中长期信息的优势进行时序预测以保证结果的稳定性.对集成SVR-CNN-GRU模型,以2015年1月1日至2020年4月10日武汉及其周边13城市的空气质量数据和地面气象数据为样本进行实例验证,结果表明,SVR-CNN-GRU在武汉市PM_(2.5)24 h预测上的表现明显优于集成之前的RNN、SVR和随机森林回归方法,而且泛化能力更强,拟合优度达到0.97,能够实现高准确度预测,达到提前24 h预警的目的.Predicting PM_(2.5)24 hours in advance will avoid the harm from serious air pollution.To enhance the efficiency and generalization ability of deep learning prediction models for PM_(2.5)24 hours in advance prediction,a sub-sampled layer based on support vector regression(SVR)was attached to traditional recurrent Neural Network(RNN)to perform nonlinear features extraction and dimension reduction.And then a multi-kernel convolutional Neural Network(CNN)was employed to enhance feature expression.At last,a gated recurrent unit(GRU)network was employed to provide high stability of time sequence prediction using its ability in long time information memory.The air quality data and meteorological data of Wuhan and its surrounding 13 cities from January 1,2015 to April 10,2020 were employed to test the SVR-CNN-GRU.The experiment results showed that SVR-CNN-GRU exceed RNN,SVR and random forest methods in higher prediction accuracy and stronger generalization ability whose R^(2) is 0.97.The proposed method would provide high accuracy prediction for early warning 24 hours in advance.

关 键 词：PM_(2.5)浓度预测 循环神经网络 支持向量机 深度学习 集成学习 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]