摘要
选取2019年1月-2020年1月PM2.5的日均值浓度对青海省海东工业园区PM2.5浓度进行预测及分析,采用Matlab神经网络工具将数据用于训练中所使用的样本数据,选用其中二十个作为测试样本。选取模型进行十次预测,将所得数据取平均值,作为最终结果进行分析和对比。本文利用三种不同网络模型预测大气中含有的细颗粒物浓度,研究真实值与预测值差异最大的天数,发现拟合相对较差时,预测值都处于波峰和波谷地段,这种情况在三种模型中都存在。Elman神经网络模型相比前两种模型预测的误差值较大,拟合效果欠佳,因此该模型存在一定记忆性,使用时需使用连续五天里的细颗粒物浓度值。
In this paper,the Matlab neural network toolbox was used to study and forest PM2.5 in Qinghai Haidong Industrial Park.The monthly air quality data are from 2019 to 2020,they were selected for the sample data used in the training,of which 20 samples were selected as test samples,10 predictions are averaged for each model,then get the final average,and then compare and analyze.In this paper,three kinds of network models are used to predict the concentration of fine particulate matter in the atmosphere,and the days with the largest relative error are studied.It is found that when large errors occur,the predicted value is in the region of the peak and trough,which has happened in all three models.Compared with the previous two models,the Elman neural network model has a large error value.Therefore,the model has a certain memory,and the concentration value of fine particles in five consecutive days should be used.
作者
韩桂兰
李雪姣
HAN Guilan;LI Xuejiao(School of Statistics and Data Science,Xinjiang University of Finance and Economics,Urumqi 830000,China)
出处
《佳木斯大学学报（自然科学版）》
CAS
2022年第3期118-121,共4页
Journal of Jiamusi University：Natural Science Edition
基金
国家社会科学基金项目(19BGL186)。
作者简介
韩桂兰(1967-),女,新疆乌鲁木齐人,教授,硕士,研究方向:资源环境统计与宏观经济分析。
