基于SARIMA模型的二氧化氮时间序列预测研究

Time series prediction of NO2 based on SARIMA Model

基于烟台主城区2010年1月至2018年6月的NO2浓度数据,利用Eviews统计软件建立了季节自回归移动平均模型(SARIMA),经过序列平稳化、模型识别及模型诊断,SARIMA(2,0,3)(1,1,1)12模型的相对误差可控制率在5%以内,拟合效果较为理想。拟合及预测结果表明,烟台市主城区NO2浓度具有季节性特征,呈递增趋势,随着预测步长的延长,预测误差逐渐增大。SARIMA模型目前适合进行短期预测,今后可结合非线性动力学方法.对其进行改进。

Based on NO2 concentrations in main urban area of Yantai from January 2010 to June 2018,a seasonal autoregressive integrated moving average model(SARIMA)was established with the statistical software Eviews.After smooth sequence,model recognition and diagnosis,the relative error of SARIMA(2,0,3)(1,1,1)12 model could be controlled within 5%and the fitting results were satisfied.The fitting and predicted results showed that NO2 concentrations in main urban area of Yantai performed seasonal characteristics and an increasing trend.With the extension of prediction time,the error of SARIMA model increases.Thus,this model was suitable for short-time forecasting.In further study,this model may be improved by combining the theory of non-linear dynamics.