基于ARIMA模型的深圳市大气PM_(2.5)浓度时间序列预测分析  

The time series prediction of PM_(2.5)in Shenzhen based on ARIMA model

作　　者：严宙宁[1] 牟敬锋[1] 赵星 严燕[1] 罗文亮[1] 胡满达 

机构地区：[1]深圳市南山区疾病预防控制中心环境卫生科,广东深圳518054 [2]四川大学华西公共卫生学院(华西第四医院),四川成都610041

出　　处：《现代预防医学》2018年第2期220-223,242,共5页Modern Preventive Medicine

基　　金：深圳市科技创新委员会科技计划项目(项目编号:JCYJ20170306103652632);深圳市卫生计生系统科研项目(项目编号:201607065)

摘　　要：目的建立深圳市大气细颗粒物(PM_(2.5))时间序列分析的自回归移动平均模型(ARIMA),预测深圳市大气PM_(2.5)浓度变化趋势,为公众健康出行提供科学依据。方法收集深圳市2016年大气PM_(2.5)逐日监测数据构建ARIMA预测模型,对建立的模型进行参数估计、模型诊断,选择最优预测模型。利用构建的最佳模型对深圳市2017年1月1日-2017年1月5日大气PM_(2.5)逐日浓度进行预测,并对预测效果进行评价。结果 ARIMA(2,1,2)模型为深圳市大气PM_(2.5)浓度最优预测模型,其最小赤池信息量准则(AIC)、贝叶斯信息准则(BIC)值分别为2 683.51、2 703.01,模型残差序列的Ljung-Box统计量χ~2=0.018,差异无统计学意义(P=0.894),提示残差为白噪声序列,模型拟合良好。深圳市2017年1月1日-2017年1月5日大气PM_(2.5)浓度监测值与预测值的平均相对误差为15.6%,实际值均在预测值95%可信区间内。结论ARIMA(2,1,2)模型能较好地模拟深圳市大气PM_(2.5)变化趋势,具有良好的预测效果。Objective To build appropriate prediction model of the PM2.5 in Shenzhen using autoregressive integrated moving average （ARIMA） model, predict of trend of PM2.5 in Shenzhen, and provide scientific basis for people＇s healthy travel. Methods Time series analysis was conducted by using the daily data of PM2.5 in Shenzhen in 2016, and a predictive model was established after parameter estimation and model diagnosis. The optimal prediction model was selected, and it was used to predict the value of PM2.5 from January 1^st,2017 to January 5^th,2017,and the prediction effect was evaluated. Results ARIMA （2,1,2） was the optimal prediction model for PM2.5 in Shenzhen, the Akaike Information Criterion （AIC） and Bayesian information Criterion （BIC） of the ARIMA（2,1,2） were 2683.51 and 2703.01, respectively. Ljung-Box statistics value χ^2=0.018, there was not significantly different （P=0.894）, suggesting a white noise sequence of residuals with good model fitting. The average relative error between the predictive value and the actual value of PM2.5 from January 1^st, 2017 to January 5^th, 2017 was 15.6%, and the actual values were within 95% CI of the predictive values. Conclusion The ARIMA （2,1,2） model could predict the change trend of PM2.5 in Shenzhen with good prediction effect.

关 键 词：PM2.5 时间序列 ARIMA模型 预测 

分 类 号：R122.2[医药卫生—环境卫生学]