基于残差优化模型的PM2.5浓度预测算法

PM2.5 Based on Residual Optimization Model Concentration Prediction Algorithm

雾霾天气的产生越来越频繁,其主要原因是大气中PM2.5浓度的增加。合理有效的PM2.5浓度预测方法对于科学治霾有着重要意义。传统的差分整合移动平均自回归(Auto-Regressive Integrated Moving Average, ARIMA)模型只适合处理线性特征,而循环神经网络预测PM2.5只是对齐非线性特征进行拟合。于是提出将ARIMA模型与循环神经网络模型组合,以陕西省西安市作为研究对象,对PM2.5进行预测。ARIMA-GRU模型的预测结果的MAE、MSE、RMSE均小于ARIMA-LSTM模型,r、R2更接近1。

Haze weather is occurring more and more frequently,mainly due to the increase in PM2.5 concentration in the atmosphere.Reasonable and effective PM2.5 concentration prediction methods are of great significance for scientific haze control.The traditional Auto-Regressive Integrated Moving Average(ARIMA)model is only suitable for dealing with linear features,while the Recurrent Neural Network predicts PM2.5 just by aligning nonlinear features for fitting.In this paper,the ARIMA model and the Recurrent Neural Network model are combined,and Xi'an City,Shaanxi Province is taken as the research object PM2.5 forecast.The MAE,MSE and RMSE of the prediction results of the ARIMA-GRU model are less than those of the ARIMA-LSTM model,and r and R2 are closer to 1.