一种基于LPP-BP神经网络的季节性PM_(2.5)预测方法

A Prediction Method of PM_(2.5) Seasonality Based on LPP-BP Neural Network

针对PM_(2.5)引起的空气污染问题,采用局部保留投影算法(Local Preserving Projection,LPP),设计一种BP神经网络模型,并基于湖州市2014—2020年的大气污染物数据和气象数据,对PM_(2.5)进行分季节预测.仿真结果表明:LPP-BP模型夏季的均方根误差为5.1978,各季节的均方根误差为10.7595,平均相关性系数R为0.857,平均运行速度为0.2696 s,远低于其他模型.通过与BP-5模型、PCA-BP模型、BP-12模型的对比分析可知,LPP-BP模型具有更高的准确率和更快的运算速度.该研究可为PM_(2.5)预警和空气污染调控提供参考.

For PM_(2.5) for the problem of air pollution caused by air pollution,this paper proposes a BP neural network model with local preserving projection(LPP)to process the input,and evaluates the PM_(2.5) based on the air pollutant data and meteorological data of Huzhou City from 2014 to 2020 forecast by season.The simulation results show that the root mean square error of LPP-BP model in summer is 5.1978,the root mean square error of each season is 10.7595,the average correlation coefficient R is 0.857,and the average running speed is 0.2696 s,which is much lower than other models.Compared with BP-5 model,PCA-BP model and BP-12 model,LPP-BP model has higher accuracy and faster operation speed.This study can be called PM_(2.5) provides a reference for early warning and air pollution control.