基于PSO算法优化BP神经网的PM2.5浓度预测模型

打开文本图片集

关键词： 浓度；预测模型；PSO算法；BP神经网络

中图分类号：TP391.4；TP183 文献标识码：A 文章编号：2096-4706（2025）07-0047-06

Abstract：Aimingat the problem thatthe traditional BPNeural Network has slowconvergence speedand iseasyto fall into local optimal solution， this paper proposes a concentration prediction model based on Particle Swarm Optimization （PSO）algorithmoptimizedBPNeuralNetwork，whichcanquicklyconvergeandgettheglobal optimalsolution.Firstly，the pollutant indexes with high correlation with concentration are selected as input variables by Pearson correlation analysis. Secondly，thePSOalgorithmisusedtooptimizetheinitialweightsandthresholdsofBPNeuralNetwork，whichovercomesthe shortcomingsofBPNeuralNetwork，suchaseasytofallintolocaloptimumandslowconvergencespeed.Finally，themodel is trained and tested using air pollutant data from July 2O21 to June 2024 in Chengdu.Theresults show that the of the test setis 0.944，theMAEofthetestsetis4.231，andtheRMSEofthetestsetis6.364.ComparedwiththeunoptimizedBPNeural Network model，thePO-BPmodelhashgherpredictionaccuracyandfasterconvergencespeed，andcaneffctivelyprdictthe concentration of the next day in Chengdu.

Keywords： concentration，prediction model，PsO algorithm，BPNeural Network

0 引言

，即细颗粒物，其尺寸微小且具有强烈的吸附能力，它们在大气中停留时间长，能远距离传输，并可通过呼吸深入人体肺部和血液[1]。（剩余6936字）