基于敏感因子的大气重污染潜势分级预报技术研究  

Research on the Grading and Forecasting Technology of Atmospheric Heavy Pollution Potential Based on Sensitive Factors

作　　者：张旭 ZHANG Xu(Bayannur Branch of Inner Mongolia Autonomous Region Environmental Monitoring Station,Bayannur 015000,China)

机构地区：[1]内蒙古自治区环境监测总站巴彦淖尔分站,内蒙古巴彦淖尔015000

出　　处：《中国资源综合利用》2024年第11期213-216,220,共5页China Resources Comprehensive Utilization

摘　　要：为了提高大气重污染潜势预报的精准性,提出基于敏感因子的大气重污染潜势分级预报技术。选取大气通风量、垂直温度梯度和风速垂直切变作为大气重污染敏感因子,采用随机森林(Random Forest,RF)算法分析大气重污染的潜在变量。依据大气重污染的日增量,进行数据预处理,建立最优多元线性回归(Multiple Linear Regression,MLR)模型,筛选出最佳变量组合对大气背景场进行聚类处理,通过最优子集回归法得到重污染潜势预报模型。试验结果表明,定量预报大气重污染潜势时,所提技术的预报正确率为92.35%;预测大气重污染潜势等级时,不同等级下的均方误差(Mean Squared Error,MSE)均低于0.2,该技术具有较高的应用价值。In order to improve the accuracy of atmospheric heavy pollution potential forecasting,a classification forecasting technology for atmospheric heavy pollution potential based on sensitive factors is proposed.Selecting atmospheric ventilation rate,vertical temperature gradient and wind speed vertical shear as sensitive factors for heavy air pollution,the Random Forest(RF)algorithm is used to analyze the potential variables of heavy air pollution.Based on the daily increase of heavy air pollution,data preprocessing is carried out to establish the optimal Multiple Linear Regression(MLR)model,and the optimal variable combination is selected for clustering of the atmospheric background field,and the heavy pollution potential prediction model is obtained through the optimal subset regression method.The experimental results show that the accuracy of the proposed technology for quantitatively predicting the potential of heavy air pollution is 92.35%;when predicting the potential level of heavy air pollution,the Mean Squared Error(MSE)at different levels is less than 0.2,indicating that this technology has high application value.

关 键 词：敏感因子 大气重污染 潜势定量预报 随机森林(RF) 垂直温度梯度 

分 类 号：P456[天文地球—大气科学及气象学]