This paper utilizes meteorological data and manually reviewed environmental air monitoring data from April to November 2020 to 2022 in Dalian urban area,where the ozone concentration is relatively high during these months,a LightGBM machine learning model was established and tested on a 10％split of the training dataset before being optimized for predicting ozone concentrations in Dalian urban area.The model can automatically rank the importance of factors,with a higher ranking for fine particulate matter concentration(PM2.5)and atmospheric pressure.PM2.5 concentration represents local pollutant emissions and background pollution,while atmospheric pressure represents seasonal changes and atmospheric diffusion conditions,which are consistent with theoretical and empirical knowledge.The correlation coefficient(R)between the model predictions and measured values reached 0.833,with a mean absolute error(MAE)of 13.068,mean absolute percentage error(MAPE)of 16.590％,and root mean square error(RMSE)of 16.424.The model performs well within the ozone concentration range of 60-120 μ g/m3,with a median relative error of 1.3％and a relative error distribution concentrated between-11.3％and 12.3％.As a young machine learning framework,LightGBM has good application prospects in forecasting atmospheric environmental pollutant concentrations.
维普
万方数据