Study On Forecast Of PM₁₀ Pollution In Daqing City

Inhalable particulates(PM10) has been the primary pollutant in the air of Daqing city for a long time.For better reflecting the pollution change tendency,enhancing pollution control and preventting serious pollution accidents,develop pollution forecasting work in time are of great significance.In this paper,the forecasting of PM10 pollution in Daqing city had been studied according to the city PM10 daily API values that monitored from 1 January 2008 to June 30,2010 and the corresponding daily meteorological data for the data base.Analyed the Daqing PM10 pollutant's variations and trends of pollution by annually, seasonal, monthly with probabilistic statistical method,with emphasis on heating period and non-heating period for comparison and analysis,at the same time to do a separate analysis on the impact of dust storms and the impact of fireworks during the Chinese New Year,and to explore the causes of pollution changes.PM10 pollution is closely-ralated with meteorological factors,through the statistical analysis of recent years meteorological elements data in Daqing city,summed up the seasonal changes of meteorological factors.And analyed the high API values day's variation of meteorological factors in order to better to find out the relationship between PM10 pollution and meteorological factors.Using SPSS software to make a correlation analysis between meteorological factors and PM10 pollution,then choose the significant impact meteorological factors as a basis for prediction.Based on the process of primary factor,the stepwise regression method and principal component analysis method are used on related factors selection.And then we tried to establish two PM10 pollution's prediction model of four seasons in Daqing city,one model is multiple regression prediction model,another prediction model is based on B-P artificial neural network.The models'predicition accuracy rates were examined in the paper and the differences between the two predicition models are also compared in the paper.By means of the simulation and practice,the multiple regression prediction model's precision of spring is 73.9%,summer's is 77.6%,antumn's is 77% and winter's is 79.7%.Four seasons forecast accuracy scores were 76.8,82.3,80.7 and 83.3.The B-P artificial neural network prediction model's precision of spring is 77.9%,summer's is 83.2%,antumn's is 75.9% and winter's is 81.6%. Four seasons forecast accuracy scores were 83.6,86.6,81.5 and 84.6.The results shows that two established PM10 pollution prediction model can meet the needs of the actual forecast.But the B-P neural network prediction model have better predictive ability than traditional statistical method,it also reflects the B-P artifical neural network have obvious advantages in dealing with non-linear problems such as pollution forecast problems.