A Study On Predicting Air Quality In Qinhuangdao City

In recent years, air pollution has attracted people’s attention which comes from the whole country widely. To improve the quality of people’s life, strengthen prevention and control of air pollution and prevent the happening of heavy air pollution days, it is significant and far-reaching to carry out air quality forecasting in time.In this paper, the relationship between API or AQI and meteorological factors are analyzed by using data of API or AQI from 2008 to 2013 and meteorological data of the same period. The relationship between API or AQI and meteorological factors is analyzed in this paper. And this paper studies on prediction methods of Qinhuangdao API.Firstly, using correlating analysis method analyses meteorological factors which have an effect on Qinhuangdao API or AQI in spring, summer, autumn, winter. The difference of meteorological conditions between AQI more than 200 and AQI less than 200 is analyzed.Secondly, based on the primary factors, this paper builds a multiple linear regression prediction model to predict API values in four seasons of Qinhuangdao by using multiple linear regression method. And four regression models were tested accuracy that the accuracy rates are 84.6%,86.8%,83.4%,83.5%.Finally, it constructs the BP neural network prediction model and BP neural network prediction model of genetic algorithm optimization which are based on principal component analysis to predict API value in the seasons. It runs 5 times for each model respectively, thus it can get the forecast average accuracy rates of the two kinds of models:85.18%,87.1%,85.72%,84.56%; 86.32%,87.66%,87.4%, 85.88%.Therefor, The three prediction model proposed in this paper can satisfy the need of practical prediction. But which can be seen by comparing the accuracy of neural network predictions is more higher than linear regression method. Stability of BP neural network method of genetic algorithm optimization is better than BP neural network. So, BP neural network method of genetic algorithm optimization is the most suitable for predicting air quality in Qinhuangdao.