The System Of Monitoring And Forecasting For Small-scale Air Quality Based On STM32

With the continuing growth of the social economy and the rapid expansion of the city population,air pollution in city is becoming a serious problem,air monitoring and its trend has attracted more and more attention.The analysis of real-time monitoring and trend of air quality have an important meaningful thing for the health of people.First of all, the article design a monitoring system for the air quality including temperature,humidity, UV index, PM2.5, PM10 and SO2 on the base of main processor of STM32. The system can not only display the current air quality parameters on the LCD, but also on the phone screen through connecting the Bluetooth module. In the mind of hardware the system mainly refer to the module circuit of DHT11, UV sensor, PM2.5, PM10, SO2 sensor and LCD display. After finishing designing the hardware circuit, in the keil integrated development environment we use the modular programming method and C language to write the programs for every module according to theirs different driving principle.Compared with the general narrow monitoring system for air quality, the monitoring system in this article is added the parameter. of AQI specially, and the parameter can make us realize the current air quality situation intuitively. In order to solve the AQI, we use SPSS software to analyze the relevance between AQI and its six among factors, and that find out the most significant correlation to build a multivariate regression model for solve the AQI. After that, we put the model into STM32 to calculate AQI by measuring several variables. On the basis of the regression model and combined with the meteorological and atmospheric environmental data from December 2014 to March 2015, we build a prediction model from the perspective of the BP neural network and RBF neural network. Then we test that which model is more accurate through the same samples. After comparing results from the two model, we find that the RBF neural network is more accurate than the BP neural network.