Research On Air Quality Prediction Based On Zigbee Wireless Sensor Network And BP Neural Network Algorithm

Air is a precious asset for human survival and development and is one of the basic elements of all life activities.The State Council has asked all departments in the country to establish a sound air quality monitoring and early warning system,to improve the accuracy of monitoring and early warning,and to issue monitoring and early warning information to the public in a timely manner.With the establishment of air quality monitoring system,the data of air quality is gradually enriched,making air quality prediction possible.In order to reflect the trend of environmental pollution better and prevent serious pollution incidents.Based on the street-lighting system of Beichen Economic and Technological Development Zone,the air quality monitoring network is constructed by Zigbee networking technology,and the BP neural network model is trained by the air quality data collected by the monitoring network.Then the trained model is transplanted to Jetson TX2 embedded Ai computing equipment to realize real time prediction and monitoring.Accordingly,the following studies are carried out:（1）Based on the existing street-lighting system network,the Zigbee node and the environmental monitoring micro-base station are installed on the street lamps of nine intersections in the park to build the Zigbee wireless sensor network,which makes the street lamps become the absolute nodes of urban air quality data collection and transmission,and realizes the real-time environmental data collection.From April 15 to September 30,2018,a total of 78418 sets of data were collected and used for training and testing the BP neural network prediction model.（2）The forward and backward propagation process of BP neural network is deduced by mathematical principle,and an improved method is put forward for the problems of BP neural network.According to the evaluation criterion and calculation method of AQI（environmental air quality index）,the influence factors of air quality are analyzed,The AQI prediction model was established by using the four meteorological data of temperature,humidity,wind direction and wind speed and the concentration values of six pollutants PM₁0、PM₂.5、SO₂、NO₂、O₃ and CO as the input parameters of BP neural network.Then the missing data is pretreated by the methods of case elimination,mean replacement,hot-card filling and k-nearest neighboring;Finally,the empirical formula and trial-and-error method are used to determine the number of hidden layers,the number of neurons in the hidden layer and the learning rate.Through analysis and comparison,the activation function is finally determined to adopt Tanh function to complete the BP neural network model design.（3）Model training and model prediction platform are constructed by GPU tower server with dual GTX1080ti graphics card and Jetson TX2,respectively,then perform model training and model prediction.The trained prediction model is transplanted to the deep learning platform of Jetson TX2 for real-time prediction,which improves the training speed and real-time prediction of air quality.（4）The collected data are divided into training set and test set,and the designed BP neural network model is trained with the training set,and the performance of the trained model is verified with the test set,and the performance of the trained model is verified by the test set,and then the average absolute error and accuracy of the predicted result are calculated.Finally,the prediction results show that the prediction accuracy of AQI is over 93.5%.Therefore,it can be proved that the air quality prediction model based on BP neural network is effective and feasible.