The Research On Urban Air Quality Assessment And Prediction Model Based On Artificial Neural Network

This work focused on Beijing’s four-year air quality data from 2015 to 2018.The main indicators include the concentration of six air pollutants and real-time air pollutant index data.At first step,we analyzed the air quality of Beijing from 2015 to 2018 on the respect of year and month,we combine the comprehensive pollutant index method with the air quality index(AQI)method in this step.The next step,we selected the main index combine the method of the comprehensive pollutant index and the Pearson Correlation Coefficient method,and then combined artificial neural network with the selected air quality index,finally the relationship between historical air pollutant data and future data can be found.We used the artificial neural network,because of it’s strong data processing ability and highly nonlinear mapping ability.We built a prediction model which is non-linearity.And the error function was used for comparative analysis.It can verify the effectiveness of the selected indicators.Analyzed the output results of the model to determine the existing problems and solve the corresponding problems.The main work of this paper was as follows:(1)Studied the present situation of Air Quality Assessment and prediction at home and abroad,and reviewed the traditional air quality assessment methods and then use the Long Short-Term Memory(LSTM)neural network in air quality prediction.(2)Studied the related concepts and calculation methods of integrated pollutant index method,AQI index method and Pearson product-moment Correlation Coefficient method,and reviewed the basic principles of optimization algorithms such as LSTM neural network and Adam.(3)Collected and standardize air quality data in Beijing from January 2015 to December 2018,that include air pollutant concentrations and the corresponding Aqi Index.(4)The air quality of Beijing from 2015 to 2018 was evaluated by comprehensive pollutant index method and Aqi index method,and then the pollutant comprehensive index method and the Pearson correlation coefficient method to evaluate the air pollution in Beijing Screening of physical indicators.(5)Through comparative analysis,the LSTM neural network based on Adam optimization algorithm was selected as the research carrier,and the air quality prediction model based on the selected model was established to verify the effectiveness of screening air pollution indicators.The prediction effect of the corresponding model was analyzed,and the problems such as over fitting and prediction time in the model prediction were analyzed and verified.The final results showed that the air quality in Beijing has improved from2015 to 2018,the effect of the state regulation and control policy was significant.The combination of Pearson product-moment Correlation Coefficient and Aqi was more effective than unfiltered index system.In addition,the model adopted in the experiment was LSTM neural network which is based on Adam optimization algorithm,it was used for prediction research,and the problems in it were analyzed.