太原市COVID-19防控前后空气质量分析及预测

为了探索特殊情况下太原市空气质量预测和评价方法,采用基于灰色关联度法的模糊综合评价方法对太原市疫情防控前后的空气质量进行评价,对相关联的污染物浓度变化进行分析,并以太原市AQI监测数据为基础,结合长短期记忆循环神经网络(LSTM)以及随机梯度下降算法(Adam)建立了太原市空气质量预测模型(即Adam-LSTM模型),...

Air quality analysis and prediction before and after the prevention and control of COVID-19 in Taiyuan

In order to explore the prediction and evaluation methods of air quality under special circumstances in Taiyuan, the air quality trend before and after the prevention and control of COVID-19 in Taiyuan was evaluated and the changes of related pollutant concentrations were analyzed by using the fuzzy comprehensive evaluation method based on grey correlation method. On the basis of the AQI monitoring data of Taiyuan, Long Short-Term Memory (LSTM) loop neural network and the stochastic gradient descent algorithm(Adam), the Taiyuan air quality prediction model (Adam-LSTM model) was established, and the prediction results were compared with that of LSTM model. The results show that the air quality of Taiyuan is improved after the start of the first level emergency response, and some pollutants are not reduced due to meteorological and Spring Festival factors.The root mean square error and training speed of LSTM and Adam-LSTM model are 0.203 s and 12.15 s, 0.183 s and 10.35 s, respectively. It shows that the proposed Adam optimization algorithm can effectively improve the training accuracy and convergence speed of LSTM model. With relatively small prediction errors, it can provide data support and prediction methods for environmental protection departments to make related air quality decisions.