统计与数据科学学院知识库

随着经济的快速持续发展，大量的能源需求和过度开发，造成植被破坏，废气排放量增多等不良影响。这直接影响空气质量的优良程度，从而影响人民的身体健康。而兰州市作为群山环绕的工业城市，大气污染物消散困难，造成空气质量下降。本文利用兰州市2014年至2020年空气污染物浓度数据，探究兰州市空气质量的变化规律，分析兰州市的空气质量的特性为后续的空气质量指数预测提供条件。在对几种模型效果对比分析的基础上，选择了组合模型对AQI空气质量指数进行预测，该模型有效的提高了预测精度。主要结果如下：首先，将兰州市2014年到2020年的空气质量数据进行年度、季度、月度的划分，分别探究兰州市空气质量的变化规律。发现兰州市空气质量呈周期性波动，季度特征较为明显，夏季空气污染程度低，冬季污染物程度高。其次，选取各个单一模型进行对比，将对比后精度较高的单一模型投入组合模型的建立中。最后运用LSTM、EEMD-LSTM、EEMD-LSTM-ARIMA模型对AQI 指数进行预测，对比各个模型的预测结果，最终得出EEMD-LSTM-ARIMA模型对AQI指数的预测最为精准，同时，选取不同的数据集验证模型的普适度，最终证明组合模型能够为空气质量预测提供相应的依据。

With the rapid and continuous economic development, the large amount of energy demand and over-exploitation has caused adverse effects such as the destruction of vegetation and increased emissions of exhaust gases. This directly affects the excellent degree of air quality and thus the health of the people. And as an industrial city surrounded by mountains, Lanzhou City has difficulties in dissipating atmospheric pollutants, which causes a decline in air quality. In this paper, we use the data of air pollutant concentration in Lanzhou City from 2014 to 2020 to explore the change pattern of air quality in Lanzhou City and analyze the characteristics of air quality in Lanzhou City to provide conditions for the subsequent air quality index prediction. Based on the comparative analysis of the effects of several models, a combined model is selected for AQI air quality index prediction, which effectively improves the prediction accuracy. The main results are as follows: firstly, the air quality data of Lanzhou City from 2014 to 2020 were divided into annual, quarterly and monthly, and the change pattern of air quality in Lanzhou City was explored separately. It was found that the air quality of Lanzhou City showed cyclical fluctuations with more obvious quarterly characteristics, low air pollution level in summer and high pollutant level in winter. Secondly, each single model was selected for comparison, and the single model with higher accuracy after comparison was put into the establishment of the combined model. Finally, the LSTM, EEMD-LSTM, and EEMD-LSTM-ARIMA models are used to predict the AQI index, and the prediction results of each model are compared, and finally the EEMD-LSTM-ARIMA model has the most accurate prediction of the AQI index. The combined model was finally proved to provide the corresponding basis for air quality prediction.