Taking Beijing as an example and using the data of air quality monitoring stations from 2015 to 2018, the impacts of temporal resolution and spatial information on the PM_2.5 concentration prediction were analyzed by a BP neural network, an LSTM network, and a CNN-LSTM hybrid model. The results show that neural network models are generally better than the multi-linear regression model. Increasing the temporal resolution of the input data can significantly improve the accuracy of the predicted daily average PM_2.5 concentration. When the temporal resolution of the input data increases from one day to 6 hours, the mean absolute error of the LSTM model reduces from 27.39 μg/m³ to 20.59 μg/m³. This improvement is more obvious when the weather is significantly getting better or getting worse. The distribution of PM_2.5 concentration in North China has distinct spatial and temporal characteristics. The first spatial mode is a uniformly increasing or decreasing mode, and the second one is a north/south dipole mode. The analysis shows that the concentration of PM_2.5 in Beijing is related to the PM_2.5 in Inner Mongolia, Hebei, and Tianjin of the previous day. The CNN-LSTM hybrid model, trained with the spatialtemporal information of PM_2.5 in North China, can further improve the predictability of PM_2.5 in Beijing. It further reduces the mean absolute error to 17.36 μg/m³.

以北京市为例, 利用2015—2018年空气质量监测站台资料, 通过BP神经网络、LSTM网络及CNNLSTM混合模型等多种模型, 分析时间精度和空间信息对PM_2.5浓度预报的影响。结果表明, 神经网络模型的效果普遍比多元线性回归模型好; 增加输入数据的时间精度能显著地提高 PM_2.5浓度日均值预报的准确率; 当输入数据的时间精度从一天提高到6小时后, LSTM模型的平均绝对误差从27.39 μg/m³降至20.59 μg/m³, 这种效果的提升在显著变好和显著变差的天气情况下更明显; 华北地区PM_2.5浓度分布有明显的时空特征, 第一空间模态为同增同减, 第二空间模态为南北反向; 北京市PM_2.5浓度与内蒙古、河北及天津等地区前一天的PM_2.5相关。利用CNN-LSTM混合模型学习华北地区PM_2.5的时空信息, 能进一步提高北京市PM_2.5浓度的预报水平, 使得误差降低至17.36 μg/m³。