College of Electronics and Information Engineering, Tongji University
Institute of New Rural Development, Tongji University
College of Environmental Science and Engineering, Tongji University

建立空气质量预报模型,预测污染物浓度对人类健康和社会经济发展具有重要意义。 然而,传统的空气质量模型 CMAQ 对污染物浓度的预报精度并不理想。 对此,本文提出了一种基于卷积神经网络(CNN)和长短期记忆神经网络(LSTM) 的空气质量预报修正模型,并使用哈里斯鹰算法(HHO)对模型的超参数进行优化;用 CMAQ 模型对上海市 2022 年 12 月六种大气污染物(SO2、NO2、PM10、PM2.5、O3、CO)浓度的预报数据以及监测站的气象数据和污染物浓度实测数据作为 HHO -CNN -LSTM 模型的输入,对 CMAQ 模型预报结果进行修正。 使用均方根误差(RMSE)、平均绝对误差(MAE)和一致性指数(IOA)作为评价指标。 结果显示,修正模型显著提高了六项污染物浓度的预测精度,RMSE 减少了 73.11% ~91.31%,MAE 减少了 67.19% ~89.25%,IOA 提升了 35.34% ~108.29%。 同时针对 HHO 算法陷入局部最优而导致修正模型对 CO 浓度预测效果不佳的问题,使用高斯随机游走策略对 HHO 算法进行改进,显著提高了 CO 浓度的预测精度。 相比于改进之前,RMSE 减少了 39.55%,MAE 减少了 45.93%,IOA 提高了 32.43%。

With rising levels of air-pollution, air-quality forecasting has become integral to the dissemi⁃nation of human health advisories and the preparation of mitigation strategies. Traditional air quality mod⁃els, such as the Community Multi-scale Air Quality (CMAQ) model, have unsatisfactory accuracy. Ac⁃cordingly, a correction model, which combines convolutional neural network (CNN) and long-shortterm memory neural network (LSTM) and optimized by harris hawks optimization algorithm (HHO) wasestablished to enhance the accuracy of CMAQ model's prediction results for six air pollutants (SO2,NO2, PM10, PM2.5, O3 and CO). The accuracy of HHO-CNN-LSTM was evaluated using root meansquare error (RMSE), mean absolute error (MAE), and the index of agreement (IOA). The resultsdemonstrated a significant improvement in the accuracy of prediction for the six pollutants using the cor⁃rection model. RMSE decreased by 73.11% to 91.31%, MAE decreased by 67.19% to 89.25%, and IOAincreased by 35.34% to 108.29%. To address the propensity of the HHO algorithm to converge on localoptima, leading to poor CO correction performance, this study proposed a method for the HHO algorithmwith a Gaussian random walk strategy to improve the CO concentration correction performance.

国家重点研发计划政府间国际合作资助项目(2022YFE0120600);国家自然科学基金面上资助项目(21978224)

郑鑫楠, 林开颜, 王孜竞, 等. 基于 HHO-CNN-LSTM 的 CMAQ 修正模型及其在上海市空气质量预报中的应用[J]. 能源环境保护, 2023, 37(6): 101-110.

ZHENG Xinnan, LIN Kaiyan, WANG Zijing, et al. Application of HHO-CNN-LSTM-based CMAQ correctionmodel in air quality forecasting in Shanghai[J]. Energy Environmental Protection, 2023, 37(6): 101-110.