结合大气污染特征的VOCs聚集区识别方法
A method for identifying VOCs aggregation areas based on the characteristics of air pollution
陆秋琴 1田园 1黄光球1
作者信息
- 1. 西安建筑科技大学 管理学院,西安 710055
- 折叠
摘要
为实现VOCs聚集区的精准感知识别,提出一种结合大气污染特征的VOCs聚集区识别方法.首先,划分区域网格,利用IDW空间插值方法得到VOCs网格数据集;其次,利用HYSPLIT计算后向气团运动轨迹并引入VGG提取轨迹特征,同数据集输入TCN-BiLSTM模型,预测各网格VOCs浓度;最后,根据预测结果进行聚集区识别.以西安市碑林区为例,对VOCs浓度值进行预测,并将聚集区识别结果可视化.结果表明:该组合预测模型能够有效提高识别精度,VOCs浓度预测结果的MAE、MSE、RMSE、R2分别为6.657、103.657、10.181、0.976,预测效果优于对比模型.消融实验证明考虑气团污染特征能提高VOCs预测准确性,实现VOCs聚集区的精准感知识别.
Abstract
To achieve precise perception and recognition of VOCs aggregation areas, this paper proposes a VOCs aggregation area recognition method combining air pollution characteristics. First, the regional grid is partitioned and the IDW spatial interpolation method is employed to obtain the VOCs grid dataset. Second, HYSPLIT is used to calculate the trajectory of the backward air mass and VGG is introduced to extract trajectory features. The same dataset is input into the TCN-BiLSTM model to predict the VOCs concentration in each grid. Finally, the clustering area is identified based on the predicted results. In Beilin District in Xi'an, the concentration values of VOCs is predicted and the identification results of aggregation areas are visualized. Our results show the combined prediction model effectively improves the recognition accuracy. The MAE, MSE, RMSE, and R2 of the VOCs concentration prediction results are 6.657, 103.657, 10.181, and 0.976 respectively, which are superior to those of the comparison model. Through ablation experiments, it proves a consideration of the characteristics of air mass pollution effectively improves the accuracy of VOCs prediction and achieves accurate perception and recognition of VOCs aggregation areas.
关键词
VOCs聚集/污染区域识别/浓度预测/大气污染特征/深度学习Key words
aggregation of VOCs/identification of contaminated areas/concentration prediction/characteristics of atmospheric pollution/deep learning引用本文复制引用
基金项目
陕西省哲学社会科学研究专项(2022HZ1555)
出版年
2024
国家科技期刊平台
NSTL
万方数据