Atmosphere (Nov 2021)

A Haze Prediction Model in Chengdu Based on LSTM

  • Xinyi Wu,
  • Zhixin Liu,
  • Lirong Yin,
  • Wenfeng Zheng,
  • Lihong Song,
  • Jiawei Tian,
  • Bo Yang,
  • Shan Liu

DOI
https://doi.org/10.3390/atmos12111479
Journal volume & issue
Vol. 12, no. 11
p. 1479

Abstract

Read online

Air pollution with fluidity can influence a large area for a long time and can be harmful to the ecological environment and human health. Haze, one form of air pollution, has been a critical problem since the industrial revolution. Though the actual cause of haze could be various and complicated, in this paper, we have found out that many gases’ distributions and wind power or temperature are related to PM2.5/10’s concentration. Thus, based on the correlation between PM2.5/PM10 and other gaseous pollutants and the timing continuity of PM2.5/PM10, we propose a multilayer long short-term memory haze prediction model. This model utilizes the concentration of O3, CO, NO2, SO2, and PM2.5/PM10 in the last 24 h as inputs to predict PM2.5/PM10 concentrations in the future. Besides pre-processing the data, the primary approach to boost the prediction performance is adding layers above a single-layer long short-term memory model. Moreover, it is proved that by doing so, we could let the network make predictions more accurately and efficiently. Furthermore, by comparison, in general, we have obtained a more accurate prediction.

Keywords