Air-pollution prediction in smart city, deep learning approach

Abstract

Over the past few decades, due to human activities, industrialization, and urbanization, air pollution has become a life-threatening factor in many countries around the world. Among air pollutants, Particulate Matter with a diameter of less than $2.5\mu m$ ( $P{M}_{2.5}$ ) is a serious health problem. It causes various illnesses such as respiratory tract and cardiovascular diseases. Hence, it is necessary to accurately predict the $P{M}_{2.5}$ concentrations in order to prevent the citizens from the dangerous impact of air pollution beforehand. The variation of $P{M}_{2.5}$ depends on a variety of factors, such as meteorology and the concentration of other pollutants in urban areas. In this paper, we implemented a deep learning solution to predict the hourly forecast of $P{M}_{2.5}$ concentration in Beijing, China, based on CNN-LSTM, with a spatial-temporal feature by combining historical data of pollutants, meteorological data, and $P{M}_{2.5}$ concentration in the adjacent stations. We examined the difference in performances among Deep learning algorithms such as LSTM, Bi-LSTM, GRU, Bi-GRU, CNN, and a hybrid CNN-LSTM model. Experimental results indicate that our method "hybrid CNN-LSTM multivariate" enables more accurate predictions than all the listed traditional models and performs better in predictive performance.

Keywords: Air-pollution; CNN-LSTM; Deep learning; Forecasting; GRU; LSTM; PM2.5.

Fig. 1

Architecture of the LSTM cell

Fig. 2

Architecture of the GRU cell

Fig. 3

Architecture of the Bi-LSTM model

Fig. 4

Architecture of the CNN model

Fig. 5

Architecture of the CNN-LSTM Model

Fig. 6

The distribution of monitoring stations in Beijing

Fig. 7

Wind Direction and Degree Values

Fig. 8

The correlation matrix of the air quality features

Fig. 9

Graphical representation of Meteorological data and $P{M}_{2.5}$

Fig. 10

The Spatiotemporal Correlation Analysis

Fig. 11

Workflow for predicting $P{M}_{2.5}$ concentrations

Fig. 12

The architecture of the proposed CNN-LSTM

Fig. 13

Comparison of the MAE for the 1day and 7day lag for the different deep learning models

Fig. 14

Comparison of the RMSE for the 1day and 7day lag for the different deep learning models

Fig. 15

Comparison of the $R^2$ for the 1day and 7day lag for the different deep learning models

Fig. 16

PM2.5 concentration forecasting results for 10 days

Fig. 17

Comparison of the RMSE and MAE for the proposed model and other model