AIR POLLUTANTS CONCENTRATION PREDICTION BASED ON TRANSFER LEARNING AND RECURRENT NEURAL NETWORK

AIR POLLUTANTS CONCENTRATION PREDICTION BASED ON TRANSFER LEARNING AND RECURRENT NEURAL NETWORK

Fong Iat Hang (Department of Computer and Information Science Faculty of Science and Technology University of Macau, Macau SAR, CHINA) and Simon Fong (Department of Computer and Information Science Faculty of Science and Technology University of Macau, Macau SAR, CHINA)
DOI: 10.4018/IJEACH.2020010106

Abstract

Air pollution poses a great threat to human health, and people are paying more and more attention to the prediction of air pollution. Prediction of air pollution helps people plan for their outdoor activities and helps protect human health. In this article, long-short term memory recurrent neural networks were used to predict the future concentration of air pollutants in Macau. In addition, meteorological data and data on the concentration of air pollutants were used. Moreover, in Macau, some air quality monitoring stations have less observed data, and some AQMSs less observed data of certain types of air pollutants. Therefore, the transfer learning and pre-trained neural networks were used to assist AQMSs with less observed data to generate neural network with high prediction accuracy. In this thesis, in most cases, LSTM RNNs initialized with transfer learning methods have higher prediction accuracy, used less training time than randomly initialized recurrent neural networks.
Article Preview
Top

Introduction

With the development of societies and industries, many countries and cities in the world have to face the problem of air pollutions, and air pollution has been bringing many undesirable effects on human health. Therefore, predict the level of air pollution in the cities and then publish the severity of air pollution to the public is important. Air pollutants mainly come from burning fossil fuels. They are mainly included: sulfur dioxide (SO2), nitrogen monoxide (NO), nitrogen dioxide (NO2), carbon monoxide (CO), inhalable particles with diameters that are generally 10 micrometers and smaller (PM10), fine inhalable particles with diameters that are generally 2.5 micrometers and smaller (PM2.5) (USA ERA, 2019). As all we know, air pollution adversely affects people's health, especially for children and the elderly; it will also make patients with respiratory diseases (such as asthma and bronchitis) or cardiovascular disease worse. In addition, prolonged exposure to traffic-related air pollution may shorten life expectancy. Moreover, people who have long-term exposure to vehicle-related air pollution may shorten their life expectancy (Hoek et al., 2019). Xi Chen et al. (2016) studied the relationship between NO2, SO2, and PM10 concentrations and lung cancer mortality in several northern cities in China, as well as the relationship between these air pollutants and patients with lung cancer. The statistical data they have researched show that the concentration of air pollutants in people's area is positively correlated with the prevalence and mortality of lung cancer.

Air pollution and the state of the atmosphere have a great relationship. For example, when the atmosphere is stable, that is to say, when the air in a certain area is not rising, the air pollutants will stay on the surface, which is unfavorable to the spread of air pollutants; However, if the atmosphere is unstable, the air will move upward vertically, which will help the air pollutants spread to the sky. The state of the atmosphere usually measures seven different elements, namely wind speed, wind direction, atmospheric temperature, relative humidity, dew point temperature, atmospheric pressure and precipitation. People usually use automatic weather stations (AWSs) (or call them meteorological monitoring stations) to automatically and periodically measure the above-mentioned seven atmospheric elements. In addition, air quality monitoring stations (AQMSs) are used to measure the concentration of air pollutants (such as PM2.5, SO2, NO, etc.) in a certain area automatically and periodically. Things measured by AWSs and AQMSs are listed in Table 1.

Table 1.
Features observed by AWS and AQMS
Name Of FeaturesUnits
air quality monitoring station
inhalable particles (PM10)μg/m3
fine inhalable particles (PM2.5)μg/m3
nitrogen monoxide (NO)ppb
Nitrogen dioxide (NO2)ppb
carbon monoxide (CO)ppm
automatic weather station
wind speedkm/h
wind runkm
station atmospheric pressurehPa
air temperature︒C
relative humidity%
precipitationmm
dew point︒C

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 2: 2 Issues (2020): 1 Released, 1 Forthcoming
Volume 1: 2 Issues (2019)
View Complete Journal Contents Listing