Abstract

BACKGROUND AND AIM: In recent years, particulate matter (PM) has become a serious air pollutant in urban areas of Taiwan and has drawn more and more attention. This study selected possible factors affecting the concentration of PM2.5 (PM with aerodynamic particle size of less than 2.5 μm) and aimed at establishing a model using machine learning for predicting air pollution. METHODS: Long-short term memory (LSTM) is a special recurrent neural network (RNN) model proposed to solve the problem of gradient dispersion of RNN model. In this study, we applied the LSTM network to identifying air pollutants (SO2, NO2) and meteorological factor (wind speed) that affect the concentration of PM. We used the information provided by the Air Quality Monitoring Network of the Environmental Protection Administration and obtained data on three air pollutants (PM2.5, SO2, NO2) from the eight monitoring stations in Kaohsiung City from 2017 to 2018. The LSTM network was used to classify, process, and train time series. The prediction model of PM2.5 was based on supervised learning that used the air pollution on the previous 3 hours to predict the level of PM2.5 every other hour. RESULTS:The average concentration of PM2.5 in Kaohsiung City from January 1, 2017 to December 31, 2018 was 25.329 μg/m3. PM2.5 had a correlation coefficient of 0.261 with SO2, and that with NO2 was 0.648 and -0.085 respectively. By integrating the above four air pollutants, an LSTM network analysis was used to establish a prediction model of PM2.5. By comparing the trained values with the test values, we obtained a root mean square error value of the model as 2.759. CONCLUSIONS:This study showed that the LSTM network approach could be applied to predict daily pollution levels. This may improve the accuracy of air quality forecasting and the warning of specific pollution elements. KEYWORDS: Long-short term memory (LSTM), deep learning, particulate matter (PM)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.