LSTM을 이용한 부산의 PM10 중 대기중금속 농도 분석과 예측

Abstract

This study is analyzed the temporal concentration change trend using the atmospheric heavy metals (Pb, Cd, Cr, Cu, Ni, As, and Mn) and PM10 data provided by the Busan Metropolitan City Institute of Health and Environment, for 7-year period (2014-2020) in four different sites(residential area; Gwangan-dong and Deokcheon-dong, commercial area; Yeonsan-dong, and industrial area; Hakjang-dong) of Busan. In addition, using data from 2014 to 2020, the 2021 trend was predicted with Exponential smoothing Methods(ETS) and Long Short-Term Memory(LSTM) and the accuracy was analyzed. From 2014 to 2020, PM10 and heavy metals concentration tended to decrease, but their extent was different. By region, the Gwangan-dong site showed the least decrease and the trend of the highest increase in Cu was confirmed through traffic analysis. Deokcheon-dong site showed a decreasing trend in all items except for Cd and Cr, which increased slightly. The Hakjang-dong site of industrial area showed a decreasing trend in all items, and it was consistent with the decrease in the number of businesses and employees, showing the highest decrease. At the Yeonsan-dong site in the commercial area, Cr, Cu, and As showed an increasing trend and among them, As showed the highest increase. For prediction using ETS and LSTM, heavy metals that can characterize regions were selected from annual average data and determined as Cu in Gwangan-dong, Pb in Deokcheon-dong, As in Yeonsan-dong, and Ni in Hakjang-dong. As the selected heavy metals, ETS(A,A,A), ETS(M,Ad,M), ETS(A,Ad,A), ETS(M,A,M) were determined, respectively, and LSTM showed better performance than ETS.