DOI QR코드

DOI QR Code

A Study on Prediction of PM2.5 Concentration Using DNN

Deep Neural Network를 활용한 초미세먼지 농도 예측에 관한 연구

  • Choi, Inho (Department of Environmental Science and Engineering, Kyung Hee University) ;
  • Lee, Wonyoung (Department of Environmental Science and Engineering, Kyung Hee University) ;
  • Eun, Beomjin (Department of Environmental Science and Engineering, Kyung Hee University) ;
  • Heo, Jeongsook (Department of Environmental Science and Engineering, Kyung Hee University) ;
  • Chang, Kwang-Hyeon (Department of Environmental Science and Engineering, Kyung Hee University) ;
  • Oh, Jongmin (Department of Environmental Science and Engineering, Kyung Hee University)
  • 최인호 (경희대학교 환경학 및 환경공학과) ;
  • 이원영 (경희대학교 환경학 및 환경공학과) ;
  • 은범진 (경희대학교 환경학 및 환경공학과) ;
  • 허정숙 (경희대학교 환경학 및 환경공학과) ;
  • 장광현 (경희대학교 환경학 및 환경공학과) ;
  • 오종민 (경희대학교 환경학 및 환경공학과)
  • Received : 2022.02.11
  • Accepted : 2022.04.20
  • Published : 2022.04.30

Abstract

In this study, DNN-based models were learned using air quality determination data for 2017, 2019, and 2020 provided by the National Measurement Network (Air Korea), and this models evaluated using data from 2016 and 2018. Based on Pearson correlation coefficient 0.2, four items (SO2, CO, NO2, PM10) were initially modeled as independent variables. In order to improve the accuracy of prediction, monthly independent modeling was carried out. The error was calculated by RMSE (Root Mean Square Error) method, and the initial model of RMSE was 5.78, which was about 46% betterthan the national moving average modelresult (10.77). In addition, the performance improvement of the independent monthly model was observed in months other than November compared to the initial model. Therefore, this study confirms that DNN modeling was effective in predicting PM2.5 concentrations based on air pollutants concentrations, and that the learning performance of the model could be improved by selecting additional independent variables.

본 연구는 국가측정망(에어코리아)에서 제공하는 2017년, 2019년 및 2020년도 대기질확정 데이터를 이용하여 Deep Neural Network(DNN) 모델을 학습하고, 2016년과 2018년도 데이터를 이용하여 학습된 모델을 평가·검증하였다. 피어슨 상관계수 0.2를 기준으로 SO2, CO, NO2, PM10 항목을 독립변수로 하여 초기 모델링을 진행하였고, 예측의 정확도를 높이기 위한 방법으로 시계열적 요소를 반영한 월별 모델링(개선모델)을 진행하여 초기모델과 비교·분석하였다. 분석에 사용한 지표는 RMSE(Root mean square error) 방법으로 오차를 계산하였으며, 예측 결과 초기모델의 RMSE값은 5.78로 국가측정망의 예측이동 평균모델의 결과(10.77)와 비교하여 초기모델에서 약 46% 오차가 감소하였다. 또한, 개선모델의 경우, 초기모델 대비 11월 모델을 제외한 모든 월별모델에서 정확도 향상이 있었다. 따라서, 본 연구에서는 DNN 모델링이 PM2.5 농도 예측에 효과적인 방법임을 제안할 수 있었으며, 향후 추가적인 독립변수 선정 및 시계열 요소를 고려한 방법으로 모델의 정확도 개선 가능성을 확인할 수 있었다.

Keywords

References

  1. Cha J, Kim J. 2018. Development of Data Mining Algorithm for Implementation of Fine Dust Numerical Prediction Model. Journal of the Korea Institute of Information and Communication Engineering, 22(4), 595-601. https://doi.org/10.6109/JKIICE.2018.22.4.595
  2. Kim DS. 2016. Characteristics in Atmospheric Chemistry between NO, NO2 and O3 at an Urban Site during MAPS (Megacity Air Pollution Study)-Seoul, Korea, Journal of Korean Society for Atmospheric Environment 32(4): 422-434. http://dx.doi.org/10.5572/KOSAE.2016.32.4.422
  3. Won DJ. 2021. Prediction of Fine Dust in Gyeonggi-do Industrial Complex using Machine Learning Methods, Journal of KIISE 48(7): 764-773. https://doi.org/10.5626/JOK.2021.48.7.764
  4. Im DY. 2021. Temporal Analyses of PM Data, Estimation of the Past Unmonitored PM2.5 Data, and Assessment of the COVID-19 Effect at the Background Areas in Korea. Journal of Korean Society for Atmospheric Environment 37(4): 670-690. https://doi.org/10.5572/KOSAE.2021.37.4.670
  5. Lee HK. 2020. A Study on the Seasonal Correlation between O3 and PM2.5 in Seoul in 2017. Journal of Korean Society for Atmospheric Environment 36(4): 533-542. https://doi.org/10.5572/KOSAE.2020.36.4.533
  6. Shin HW. 2021. Changes of Air Pollutants Concentrations Associated with the Rates of Rainfall and Its Duration during Summertime in Korea. Journal of Korean Society for Atmospheric Environment 2014(10): 175.
  7. Lee HW. 1999. The Effect of Meteorological Factors on Variation and Temporal and Spatial Characteristics of NO2 Concentration in Pusan Area. Journal of Environmental Science International 8(4): 465-471.
  8. Hwang SM, Shin DS, Lee BM, Kim HH, Cho HG, Moon GJ, Jeong SH. 2007. A Study on the Distribution Characteristics of Air Pollutants at PAMS in Seoul Metropolitan Area, Proceeding of 43th meeting of Korean Society for Atmospheric Environment in 2007, pp. 1396-1399.
  9. Lim IH. 2013. A Study on the hourly Characteristics of Air Pollution in the Gwangyang Bay. 環境硏究論文集 13: 29-39.
  10. Jeon BI. 2010. Characteristics of Spacio-Temporal Variation for PM10 and PM2.5 Concentration in Busan, Journal of Environmental Science International 19(8): 1013-1023. https://doi.org/10.5322/JES.2010.19.8.1013
  11. Kim JS, Bais AL, Kang SH, Lee J, Park K. 2011. A semi-continuous measurement of gaseous ammonia and particulate ammonium concentrations in PM2.5 in the ambient atmosphere. J. Atmos. Chem. 68(3): 251-263. doi:10.1007/s10874-012-9220-y
  12. Lee BK, Lee DS, Kim MG. 2001. Rapid time variations in chemical composition of precipitation in South Korea. Water, Air, and Soil Pollution 130(1-4): 427-432. doi:10.1023/A:1013845620642.
  13. Sung SH, Kim SJ, Ryu MH. 2020. A Comparative Study on the Performance of Machine Learning Models for the Prediction of Fine Dust: Focusing on Domestic and Overseas Factors. Korea Innovation Studies 15(4): 339-357. https://doi.org/10.46251/INNOS.2020.11.15.4.339
  14. Son SH, Kim JS. 2020. Evaluation and Predicting PM10 Concentration Using Multiple Linear Regression and Machine Learning. Korean Journal of Remote Sensing 36(6-3): 1711-1720. https://doi.org/10.7780/KJRS.2020.36.6.3.7
  15. Yoon SY. 2014. Effect of Precipitation Cleaning on Air Pollutants in Summer and Winter. Korean Meteorological Society, Proceedings of the Autumn Meeting of KMS, 2014: 146-147.
  16. Ghim YS. 2013. Regional Trends in Short-Term High Concentrations of Criteria Pollutants from National Air Monitoring Stations. Journal of Korean Society for Atmospheric Environment 29(5): 545-552. https://doi.org/10.5572/KOSAE.2013.29.5.545