Abstract
This paper proposes a deep learning model that integrates a convolutional neural network with a gate circulation unit that captures patterns of high-peak PM2.5 concentrations. The purpose is to accurately predict high-peak PM2.5 concentration data that cannot be trained in general deep learning models. For the training of the proposed model, we used all available weather and air quality data for three years from 2015 to 2017 from 25 stations of the National Institute of Environmental Research (NIER) and the Korea Meteorological Administration (KMA) observatory in Seoul, South Korea. Our model trained three years of data and predicted high-peak PM2.5 concentrations for the year 2018. In addition, we propose a Gaussian filter algorithm as a preprocessing method for capturing high concentrations of PM2.5 in the Seoul area and predicting them more accurately. This model overcomes the limitations of conventional deep learning approaches that are unable to predict high peak PM2.5 concentrations. Comparing model measurements at each of the 25 monitoring sites in 2018, we found that the deep learning model with a Gaussian filter achieved an index of agreement of 0.73–0.89 and a proportion of correctness of 0.89–0.96, and compared to the conventional deep learning method (average POC = 0.85), the Gaussian filter algorithm (average POC = 0.94) improved the accuracy of high-concentration PM2.5 prediction by an average of about 9%. Applying this algorithm in the preprocessing stage could be updated to predict the risk of high PM2.5 concentrations in real time.
Highlights
The World Health Organization (WHO) guidelines for ensuring sufficient air quality recommend that concentrations of particulate matter with diameters of 2.5 μm and smaller (PM2.5 ) and PM10 should not exceed 24-h average concentration thresholds of 15 μg/m3 and 45 μg/m3, respectively [1]
To improve the accuracy of PM2.5 prediction, we used a Gaussian filter that removes noise from the original image in the image-processing research field [26]; we propose a method of improving the prediction accuracy of daily maximum concentrations of PM2.5, which is currently a limitation of the already developed deep learning prediction system [7,18]
If the deep learning architecture was configured in this way, the gated recurrent unit (GRU)
Summary
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. To address concerns about air quality and the risks posed by high emissions, air pollution researchers have begun to apply various machine learning and deep learning algorithms to various fields of atmospheric sciences to more accurately predict PM2.5 concentrations [7,15,16,17]; and several studies have improved numerical models [18]. As the GRU has fewer parameters than LSTM, its learning rate is faster than that of LSTM [24,25] In fields such as air pollution concentration prediction, which uses time sequences, it is suitable to apply the classification or prediction of time series data using GRU networks among RNN architectures [7]. We propose a method of capturing maximum daily concentrations of PM2.5 using a Gaussian filter, which will improve the prediction accuracy of the captured maximum daily concentrations of PM2.5
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
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
