Analysis of Correlation Between Secondary PM2.5 and Factory Pollution Sources by Using ANN and the Correlation Coefficient

• Published in: IEEE access Vol. 5; pp. 22812 - 22822
• Main Authors: Chang, Jui-Hung, Tseng, Chien-Yuan
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2017; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Air monitoring; Air pollution; Ammonia; Artificial neural network; Artificial neural networks; Atmospheric modeling; Chemical reactions; Classification; Correlation; Correlation analysis; Correlation coefficients; Exhaust gases; Factories; Farming; Industrial plants; Industries; mean square error methods; Meteorological data; Monitoring; Pollution monitoring; Pollution sources; Production facilities; Stations; tensorflow
• DOI: 10.1109/ACCESS.2017.2765337
• ISSN: 2169-3536

Industry 4.0 is gaining more attention from the public, and thus the correlation between factories and nearby environmental pollution sources is a subject worth in-depth research. Among environmental issues, Particulate Matter2.5 (PM2.5) has received considerable attention in recent years from academic units and governments, and one of the secondary PM2.5 sources is the complex chemical reaction of exhaust gases emitted from factories and ammonia (NH3), with NH3 mostly coming from stock farming. Therefore, the correlation between stock farming data and pollutionsources emitted from factories can be examined by using an artificial neural network (ANN). The first target of this study is to investigate the correlation of factory air pollution source data and stock farming data nearby air monitoring stations to the annual mean PM2.5 concentration of nearby air monitoring stations. Second, the study uses Tensorflow to build an ANN model to analyze whether the industrial and stock farming data have an effect on the PM2.5 concentration. Weather data are taken in this experiment to learn about the correlation. The experimental results show that the Spearman's correlation coefficient of the factory emitted air pollution data and stock farming data nearby air monitoring stations for the annual mean PM2.5 concentration is 0.6 to 0.9, representing positive correlation. The ANN experiment shows the annual mean PM2.5 concentration classification model with industrial data plus stock farming data plus weather data, in which the ANN classification accuracy is 0.75 as validated by mean square error (MSE) methods. Compared with the ANN classification model only with weather data, the MSE classification accuracy is 1.5. According to the two experiments, the industrial factor and stock farming factor are items that may influence the PM2.5 concentration change.