مدل‌سازی غلظت مونواکسید کربن، ازن و ذرات معلق کوچکتر از ۱۰ میکرون در هوای کرج با استفاده از شبکه عصبی مصنوعی

نوع مقاله : مقاله پژوهشی

نویسندگان

1 گروه عمران. دانشگاه پیام نور

2 پیام نور

10.22034/jess.2022.144413

چکیده

آلودگی هوا یک چالش‌ مهم برای زندگی در شهرهای بزرگ است و باعث افزایش مراجعات به مراکز درمانی و تعداد مرگ و میر سالمندان و مبتلایان به بیماری‌های قلبی و ریوی در مقاطعی از سال می‌شود. از این رو یافتن عوامل تاثیرگذار بر آلودگی هوای شهرها و دستیابی به مدلی معتبر برای پیش ‌بینی کیفیت هوا اجتناب ناپذیر می‌نماید. در این تحقیق تاثیر متغیرهای هواشناسی همچون درجه حرارت، سرعت باد، رطوبت نسبی، بارش باران و ابرناکی، بر غلظت مونواکسید کربن، ازن سطحی و ذرات معلق کوچکتر از 10 میکرون در کلان‌شهر کرج مورد بررسی قرار گرفته است. همچنین مدل سازی توسط شبکه عصبی مصنوعی و با استفاده از مقادیر پارامترهای هواشناسی شهر کرج و غلظت آلاینده‌ها در سطح این شهر در دوره زمانی 1391 الی 1398 انجام شده و داده‌های مربوط به سال 1399 برای آزمایش مدل ساخته شده مورد استفاده قرار گرفته است. نتایج بررسی‌ها نشان می‌دهد که قوی‌ترین همبستگی مونواکسید کربن به ترتیب با سرعت باد و دما به میزان 216/0- و 146/0- است. بیشترین همبستگی ازن به ترتیب با رطوبت نسبی، ابرناکی و دما به مقدار 328/0- ، 167/0- و 411/0 است. همچنین ذرات معلق به ترتیب با رطوبت نسبی، بارش و دما به اندازه 249/0- ، 174/0- و 211/0 همبستگی دارد. به علاوه ضریب همبستگی بین غلظت‌های واقعی و مقادیر پیش بینی شده توسط مدل برای مونواکسید کربن، ازن و ذرات معلق به ترتیب برابر با 909/0، 856/0 و 854/0 است.

کلیدواژه‌ها


عنوان مقاله [English]

modeling of carbon monoxide, ozone and PM10 airborne pollutants in the air of Karaj using artificial neural network

نویسندگان [English]

  • Elham Asrari 1
  • abolfazl Rock 2
1 Payame BPayame Noor University
2 payame noor
چکیده [English]

1. Introduction
Air pollution is an important challenge for life in large cities and increases the number of visits to medical centers and the number of deaths of the elderly and people with heart and lung diseases at certain times of the year. Therefore, finding the factors affecting air pollution in cities and achieving a valid model for predicting air quality is inevitable. In this study, the effect of meteorological parameters such as temperature, wind speed, relative humidity, rainfall and cloudiness on the concentration of carbon monoxide, surface ozone and particulate matter smaller than 10 microns in the metropolis of Karaj has been investigated.

2. Methodology
In this study, data on carbon monoxide, ozone and particulate matter smaller than 10 microns in Karaj for the period of 2012 – 2019 were received from the Air Quality Monitoring Center and data on meteorological parameters for the same time were received from the Meteorological Organization. The temporal and spatial distribution of pollutants in the city of Karaj has been studied. In addition, correlation coefficients were obtained between each of the pollutants and each of the meteorological parameters. Finally, using artificial neural network in MATLAB software, a nonlinear regression model was constructed between each of the pollutants with meteorological parameters including wind speed, temperature, relative humidity, precipitation and cloud intensity.

3. Results and discussion
3.1. Temporal distribution of pollutants
The maximum average concentration of 8-hour carbon monoxide for three stations take place in the cold season of the year and specifically in February at 2.30 PPM and the minimum in the warm season in September and equal to 1.27 PPM. The maximum average concentration of hourly ozone for three stations has happened in the summer season and in June, equal to 0.027 PPM and the minimum concentration in November at 0.018 PPM. Also, the maximum daily concentration of suspended particles smaller than 10 microns for three stations was in July at 75.33 μg/m3 and the minimum concentration was in April and equal to 43.71 μg/m3.

3.2. Spatial distribution of pollutants
By examining the average concentrations of 8-hour carbon monoxide, 1-hour ozone and 24-hour suspended particles smaller than 10 microns in different stations of Karaj and analyzes performed using SPSS software, statistical indicators related to the concentrations of the above pollutants in different stations of the city is calculated according to the Iranian Clean Air Act Standard approved in 2016 and result is shown in Table 1. In the above standard, the average concentrations of carbon monoxide, ozone and suspended particles smaller than 10 microns are set to 9 PPM, 0.125 PPM and 3150 μg/m3, respectively. As can be seen in Table 1, in Farhangsara station, 100% of the recorded concentrations are less than the 8-hour standard, and in the two metro stations and the Danshkadeh, they are 99.97 and 99.99% lower than the maximum allowable, respectively. Regarding 1-hour ozone in the two stations of Farhangsara and Danshkadeh, only one case of observations was more than the above-mentioned concentration, and in the metro station, no case was exceeded. Also, the conditions are different in relation to suspended particles smaller than 10 microns, and although the permissible limit of this pollutant has increased from 50 μg/m3 in the standard approved in 2009 to 315 μg/m3 in the standard in 2016, but the concentration of this pollutant in different parts of Karaj, especially at the Metro station is more than the other two pollutants.

Table 1. Statistical information on the spatial distribution of air pollutants in the city of Karaj
Pollutant Index Station Name
Daneshkadeh Metro station Farhangsara
CO Average Concentration (PPM) 1.8871 1.7447 2.0588
Minimum Concentration (PPM) 0.0360 0.1300 0.0962
Maximum Concentration (PPM) 9.1762 11.7400 7.1100
Cumulative frequency percentage below than Standard 99.99 99.97 100
O3 Average Concentration (PPM) 0.0340 0.0154 0.0194
Minimum Concentration (PPM) 0.0033 0.0004 0.0003
Maximum Concentration (PPM) 0.1366 0.1064 0.1321
Cumulative frequency percentage below than Standard 100 100 99.996
PM10 Average Concentration (μg/m3) 55.5447 76.3733 61.4949
Minimum Concentration (μg/m3) 9.07 26.63 7.14
Maximum Concentration (μg/m3) 500.61 188.57 210.42
Cumulative frequency percentage below than Standard 99.2 97.5 99

3.3. Correlation coefficient between pollutant and meteorological parameters
SPSS software was used in order to determine the correlation coefficient between each of the pollutants with meteorological parameters including temperature, relative humidity, rainfall, cloudiness and wind speed. It is observed that the average 8-hour concentration of CO has the highest negative correlation with wind and temperature -0.216 and -0.146, respectively. In addition, this pollutant has a significant but very weak positive correlation with relative humidity and cloudiness of 0.087 and 0.057, respectively. Although carbon monoxide has a very small positive correlation with precipitation, this small correlation is not even significant. Regarding ozone, it was noted that this pollutant has the highest negative correlation with relative humidity at -0.328 and weakly negative correlation with cloudiness at -0.167. Also, the correlation of this pollutant with precipitation is very weak -0.112. This pollutant has a significant positive correlation 0.41 with temperature and a weak positive correlation 0.185 with wind speed. Suspended particles have the highest negative correlation with relative humidity at -0.249. Also, a relatively weak negative correlation of -0.174 is observed between rainfall and this pollutant, but the correlation of this pollutant with cloudiness and wind is negligible. Among the above meteorological parameters, this pollutant shows only a positive correlation with temperature of 0.211.

3.4. Modeling of air pollutants in Karaj
To build a model related to carbon monoxide, average 8-hour concentrations of carbon monoxide have been used in three stations in Karaj. Also, meteorological parameters including relative humidity, temperature, cloudiness and wind speed were selected for use in the model according to the correlation coefficients. In constructing the artificial neural network model for ozone, values of average 1-hour concentrations of ozone for three stations and meteorological parameters including temperature, wind speed, relative humidity and cloudiness have been exploited. In order to build an artificial neural network model for suspended particles, 24-hour values of meteorological parameters including relative humidity, precipitation, wind speed, cloudiness were used due to the correlation of these parameters with the desired pollutant as well as average 24-hour concentrations of PM10 for three stations. A summary of the neural network performance including values for the total predicted squares, total error squares, total squares, R2 (R square), and error percentage for each pollutant is shown in table 2. It is worth noting that:

- The sum of the error squares is equal to the product of the mean square error (MSE) multiplied by the number of predictions.
- The coefficient of determination or (R2) is equal to the sum of the predicted squares on the total squares.
- The percentage of error is equal to (1-R2) or the sum of the squares of the error on the sum of the total squares.

As can be seen in Table 2, the percentage of error in prediction by the artificial neural network models for carbon monoxide, ozone and suspended particles is 16.7%, 7% and 25.3%, respectively, or in other words, the accuracy of the network for prediction of the concentrations of pollutants are 83.3%, 93% and 74.7%, respectively.

Table 2. Summary of the neural network performance
Pollutant sum of the predicted squares Sum of the total squared Sum of Square error R2 Error percentage
CO (PPM) 6870.085 8239.032 1395.647 0.833 16.7
O3 (PPM) 2.898 3.117 0.219 0.930 7
PM10 (μg/m3) 1583995.239 2121810.096 537814.857 0.747 25.3


4. Conclusions
Among the three pollutants studied in this research, particulate matter smaller than 10 microns play the largest role in air pollution in the metropolis of Karaj and Karaj metro station has the worst situation in this regard. After particulate matter, ozone is the second largest polluter in the city, which plays a major role in the pollution of Daneshkadeh station. In addition to that, in terms of time, July is the worst pollution situation for both of these pollutants, while in terms of carbon monoxide pollution, July is the best and February is the worst situation in Karaj. Based on this, it can be said that despite the fact that many believe that the air quality in Karaj is worse in winter than in summer, the results of this study not only do not confirm this belief, but even prove the opposite. Secondly, the use of artificial neural network provides an appropriate model for predicting air quality in the city of Karaj to city officials and managers. However, it should be noted that the ability of the neural network in predicting the concentration of pollutants in Karaj is not the same for different pollutants, and in this regard it can be said that the prediction of ozone with this type of network has maximum reliability, followed by carbon monoxide next.

کلیدواژه‌ها [English]

  • Air Pollutants
  • Meteorological parameters
  • Karaj
  • Artificial Neural Network