PM10.0 AIR POLLUTION MODELING AND ESTIMATION USING ARTIFICIAL NEURAL NETWORK (ANN)

Abstract

Air pollution is a huge challenge to the residents of highly populated cities and their municipal managers over the years because of the serious threats it poses to human health and environment. Several smart Artificial Intelligence based techniques such as the use of Machine learning and deep learning algorithms methods such as Artificial Neural Network (ANN) can be deployed to predict or estimate the levels of emissions of pollutants in the ambient environment within a particular locality or city provided past historical dataset  of other air and meteorological parameters are available to use to identify patterns in an occurrence in the dataset which can be used to predict or forecast future occurrences of air pollution in that particular location. This paper used historical dataset of air pollutants and meteorological parameters such as PM2.5, PM10.0, PM1.0, and other parameters obtained in Awka Metropolis from October 14th 2021 to December 4, 2021 to model the proposed models. In this work, the particulate matter PM10.0 prediction modeling were carried out using Artificial Neural Network (ANN) and eight other machine learning models using the Train-Test Data Split method. The prediction performances of these machine learning models were evaluated using statistical performance metrics such as MAE, RMSE and R².  After more than 100 experimental runs, Artificial Neural Network (ANN) algorithm using Multi Layer Perceptron (MLP) with 200 hidden neurons, we obtained an RMSE value of 1.2637 µg/m³, MAE of 0.5885 µg/m³Â and R²Â of 0.9828 or 98.28%, which is a high accurate prediction of the measured value. Other machine learning algorithms tried on the training and testing data gave different results of RMSE, MAE and R². This shows that apart from ANN, other machine learning algorithms can accurately predict or forecast in advance time the levels of dispersion of air pollutants such as PM10.0 in the ambient environment of a particular city or location.

 

KEYWORDS: PM10.0, PM2.5, particulate matters, Artificial Neural Network (ANN), smart city