Development of Models for the Prediction of the Level of Concentration of Gas Pollutants in Utorogu Gas Plant in Delta State, Nigeria

Abstract

In adequate modelling of process parameters for the prediction of

the level of concentration of gaseous pollutants in gas flaring

during the extraction of crude oil poses great challenges to human

health. This study focused on the development of models for the

prediction of the level of concentration of gas pollutants in Utorogu

gas plant in Delta State, Nigeria. Aeroqual multi-parameter

environmental monitor (series 500), was employed to monitor the

concentrations of volatile organic compounds (VOCs), oxides of

nitrogen (NO2), oxides of Sulphur (SO2), ozone (O3) and methane

(CH4). The concentrations of the particulate matter (PM2.5 of these

gases were obtained at each monitoring point on daily bases for a

period of twelve weeks using Aerocet-531 SPM meter. Sky master

thermo anemometer (SM-28) was used to obtain the important

climatic variables (wind speed, atmospheric pressure, ambient

temperature and relative humidity) which affect the dispersion of

gaseous pollutants. The maximum concentration of each monitored

gaseous pollutant during the twelve weeks (12) monitoring period

was selected and recorded for data processing. In this study,

mathematical models were developed for predicting each gaseous

pollutant such as volatile organic compounds (VOCs), oxides of

nitrogen (NO2), oxides of Sulphur (SO2), ozone (O3) and methane

(CH4). The curve fitting tool in Matrix Laboratory {MATLAB

(2016a)} was employed to select models and to model the exact

mathematical relationship between the pollutant concentrations

and the flare distance; then the pollutants concentrations were

predicted beyond the experimental distance of 500m from the flare

point, the models were validated using coefficient of determination

(R2

), and root mean square error (RMSE). Based on the

parameters, it was observed that the Fourier function model had

the lowest root mean square error value of 0.7694 and coefficient

of determination r2

value of 0.9927. The results obtained satisfy

good model predictability.