Estimating the chemical oxygen demand of petrochemical wastewater treatment plants using linear and nonlinear statistical models – A case study

Abstract

In this research, twelve linear and nonlinear regression models were performed and evaluated to formulate the best one for the estimation of chemical oxygen demand level in the effluent of the clarifier unit of a petrochemical wastewater treatment plant. The input variables measured twice a day in the influent of the biological unit over a period of 13 months using standard methods. The piece-wise linear regression with breakpoint method, with a mean squared error value equal to 0.041, mean absolute error of 0.144, and correlation coefficient equal to 0.835 was found to estimate the output chemical oxygen demand parameter more sustainable rather than other linear and nonlinear methods. However, some of the other applied models such as radial basis function neural network and gene expressing programming models achieved good performance considering their correlation coefficient, robustness in presence of outliers, mean squared error and mean absolute error test. Mathematical and intelligent modeling proved useful as an accurate alternative to estimate the amount of chemical oxygen demand rather than spending time and cost for its laboratory tests.