Evaluation of auto regressive integrated moving average (arima) and artificial neural networks (ann) in the prediction of effluent quality of a wastewater treatment system

Abstract

The main objective of wastewater treatment is to purify the water by degradation of organic matter in the water to an
environmentally friendly status. To achieve this objective, some effluent (waste water) quality parameters such as
Chemical oxygen demand (COD) and Biochemical oxygen demand (BOD5) should be measured continuously in order
to meet up with the said objective and regulatory demands. However, through the prediction on water quality
parameters, effective guidance can be provided to comply with such demand without necessarily engaging in rigorous
laboratory analysis. Box-Jenkin’s Auto Regressive Integrated Moving Average (ARIMA) technique is one of the most
refined extrapolation techniques for prediction while Artificial Neural Network (ANN) is a modern non-linear method
also used for prediction. The Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Root Mean
Square Error (RMSE) and Correlation coefficient (r) are used to evaluate the accuracy of the above-mentioned
models. This paper examined the efficiency of ARIMA and ANN models in prediction of two major water quality
parameters (COD and BOD5) in a wastewater treatment plant. With the aid of R software, it was concluded that in all
the error estimates, ANNs models performed better than the ARIMA model, hence it can be used in the operation of
the treatment system.