DEVELOPMENT OF PREDICTIVE MODEL FOR THE SIMULATION OF INDUSTRIAL WASTEWATER TREATMENT

Abstract

In this work, serial hybrid model, which statistical model or machine learning model was used in conjunction with mechanistic model, was used to predict petrochemical industrial wastewater treatment. To determine the accuracy of the different models, two lab-scale systems were set up. The two systems were fed with different composition of petrochemical industrial wastewater and operated for a period of 150 days. Results from the systems were compared to that of the hybrid model. The machine learning model was used to predict the biodegradability of the wastewater collected within the period of study. Subsequently, the output of the model was fed into a mechanistic model to predict the effluent COD of the lab-scale systems. Results from the hybrid machine learning model followed the results trend of both lab-scale systems, indicating that serial hybrid modeling is a useful tool for the prediction of petrochemical wastewater treatment performance.