Intelligent Optimization of Common Water Treatment Plant for the Removal of Organic Carbon

Intelligent model optimization is a key factor in the improvement of water treatment. In the current study, we applied artificial neural networks modelling for the optimization of the coagulation and flocculation processes to achieve sufficient water quality control over the total organic carbon parameter. The ANN network consisted of a multilayer feed-forward structure with a back propagation learning algorithm with the output layer of ferric chloride and cationic polymer dosages. The results were simultaneously compared with the nonlinear multiple regression model. The model validation phase was performed using 94 unknown samples for which the prediction result was in good agreement with the observed values. Analysis of the results showed a determination coefficient of 0.85 for the cationic polymer and 0.97 for the ferric chloride models, respectively. He mean absolute percentage error and root mean square errors were calculated, consequently, as 5.8% and 0.96 for the polymer and 3.1% and 1.97 for the ferric chloride models, respectively. According to the results, artificial neural networks proved to be very promising for the optimization of water treatment processes.