Modeling and Optimization of Coagulant Dosage in Drinking Water Treatment Plant Using Artificial Intelligence. Case of the Ain Tin Station in Mila, Algeria

This study presents the modeling and optimization of the coagulant dose of a drinking water treatment plant by the use of neuron networks based on the characteristics of raw water. The determination of the optimal dose of the coagulant is particularly important in the treatment plant and requires so-called Jar-Test laboratory tests. This type of approach (Jar Test) has the disadvantage of requiring a relatively long response time, in addition, it does not allow for detailed monitoring of the evolution of raw water quality. To this end, the objective is to provide a preliminary tool for the automated management of the said station by improving its quality of service accordingly. For this, neural models were used where modelling attempts were used to link the value of the optimal dose of the coagulant (aluminium sulphate) to the quality of the raw water at the inlet of the treatment plant (turbidity, temperature, pH, conductivity and dissolved oxygen) of the surface waters of Ain Tin. The results we arrived at are very satisfactory, where the correlation coefficient R=0.998, NSE=0.996, RMSE=0.187 and MAE=0.016.