Modelling coagulant dosage in drinking water treatment plant using advance machine learning model: Hybrid extreme...

Area of Science:

• Environmental Engineering
• Water Treatment Technologies
• Computational Intelligence

Background:

• Coagulation is crucial in drinking water treatment, but linking raw water quality to coagulant dosage remains challenging.
• Accurate coagulant dosage is essential for effective water purification and operational efficiency in treatment plants.

Purpose of the Study:

• To develop an integral modelling framework for predicting coagulant dosage rate based on raw water quality parameters.
• To evaluate the performance of a hybrid extreme learning machine and Bat metaheuristic algorithm (ELM-Bat) against other machine learning models.

Main Methods:

• A hybrid ELM-Bat model was developed to predict coagulant dosage using water temperature, pH, specific conductance, dissolved oxygen, and turbidity.
• The model was trained and validated using statistical metrics including RMSE, MAE, R, and NSE.
• Performance was compared against standalone ELM, ORELM, OSELM, OPELM, and KELM.

Main Results:

• The hybrid ELM-Bat model demonstrated significantly higher accuracy compared to all other evaluated models.
• During validation, ELM-Bat achieved R ≈ 0.959 and NSE ≈ 0.918, showing substantial improvements over other methods.
• The model exhibited strong predictive capabilities for determining optimal coagulant dosage.

Conclusions:

• The ELM-Bat model offers a highly accurate and efficient solution for optimizing coagulant dosage in drinking water treatment plants.
• This approach effectively addresses the challenge of linking raw water quality to treatment parameters.
• The findings suggest the ELM-Bat model can be optimally utilized for precise coagulant dosing, enhancing water treatment processes.