Wastewater treatment process enhancement based on multi-objective optimization and interpretable machine learning.

Area of Science:

• Environmental Engineering
• Computational Science

Background:

• Wastewater treatment process (WTP) optimization is crucial for cost reduction and efficiency.
• Intelligent operation and maintenance management of wastewater treatment plants (WWTPs) are essential for environmental protection.

Purpose of the Study:

• To propose a wastewater treatment process multi-objective optimization (WTPMO) framework for decision-making in WTP parameter settings.
• To develop accurate prediction models for effluent water quality (EQ) and energy consumption (EC).
• To optimize WTP operations for reduced energy consumption while maintaining water quality standards.

Main Methods:

• Developed Extreme Gradient Boosting (XGB) prediction models optimized with Bayesian optimization (BO) for EQ and EC.
• Employed SHapley Additive exPlanations (SHAP) for model interpretability, analyzing feature impacts on predictions.
• Utilized Non-dominated Sorting Genetic Algorithm II (NSGA-II) combined with Technique for Ordering Preferences on Similarity of Ideal Solutions (TOPSIS) for multi-objective optimization.

Main Results:

• BOXGB models achieved high accuracy for EQ (R²=0.923) and EC (R²=0.965) predictions.
• SHAP analysis provided clear insights into how influent quality and process variables influence EQ and EC.
• Achieved a 1.552% energy consumption optimization rate while ensuring effluent water quality compliance on Benchmark Simulation Model 1 (BSM1).

Conclusions:

• The proposed WTPMO framework effectively optimizes WTP operations.
• The research supports intelligent management of WWTPs, contributing to environmental protection and sustainable development goals.