Wastewater treatment plant performance analysis using artificial intelligence - an ensemble approach.

Area of Science:

• Environmental Engineering
• Artificial Intelligence
• Wastewater Treatment

Background:

• Wastewater treatment plant (WWTP) performance prediction is crucial for environmental management.
• Artificial intelligence (AI) offers advanced methods for complex system modeling.
• Traditional methods may not fully capture the non-linear dynamics of WWTPs.

Purpose of the Study:

• To compare the predictive performance of AI models (FFNN, ANFIS, SVM) and MLR for Nicosia WWTP.
• To develop and evaluate single and ensemble AI models for enhanced prediction accuracy.
• To identify the most robust ensemble method for predicting effluent quality parameters.

Main Methods:

• Application of Feed Forward Neural Network (FFNN), Adaptive Neuro Fuzzy Inference System (ANFIS), Support Vector Machine (SVM), and Multi-Linear Regression (MLR).
• Development of single and ensemble models using daily WWTP operational data.
• Implementation of Simple Averaging Ensemble (SAE), Weighted Averaging Ensemble (WAE), and Neural Network Ensemble (NNE) techniques.

Main Results:

• ANFIS demonstrated superior performance among single AI models.
• Ensemble models significantly improved prediction accuracy, especially for effluent biological oxygen demand (BOD_eff).
• Neural Network Ensemble (NNE) achieved the highest performance improvement (up to 24%) in the verification phase for BOD_eff.

Conclusions:

• AI-based models, particularly ANFIS, are effective for predicting WWTP performance.
• Ensemble modeling, specifically NNE, offers a robust and reliable approach for enhancing prediction accuracy.
• The NNE model's non-linear averaging kernel contributes to its superior performance in wastewater treatment prediction.