Integrated data-driven strategy to optimize the processes configuration for full-scale wastewater treatment plant...

Area of Science:

• Environmental Engineering
• Data Science
• Water Treatment Technology

Background:

• Wastewater treatment plants (WWTPs) are crucial for pollutant removal and water recycling.
• Current WWTP configuration selection relies on standards and experience, not data-driven methods.

Purpose of the Study:

• To develop an intelligent, data-driven strategy for optimizing WWTP process configurations during predesign.
• To leverage machine learning for more efficient and effective WWTP planning.

Main Methods:

• Utilized t-distributed stochastic neighbor embedding (t-SNE) for clustering WWTP data based on influent characteristics.
• Developed deep neural network (DNN) classification models for four identified clusters.
• Trained and tested DNN models on a dataset of 14,647 samples from 10 full-scale WWTPs.

Main Results:

• t-SNE successfully clustered WWTP data into four distinct groups.
• DNN models achieved high classification accuracy, with a maximum testing accuracy of 0.9505.
• The models demonstrated capability in identifying optimal WWTP process configurations for specific scenarios.

Conclusions:

• The integrated t-SNE and DNN approach effectively optimizes WWTP predesign.
• This data-driven strategy enhances the selection of processes configuration by utilizing relationships between parameters and plant design.
• The findings support engineers in predesigning WWTPs with optimal configurations.