Simulation of aerated lagoon using artificial neural networks and multivariate regression techniques

Area of Science:

• Environmental Engineering
• Water Quality Management
• Industrial Wastewater Treatment

Background:

• Pulp and paper mills generate significant wastewater effluent.
• Accurate prediction of biochemical oxygen demand (BOD) is crucial for environmental compliance and process optimization.
• Aerated lagoons are common in industrial wastewater treatment, but their effluent quality can vary.

Purpose of the Study:

• To develop an empirical model for predicting the biochemical oxygen demand (BOD) of treated wastewater.
• To compare the performance of different predictive modeling techniques for industrial effluent.
• To assess the effectiveness of data preprocessing on model accuracy.

Main Methods:

• Development of predictive models using functional link neural networks (FLNNs).
• Comparison with multiple linear regression, principal components regression, and partial least-squares regression (PLSR).
• Evaluation of FLNN performance with and without PLSR data preprocessing.

Main Results:

• FLNN models showed improved predictive capability when data was preprocessed using PLSR.
• PLSR demonstrated effectiveness as a linear regression technique for this industrial wastewater data.
• The study identified limitations in operational data that influenced model development.

Conclusions:

• Data preprocessing, particularly with PLSR, significantly enhances the performance of FLNN models for BOD prediction.
• PLSR is a robust linear regression method suitable for modeling industrial wastewater with data limitations.
• The developed empirical models can aid in managing and optimizing wastewater treatment at pulp and paper facilities.