Boosting biogas production through innovative data-driven modeling and optimization methods at NJWTP.

Area of Science:

• Environmental Engineering
• Biotechnology
• Data Science

Background:

• Wastewater treatment plants (WWTPs) generate significant sludge, presenting disposal challenges and opportunities for energy recovery.
• Anaerobic digestion is a key process for converting organic waste into biogas, a renewable energy source.
• Optimizing anaerobic digestion requires sophisticated modeling to manage complex biological and operational variables.

Purpose of the Study:

• To develop and evaluate data-driven models for enhancing biogas production via anaerobic digestion of wastewater sludge.
• To compare the performance of Deep Belief Network (DBN), DBN with Osprey Optimization Algorithm (DBN-OOA), and DBN with Boosted Osprey Optimization Algorithm (DBN-BOOA).
• To identify optimal operational parameters for maximizing biogas yield and minimizing sludge production.

Main Methods:

• Collected 180 data points from Nanjing Jiangnan Wastewater Treatment Plant (NJWTP) between 2016 and 2018.
• Developed and compared three machine learning models: DBN, DBN-OOA, and DBN-BOOA.
• Utilized statistical metrics including correlation coefficient (R), root mean square error (RMSE), and index of agreement (IA) for model evaluation.

Main Results:

• The DBN-BOOA model achieved superior performance with R=0.98, RMSE=0.41 m³/min, and IA=0.99.
• DBN-BOOA significantly outperformed standalone DBN and DBN-OOA models in accuracy and optimization.
• Identified optimal parameters leading to a maximum biogas production of 31.35 m³/min.

Conclusions:

• The DBN-BOOA model offers a highly accurate and efficient method for optimizing anaerobic digestion and biogas production.
• This data-driven approach requires no input variable pre-processing, making it practical for real-world applications.
• The DBN-BOOA model provides a user-friendly solution for wastewater treatment plant operators to enhance biogas yields and manage sludge effectively.