Electrical energy recovery from wastewater: prediction with machine learning algorithms

Area of Science:

• Environmental Science
• Energy Engineering
• Computer Science

Background:

• Biomass is a significant renewable energy source, with biogas production from waste gaining popularity.
• Wastewater treatment plants offer potential for energy recovery through waste conversion.
• Estimating electrical energy from wastewater recovery using machine learning is an underexplored area.

Purpose of the Study:

• To predict the electrical energy recovery potential of sewage sludge from Kahramanmaraş Advanced Biological Wastewater Treatment Plant (KABWWTP).
• To evaluate energy recovery through incineration and anaerobic digestion processes.
• To apply and compare six distinct machine learning algorithms for this prediction task.

Main Methods:

• Utilized six machine learning algorithms: linear regression (LR), extreme gradient boosting (XGB), Gaussian process regression (GPR), ridge regression (RR), Lasso regression (LASReg), and Bayesian ridge regression (BR).
• Employed a novel approach with only three input parameters: gas flow, conductivity, and total suspended solids (TSS) to predict electrical energy output.
• Applied heat mapping and correlation analyses to understand parameter relationships.

Main Results:

• The extreme gradient boosting (XGB) algorithm demonstrated the highest performance with a Mean Absolute Percentage Error (MAPE) of 1.032%.
• Identified key relationships between input parameters (gas flow, conductivity, TSS) and electrical energy output.
• Performance metrics for all tested algorithms were presented for comparison.

Conclusions:

• Machine learning, particularly XGB, is effective for predicting electrical energy recovery from sewage sludge.
• The study successfully demonstrated a method for estimating energy potential using limited input parameters.
• This research contributes to sustainable energy generation through efficient waste management and wastewater recovery.