Related Experiment Video
Updated: Nov 3, 2025
06:11
Author Spotlight: Advancing Anaerobic Microbiota Research Using a Novel Respirometry Protocol
Published on: April 26, 2024
1.5K
Biogas maximization using data-driven modelling with uncertainty analysis and genetic algorithm for municipal
Mohsen Asadi1, Kerry McPhedran1
1Department of Civil, Geological & Environmental Engineering, University of Saskatchewan, Saskatoon, Saskatchewan, Canada.
Journal of Environmental Management
|June 1, 2021
Summary
Data-driven models for anaerobic digestion can boost biogas energy production. A non-linear regression model using unprocessed data best estimated biogas rates, outperforming complex models and aiding optimization for increased energy yield.
Area of Science:
- Environmental Engineering
- Biotechnology
- Renewable Energy
Background:
- Anaerobic digestion (AD) is crucial for biogas production, a valuable energy source.
- Optimizing AD processes through data-driven modeling can enhance biogas yield and energy output.
- Municipal wastewater treatment plants (MWTPs) utilize AD for waste management and energy recovery.
Purpose of the Study:
- To develop and compare 'white-box' (regression) and 'black-box' (ANN, ANFIS) models for estimating biogas production rates in MWTP AD.
- To assess the impact of data processing (correlation tests, PCA) on model performance.
- To integrate developed models with a genetic algorithm (GA) for optimizing AD operating parameters to maximize biogas production.
Main Methods:
- Collected AD operating parameters (volatile fatty acids, solids, pH, inflow rate) from an MWTP.
- Developed regression models using processed and unprocessed input variables.
- Compared regression models against artificial neural network (ANN) and adaptive network-based fuzzy inference system (ANFIS) models.
- Utilized Monte Carlo Simulation for uncertainty analysis.
- Integrated models with a genetic algorithm (GA) for optimization.
Main Results:
- Data processing did not enhance regression model performance; unprocessed data yielded better results.
- The non-linear regression model with unprocessed inputs showed strong performance (R=0.81, RMSE=0.95, IA=0.89).
- Black-box models (ANN, ANFIS) were more accurate but exhibited higher uncertainty compared to the regression model.
- GA optimization predicted maximum biogas production rates of 22.0 m³/min (ANN), 23.1 m³/min (ANFIS), and 28.6 m³/min (non-linear regression).
Conclusions:
- A non-linear regression model using unprocessed AD data provides a robust and less uncertain method for estimating biogas production.
- There is a trade-off between model accuracy and uncertainty, with simpler models potentially offering more reliable uncertainty estimates.
- Optimizing AD operating parameters via GA integration can significantly enhance biogas production rates for increased energy recovery.
