A novel intensified anaerobic co-digestion process for high-quality biomethane production using ultrasound energy: Optimization and robustness tests with ML-RSM-TLBO algorithm

Biogas, a viable substitute for petroleum-based fuels, faces limitations in conventional production methods, which are slow and yield low outputs. Process intensification techniques, offer faster conversion and higher yields compared to conventional anaerobic digestion. However, the nonlinear dynamics and complex reaction kinetics involved in anaerobic digestion present significant challenges for traditional modeling approaches. To address this, machine learning (ML) has emerged as a powerful tool for capturing these complex behaviors. This study investigates the application of tree-based ensemble ML models to analyze biogas production from food waste co-digested with cow dung using an ultrasonic reactor across varying input parameters and operational conditions. Ten ML regression algorithms were assessed based on statistical performance metrics, including mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE). The gradient-boosting regressor achieved superior performance with an MAE of 1.359, RMSE of 1.9932, and MAPE of 2.44 %, outperforming other models in predicting biogas yields. Further optimization of input parameters was conducted using response surface methodology (RSM) with desirability analysis, while the teaching-learningbased optimization (TLBO) algorithm tested model robustness. Optimal conditions for maximizing methane yield (69.831 %) were identified as 1500 W ultrasonic power, 52.21 min of sonication, and a reactor temperature of 56.73 degrees C. These findings were validated through confirmatory experiments, aligning closely with the RSM results, further confirming the robustness of the developed model. A comparative analysis between conventional and ultrasonic techniques demonstrated that the ultrasonic method achieved a 32.37 % higher methane yield compared to the conventional anaerobic digestion process. This study is significant for its potential to improve biogas production efficiency via process optimization and machine learning, thereby decreasing costs and energy consumption and operating as an excellent resource for sustainable energy production.