Renewable energy is significant for sustainable development as it decreases dependence on fossil fuels, manages climate change, and ensures long-term energy security. The research employs a novel pretreatment method for the different forms of sweet sorghum biomass using a bioenzyme (BE). Subsequently, Anaerobic digestion (AD) experiments were conducted under varying conditions to assess their impact on methane yield. To model and optimize the complex AD process, the study integrates a physical model with the machine learning (ML) model, leading to a novel hybrid modelling approach. A weighted ensemble method has been utilized by the integration layer with adaptive weighing based on the production phases, i.e., Lag, Exponential, and Stationary. Initially, dataset is plotted and divided into different regimes or phases. Then, according to these regimes, different models are trained and applied to regime-specific data. Further, physical models and the regime-specialized model's results are passed through an inverse-error weighted integration layer, which provides the weights. After this, the prediction is made with uncertainty. The hybrid model shows an R 2 value of 0.89, which outperforms the physical model (true hybrid), which has an R 2 score of 0.86, and the ML models with an R 2 of 0.84. The hybrid model also shows a decrease in the root mean square error (RMSE) by 18% as compared to the physical model and by 12% as compared to the ML model. The findings demonstrate that the combined approach of enzymatic pretreatment and ML-based optimization (phase-specific inverse-error weighted hybrid modelling) significantly enhances biomethane production, offering a sustainable and efficient strategy for renewable energy generation from agricultural residues. • Novel bioenzyme pre-treatment boosts sweet sorghum biomass efficiency. • Hybrid physical–ML model with phase-specific Bayesian weighting. • Achieved R 2 of 0.89 with 18% lower RMSE than physical model. • Enhanced biomethane yield supports sustainable energy generation.
Aggarwal et al. (Mon,) studied this question.