Improved beetle antennae search algorithm-optimized machine learning model for predicting methane production in anaerobic digestion

Anaerobic digestion is a promising approach for converting lignocellulosic waste into methane, yet variability in feedstock composition as well as process and operational conditions poses significant challenges for accurately predicting methane yield. While traditional biochemical methane potential assays are widely used, they are resource-intensive and impractical for large-scale applications. To advance machine learning applications, this study introduces an innovative predictive framework driven by an improved quantum beetle antennae search algorithm. By integrating Bloch sphere quantum coding and a nonlinear step size, the improved quantum beetle antennae search algorithm significantly enhances global exploration, demonstrating superior optimization performance on the 2020 IEEE Congress on Evolutionary Computation benchmark. A robust stacking model was developed utilizing a comprehensive dataset comprising 157 samples and 4538 data entries. This model characterizes the relationship between cumulative methane yield, key feedstock characteristics, and digestion time. It achieves high predictive accuracy, with determination coefficients of 0.990 for training set and 0.912 for test set, and corresponding root mean square errors of 9.55 and 29.07. Crucially, by employing Wasserstein Gradient Boosting as the terminal estimator, the proposed framework advances beyond traditional deterministic approaches to enable probabilistic forecasting and uncertainty quantification, thereby providing risk-aware insights for managing highly heterogeneous biomass. Concurrently, SHapley Additive exPlanations methodology was incorporated to elucidate feature importance and the underlying mechanisms driving methane production. This framework transcends improved prediction accuracy to function as a prototype for an end-to-end intelligent decision support system, paving the way for efficient, sustainable waste management and optimized large-scale anaerobic digestion systems.