Biomethane Production from the Mixture of Sugarcane Vinasse, Solid Waste and Spent Tea Waste: A Bayesian Approach for Hyperparameter Optimization for Gaussian Process Regression

Abstract

In this work, sugarcane vinasse combined with organic waste (food and wasted tea) was demonstrated to be an excellent source of biomethane synthesis from carbon-rich biowaste. The discarded tea trash might be successfully used to generate bioenergy. The uncertainties and costs associated with experimental testing were recommended to be decreased by the effective use of contemporary machine learning methods such as Gaussian process regression. The training hyperparameters are crucial in the construction of a robust ML-based model. To make the process autoregressive, the training hyperparameters were fine-tuned by employing the Bayesian approach. The value of R2 was found to be greater during the model test phase by 0.72%, assisting in the avoidance of model overtraining. The mean squared error was 36.243 during the model training phase and 21.145 during the model testing phase. The mean absolute percentage error was found to be under 0.1%, which decreased to 0.085% throughout the model’s testing phase. The research demonstrated that a combination of wasted tea trash, sugarcane vinasse and food waste may be a viable source for biomethane generation. The contemporary methodology of the Bayesian approach for hyperparameters tuning for Gaussian process regression is an efficient method of model prediction despite the low correlation across data columns. It is possible to enhance the sustainability paradigm in the direction of energy security via the efficient usage of food and agroforestry waste.