Lignocellulosic biowaste hydrothermal humification: Is there a method for predicting humic acid?

Biomass-derived hydrochar exhibits significant potential for producing hydrothermal humic acid (HHA) in an alkaline hydrothermal environment. However, the quantitative relationship between the chemical characteristics of hydrochar and HHA yield remains unclear, and a method for assessing the hydrothermal humification potential of lignocellulosic biowaste under complex conditions is lacking. This study addressed this knowledge gap by identifying three scenarios based on the key indicator “the unsaturation degree of lignocellulosic biowaste’s and hydrochar’s average molecular formula (defined as Ω 1 and Ω 2 , respectively)”: (1) insufficient hydrothermal carbonization (Ω 2  − Ω 1  < 1), (2) sufficient hydrothermal carbonization (Ω 2  − Ω 1  ≥ 1 and Ω 2  ≤ 4), and (3) excessive hydrothermal carbonization (Ω 2  ≥ 4). Using these scenarios, the study compiled data from literature and established a database (approximately 500 data sets) on the hydrothermal humification of lignocellulosic biowaste via hydrochar. Subsequently, the study applied a machine learning approach—Particle Swarm Optimization and Least Squares Support Vector Machine (PSO-LS-SVM)—to predict HHA yield based on different raw materials and complex hydrothermal conditions. The MAE, MSE, RMSE, and R 2 values from 50 random PSO-LS-SVM models demonstrated high accuracy and stability in predictions, with the highest R 2 values reaching 0.96 (scenario 1) and 0.97 (scenario 2). This approach can significantly reduce costs and effort for users conducting biowaste hydrothermal humification experiments.