---

A study published in the Journal of Analytical and Applied Pyrolysis explored the use of machine learning to predict the physicochemical properties of biochar, a carbon-rich product derived from biomass through pyrolysis. This research aligns with the circular economy’s principles by potentially reducing the time, energy, and resources typically required for biochar production.

The study examined various machine learning algorithms, including support vector machines, multiple linear regression, nearest neighbor algorithm, random forest, gradient boosting regression, and eXtreme Gradient Boosting (XGB). Among these, XGB stood out for its excellent predictive performance, achieving an R2 value greater than 0.99 for biochar yield and elemental distribution. This high level of accuracy suggests that XGB can effectively predict biochar properties from different biomass sources without the need for extensive and costly pyrolysis experiments.

Further analysis using Pearson correlation coefficient and SHAP (Shapley Additive exPlanetions) techniques highlighted a strong positive correlation between pyrolysis temperature and the degree of biochar aromatization. This insight is crucial for optimizing biochar production processes, as it indicates that higher temperatures can enhance certain desirable properties of biochar.

The study demonstrates the potential of machine learning to streamline biochar production by accurately predicting its properties based on the characteristics of the biomass feedstock and the conditions under which it is pyrolyzed. This advancement could lead to more efficient and sustainable practices in biomass management and biochar utilization, supporting environmental goals such as waste management, carbon sequestration, and soil restoration.