Chiew C.H., Lim L.Y., Ong P.Y., Li C., Fan Y.V., 2024, Regression Models for Predicting Physicochemical Properties of BiocharBiochar is a carbon-rich material created from biomass decomposition in low-oxygen conditions. It has important applications in environmental remediation, soil improvement, agriculture, carbon sequestration, energy storage, and sustainable materials, promoting efficiency and reducing waste in various contexts while addressing climate change challenges. More, Chemical Engineering Transactions, 113, 553-558 DOI:10.3303/CET24113093

Predicting the physicochemical properties of biochar is critical for optimizing its use in sustainable agriculture and environmental remediation. A recent study compared the performance of various regression models—linear, quadratic, non-linear regression (NLR), and multiple linear regression (MLR)—for this purpose. The results revealed that MLR consistently outperformed others, achieving high predictive accuracy with R² values exceeding 0.92. This model excelled in predicting properties such as cation exchange capacity (CEC) and electrical conductivity (EC), demonstrating its utility for practical biochar applications.

NLR models also showed strong performance, especially in predicting biochar’s high heating value (HHV), with R² values as high as 0.98. However, certain properties, like CEC and specific surface area (SSA), exhibited inconsistencies between R² and root mean square error (RMSE) values, suggesting challenges in capturing the variability of these parameters. Municipal Solid Waste (MSW)-derived biochar posed the greatest prediction challenges due to its heterogeneous composition.

PyrolysisPyrolysis is a thermochemical process that converts waste biomass into bio-char, bio-oil, and pyro-gas. It offers significant advantages in waste valorization, turning low-value materials into economically valuable resources. Its versatility allows for tailored products based on operational conditions, presenting itself as a cost-effective and efficient More temperature emerged as a significant predictor across models, particularly for nitrogen content and EC. Despite these advances, integrating MLR with non-linear techniques could enhance predictive accuracy and practical usability, addressing limitations like multicollinearity and data variability.

The study highlights the potential of hybrid models combining traditional regression methods and machine learning. Such approaches can improve the reliability of biochar property predictions, facilitating its broader adoption in agriculture and environmental management.