In a recent study published in Energy Exploration & Exploitation, researchers Walid Abdelfattah, Munthar Kadhim Abosaoda, Krunal Vaghela, Gowrishankar J, Prabhat Kumar Sahu, Kamred Udham Singh, R Sivaranjani, and Samim Sherzod tackled the persistent challenge of accurately predicting 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 yield from biomassBiomass is a complex biological organic or non-organic solid product derived from living or recently living organism and available naturally. Various types of wastes such as animal manure, waste paper, sludge and many industrial wastes are also treated as biomass because like natural biomass these More 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. Their work introduces a robust machine learning framework that significantly enhances our ability to forecast biochar output, a critical step for optimizing production in sustainable agriculture and energy applications.

Biochar offers numerous benefits. It improves soil fertility and water retention, acts as a carbon sequestration tool, and serves as an effective solid fuel with higher carbon content and energy density than untreated biomass. Given its diverse applications, from environmental remediation to agricultural enhancement, precisely predicting its yield is paramount for efficient and sustainable production. The research team developed a predictive model using a comprehensive dataset of 211 biomass samples, each characterized by 14 normalized input features, including chemical, physical, and operational parameters. The target variable for prediction was biochar yield, measured as a weight percentage. To build their predictive framework, the researchers employed Gradient Boosted Decision Trees (GBDT), an ensemble machine learning algorithm known for its high predictive accuracy and flexibility.

A key aspect of this study was the exhaustive tuning of GBDT hyperparameters using four advanced optimization algorithms: Gaussian Processes Optimization (GPO), Evolutionary Strategies (ES), Bayesian Probability Improvement (BPI), and Batch Bayesian Optimization (BBO). These algorithms systematically searched for optimal parameter settings to maximize the model’s performance. The models were rigorously evaluated using a 90% training and 10% testing split of the dataset.

The results demonstrated that the GBDT-BPI model achieved the best performance. Beyond predictive accuracy, the study also identified the most influential factors affecting biochar yield through sensitivity analysis and SHapley Additive exPlanations (SHAP) values. The analysis revealed that residence timeResidence time refers to the duration that the biomass is heated during the pyrolysis process. The residence time can influence the properties of the biochar produced.   More, pyrolysis temperature, and fixed carbon content are the top three parameters significantly influencing biochar production. Specifically, longer residence times were associated with higher biochar yields, while increased temperatures led to decreased yields, a finding consistent with the theory that higher temperatures promote the release of volatile compounds rather than solid biochar formation.

This research marks a significant advancement in the accurate modeling of biochar yield. By integrating advanced machine learning with sophisticated hyperparameter optimization techniques and providing clear interpretability of feature importance, the study offers valuable insights for optimizing biomass pyrolysis workflows. These findings are crucial for both researchers and industry professionals seeking to enhance the efficiency and sustainability of biochar production for agricultural and energy applications. Future work will focus on expanding the dataset and exploring additional modeling paradigms to further improve generalizability and applicability.

Source: Abdelfattah, W., Abosaoda, M. K., Vaghela, K., J, G., Sahu, P. K., Singh, K. U., … & Sherzod, S. (2025). Accurate modeling of biochar yield based on proximate analysis. Energy Exploration & Exploitation, 1-27.