Machine learning prediction of higher heating value of biochar based on biomass characteristics and pyrolysis conditions

Near-infrared spectroscopy as a green analytical tool for sustainable biomass characterization for biofuels and bioproducts: An overview

Biomass, a widely used renewable energy source, requires characterization to optimize biofuel and bioproduct processes, customize feedstocks, and ensure economic and environmental sustainability. Conventional wet-chemistry methods for biomass analysis are slow, expensive, and require significant reagents and skilled personnel. In contrast, near-infrared (NIR) spectroscopy, a faster, cost-effective, and reagent-free green technology, enables non-destructive biomass analysis with minimal sample preparation. This study provides an overview of the fundamentals of NIR spectroscopy and explores its recent applications for analyzing various biomass properties important to the biofuel and bioproduct industry. The study also critically evaluates the challenges and opportunities of using NIR spectroscopy for biomass analysis. This review aims to guide future research for rapid and high throughput characterization of biomass in the biomass industry, supporting the United Nations' sustainable development goal (SDG) 7: producing affordable and sustainable energy.

Targeted conversion of cellulose and hemicellulose macromolecules in the phosphoric acid/acetone/water system: An exploration of machine learning evaluation and product prediction

The simultaneous hydrolysis of cellulose and hemicellulose involves trade-offs, making precise control of hydrolysis products crucial for sustainable development. This study employed three machine learning (ML) models-Random Forest (RF), Extreme Gradient Boosting (XGB), and Support Vector Machines (SVM)-to simulate and predict the yields of xylose (Xyl), furfural (FF), glucose (Glu), 5-hydroxymethylfurfural (5-HMF), and levulinic acid (LA) in a phosphoric acid/acetone/water system. The RF model demonstrated the highest accuracy, with R2 values between 0.782 and 0.887, and RMSE from 1.740 to 3.370. Key factors affecting the targeted conversion of macromolecules were identified as the solid-liquid ratio, reaction temperature, and acid dosage, with 160 °C recognized as a critical threshold for converting sugars derived from cellulose and hemicellulose into aldehydes and acids. The presence of metal chlorides, particularly AlCl3, significantly enhanced the selectivity of reactions and affected the distribution of products. It was found that corncobs are more efficient than bagasse in producing Glu. This study supports precise control over a multivariate system for producing multiple hydrolysis products from hemicellulose and cellulose, paving the way for data-driven optimization of lignocellulosic biomass conversion to high-value chemicals.

Thermal behavior and conversion of agriculture biomass residues by torrefaction and pyrolysis

Vegetal biomass is an abundant, readily available and easy to collect resource which can be converted into energy and materials. Biomass residues from agriculture and fruit crop activities, grouped in four classes (stalks, hulls, shells, pits), were subjected to thermal analysis and valorization. Thermogravimetry revealed high homogeneity between shells, large heterogeneity of stalks, and presence of thermally sensitive compounds in hulls. The Fisher weight variable selection analysis indicates that the differences in thermal behavior of biomass residues come from the components with specific biological functions (e.g. light volatiles and oils), while the structural components (hemicelluloses, cellulose and lignin) provide the general trend. This allows sample classification prior deciding on further waste management procedures. Torrefaction at 250 °C concentrated most part of the energy content into solids, with energy yield approaching 100%. Pyrolysis at 550 °C produces biochars with calorific values above 30 kJ/g from shells and pits. Most part of the energy input is used to produce oils with various compositions. Shells can be used to obtain phenolic compounds, hulls for production of aromatics and stalks for furans and ketones. Pits, on the other hand, are suitable raw material when fatty acids are targeted as pyrolysis compounds.

Cd adsorption prediction of Fe mono/composite modified biochar based on machine learning: Application for controllable preparation

In this study, artificial neural network (ANN) and random forest (RF) were constructed to predict the Cd adsorption capacity of Fe-modified biochar. The RF model outperformed ANN model in accuracy and predictive performance (R2 = 0.98). Through the contribution factors analysis of SHAP, structural characteristics (55.44%) were most important of Fe composite-modified biochar (CBC). And CBC have the best adsorption performance when C, Fe, O, H, N, and pH content were <50%, 10-20%, 10-20%, 0.5-1%, 0-2%, and >10, respectively. The Fe-Ca modified biochar (FeCa-BC) of different raw materials (wheat straw, corn straw and walnut shell) were successfully prepared according to the ML results, and the experimental data of FeCa-BC verified the accurate predictive ability of RF model (R2 = 0.89). The developed GUI toolbox results showed that the error between predicted and actual values was less than 5% based on the training set, testing set, and experimental validation set. The analysis of FTIR, XRD and XPS indicated that surface complexation, precipitation, and ion exchange were the main Cd adsorption mechanisms of FeCa-BC. This work presents new insights for the targeted preparation of functional biochar and its application in contaminated water through ML.

Machine learning prediction of fundamental sewage sludge biochar properties based on sludge characteristics and pyrolysis conditions

Sewage sludge biochar (SSBC) has significant potential for resource recovery from sewage sludge (SS) and has been widely studied and applied across various fields. However, the variability in SSBC properties, resulting from the diverse nature of SS and its intricate interaction with pyrolysis conditions, presents notable challenges to its practical use. This research employed machine learning techniques to predict fundamental SSBC properties, including elemental content, proximate compositions, surface area, and yield, which are essential for assessing the applicability of SSBC. The models achieved robust predictive accuracy (test R2 = 0.82-0.95), except for surface area. Notably, analysis of the optimal models revealed SS characteristics had a significant impact on SSBC properties, particularly total and fixed carbon content (combined importance exceeding 80%). This underscores the needs of source analysis and preparation optimization in targeted SS recovery or SSBC applications. To facilitate this, a graphical user interface was developed for strategic analyzation of sludge sources and pyrolysis settings.

Critical insights into ensemble learning with decision trees for the prediction of biochar yield and higher heating value from pyrolysis of biomass

Pyrolysis is an efficient thermochemical conversion process, but accurate prediction of yield and properties of biochar presents a significant challenge. Three prominent ensemble learning methods, viz. Random Forest (RF), eXtreme Gradient Boosting (XGB), and Adaptive Boosting (AdaBoost) were utilized to develop models to predict yield and higher heating value (HHV) of biochar. Dataset comprising 423 observations from 44 different biomasses was curated from peer-reviewed journals for predicting biochar yield. RF regressor achieved a test R2 of 0.86 for biochar yield, while XGB regressor achieved a test R2 of 0.87 for biochar HHV prediction. The SHapley Additive exPlanations (SHAP) analysis was conducted to assess influence of each feature on the model's output. Pyrolysis temperature and ash content of biomass were identified as the most influential features for the prediction of both yield and HHV of biochar. The partial dependence plots (PDPs) revealed nonlinear relationships, interpreting how the model formulates its predictions.

Accelerated screening of active sites on biochar for catalysis and adsorption via multidimensional fingerprint factor descriptors

Highly active biochar has great application potential in heterogeneous catalysis and adsorptive processes. The complexity of carbonization process makes it difficult to construct target active sites. This work put forward a reactive descriptor based on pyrolysis parameters and intrinsic composition of biomass. Results show that the model showed better predictive performance for C-C/C=C (R2 = 0.85), C=O (R2 = 0.85) and defect (R2 = 0.91) sites. The SHapley Additive exPlanation analysis shows that the pyrolysis parameters and the higher heating values are equally important for the active sites. The predictive performance and guiding role of the descriptor were validated by experiments. The descriptors proposed in this study integrated significant advantages of simplicity and easy accessibility, which would break the bottleneck of accurate construction of active sites and provide a theoretical basis for high-value resource utilization of biomass.

Thermogravimetric experiments based prediction of biomass pyrolysis behavior: A comparison of typical machine learning regression models in Scikit-learn

A variety of machine learning (ML) models have been extensively utilized in predicting biomass pyrolysis owing to their prowess in deciphering complex non-linear relationships between inputs and outputs, but there is still a lack of consensus on the optimal methods. This study elaborates on the development, optimization, and evaluation of three ML methodologies, namely, artificial neural networks, random forest (RF), and support vector machines, aimed to determine the optimal model for accurate prediction of biomass pyrolysis behavior using thermogravimetric data. This work assesses the utility of thermal data derived from these models in the computation of kinetic and thermodynamic parameters, alongside an analysis of their statistical performance. Eventually, the RF model exhibits superior physical interpretability and the least discrepancy in predicting kinetic and thermodynamic parameters. Furthermore, a feature importance analysis conducted within the RF model framework quantitatively reveals that temperature and heating rate account for 98.5 % and 1.5 %, respectively.