Biomass Dataset for Higher Heating Value Prediction: Proximate and Ultimate Analysis of 315 Heterogeneous Biomass Samples Across Six Feedstock Categories

Abstract

This dataset is a collection of 315 biomass samples compiled from literature sources worldwide. Each sample is characterised by proximate and ultimate analysis variables and the Higher Heating Value (HHV). Proximate properties include ash content (Ash), volatile matter (VM), and fixed carbon (FC). Ultimate properties include elemental carbon (C), hydrogen (H), oxygen (O), nitrogen (N), and sulphur (S). All compositional variables are reported as percentage (wt%) on a dry basis while HHV is presented in MJ/kg. Samples are classified into six feedstock categories following the TNO taxonomy: untreated wood (n=50), grass/plant (n=60), husk/shell/pit (n=64), organic residue/product (n=50), manure (n=41), and straw/stalk/cob/ear (n=50). The column "Material" names the biomass as described in the original source. The column "Reference" provides the bibliographic source from which each observation was extracted. Missing values in nitrogen and sulphur were imputed as zero, reflecting their typically negligible or unreported concentrations in biomass. Imputed values are indicated in bold in the deposited file. No synthetic or generated samples are included. This dataset supports the accompanying manuscript evaluating machine learning models and synthetic data augmentation strategies for biomass HHV prediction.