Statistical prediction of biomethane potentials based on the composition of lignocellulosic biomass

Mixture models are introduced as a new and stronger methodology for statistical prediction of biomethane potentials (BPM) from lignocellulosic biomass compared to the linear regression models previously used. A large dataset from literature combined with our own data were analysed using canonical linear and quadratic mixture models. The full model to predict BMP (R²>0.96), including the four biomass components cellulose (x_C), hemicellulose (x_H), lignin (x_L) and residuals (x_R=1-x_C-x_H-x_L) had highly significant regression coefficients. It was possible to reduce the model without substantially affecting the quality of the prediction, as the regression coefficients for x_C, x_H and x_R were not significantly different based on the dataset. The model was extended with an effect of different methods of analysing the biomass constituents content (D_A) which had a significant impact. In conclusion, the best prediction of BMP is pBMP=347x_C+H+R-438x_L+63D_A.