Optimization of specific methane yield prediction models for biogas crops based on lignocellulosic components using non-linear and crop-specific configurations

• Published in: Industrial crops and products 2018-09, Vol.120, p.330-342
• Main Authors: von Cossel, Moritz, Möhring, Jens, Kiesel, Andreas, Lewandowski, Iris
• Format: Article
• Language: English
• Subjects: Anaerobic digestion; Biogas yield; Crop biomass; Lignocellulose; Meta-analysis; Prediction models
• ISSN: 0926-6690; 1872-633X
• DOI: 10.1016/j.indcrop.2018.04.042

•678 observations taken from 17 studies covering 13 crops were used for meta-analysis.•Seven available models for specific methane yield prediction were less precise than expected.•New models improved the correlation from r = 0.51 up to r = 0.66.•Crop-specific intercepts brought about the largest improvement.•Improvements via crop-specific slopes and two-fold interaction terms were minimal. Basing the prediction of specific methane yield (SMY) of crop biomass on lignocellulosic components has become a promising tool in biogas plant management and bioenergy policies. Most studies on SMY prediction provide linear or non-linear models across crops with lignin content as major regressor variable. To determine the effect of crop-specific regressions, a meta-analysis was conducted using data from 14 published studies (518 observations) and three of the authors' own experiments (160 observations). In total, 678 observations of biomass components and SMY from 13 potential biogas crops were included. The data were used to validate seven published models and both develop and cross-validate new linear and non-linear models with and without crop-specific regressions. Available models showed correlations between r = 0.12 and 0.51. New models reached correlations of up to r = 0.66. Both crop-specific intercepts and slopes as well as non-linear regressions significantly increased model predictability. Of these, crop-specific intercepts brought about the largest improvement but still allowed easy use and interpretability. Therefore, it was shown that the use of biomass source information can help optimize the precision of SMY prediction.