Parametric and Nonparametric Approaches for Detecting the most Important Factors in Biogas Production
More details
Hide details
Panvita d.d., Rakičan, Murska Sobota, Slovenia
Faculty of Agriculture and Life Sciences, Hoče, Slovenia
Peter Vindiš   

University of Maribor, Faculty of agriculture and Life sciences
Submission date: 2017-11-30
Final revision date: 2018-01-15
Acceptance date: 2018-01-24
Online publication date: 2018-08-06
Publication date: 2018-11-20
Pol. J. Environ. Stud. 2019;28(1):291–301
The aim of this paper is to compare results obtained via the well-known regression method ordinary least squares (OLS) and the alternative regression method called multiple model regression estimation (MM-estimation). This is motivated by the fact that exceptional crop yield observations (outliers and leverage points) can cause misleading results if least squares regression is applied. The paper demonstrates that in this case, robust regression is a more appropriate approach, with higher adjusted R-squared value. With both methods, several models have been proposed for predicting the production of biogas where various explanatory variables have been considered, such as the parameters of Weende analysis, C/N ratio, pH value, and the value of volatile fatty acids. Anaerobic digestion was carried out with a basic substrate of pig slurry and with different combinations of co-substrates, where co-substrate maize (main crop), maize (stubble crop), triticale (main crop), sorghum (main crop), a mixture of plants for biomass production (main crop), and grain maize (grain at the wax ripeness stage) were used. To optimize the anaerobic process of fermentation of substrate with co-substrate, the experimental reactor of the Nemščak biogas plant was applied. The average yield of biogas ranged from 384 Nl/kg VS to 635 Nl/kg. The resulting models revealed that crude protein (XP), starch (XS), nitrogen-free extracts (NFE), C/N ratio, volatile fatty acids (VFA), and pH value were the most important predictors affecting biogas production from different substrates. These models are helpful tools in optimising and predicting biogas production from energy crops.