ABSTRACT
One of the simplest methods for increasing productivity in biogas production is co-digestion. Co-digestion leads to more biogas yield as well as more nutrient bioavailability. This study reports the modeling and optimization of biomethane potential tests from different combinations of organic fraction of municipal solid wastes (OFMSW), cow manure (CM), and municipal sewage sludge (MSS) in solid-state. The simplex-centroid mixture design (SCMD) and artificial neural network (ANN) models were validated with a high correlation to the real data (R2 = 0.99). Experimental results indicated that the maximum amount of CH4 production of 436 mL CH4/g VS occurs at the identical weight percent of the three substances (33.33%). The maximum methane yield was found to be 445.9 mL CH4/g VS as obtained by the genetic algorithm (GA) optimization process, while 448.5 mL CH4/g VS (448.5 ± 3.05) of methane was produced experimentally. Therefore, integration of the SCMD and ANN model with the GA optimization is useful in the prediction of biomethane production. This research is the first in providing the best combination of co-treating and co-utilizing OFMSW with CM and MSS in the solid-state for biowaste management.
Nomenclature
AAD | = | Absolute average deviation MLR Multiple linear regression |
AD | = | Anaerobic digestion MSE Mean square error |
AI | = | Artificial intelligence MSS Municipal sewage sludge |
ANN | = | Artificial neural network OFMSW The organic fraction of municipal solid wastes |
ANOVA | = | Analysis of variance RBF Radial basis function |
BOD | = | Biochemical oxygen demand RMSE Root mean square error |
BPNN | = | Back-propagation neural network RSM Response surface methodology |
CCD | = | Central composite design RTMO Recycling and transformation of materials organization |
CM | = | Cow manure SCMD Simplex-centroid mixture design |
CSTR | = | Continuous stirred-tank reactor SSAD Solid-state anaerobic digestion |
FNN | = | Fuzzy neural networks trainbr Bayesian regularization back-propagation |
GA | = | Genetic algorithm trainlm Levenberg-Marquardt back-propagation |
IF | = | Interaction factor TC Total carbon |
LOF | = | Lack-of-fit TN Total nitrogen |
MAPE | = | Mean absolute percentage error TS Total solids |
MC | = | Moisture content VS Volatile solids |
MLP | = | Multi-layer perceptron R2 Coefficient of determination |
Acknowledgments
The authors would like to appreciate the financial supports provided by the University of Tabriz. We also thank the kindly cooperation of the biogas lab in the Department of Biosystems Engineering, Ferdowsi University of Mashhad.