Response surface methodology and artificial neural network-genetic algorithm for modeling and optimization of bioenergy production from biochar-improved anaerobic digestion

Abstract

Biochar can be used to improve the anaerobic digestion (AD) of agricultural wastes for higher methane production. However, the interaction of biochar addition with other factors of the anaerobic co-digestion (Co-AD) process has rarely been investigated. In this study, process models based on response surface methodology (RSM) and artificial neural network (ANN) were compared in modeling the methane yield (MY, mL CH4/g VS added) from the Co-AD of poultry litter and wheat straw with biochar addition. Box-Behnken design was applied, with the controlling parameters being carbon to nitrogen ratio (C/N), total solids (TS, %), and biochar addition (Biochar, % TS). Numerical optimization and genetic algorithm (GA) were used as optimization tools for RSM and ANN, respectively. A significant second-order quadratic model was built by RSM (R2 = 0.9981 and RMSE = 0.91), which demonstrated significant interactions between C/N and TS (p < 0.0001), and between C/N and Biochar (p < 0.05). The trained ANN (3−3−1) was less accurate (R2 = 0.9926, RMSE = 1.80) compared to RSM. The optimization results by RSM and ANN coupled with GA (ANN-GA) were both validated with prediction errors <0.5%. The optimization results by ANN-GA should be used since it generated a higher maximum MY of 290.7 ± 0.2 mL CH4/g VS added, under the optimal conditions of C/N ratio 24.46, TS 5.03%, and Biochar 8.73% TS, showing an improvement of 20.6% (compared to the control) through process optimization. The methods can also be applied in other scenarios for process modeling and optimization. The optimized results could support real applications of using additives including biochar, active carbon, nanoparticles, etc., to promote the bioenergy production from AD of agricultural wastes.