Application of new hybrid models based on artificial neural networks for modeling pyrolysis yields of Atriplex nitens S.

Biochar, bio-oil, and synthesis gas are products obtained from the pyrolysis process, which have alternative usage areas. Biochar is used for adsorption of pollutants, soil conditioner, and as an alternative heating source. Bio-oil is used as a fuel for conventional engines after various upgrading methods, and the synthesis gas is a heat source in energy production. Biochar, bio-oil, and synthesis gas yields obtained from the Atriplex nitens S. plant at the end of the pyrolysis process were modeled using artificial neural networks (ANNs) and hybrid models in this study. Multiple linear regression, ANNs, principal component analysis + multiple linear regression, and principal component analysis + ANN models were used. In addition, 48 different network architectures in ANNs were tested. At the end of the study, the best prediction results of biochar, bio-oil, and synthesis gas were obtained from the ANN 35 (R² = 0.977), ANN 17 (R² = 0.985), and ANN 44 (R² = 0.969) ANN architecture, respectively. The sensitivity analyses were performed using these best models for biochar, bio-oil, and synthesis gas. As a result of the sensitivity analysis, it was determined that the most effective factor in the production of biochar and bio-oil was the gas flow rate, and this was followed by the holding time and pyrolysis temperature. However, the most effective factor in synthesis gas yield was determined as the holding time, and this was followed by gas flow rate and pyrolysis temperature parameters.