Appraisal of methane production and anaerobic fermentation kinetics of livestock manures using artificial neural networks and sinusoidal growth functions

Abstract

This study aimed to perform a comparative analysis of the performance of five models (Gompertz, logistic, Richards, the first-order, artificial neural networks) in predicting methane production rate from anaerobic digestion of livestock manures. The input variables were fermentation time, digestion temperature, biogas temperature, ambient temperature, pH, and specific biogas production rate. The physicochemical compositions of cow manure and sheep manure showed that volatile solid (VS) contents were close to each other in manure compositions (77.6% and 64.7%, respectively), while the potential of methane production from cow manure (673.44 mL CH4/g VS) was greater than that from sheep manure (320.32 mL CH4/g VS). The determination coefficients (R-2) for logistic function, Gompertz, Richards, the first-order, and ANN models were obtained as 0.968, 0.967, 0.975, 0.825, and 0.995 for the cow manure, respectively. In case of the sheep manure, the R-2 values obtained from these models were 0.976, 0.979, 0.981, 0.968 and 0.991, respectively. Although the determination coefficients of all models were in satisfactory agreement with the experimental data, the ANN model showed competitive lower RMSE values of 0.111 and 0.164 for cow and sheep manure data sets, respectively, indicating its superior performance than other models.