Influence of Molasses Residue on Treatment of Cow Manure in an Anaerobic Filter with Perforated Weed Membrane and a Conventional Reactor: Variations of Organic Loading and a Machine Learning Application

This study highlighted the influence of molasses residue (MR) on the anaerobic treatment of cow manure (CM) at various organic loading and mixing ratios of these two substrates. Further investigation was conducted on a model-fitting comparison between a kinetic study and an artificial neural network (ANN) using biomethane potential (BMP) test data. A continuous stirred tank reactor (CSTR) and an anaerobic filter with a perforated membrane (AF) were fed with similar substrate at the organic loading rates of (OLR) 1 to OLR 7 g/L/day. Following the inhibition signs at OLR 7 (50:50 mixing ratio), 30:70 and 70:30 ratios were applied. Both the CSTR and the AF with the co-digestion substrate (CM + MR) successfully enhanced the performance, where the CSTR resulted in higher biogas production (29 L/d), SMP (1.24 LCH4/gVSadded), and VS removal (>80%) at the optimum OLR 5 g/L/day. Likewise, the AF showed an increment of 69% for biogas production at OLR 4 g/L/day. The modified Gompertz (MG), logistic (LG), and first order (FO) were the applied kinetic models. Meanwhile, two sets of ANN models were developed, using feedforward back propagation. The FO model provided the best fit with Root Mean Square Error (RMSE) (57.204) and correlation coefficient (R2) 0.94035. Moreover, implementing the ANN algorithms resulted in 0.164 and 0.97164 for RMSE and R2, respectively. This reveals that the ANN model exhibited higher predictive accuracy, and was proven as a more robust system to control the performance and to function as a precursor in commercial applications as compared to the kinetic models. The highest projection electrical energy produced from the on-farm scale (OFS) for the AF and the CSTR was 101 kWh and 425 kWh, respectively. This investigation indicates the high potential of MR as the most suitable co-substrate in CM treatment for the enhancement of energy production and the betterment of waste management in a large-scale application.