Biogas production from Udara seeds inoculated with food waste digestate and its optimal output for energy utilities: Central composite design and machine learning approach

Abstract

Anaerobic digestion of abundant feedstock from biomaterials is a good innovative fossil fuel alternative approach for the synthesis of green fuel (biogas). Rotatable central composite design (CCD) and machine learning (ML) via Python coding were successfully used to design, optimize, and predict the rate of biogas production from stew-rice and eggs digestate with Udara seeds in an anaerobic unit. Two-input parameters, such as inoculation ratio (S/I) and hydraulic reaction time (HRT) were considered, resulting in 13 experimental setups under mesophilic surroundings of 25–34°C. Mixture ratios of substrate/inoculum (S/I) of 0.98:1, 1.5:1, 2.75:1, 2.75:1, 4:1, 1.5: 1, and 4.52:1 were used against 30, 20, 44.14, 15.86, 40, 40, and 30 days HRT as modeled by CCD rotatable to optimize biogas production from crushed Udara seeds with spoilt stew-rice and eggs digestate. From the results, it was observed that the coefficient of determination (R²) of 0.9573 was generated via CCD rotatable whereas, the R² of 1 was generated from the multivariate regression of ML approach. Also, the data and graphs derived via ML were superior to the ones derived from CCD rotatable. However, the maximum output of 4.84 L at 4 mixing ratio and 40 days HRT from CCD rotatable is close to the ML value of 4.89 L under the same input factors, yet ML yielded more. Thus, it is clear that the Python-based ML algorithm approach has the potential to predict biogas output better than CCD rotatable. However, the Gas Chromatography Mass Spectrometry analysis of the highest output produced generated 63.29% biomethane and 26.71% CO₂ by volume and produced a flashpoint of −167°C which is flammable. Thus, the generated biogas via an anaerobic unit can be transmitted into large-scale commercial applications for the betterment of mankind.