H2SO4-treated and raw watermelon waste bio-briquettes: Comparative, eco-friendly and machine learning studies

Abstract

As the world faces the challenges of climate change and diminishing fossil fuel reserves, researchers and scientists have increasingly explored the use of plant waste as a viable biofuel source. Among the various plant waste materials, watermelon peel waste, which is typically discarded during processing of beverages or consumption, has gained attention due to its high cellulose content. This study focuses on comparing the properties and environmental sustainability of watermelon peel waste biofuel treated with H2SO4 to raw watermelon peel waste biofuel. From a sustainable energy perspective, maximizing the amount of energy recovered from solid bio-fuel is a critical consideration. Thus, the adaptive neuro-fuzzy inference system (ANFIS) was developed in this study to predict the calorific value (a significant indicator of the energy values) of both H2SO4-treated and raw watermelon waste briquettes while providing useful insights into the effect of the H2SO4 treatment on the energy value of the briquette. The results indicate that the watermelon peel waste treated with acid performed better compared to the untreated sample. It exhibited a higher mean fixed carbon content and calorific value of 22.70 ± 0.16 % and 14.62 ± 0.21 MJ/kg, respectively. Additionally, the treated sample showed a significantly lower mean ash content of 11.20 ± 0.13 %. The EDXRF analysis revealed that the treated samples had reduced proportions of Nickel, Sulfur, and Arsenic, along with a higher carbon count. The FTIR analysis confirmed surface modification in the acid-treated watermelon peel waste (ATWB) through a decrease in lignin content (8.41 ± 0.014 wt%). It also identified a shift in the C-O vibrational stretch from 1021 cm−1 in the untreated watermelon peel waste (UWB) sample to 1028 cm−1 in the ATWB. Moreover, the SEM micrographs displayed a well-packed surface, indicating a reduction in fiber diameter (porosity) in the treated sample. The recorded RMSE, MAD, MAE, and MAPE values of the model are 0.1427, 0.0883, 0.1246 and 1.2313 at the training phases and 0.1581, 0.0997, 0.1484, and 1.9011 at the testing. The regression plot of calorific value with a R2-value of 0.9065, depicts a strong positive linear relationship between the predicted and real values. This outcome of the study suggests an improvement in thermal energy release and enhanced combustibility while the ANFIS model's ability to optimize energy values aligns with the objective of developing eco-friendly bio-briquettes.