Optimization of banana peel waste based microbial fuel cells by machine learning.

Banana is a widely cultivated fruit, 114 million tonnes produced worldwide every year. Almost one-third of the weight of banana fruit is made up of banana peel, which is a kind of agro-waste. The weight of banana peel trash generated each year globally is around 39.9 million tonnes. In the present investigation, the microbial fuel cell (MFC) was operated using banana peel slurry as substrate with baker's yeast as external inoculum. Decision tree algorithms were applied for optimizing the input parameters of MFC. Input variables including temperature, pH, resistance, pretreatment of slurry and slurry concentration were optimized for achieving the maximum power density in MFC. Total five combination were obtained by the decision tree model that led to high power density. All five combinations were also tested for validation. Experimental validation of decision tree models showed accuracy in range of 77 to 99%. In order to obtain high power density, the best combination was determined by accuracy level and experimental validation. The best set of rules for high power density was 41.47 mL/L < slurry concentration < 87.5 mL/L, 22.44 °C ≤ temperature < 36.25 °C, 5.23 ≤ pH < 7.25, slurry was pretreated and resistance < 285 Ω. It was also observed that temperature, resistance and pretreatment were the most influential input parameters to achieve high power density. Results obtained in this work can be directly implemented at pilot and industrial scale without further experimental trails. [ABSTRACT FROM AUTHOR]