Data-Driven Assessment Of Edible And Non-Edible Oil Suitability For Sustainable Biodiesel Production Using Machine Learning Classification Models

Abstract

Biodiesel has the potential to make transportation fuels more sustainable than traditional fossil fuels. In recent years alternate sustainable energy sources are gaining popularity because it can reduce emission of greenhouse gas than conventional fuels. Various feed stocks are being explored for bio diesel production including waste cooking oil which proofs to be cost effective implementing waste valourization. Production of good quality bio diesel requires proper selection of raw oil feedstock. The suitability of raw oil for biodiesel production is critically influenced by its physicochemical characteristics. In this research, machine learning techniques such as K- nearest Neighbour (KNN), Support Vector Machine (SVM) and Random Forest (RF) are employed to find the suitability of the waste cooking oil for the production of Biodiesel from the properties of the raw oil. Based on the quality of the produced biodiesel the suitability is classified as not suitable, low, medium and highly suitable category. RF classifier is able to produce the classification accuracy of 99.15 % whereas, KNN and SVM classifier produced 97 % and 95.76% respectively. This research is significant in biodiesel production, which can considerably reduce the time, effort and resources by avoiding unnecessary processing of unsuitable oil. Predicting the suitability in advance supports more efficient, sustainable biodiesel production.