Prediction of the cold flow properties of biodiesel using the FAME distribution and Machine learning techniques

Burning fossil fuels is a significant contributor to global warming due to CO2 emissions. To mitigate these emissions, alternative bio-based fuels, such as biodiesel, have been developed. The cold flow properties of biodiesel, including pour point (PP), cold filter plugging point (CFPP), and cloud point (CP), are crucial. Predicting these properties can aid in selecting bio-oils for biodiesel production. Machine learning techniques were utilized to reveal intricate connections between the content of fatty acid methyl esters (FAME) in biodiesel and its cold flow properties. This study created three machine learning models based on a database of over 200 biodiesel samples to predict the aforementioned cold flow properties. The models' performance was assessed using three standard regression metrics: mean absolute error, mean squared error, and coefficient of determination. The experimental results show that the optimal algorithm for PP, CFPP, and CP has an average error of 4.51 degrees C, 3.56 degrees C, and 4.17 degrees C, respectively. The study also investigated the significance of various biodiesel attributes in making precise predictions, revealing that the distribution of FAME and the number of double bonds in the biodiesel are crucial factors for accurate predictions.