Predictive model for the determination of the hydrotreated vegetable oil (HVO) content in HVO/fossil diesel blends using gas chromatography coupled to mass spectrometry and multivariate analysis

Abstract

Low-carbon fuels, emitting less carbon than fossil fuels, are proposed to help in the transition to a sustainable, decarbonized transport sector. The new biofuels being studied and developed in this context include hydrotreated vegetable oils (HVO). Its chemical composition, which is the same as fossil diesel (primarily composed of linear chain hydrocarbons C12–C24), makes HVO (more homogeneous mixtures of paraffinic hydrocarbons C10–C20, containing no sulfur or aromatics) a fuel with slightly lower density than fossil diesel due to these characteristics. Therefore, it can be easily used in place of fossil diesel without the need for modifications to vehicle engines. However, this similarity means that when these fuels are mixed, differentiating and determining their respective concentrations becomes more challenging. ASTM D6866-22 is a standard that describes two tests for determining the carbon-14 (C14) in fossil and renewable fuels; however, both methodologies are expensive and time-consuming. To develop an alternative method to determine HVO in fossil diesel mixtures, a predictive model was created from the chemometric processing of total ion chromatograms obtained from gas chromatography-mass spectrometry (GC-MS). Forty-one mixtures were prepared in triplicate, with concentration levels ranging from 0% to 100% (v/v) HVO in fossil diesel. The samples were divided into training (96 samples) and testing (27 samples) datasets. The raw data obtained from the chromatograms was processed using Orange Data Mining using partial least squares regression. The validation parameters obtained from the predictive model generated by this method had a coefficient of determination of 0.991, mean absolute error of 1.714, and mean squared error of 12.648. The model developed obtained very satisfactory results for the HVO used in this study, however, it is always important to consider HVO's from other sources to strengthen the model.