Multiparameter Modeling for 10 Crude Oil Properties Using Comprehensive Two-Dimensional Gas Chromatography and Pixel-Based Chemometrics

This is the authors' abstract. We don't add key points for this paper.

Compositional petroleum analysis may be used to estimate important macroscopic properties of oil, determining its quality parameters. Such parameters are currently determined individually using standard methods. Chemometric models may be used to estimate quality parameters from compositional data as an alternative to otherwise time-consuming and sample-demanding methods. We present the analysis of the 48 crude oils analyzed by comprehensive two-dimensional gas chromatography high-resolution mass spectrometry (GC × GC-HRMS) using a Fourier transform Orbitrap mass analyzer. The premise of this study is that the information-rich data tensors generated by GC×GC may allow the modeling of multiple oil quality parameters since the detailed hydrocarbon composition is directly and indirectly related to the quality parameters. Following data collection, unfolded and aligned GC×GC chromatograms were analyzed by partial least-squares regression (PLS). The PLS models successfully predicted 10 crude oil quality parameters, including American Petroleum Institute (API) gravity, wax content, carbon residue, sulfur and nitrogen contents, and saturate, aromatic, resin, and asphaltene (SARA) fractions. Based on the external validation data set, most quality parameters showed acceptable determination coefficients (R2), ranging from 0.840 to 0.931. Limited performance was observed for low sulfur contents, but it might be related to the intrinsic errors of the reference data used for PLS training. It is expected that increasing the number of observations and improving the model coverage over time should mitigate such findings. We hope to illustrate the potential of GC×GC for modeling macroscopic parameters in the oil and gas industry.