Predictive Petroleomics: Measurement of the Total Acid Number by Electrospray Fourier Transform Mass Spectrometry and Chemometric Analysis Boniek G. Vaz,* ,†,‡ Patrícia V. Abdelnur, § Werickson F. C. Rocha, ∥ Alexandre O. Gomes, ⊥ and Rosana C. L. Pereira* ,⊥ † Pontifical Catholic of Rio de Janeiro, Rio de Janeiro (RJ) 22451-900, Brazil ‡ Chemistry Institute, Federal University of Goia ́ s, Goiâ nia, (GO) 74001-970, Brazil § Embrapa Agroenergia, Brasília, Distrito Federal (DF) 70770-901, Brazil ∥ National Institute of Metrology, Quality and Technology (Inmetro), Directorate of Industrial and Scientific Metrology (DIMCI), Chemical Metrology Division (DQUIM), , Xere ́ m, Duque de Caxias (RJ), 25250-020, Brazil ⊥ CENPES, Petró leo Brasileiro S.A. (Petrobras), Rio de Janeiro, Rio de Janeiro (RJ) 28999-999, Brazil ABSTRACT: Crude oil samples are uniquely complex because of the number of compounds present that can only be resolved using Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS). The FT-ICR MS technique has been redefined for examining the composition of crude oil and its products, which has led to a new field called “petroleomics”. The chemical composition ultimately determines the chemical and physical properties and the behavior of petroleum and its products. “Petroleomics” predicts the properties and behavior of petroleum using its composition to solve production and processing problems. This paper correlates the chemical composition of crude oil with the total acid number (TAN), which enables the development of prediction models using partial least squares (PLS) and support vector machines (SVMs) as alternative multivariate calibration methods that allow for the application of FT-ICR MS analysis in direct measurements. The prediction models using PLS and SVM demonstrated low prediction errors and superior performance in relation to the univariate method. These results support the development of robust models to predict crude oil properties based on the vast quantity of information provided by FT-ICR MS using PLS and SVM as multivariate calibration procedures. 1. INTRODUCTION Crude oil, which is currently considered the most complex mixture in nature, 1 has challenged the analytical and petrochemical community for decades to unravel its complexity and describe its individual constituents on a molecular level. Such characterization is vital to understanding the functionality and the various properties of crude oil, which require the identification of tens of thousands of chemical components in a typical oil sample. Among the mass spectrometry techniques used for the analysis of crude oil, Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS) is most able to achieve the peak capacity needed to resolve the individual components with a minimal sample preparation. 2 Various ionization methods can be coupled to a FT-ICR mass spectrometer, allowing the user to generate ions from their samples using different methods, each of which has advantages and disadvantages. The use of electrospray ionization (ESI), 3 for example, is widespread; however, this technique only detects strongly polar or ionic species; therefore, it is best suited to studying the acidic or basic components of petroleum. In addition, the high resolution and accuracy provided by FT- ICR MS allows for unique elemental compositions (C c H h N n O o S s ) to be determined. 4 Typically, more than 20 000 chemically distinct components can be detected and analyzed from a single sample. 5 On the basis of these data, the sample can be further characterized according to the distribution of the heteroatom classes or the degree of aromaticity. 6 The ability of FT-ICR MS to evaluate crude oil components has led to the term “petroleomics”, which refers to the principle that the properties and behavior of the organic components of petroleum and its derivatives and products can be correlated (and ultimately predicted) through sufficiently complete characterizations. 7 The petroleomic MS characterization of crude oils has highlighted the compositional trends to elucidate important crude oil properties. A fundamental goal of petroleomics is to link such detailed crude oil compositions to its properties. However, to our knowledge, no systematic studies have yet been published on how to predict crude oil properties using the spectral information obtained from FT- ICR MS. 8 To relate the measured spectra to specific parameters, uni- and multivariate calibration are often used, which are especially useful with parameters that are difficult to measure directly. 9 Many properties of crude oil and its products can only be determined through laborious means using uni- or multivariate Special Issue: 13th International Conference on Petroleum Phase Behavior and Fouling Received: September 17, 2012 Revised: January 23, 2013 Published: January 27, 2013 Article pubs.acs.org/EF © 2013 American Chemical Society 1873 dx.doi.org/10.1021/ef301515y | Energy Fuels 2013, 27, 1873−1880