A method for predicting the composition information of crude oil samples using a computer.

A prediction model using FT-IR and FT-ICR MS data through machine learning addresses the inefficiencies in crude oil composition prediction, enabling precise and timely analysis for optimized refinery operations and reduced emissions.

JP2026181860APending Publication Date: 2026-11-19CARBON NEUTRAL FUEL TECH CENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2025077609
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2025-05-07
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Current methods for predicting the composition of crude oil samples are time-consuming and imprecise, hindering the optimization of refinery operations and contributing to high CO2 emissions, as they require extensive evaluation and energy-intensive distillation processes.

Method used

A prediction model using FT-IR measurement data as explanatory variables and FT-ICR MS data as objective variables, trained through machine learning, to quickly and accurately predict the component information of crude oil samples.

Benefits of technology

Enables rapid and precise prediction of crude oil sample composition, allowing for optimized refinery operations and reduced CO2 emissions by improving the accuracy of refining processes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2026181860000001_ABST
    Figure 2026181860000001_ABST
Patent Text Reader

Abstract

This invention provides a technical means for rapidly and accurately predicting the component information of crude oil samples using a computer. [Solution] A method for predicting the component information of a target crude oil sample, Step (1): A step of preparing a computer that stores a predictive model of the component information of a crude oil sample, wherein the predictive model is generated by machine learning using training data that includes explanatory variables including measurement data of the crude oil sample by Fourier transform infrared spectroscopy (FT-IR) and a target variable including component information of the crude oil sample classified by compound group by ramping measurement data of crude oil sample fractions by a mass spectrometer using Fourier transform ion cyclotron resonance, and Step (2): Using the computer described above, predict the component information of the target crude oil sample based on the FT-IR measurement data of the target crude oil sample. Methods that include...
Need to check novelty before this filing date? Find Prior Art

Description

[Technical Field]

[0001] This disclosure relates to a method for predicting the composition information of crude oil samples using a computer. More specifically, it relates to a method for predicting the composition information of crude oil samples using a computer, which can be used to appropriately set the operating conditions of petroleum-related equipment. [Background technology]

[0002] Crude oil is believed to have been formed over long periods of time in geological strata by aquatic and terrestrial organisms through thermal decomposition and maturation, and it consists of hundreds of thousands of compounds. Its main component is hydrocarbons, but it also contains heterocompounds including sulfur, nitrogen, and oxygen, as well as trace metal components such as vanadium and nickel. Crude oil is a mixture of compounds with extremely diverse compositions and structures. Furthermore, the relationship between the API (API Baume degree: American Petroleum Institute's Baume degree) and sulfur / metal content (vanadium) of crude oil produced in various countries around the world is not uniform, and the composition of crude oil usually differs depending on the region of origin.

[0003] In crude oil importing countries, including Japan, various types of crude oil are imported from around the world and processed at refineries. However, since the composition of crude oil differs depending on the origin, when purchasing crude oil and creating processing plans at refineries, it is essential to have information that contributes to considering the reactivity of heavy fractions in secondary equipment and the impact on catalytic activity, as well as crude oil assays, in addition to prior evaluation of product yield and product quality in atmospheric distillation units and vacuum distillation units (hereinafter also referred to as "atmospheric distillation units, etc.").

[0004] A crude oil assay refers to the chemical evaluation data of crude oil conducted by a petroleum testing laboratory. Ideally, crude oil assays should be created and updated each time crude oil is received at a refinery, but with current technology, this requires considerable effort and time. The general procedure for obtaining a crude oil assay involves separating the crude oil into various fractions—LPG (Liquefied Petroleum Gas), naphtha, kerosene, diesel fuel, atmospheric pressure residue oil, vacuum diesel fuel, and vacuum pressure residue oil—using batch distillation and vacuum distillation units. Then, general analytical values ​​of the crude oil, the yield of each fraction, and the general properties of each fraction are measured as evaluation data. This evaluation data is then input as constituent data for the crude oil assay, thus creating the assay. The total amount of crude oil required for the entire evaluation is typically several tens of liters, and the evaluation period, including the final product, is several months. Because the creation of a crude oil assay takes a long time to complete, past crude oil assay data is actually used as input data in simulations performed for the optimization control of atmospheric pressure distillation units, etc. (Non-Patent Literature 1, etc.).

[0005] Furthermore, the distillation process in each unit of an oil refinery requires a large amount of energy. Atmospheric distillation units are among the largest emitters of CO2 in the refinery's equipment. Therefore, if the operation of atmospheric distillation units and other equipment can be optimized with greater precision than currently possible, it is expected to make a significant contribution to reducing CO2 emissions at the refinery. Rapid acquisition of high-precision crude oil assay data is essential for understanding the delicate characteristics of crude oil in advance, optimizing the operating conditions of each unit, and preventing problems related to yield, quality, and the environment. [Prior art documents] [Non-patent literature]

[0006] [Non-Patent Document 1] JXTG Technical Review Volume 59, Issue 2 (July 2017) [Overview of the Initiative] [Problems that the invention aims to solve]

[0007] The present inventors have now created a prediction model that predicts component information of a crude oil sample from measurement data of the crude oil sample obtained by Fourier Transform Infrared Spectroscopy (hereinafter also referred to as "FT-IR"). However, in the studies conducted by the present inventors, it has been clarified that crude oil contains more than 100,000 components, it is not realistic to predict all component information quickly and accurately, and even if such prediction were possible, it would be difficult to adjust the operating conditions of each unit in a refinery based on a wide range of component information.

[0008] Accordingly, as a result of further intensive studies, the present inventors used FT-IR measurement data of crude oil samples as explanatory variables, set component information of a plurality of crude oil samples obtained using measurement techniques including a Fourier Transform Ion Cyclotron Resonance Mass Spectrometer (FT-ICR MS) as objective variables, and created a prediction model for component information of crude oil samples by machine learning. As a result, the inventors have found that component information of a crude oil sample can be predicted with high accuracy and speed by the prediction model. The present disclosure is based on such findings.

[0009] An object of the present disclosure is to provide a new technical means capable of predicting component information of a crude oil sample with high accuracy and speed by a computer.

[0010] According to one embodiment of the present disclosure, there is provided a method for predicting component information of a target crude oil sample by a computer, the method comprising: Step (1): a step of preparing a computer storing a prediction model for component information of a crude oil sample, wherein the prediction model comprises explanatory variables including measurement data of the crude oil sample obtained by FT-IR, and objective variables including component information of the crude oil sample classified for each compound group by ramping of measurement data of a crude oil sample fraction obtained by FT-ICR MS a step generated by performing machine learning using training data including, and Step (2): a step of using the computer to predict the component information of the target crude oil sample based on FT-IR measurement data of the target crude oil sample A method is provided, comprising

[0011] According to the present disclosure, it is possible to predict the component information of a crude oil sample with high accuracy and speed by a computer. Since the present disclosure can predict the component information of a crude oil sample with high accuracy and speed, it can be advantageously used to predict the property information of the crude oil sample with high accuracy and speed based on the component information. [BRIEF DESCRIPTION OF THE DRAWINGS]

[0012] [Figure 1] It is a flowchart showing one embodiment of the method for predicting component information of a crude oil sample. [Figure 2] It is a flowchart showing one embodiment of creating a prediction model for component information of a crude oil sample. [Figure 3] It is a flowchart illustrating an example of processing by FT-ICR MS. [Figure 4] It is a diagram illustrating an example of a lamping method for non-heterocyclic molecule groups among single-core molecule groups. [Figure 5] It is a diagram illustrating an example of a lamping method for sulfur atom-containing heterocyclic molecule groups among single-core molecule groups. [Figure 6] It is a diagram illustrating an example of a lamping method for heterocyclic molecule groups containing heteroatoms other than sulfur atoms among single-core molecule groups. [Figure 7] It is a diagram illustrating an example of a lamping method for non-heterocyclic molecule groups among double-core molecule groups. [Figure 8] It is a diagram illustrating an example of a lamping method for sulfur atom-containing heterocyclic molecule groups among double-core molecule groups. [Figure 9]This figure illustrates an example of a ramping method for heterocyclic molecules containing heteroatoms other than sulfur atoms, which are part of the double-core molecule group. [Figure 10] This is a functional block diagram showing one embodiment of a device for predicting the component information of crude oil samples. [Figure 11] This graph shows the relationship between API and sulfur content (mass%) of the crude oil samples (55 types) used in Test Example 1. [Figure 12] This graph shows the relationship between the abundance of n-paraffins with 8 to 30 carbon atoms in crude oil and the cloud point. The white triangles represent the correlation coefficient R2 between n-paraffins and the cloud point, and the white circles represent the number of carbon atoms and their abundance in the diesel fraction. [Figure 13] This graph shows the relationship between the abundance of cyclohexane with 8 to 30 carbon atoms and the total acid value. The white triangles represent the correlation coefficient R2 between normal paraffin and the total acid value, and the white circles represent the number of carbon atoms and abundance in the diesel fraction. [Modes for carrying out the invention]

[0013] The embodiments of this disclosure will be described in detail below with reference to the figures as appropriate. However, this disclosure is not limited to the embodiments described below, and various modifications are possible as long as they do not deviate from the gist of this disclosure.

[0014] In this disclosure, when a numerical range is represented using "~", the numbers on both sides of the "~" are included in that numerical range.

[0015] [Terminology and expressions] In describing embodiments of this disclosure, we will first explain the terms and expressions used herein.

[0016] In this disclosure, "crude oil" refers to unrefined petroleum extracted from an oil field, from which gas, water, and other impurities have been roughly removed. Crude oil reserves can be found in oil fields (underground), oil shale (contained in underground shale), oil sands, lake asphalt, etc.

[0017] Crude oil is a black, viscous liquid, primarily composed of a mixture of hydrocarbons of various molecular weights. It also contains small amounts of compounds including sulfur, oxygen, and nitrogen. The elemental composition of crude oil is, for example, 83-87% by weight of carbon, 11-14% by weight of hydrogen, less than 5% by weight of sulfur, and less than 2% by weight of other elements. The specific gravity of crude oil at 15°C is, for example, 0.75-0.98 g / cm³. 3 The API score is approximately 13-57.

[0018] Furthermore, in this disclosure, "petroleum" is a general term that includes crude oil, as well as the various fractions obtained by distilling crude oil, and fractions obtained by applying secondary treatments such as reforming and cracking to these fractions. Alternatively, it may refer to fractions obtained by further separating a certain fraction obtained by distilling crude oil into components such as saturated hydrocarbons and aromatic hydrocarbons.

[0019] In this disclosure, "component" means "a group of a mixture based on certain physical or chemical properties," that is, "a fraction separated based on certain physical or chemical properties." Methods for grouping based on specific physical or chemical properties include, for example, identifying a boiling point range in a distillation test and separating those within that temperature range as a single component. In this case, the mixture becomes a "collection of fractions." Alternatively, "component" may be understood as each individual component constituting a multi-component mixture, "a collection of molecules recognized as belonging to the same molecular species." Here, "identical" may be understood as "identical in terms of perfectly identified molecular structure," or "isomers in molecular structure (same molecular formula but different structure) are considered identical," or, for example, "identical in the structure identified by a method such as JACD, as described later." Furthermore, it may be broadly understood as "a collection of molecules grouped together based on arbitrarily defined criteria."

[0020] In this disclosure, "constituting" does not necessarily mean all 100% of the components present in a multi-component mixture. Depending on how the molecular structures of each component identified in this disclosure are used, the level of detail required for identifying the molecular species as a component may be appropriately determined. For example, only molecular species present in a certain amount (proportion) or more in a multi-component mixture may be considered as "constituting components." It is not always necessary to identify the molecular structures of all of the vast number of molecular species, such as petroleum, and molecular species present in trace amounts may be ignored as needed. For example, when considering polycyclic aromatic resin (PA) as a "multi-component mixture," the presence of paraffinic compounds and olefinic compounds as components constituting PA may be ignored.

[0021] In this disclosure, "fraction" refers to any value indicating the proportion of existence, such as mass fraction, volume fraction, or mole fraction, and is a concept that includes all of these. When calculating the average Hansen solubility index value for the entire liquid phase, volume fraction is preferably used, and it is calculated as a weighted average value weighted by the volume fraction of each component in the liquid phase.

[0022] In this disclosure, "identifying the molecular structure" encompasses any act of identifying any information regarding the structure of a molecule in the "component" described above. The degree and method of representation may be appropriately selected depending on the purpose and necessity. It is not limited to identifying the structure of the entire molecule; information regarding the structure of a part of the molecule may also be incorporated. For example, the structure of only the core portion may be identified, while the structures of the side chains and cross-linking portions may be left as molecular formulas without identification. In this disclosure, preferably, the molecular structure of a molecule identified by "JACD" as described later is a concept that includes all isomers resulting from differences in the binding positions of the attributes described later. In this disclosure, the term "molecule" may be understood as a concept that includes all isomers.

[0023] In this disclosure, "identifying the proportion of each component" encompasses any action that identifies the ratio in which each component constituting the mixture exists. Furthermore, this does not mean that the proportion of every component constituting the mixture must be identified. It is not the case that "the proportion of each component has been identified" only after identifying the proportion of every component, including components present in amounts too small to be detected by analytical techniques or components that do not need to be identified. Such trace components may be treated together as "other components." Moreover, these may be excluded from the scope of "each component constituting the mixture" and may not be included in the denominator when calculating the proportion of other components.

[0024] In this disclosure, "all" does not necessarily mean "100% of all." For example, where the phrase "all peaks" is used with respect to the mass spectrum, it may be interpreted not only as literally meaning "100% of all peaks," but also as referring to the peaks remaining after appropriately excluding, for example, peaks related to molecules that are not necessarily necessary for the purpose of the study in that context, or peaks that are difficult to distinguish.

[0025] In this disclosure, the horizontal axis of the “peak” obtained in mass spectrometry represents the m / z for the molecular ion or pseudomolecular ion of each component constituting the multicomponent mixture. Since the numerical value of this m / z corresponds to the mass of the molecular ion or pseudomolecular ion, it generally represents the molecular weight of the molecule attributed to that peak. In this disclosure, this “m / z peak for molecular ions or pseudomolecular ions obtained in mass spectrometry” may be referred to as the “peak obtained in mass spectrometry” or simply as the “peak.” Furthermore, the height of the peak indicates the relative abundance of the molecule attributed to that peak.

[0026] In this disclosure, "molecular formula" refers to a formula that shows only the types and number of elements (C, H, N, O, S, etc.) that make up a molecule, and does not specify the structure. Since the types and number of elements that make up the molecule are known, information such as molecular weight and the DBE value described later can be obtained.

[0027] In this disclosure, "JACD" (Juxtaposed Attributes for Chemical-structure Description) refers to a novel labeling system for molecular structure that indicates the structure of a molecule by the type and number of attributes. It does not indicate the position in which an attribute is bound to another attribute. "Attributes" refer to the chemical structural components (parts) that make up a molecule. In aromatic compounds, this specifically refers to the aforementioned "core," "crosslinks," and "side chains." This labeling method allows for the identification of the structure of each of the vast number of molecules that make up petroleum, to a necessary and sufficient degree. A molecule represented in JACD format, with its structure determined, is a concept that includes all isomers resulting from differences in attribute binding positions.

[0028] Using saturated hydrocarbon compounds as an example, the structural information (JACD, 6 digits × 8 = 48 digits of alphanumeric characters) and quantity information (mole fraction) in JACD format are shown below. [ka]

[0029] <Method for predicting the component information of a target crude oil sample using a computer> According to one embodiment of this disclosure, a method for predicting the composition information of a target crude oil sample using a computer is: Step (1): A step of preparing a computer that stores a predictive model of the component information of a crude oil sample, The above prediction model includes explanatory variables, including measurement data of crude oil samples by FT-IR, Ramping of measurement data from crude oil sample fractions obtained by FT-ICR MS allows for the determination of a target variable containing component information of crude oil samples classified by compound group. Steps generated by machine learning using training data including, Step (2): Using the computer described above, predict the component information of the target crude oil sample based on the FT-IR measurement data of the target crude oil sample. Includes.

[0030] One embodiment of this disclosure is shown in the flowchart of Figure 1. Hereinafter, one embodiment of this disclosure will be described step by step.

[0031] [Step (1): Prepare a computer that stores a predictive model of the characteristics of the target crude oil sample.] According to one embodiment of the present disclosure, step (1) is to prepare a computer that stores a predictive model of the characteristics of the target crude oil sample.

[0032] The crude oil sample may be crude oil, a fraction of crude oil, or a mixture thereof, but preferably it is crude oil or a fraction of crude oil. Furthermore, the crude oil and the fraction of crude oil may each be one type or a mixture of two or more types, but from the viewpoint of predicting more accurate characteristic information, it is preferable to have one type.

[0033] Crude oils used as crude oil samples include, but are not limited to, Arabian Extra Light (Saudi Arabia), Arabian Light (Saudi Arabia), Maban (UAE), Das (UAE), Kuwait (Kuwait), Arabian Heavy (Saudi Arabia), Sokol (Russia), Oman (Oman), Banoko Arab Medium (Bahrain), Espo Blend (Russia), WTI Midland (United States), Arabian Super Light (Saudi Arabia), Arabian Medium (Saudi Arabia), Napo (Ecuador), Bazla Light (Iraq) and Upper Zakum (UAE), Qatar Marine (Qatar), and others.

[0034] Furthermore, while there are no particular limitations on the crude oil sample fractions used as crude oil samples, examples include fractions having a boiling point below the boiling point of diesel fuel (360°C) in accordance with JIS K2254. Specific examples include naphtha (boiling point: approximately 30 to 150°C), such as light naphtha (boiling point: approximately 35 to 80°C) and heavy naphtha (boiling point: approximately 80 to 150°C), as well as kerosene (boiling point: approximately 150 to 250°C) and diesel fuel (boiling point: approximately 250 to 360°C), among other kerosene fractions (boiling point: approximately 150 to 360°C).

[0035] Furthermore, the crude oil sample fraction may be the residual oil from distillation. Examples of such residual oil include fractions with a boiling point higher than the boiling point of diesel fuel (360°C) according to JIS K2254, and specific examples include atmospheric residue (AR), vacuum gas oil (VGO), and vacuum residue (VR).

[0036] The crude oil sample fraction may be a sample after PONA analysis, two-dimensional gas chromatography analysis, type-specific fractionation, TLC (thin-layer chromatography), or a sample obtained by mass spectrometry using Fourier transform ion cyclotron resonance.

[0037] In PONA analysis, a gas chromatograph using a high-resolution capillary column is used to classify and quantify each component in the distillate obtained from atmospheric distillation according to its carbon number and structure into types such as n-paraffins, isoparaffins, olefins, naphthenes, and aromatics. According to PONA analysis, the composition of the distillate obtained from atmospheric distillation can be calculated by measuring the volume concentration, weight concentration, and molar ratio of each component, and it is preferably applied to fractions lighter than naphtha. PONA analysis can be performed in accordance with the "Japanese Industrial Standard, Method for Analyzing Total Composition by Gas Chromatography (JIS K-2536)".

[0038] Two-dimensional gas chromatography (GC×GC) is a well-known analytical method and can be performed in accordance with the description in the Journal of Chromatography A 892(2000)29, as exemplified by Schoenmkers et al. Two-dimensional gas chromatography allows for the separation of components that could not be separated using one-dimensional chromatography due to their similar boiling points, based on differences in polarity, etc.

[0039] Furthermore, when crude oil sample fractions are fractionated into multiple types, they are preferably fractionated into three or more fractions, more preferably into seven fractions. These seven fractions include asphaltenes (As), saturated components (Sa), mono-ring aromatics (1A), bi-ring aromatics (2A), three-ring or more aromatics (3A+), polar resins (Po), and polycyclic aromatic resins (PA). When separating each fraction, there is no particular criterion for the boundary between fractions or a method for fractionation, and known methods such as column chromatography fractionation, Soxhlet extraction, or fast solvent extraction may be used. In this specification, the seven-fractionation method can be performed by applying a column chromatography fractionation method as described in Japanese Patent Publication No. 5508837.

[0040] Furthermore, the separation method using TLC (thin-layer chromatography) can be carried out in accordance with the simplified method described in the Japan Petroleum Institute standard JPI-5S-70-10.

[0041] The following describes the flow steps S1 to S3 for creating a predictive model of crude oil sample component information in one embodiment of this disclosure, with reference to Figure 2. Such a predictive model can preferably be created using machine learning with a computer.

[0042] [Step S1] In one embodiment of this disclosure, FT-IR measurement data (explanatory variables) of multiple crude oil samples are obtained as training data for creating a predictive model. Here, FT-IR is a type of infrared spectroscopy that performs structural analysis and quantification of a sample by irradiating a substance with infrared light and measuring the transmitted or reflected light. The data obtained by FT-IR reflects various information, including the type and amount of functional groups, and can be measured quickly. Therefore, it is considered advantageous for improving the accuracy and analysis speed of the predictive model by applying the rapid computational processing power of a computer, machine learning methods, and multivariate analysis.

[0043] When measuring a sample using FT-IR, infrared light emitted from the light source passes through a Michelson interferometer along its optical path. The Michelson interferometer consists of a semi-transparent mirror (beam splitter), a moving mirror, and a solid mirror. The light, split into two by the semi-transparent mirror, is reflected by the moving and fixed mirrors and then recombined by the semi-transparent mirror. As the moving mirror reciprocates, a path difference is created between it and the fixed mirror, resulting in the Michelson interferometer producing interference light from the two beams with varying phase differences over time. The intensity of this interference light is recorded with the path difference on the horizontal axis to obtain an interferogram. The data directly acquired by FT-IR is this interferogram, recording the infrared light transmitted through the sample. By performing a Fourier transform on this interferogram, the horizontal axis represents the wavenumber (cm²). -1 An infrared absorption spectrum is obtained with the following characteristics.

[0044] Since FT-IR measurements are single-beam measurements, an FT-IR spectrophotometer cannot directly measure the transmittance (%) of a sample. Therefore, a background measurement is first performed without the sample, and then the sample measurement is performed. The transmittance of the sample is obtained by calculating the ratio of both the background and the sample based on the following formula. IR spectrum = -log(samples / background)

[0045] In FT-IR, it is preferable to set a wide wavelength range from the viewpoint of rapidly predicting the component information of many crude oil samples by using measurement data from numerous FT-IRs as explanatory variables. The specific wavelength range for FT-IR is usually 350 to 7,500 cm². -1 And preferably 500 to 7,000 cm -1 And more preferably 500~6,500cm -1 That is the case.

[0046] FT-IR measurement methods include transmission, total internal reflection, diffuse reflection, normal incidence transmission, and micro-infrared, but from the viewpoint of simple and rapid measurement, transmission or total internal reflection is preferred.

[0047] The transmission method involves irradiating a sample with infrared light and detecting the transmitted light. Because the transmission method directly detects the transmitted light after it passes through the sample, it is advantageous for obtaining high-intensity data.

[0048] The total internal reflection method is also called the attenuated total internal reflection method (ATR method; hereinafter simply referred to as "ATR"). This measurement method utilizes the principle that when light (in this case, infrared light) is incident from a medium with a high refractive index (first medium) to a medium with a low refractive index (second medium), total internal reflection occurs depending on the angle of incidence, and the light passing through the evanescent field generated along the interface between the first and second media is absorbed by the material present in the evanescent field. By incidenting infrared light from a high refractive index ATR prism onto a material to be measured with a low refractive index, and causing total internal reflection of the infrared light by the material to be measured, the infrared absorption spectrum obtained according to the material of the material to be measured is measured. The total internal reflection method allows for easy and repeated measurements and is advantageous for measuring absorption spectra accurately over a wide wavelength range in a short time.

[0049] FT-IR measurements can be performed in accordance with the conditions described in the examples below.

[0050] [Step S2] Furthermore, according to one embodiment of this disclosure, as training data for creating a predictive model, component information (target variable) of crude oil samples classified by compound group is obtained by ramping measurement data of crude oil sample fractions by FT-ICR MS. Using ramped component information obtained by FT-ICR MS as the target variable is advantageous in significantly improving the prediction accuracy of the component information of the target crude oil sample.

[0051] The crude oil sample fraction used for FT-ICR MS measurement in step S2 is not particularly limited, but atmospheric residue is preferred.

[0052] FT-ICR MS measurement of crude oil sample fractions yields a mass spectrum with m / z on the x-axis and peak intensity on the y-axis. From the mass spectrum, the structure (molecular formula) and quantity (mole fraction) of the compound corresponding to each peak can be obtained.

[0053] (Processed by FT-ICR MS) Figure 3 is a flowchart illustrating an example of processing by FT-ICR MS in step (S2). After separation pretreatment (S11), the crude oil sample is subjected to mass spectrometry by FT-ICR MS (S12), and structural and quantitative information is assigned to the compounds corresponding to the peaks in the mass spectrometry (S13).

[0054] (Step S11) Crude oil samples contain a diverse range of molecular species. The ionization efficiencies of these molecular species in mass spectrometry vary considerably, and without pretreatment, this can lead to overestimation or underestimation of specific components. In the FT-ICR MS measurements described herein, it is preferable to pretreat the crude oil sample based on molecular polarity and structural information (molecular size, number of aromatic rings). Pretreatment (S11) is performed, for example, based on polarity and structural information (molecular size, number of aromatic rings). This pretreatment method can be carried out by the seven-fractionation method described above.

[0055] (Step S12) Molecular-level analysis can be performed on crude oil sample fractions (or fractionated crude oil samples if pre-treatment has been performed) using FT-ICR MS.

[0056] In FT-ICR MS analysis, the sample is ionized by an ionization method appropriately selected from methods such as atmospheric pressure photoionization (APPI), electrospray ionization (ESI), and laser desorption ionization (LDI), depending on the sample. The APPI method can efficiently ionize nonpolar and aromatic molecules, while the ESI method can ionize mainly polar molecules. The LDI method has high sensitivity to condensed aromatic rings and nitrogen-containing compounds. Furthermore, silver ions (Ag + By using ) as a supporting substance, it can also be applied to the detection of saturated components that cannot be ionized by the AIPP method and the ESI method.

[0057] FT-ICR MS can determine m / z up to four decimal places, and by performing number matching of the exact mass of each element( 12 when C is exactly 12, the monoisotopic mass up to six decimal places is 1 H is 1.007825, 32 S is 31.972071.), the molecular formula of a compound can be determined almost uniquely.

[0058] Once the molecular formula is determined, it is possible to calculate the DBE (Double Bond Equivalence) value, which is an index of the degree of unsaturation. If the molecular formula is C c H h N n O o S s (where c, h, n, o, s respectively represent the number of carbon atoms, the number of hydrogen atoms, the number of nitrogen atoms, the number of oxygen atoms, and the number of sulfur atoms), the DBE value can be calculated by formula (1).

[0059]

Formula

[0060] As can be seen from formula (1), even if S is contained in the molecular formula, it does not affect the DBE value. The DBE value generally indicates the degree of unsaturation in a molecule, particularly the extent of the presence of double bonds and rings.

[0061] By performing analysis with FT-ICR MS, a molecular formula can be assigned almost uniquely to the detected component, but since a plurality of isomers can correspond to one molecular formula, more in-depth analysis is required to obtain specific structural information.

[0062] Furthermore, due to the characteristics of FT-ICR MS, even if the molecular formula is the same, the mass may differ from the original molecular ion due to, for example, the addition of hydrogen ions to the molecular ion, and therefore it may appear as a separate peak. Thus, even if they appear as separate peaks in the measurement, if the types and number of elements constituting the molecular formula are the same, they can be considered as the "same molecular formula." In the phrase "the molecule corresponding to that molecular formula," "that molecular formula" can be understood to mean this "same molecular formula." Also, when referring to "a certain peak," it can be considered a concept that encompasses all the various m / z peaks that were considered to represent the "same molecular formula" in the sense described above.

[0063] In this disclosure, it is preferable to perform collision-induced dissociation (CID) at low energy inside the FT-ICR MS instrument and observe the structural blocks (cores) constituting the petroleum molecule as fragment ions. CID cleaves the side chains and crosslinks of the molecule, reducing the number of carbon atoms per compound and decreasing the DBE value. The conditions for performing CID (Collision Injection) include a collision energy that can effectively cleave crosslinks and side chains in the molecule, preferably 10-50 kcal / mol, and more preferably 20-40 kcal / mol. Note that 40 kcal / mol corresponds to 32 eV when the molecular weight is 700.

[0064] By using FT-ICR MS, the resulting molecular composition is based on actual molecular formulas and their abundances, resulting in a more accurate molecular composition. In this disclosure, it is preferable to assign a novel chemical formula JACD (Juxtaposed Attributes for Chemical-structure Description) to the peak (parent molecule) before decomposition by CID, based on the data before and after CID using FT-ICR MS. In this case, it is preferable to use analytical information such as fractionation yield as input in addition to the FT-ICR MS data.

[0065] Identification of core structure After CID, the observed fragment ions are compared and matched against a pre-defined list of core structures, and the observed peaks are assigned to core structures. The list of core structures is preferably created by considering the core structure as divided into three parts (benzene ring, naphthene ring, and heterocycle) and comprehensively combining them, as shown in the formula below.

[0066] [ka]

[0067] One method for identifying the core structure after CID is to first create a "core structure list" containing various cores that can be assumed to constitute each component molecule of the other component mixture as models. Then, the molecular weight, type and number of heteroatoms, and other information of the cores stored in this list are compared with the core information obtained above, and the most appropriate core model is selected from this list and assigned as the core in question. This method allows for the assignment of a core to every peak obtained from FT-ICR MS analysis after collision-induced dissociation, making it possible to determine its structure. The types of cores to be included in the above core structure list are not particularly limited and can be any type; however, the appropriateness of the selection of cores to be included will directly affect the appropriateness of the structure of each core.

[0068] In this disclosure, "core" is a type of "attribute" in JACD, specifically, one selected from the group consisting of benzene rings, naphthene rings, and heterocycles, or two or more selected from the group consisting of benzene rings, naphthene rings, and heterocycles with overlapping properties, that are directly bonded rather than being cross-linked. Since cross-linking and side chains are attributes separate from "core," "core" means a component that does not have any cross-linking or side chains.

[0069] On the other hand, "single-core" refers to a molecule that has only one of the above-mentioned cores. As it is a concept referring to a molecule, it also includes molecules in which side chains are attached to the core. A molecule in which two or more of the above-mentioned cores are cross-linked is called a "multi-core." As "multi-core" also refers to a molecule, it also includes molecules in which side chains are attached to the core. A molecule in which two cores are cross-linked is called a "double-core." For example, the naphthalene molecule is a "single-core" molecule and not a "double-core" molecule because it has only one naphthalene ring in which two benzene rings are directly bonded rather than cross-linked. Note that a core containing a heteroatom is also called a "heterocore."

[0070] Core structure combination estimation The number of cores that make up a molecule is not always one (single core). If the degree of unsaturation after CID decreases compared to before CID, it means that the cross-linking sites connecting the cores have been cleaved. The information obtained from core structure analysis (type and quantity of core structure) alone is insufficient to estimate specific core combinations. It is preferable to consider both single-core and multi-core molecular structures, and to assume that heteroatoms are contained only within the core structure. The abundance of multi-cores is preferably the product of the abundance of each core that constitutes them. Furthermore, since the abundance of single-cores decreases as the degree of unsaturation of the parent molecule increases, it is preferable to add a factor that increases or decreases in accordance with the increase in the degree of unsaturation of the parent molecule.

[0071] Estimation of side chains and bridges and structural information of JACD type For side chains and crosslinks, the difference between the molecular formula of the molecule and the molecular formula of the core (estimated by combination estimation of the molecular formula and the core) is used as the total amount of side chains and crosslinks, and the amounts of side chains and crosslinks are allocated according to the allocation rules. Combined with the identification and combination estimation of the core structure, it becomes possible to assign JACD-style structural information to each peak observed before CID. In the JACD format, the core (core1, core2, core3), crosslinks (crosslink1, crosslink2), and side chains (side chain1, side chain2, side chain3) are each represented by a 6-digit ID, and the compound is represented by a 48-digit ID. For example, the JACD format structural information for a compound represented by the following structural formula is as follows:

[0072] [ka]

[0073] Using an ID in JACD format, it is possible to quickly search for physical property information associated with the molecular structure in the complete petroleum molecular database (ComCat: Compounds Catalogue). ComCat contains registered information on over 25 million compounds that may be found in petroleum, along with their JACD (Japan Agricultural Code), representative structural formulas, and more than 20 different physical properties.

[0074] Optimization of molecular composition The sensitivity of individual peaks in mass spectrometry is not uniform and depends on the structural type and molecular weight of the molecule. While the problem of sensitivity differences due to type can be largely resolved with high-precision separation by type, correction may be necessary for sensitivity differences due to molecular weight. In fractionated crude oil samples, the most straightforward physical property that expresses its characteristics is the hydrogen / carbon (H / C) atomic ratio, and if this value is the same, other physical properties will also be approximately the same. Therefore, the correction of the molecular weight distribution is performed using the H / C atomic ratio as a parameter. If the H / C atomic ratio matches the estimated value from the mass spectrometry results and the results of the elemental analysis of the sample, the peak sensitivity in mass spectrometry can be considered to be the same. However, if there is a significant difference in this value, it can be said that the peak sensitivity differed in mass spectrometry, that is, there was a decrease in sensitivity in the high molecular weight region. The molecular weight at which this decrease in sensitivity begins and the degree of the decrease in sensitivity differ depending on the sample and analytical conditions. For example, in the APPI method, a decrease in sensitivity may occur at molecular weights of approximately 600 or higher. Therefore, the H / C atomic ratio is corrected by multiplying the peak with decreased sensitivity by a set factor to correct the peak sensitivity.

[0075] Correction coefficient for peak i (f i ) is the molecular weight (M) at which sensitivity begins to decrease. c Below this point, equation (2) is used, and for molecular weights exceeding this, the decrease in sensitivity is corrected by equation (3).

[0076]

number

[0077] In equation (2), α is f i This is the rate of change with respect to molecular weight. From this, the abundance (mole fraction) of peak i obtained by mass spectrometry is given by I i Therefore, the number of carbon atoms in each peak (C i ), number of hydrogen atoms (H i Since the total number of carbon atoms in the sample (C) is known, m ) and number of hydrogens (H m ) is calculated using equations (4) and (5).

[0078]

number

[0079] Next, let δ be the difference from the results obtained by elemental analysis, and set α and M so that equation (6) is minimized. i To decide.

[0080]

number

[0081] The final result is obtained by correcting the sum of the abundances of each peak calculated in this way so that it equals 100%. This correction ensures that the abundance in the high molecular weight region is more in line with the actual sample.

[0082] By performing the above process, it is possible to calculate molecular structure information and quantity information of compounds contained in the crude oil sample based on measurement data from the crude oil sample obtained by FT-ICR MS.

[0083] The group of compounds contained in the crude oil sample obtained in step (1) described above is classified by ramping. The ramping of compounds contained in crude oil samples is preferably classified into at least acyclic saturated aliphatic molecules, single-core molecules, and double-core molecules.

[0084] Acyclic saturated aliphatic molecule group The above acyclic saturated aliphatic molecules are linear or branched molecular formula C n H 2n+2 This is a group of molecules consisting of compounds represented by (where n is an integer greater than or equal to 1).

[0085] Single-core molecule group The single-core molecule group described above consists of molecules that have only one "core." A "core" is a type of JACD attribute, and specifically refers to an aromatic ring, naphthenic ring, or heterocycle itself, or a fused ring in which two or more rings, selected with overlap from the group consisting of aromatic rings, naphthenic rings, and heterocycles, are directly bonded rather than being bridged. The above-mentioned aromatic rings are aromatic hydrocarbon rings, and non-limiting examples of aromatic rings include benzene rings and naphthalene rings. Note that bridges and side chains are attributes separate from the "core," and therefore the "core" does not include bridges or side chains.

[0086] The above single-core molecule group preferably includes at least one selected from the group consisting of non-heterocyclic molecules, sulfur atom-containing heterocyclic molecules, and heterocyclic molecules containing heteroatoms other than sulfur atoms.

[0087] The above-mentioned group of non-heterocyclic molecules consists of molecules having exactly one core, which is made up of at least one hydrocarbon ring selected with overlap from the group consisting of naphthene rings and benzene rings. When the core consists of two or more hydrocarbon rings, the hydrocarbon rings are directly bonded to each other, rather than being cross-linked.

[0088] A non-restrictive example of a naphthenic ring is shown below. [ka]

[0089] The above group of sulfur atom-containing heterocyclic molecules consists of molecules having only one core containing a heterocyclic ring (excluding rings containing heterocyclic rings other than sulfur atoms; hereinafter also referred to as "sulfur atom-containing heterocyclic ring") which contains at least one sulfur atom as a heteroatom. The above core may also contain a hydrocarbon ring in addition to the sulfur atom-containing heterocyclic ring. The above hydrocarbon ring is preferably selected from the group consisting of naphthene rings and benzene rings, with some overlap allowed.

[0090] Non-restrictive examples of sulfur atom-containing heterocycles are shown below. [ka]

[0091] The group of heterocyclic molecules containing heteroatoms other than sulfur atoms is a group consisting of molecules having only one core containing a heterocyclic ring with heteroatoms other than sulfur atoms. The core may also contain rings selected with overlap from sulfur atom-containing heterocyclic rings and hydrocarbon rings, in addition to the heterocyclic ring containing heteroatoms other than sulfur atoms. The hydrocarbon rings are preferably selected with overlap from the group consisting of naphthene rings and benzene rings.

[0092] Examples of heterocycles containing heteroatoms other than the sulfur atom mentioned above include heterocycles containing only nitrogen atoms as heteroatoms, heterocycles containing only oxygen atoms as heteroatoms, heterocycles containing both nitrogen and oxygen atoms as heteroatoms, and heterocycles containing both sulfur and nitrogen atoms as heteroatoms.

[0093] The following are non-restrictive examples of heterocycles containing heteroatoms other than sulfur atoms. [ka]

[0094] Double-core molecular group The double-core molecule group described above consists of molecules that have only two "cores." The "cores" are as described in the single-core molecule group. The compounds included in the double-core molecule group have "crosslinks" that connect the "cores" to each other. The "cores" in the double-core molecule group can be a combination of two cores containing only hydrocarbon rings, a combination of one core containing only hydrocarbon rings and one core containing a heterocycle, or a combination of two cores containing heterocycles. Heterocycles further include sulfur-containing heterocycles, which contain only sulfur atoms as heterocycles, and heterocycles containing heterocycles other than sulfur atoms. Among the double-core molecule group, those in which both cores contain only hydrocarbon rings are called the non-heterocyclic molecule group, those in which the two cores contain a sulfur atom-containing heterocyclic ring and consist of one core containing only hydrocarbon rings and one core containing a sulfur atom-containing heterocyclic ring and one core containing only hydrocarbon rings, or those in which the two cores contain a sulfur atom-containing heterocyclic ring and do not contain heterocyclic rings containing sulfur atom-containing heterocyclic molecule group.

[0095] (Ramping of compounds) An example of a method for ramping compounds will be described. Based on the molecular structure information obtained in step (1), the compounds are classified into acyclic saturated aliphatic molecules, single-core molecules, and double-core molecules. The acyclic saturated aliphatic molecules, single-core molecules, and double-core molecules are as described above.

[0096] Ramping of acyclic saturated aliphatic molecules Acyclic saturated aliphatic molecules are ramped based on the number of carbon atoms in the molecule. The number of components to ramp the acyclic saturated aliphatic molecules can be set as appropriate. For example, they can be ramped to around 10-100 components, 30-80 components, or around 50 components.

[0097] Ramping of single-core molecules The single-core molecule group is further classified into non-heterocyclic molecules, sulfur atom-containing heterocyclic molecules, and heterocyclic molecules containing heteroatoms other than sulfur atoms.

[0098] Ramping of nonheterocyclic molecules Of the single-core molecules described above, the non-heterocyclic molecules are preferably divided according to the composition and number of rings constituting the core, and further ramped by the number of side chains. The number of components to be ramped for the non-heterocyclic molecule group can be set as appropriate. For example, the entire single-core molecule group can be ramped to approximately 50-500 components, 100-400 components, 100-300 components, or 150 components.

[0099] In the example shown in Figure 4, the non-heterocyclic molecules are ramped as described below. Among the group of molecules whose core contains a total of one benzene ring and one naphthene ring, the group of molecules containing one benzene ring is ramped into groups based on the number of side chains: 0-4, 5-9, 10-14, 15-19, 20-24, 25-29, 30-39, 40-49, and 50 or more. The group of molecules containing one naphthene ring is ramped into groups based on the number of side chains: 0-4, 5-9, 10-14, 15-19, 20-24, 25-29, 30-39, 40-49, and 50 or more. Among the group of molecules whose core contains a total of two benzene rings and naphthene rings, the group containing two benzene rings is ramped into groups based on the number of side chains: 0-4, 5-9, 10-14, 15-19, 20-24, 25-29, 30-39, 40-49, and 50 or more. The group of molecules containing 0-1 benzene rings and 1-2 naphthene rings (where the total number of benzene and naphthene rings is 2) is ramped into groups based on the number of side chains: 0-4, 5-9, 10-14, 15-19, 20-24, 25-29, 30-39, 40-49, and 50 or more. Among the group of molecules whose core contains a total of three benzene rings and naphthene rings, the group containing three benzene rings is ramped into three groups based on the number of side chains: 0-19, 20-39, and 40 or more. The group containing 0-2 benzene rings and 1-3 naphthene rings (where the total number of benzene and naphthene rings is 3) is also ramped into three groups based on the number of side chains: 0-19, 20-39, and 40 or more. Among the group of molecules whose core contains a total of four benzene rings and naphthene rings, the group containing four benzene rings is ramped into three groups based on the number of side chains: 0-19, 20-39, and 40 or more. The group containing 0-3 benzene rings and 1-4 naphthene rings (where the total number of benzene and naphthene rings is 4) is also ramped into three groups based on the number of side chains: 0-19, 20-39, and 40 or more. For molecules whose core contains a total of 5 to 11 benzene and naphthene rings, ramping is performed based on the total number of rings. The number of side chains is not considered.

[0100] Ramping of sulfur atom-containing heterocyclic molecules Of the single-core molecule group described above, the sulfur atom-containing heterocyclic molecule group is preferably divided according to the composition and number of rings constituting the core, and further ramped by the number of side chains. The number of components to be ramped among the sulfur atom-containing heterocyclic molecules can be set as appropriate. For example, the entire single-core molecule group can be ramped to approximately 50-500 components, 100-400 components, 100-300 components, or 150 components.

[0101] In the example shown in Figure 5, the group of sulfur atom-containing heterocyclic molecules is ramped as described below. The group of molecules whose core contains one sulfur atom-containing heterocycle is ramped into the following groups based on the number of side chains: 0-4, 5-9, 10-14, 15-19, 20-24, 25-29, 30-39, 40-49, and 50 or more. For molecules whose core contains 1-2 sulfur atom-containing heterocycles and 0-1 hydrocarbon rings (benzene rings and naphthene rings) (the total number of sulfur atom-containing heterocycles and hydrocarbon rings is 2), we ramp them into groups based on the number of side chains: 0-4, 5-9, 10-14, 15-19, 20-24, 25-29, 30-39, 40-49, and 50 or more. For molecules whose core contains 1 to 3 sulfur-containing heterocycles and 0 to 2 hydrocarbon rings (benzene rings and naphthene rings) (the total number of sulfur-containing heterocycles and hydrocarbon rings is 3), we ramp them into groups based on the number of side chains: 0 to 19, 20 to 39, and 40 or more. For molecules whose core contains 1 to 4 sulfur-containing heterocycles and 0 to 3 hydrocarbon rings (benzene rings and naphthene rings) (the total number of sulfur-containing heterocycles and hydrocarbon rings is 4), we ramp them into groups based on the number of side chains: 0 to 19, 20 to 39, and 40 or more. For molecules whose core contains a total of 5 to 11 sulfur atom-containing heterocycles and hydrocarbon rings (benzene rings and naphthene rings), ramping is performed based on the total number of rings. The number of side chains is not considered.

[0102] Ramping of heterocyclic molecules containing heteroatoms other than sulfur atoms Of the single-core molecules described above, heterocyclic molecules containing heteroatoms other than sulfur atoms are preferably divided according to the composition and number of rings constituting the core, and further ramped by the number of side chains. The number of components to ramp for heterocyclic molecules containing heteroatoms other than sulfur atoms can be set as appropriate. For example, the entire single-core molecule group can be ramped to approximately 50-500 components, 100-400 components, 100-300 components, or 150 components.

[0103] In the example shown in Figure 6, a group of heterocyclic molecules containing heteroatoms other than sulfur atoms are ramped as described below. For a group of molecules whose core contains one heterocycle with a heteroatom other than a sulfur atom, we perform ramping as a single group. For molecules whose core contains 1-2 heterocycles with heteroatoms other than sulfur atoms and 0-1 other rings (benzene rings, naphthene rings, and sulfur-containing heterocycles) (total number of rings is 2), we ramp them into groups based on the number of side chains: 0-19 and 20 or more. For molecules whose core contains 1 to 3 heterocycles with heteroatoms other than sulfur atoms and 0 to 2 other rings (benzene rings, naphthene rings, and sulfur-containing heterocycles) (total number of rings is 3), we ramp them into groups based on the number of side chains: 0 to 9, 10 to 19, and 20 or more. For molecules whose core contains 1 to 4 heterocycles with heteroatoms other than sulfur atoms and 0 to 3 other rings (benzene rings, naphthene rings, and sulfur-containing heterocycles) (total number of rings is 4), we ramp them into groups based on the number of side chains: 0 to 4, 5 to 9, 10 to 19, and 20 or more. For molecules whose core contains 1 to 5 heterocycles with heteroatoms other than sulfur atoms and 0 to 4 other rings (benzene rings, naphthene rings, and sulfur-containing heterocycles) (total number of rings is 5), we ramp them into groups based on the number of side chains: 0 to 9, 10 to 19, and 20 or more. For molecules whose core contains 1 to 6 heterocycles with heteroatoms other than sulfur atoms and 0 to 5 other rings (benzene rings, naphthene rings, and sulfur-containing heterocycles) (total number of rings is 6), we ramp them into groups based on the number of side chains: 0 to 19 and 20 or more. For molecules whose core contains a total of 7 to 11 heterocycles containing heteroatoms other than sulfur atoms, hydrocarbon rings (benzene rings and naphthene rings), and sulfur-containing heterocycles, ramping is performed based on the total number of rings. The number of side chains is not considered.

[0104] Ramping of double-core molecules The double-core molecule group is further classified into non-heterocyclic molecules, sulfur atom-containing heterocyclic molecules, and heterocyclic molecules containing heteroatoms other than sulfur atoms.

[0105] Ramping of nonheterocyclic molecules Of the double-core molecules described above, the non-heterocyclic molecules are preferably divided according to the composition and number of rings constituting the core, and further ramped by the number of side chains. The number of components to be ramped for the non-heterocyclic molecule group can be set as appropriate. For example, the entire double-core molecule group can be ramped to approximately 50-500 components, 100-400 components, 100-300 components, or 150 components.

[0106] In the example shown in Figure 7, the non-heterocyclic molecules are ramped as described below. For the group of molecules containing a total of 2 to 5 benzene and naphthene rings in two cores, we ramp them into three groups based on the total number of rings: 0 to 19, 20 to 39, and 40 or more side chains. For molecules containing a total of 6 to 20 benzene and naphthene rings in two cores, ramping is performed based on the total number of rings. The number of side chains is not considered.

[0107] Ramping of sulfur atom-containing heterocyclic molecules Of the double-core molecule group described above, the sulfur atom-containing heterocyclic molecule group is preferably divided according to the composition and number of rings constituting the core, and further ramped according to the number of side chains. The number of components to ramp in the sulfur atom-containing heterocyclic molecule group can be set as appropriate. For example, the entire double-core molecule group can be ramped to approximately 50-500 components, 100-400 components, 100-300 components, or 150 components.

[0108] In the example shown in Figure 8, the group of sulfur atom-containing heterocyclic molecules is ramped as described below. For the group of molecules containing two cores with a total of 2 to 5 sulfur atom-containing heterocycles, benzene rings, and naphthene rings, the molecules are ramped into groups based on the total number of rings: 0 to 19, 20 to 39, and 40 or more side chains. For molecules containing a total of 6 to 20 sulfur-containing heterocycles, benzene rings, and naphthene rings in two cores, ramping is performed based on the total number of rings. The number of side chains is not considered.

[0109] Ramping of heterocyclic molecules containing heteroatoms other than sulfur atoms Of the double-core molecule group described above, the heterocyclic molecule group containing heteroatoms other than sulfur atoms is preferably divided according to the composition of the rings constituting the core and the number of side chains. The number of components to ramp for heterocyclic molecules containing heteroatoms other than sulfur atoms can be set as appropriate. For example, the entire double-core molecule group can be ramped to approximately 50-500 components, 100-400 components, 100-300 components, or 150 components.

[0110] In the example shown in Figure 9, a group of heterocyclic molecules containing heteroatoms other than sulfur atoms are ramped as described below. For a group of molecules containing heterocyclic rings with heteroatoms other than sulfur atoms in the two cores, sulfur atom-containing heterocyclic rings, benzene rings, and naphthene rings in total, the molecules are ramped into two groups based on the total number of rings: those with 0 to 19 side chains and those with 20 or more side chains. For molecules containing 6 to 20 heterocycles (including heteroatoms other than sulfur atoms in the two cores), sulfur-containing heterocycles, benzene rings, and naphthene rings in total, ramping is performed based on the total number of rings. The number of side chains is not considered.

[0111] Furthermore, according to one embodiment of this disclosure, the target variable of the training data may further include component information of crude oil sample fractions classified by compound group based on measurement data of crude oil sample fractions by PONA analysis. From the viewpoint of obtaining component information of crude oil samples without omission, it is preferable that the crude oil sample fractions measured by PONA analysis are different from the crude oil sample fractions measured by FT-ICR MS. In addition, it is preferable that the crude oil sample fractions used in PONA analysis for predictive model creation are naphtha fractions.

[0112] According to one embodiment of the present disclosure, in the classification of compound groups in PONA analysis, crude oil sample fractions can be classified into five categories: normal paraffins, isoparaffins, olefins, naphthenes, and aromatics, and these five categories can be further classified by the number of carbon atoms. The range of carbon atoms to be classified is not particularly limited, but for example, it can be classified into categories with 3 to 13 carbon atoms.

[0113] Furthermore, according to one embodiment of this disclosure, the target variable of the training data may further include component information of the crude oil sample classified by compound group based on measurement data of the crude oil sample fraction by two-dimensional gas chromatography analysis. From the viewpoint of obtaining component information of the crude oil sample without omission, it is preferable that the crude oil sample fraction to be measured by two-dimensional gas chromatography analysis is different from the crude oil sample fraction to be measured by PONA analysis or FT-ICR MS. In addition, it is preferable that the crude oil sample fraction used in the two-dimensional gas chromatography analysis for creating the predictive model is a kerosene fraction.

[0114] According to one embodiment of the present disclosure, in the classification of compound groups in two-dimensional gas chromatography analysis, crude oil sample fractions can be classified into compound groups such as linear aliphatic hydrocarbons (straight-chain alkanes, branched alkanes), cyclic aliphatic hydrocarbons, monocyclic aromatic hydrocarbons, dicyclic aromatic hydrocarbons (naphthalene type, biphenyl type), tricyclic aromatic hydrocarbons, tetracyclic aromatic hydrocarbons, sulfur-containing compounds, nitrogen-containing compounds, and oxygen-containing compounds, and each compound group can be further classified by the number of carbon atoms. The range of carbon atoms to be classified is not particularly limited, but for example, it can be classified into 0 to 12, or more specifically, into 0, 1, or 2 or more.

[0115] (Step S13) Based on the structural and quantitative information for each compound group classified by ramping, obtained in the above step (S12), the component information of the crude oil samples classified for each compound group is calculated by computer. Preferably, the quantitative information for each compound group classified by ramping is the sum of the amounts (mole fractions) of the compounds contained in that compound group. Depending on the performance of the computer and other equipment, FT-ICR MS suggests that it is possible to use more than 100,000 data points before ramping.

[0116] [Step S3] Furthermore, according to one embodiment of the present disclosure, a predictive model for the component information of a target crude oil sample is created using the above-mentioned training data and based on a regression analysis method selected for each group of compounds in the crude oil sample. Accordingly, according to a preferred embodiment, the predictive model of the present disclosure is a regression analysis model that predicts the component information of a crude oil sample from measurement data of the crude oil sample by FT-IR according to the present disclosure using a regression analysis method set for each group of compounds.

[0117] The regression analysis methods mentioned above are not particularly limited, but include OLS (Ordinary Least Squares), PLS (Partial Least Squares Regression), RF (Random Forests), RR (Ridge Regression), LASSO (Least Absolute Shrinkage and Selection Operator), GPR (Gaussian Process Regression), GPR0 (Gaussian Process Regression (linear-Kernel)), LSVR (Linear Support Vector Regression), NSVR (Non-Linear Support Vector Regression (Gaussian Kernel)), DT (Decision Tree), GPR1~10 (Gaussian Process Regression (Nonlinear-Kernel)), RF (Random Forests), GBDT (Gradient Boosting Decision Tree), XGB (XGBoost), LGBM (Light Grandient Boosting Model), and RR (Ridge Regression).

[0118] As described above, the regression analysis method used to create the predictive model is preferably selected so as to fall within the range of predictive accuracy pre-set for each component information, taking into account the desired predictive accuracy. Therefore, according to one embodiment of this disclosure, the regression analysis method is selected for each component information of the compound group of the crude oil sample by referring to a threshold of predictive accuracy. For determining such a threshold of predictive accuracy, for example, the standard deviation, coefficient of determination, root mean square error, or inter-laboratory recall tolerance may be used, but preferably the coefficient of determination (R 2 ), is determined by referring to the root mean square error or inter-laboratory recall tolerance. According to one preferred embodiment, the regression analysis method is determined by the coefficient of determination (R) for each component to be predicted. 2 The method that maximizes ) is selected.

[0119] Coefficient of determination (R2 The coefficient of determination (R) is the ratio of the sum of squares of the differences between the mean of the dependent variable and the data to the sum of squares of the residuals. It is an indicator that shows how well the predicted value of the dependent variable obtained by regression analysis matches the actual value of the dependent variable. 2 A value closer to 1 indicates a smaller error. The coefficient of determination (R) 2 The threshold of the coefficient of determination (R) can be set appropriately according to the desired prediction accuracy, and the coefficient of determination (R) 2 ) is preferably 0.7 or higher, more preferably 0.8 or higher, and even more preferably 0.9 or higher.

[0120] The root mean squared error (RMSE) indicates how much the predicted value obtained from the regression equation deviates from the true value. A value closer to 0 for RMSE indicates a smaller error. The RMSE threshold can be set appropriately depending on the desired prediction accuracy.

[0121] The interlaboratory reproducibility limit is the value at which the absolute difference between two measurement results obtained under reproducible conditions is expected to be less than or equal to that value with a 95% probability. For predictive items in the test method that do not have an interlaboratory reproducibility limit, the threshold can be set appropriately according to the desired predictive accuracy, but it is preferable that it be within ±10% of the measured value.

[0122] The prediction accuracy thresholds described above can be adjusted as appropriate depending on the type and characteristics of each sample.

[0123] In this disclosure, by referring to the prediction accuracy threshold mentioned above and setting a regression analysis method for each component of information, it is possible to incorporate combinations of explanatory variables, predicted component information (dependent variable), and regression analysis with high prediction accuracy into the prediction model.

[0124] The creation of the above prediction model and the subsequent process of predicting component information using it can be carried out using publicly known software and hardware.

[0125] [Step (2): Using a computer to predict the component information of the target crude oil sample] According to one embodiment of the present disclosure, a computer storing the predictive model described in step (1) is used to predict the component information of the target crude oil sample.

[0126] As described above, in this disclosure, FT-IR measurement data can be used as explanatory variables to perform highly accurate predictions of the component information of crude oil samples. Accordingly, according to one embodiment of this disclosure, step (2) includes inputting the data of the target crude oil sample measured by FT-IR into a prediction model stored in a computer to predict the component information of the target crude oil sample.

[0127] [Step (3): Store the component information of the target crude oil sample, predicted by the predictive model, in a computer as at least part of the crude oil assay data.] The component information of the target crude oil sample obtained as described above can be used as data constituting a crude oil assay and can be advantageously utilized for the operation of highly accurate petroleum-related equipment. Accordingly, according to one embodiment, the method of the present disclosure further includes, as step (3), the step of storing the component information of the target crude oil sample predicted by a predictive model in a computer as at least part of the crude oil assay data.

[0128] <Method for predicting the characteristics of a target crude oil sample using a computer> Furthermore, since the component information of a crude oil sample includes information that also encompasses the structure and quantity of minute compounds through classification, it can be advantageously used to predict with high accuracy characteristic information such as physical properties and characteristics that are affected by the composition of the crude oil sample. Accordingly, according to one embodiment of this disclosure, a method is provided for predicting the characteristic information of a target crude oil sample using a computer, which includes the step of estimating the characteristic information of the target crude oil sample using a computer based on the component information of the target crude oil sample predicted by any of the above methods of this disclosure.

[0129] According to one embodiment of this disclosure, the prediction of the characteristic information of a target crude oil sample can be performed using a prediction model created based on a regression analysis method selected for each component of the crude oil sample, using training data in which the component information of the crude oil sample is used as an explanatory variable and the characteristic information of the crude oil sample is used as the dependent variable. The prediction model for the characteristic information of the crude oil sample can be created in accordance with the description in step S3. Alternatively, the prediction of the characteristic information of a target crude oil sample may be performed by creating a correlation formula in advance based on measured values ​​between the component information of the crude oil sample and the characteristic information, and using that correlation formula.

[0130] The characteristic information of the crude oil samples mentioned above includes the cloud point, clogging point, pour point, asphaltene content, total acid number, and mercaptan sulfur content for each fraction.

[0131] <How to operate equipment related to petroleum> The method for predicting the component information of a crude oil sample described herein forms the basis for predicting the characteristic information of a crude oil sample, and is therefore suitably used in adjusting the operating conditions of petroleum-related equipment. Accordingly, in a preferred embodiment, a method for operating petroleum-related equipment is provided, which sets the operating conditions based on the component information or characteristic information of the target crude oil sample obtained by the above method described herein.

[0132] Furthermore, equipment related to petroleum includes all equipment related to the processing of petroleum, including distillation and extraction equipment, as well as equipment involving chemical reactions such as reforming equipment, hydrogenation reactors, and desulfurization equipment. Equipment related to petroleum is collectively also called "petroleum refining equipment," and includes atmospheric distillation equipment, vacuum distillation equipment, etc., with atmospheric distillation equipment being preferred.

[0133] Next, with reference to Figure 10, an embodiment of the crude oil sample component information prediction device of this disclosure will be described. Figure 10 is a functional block diagram of the crude oil sample component information prediction device, and the interfaces for inputting and outputting information are omitted from the illustration.

[0134] In Figure 10, the crude oil sample component information prediction device 1 is: (1) FT-IR data receiving unit 10 that receives data of crude oil samples measured by Fourier transform infrared spectroscopy (FT-IR), (2) Crude oil sample component information prediction unit 20 predicts the component information of the target crude oil sample based on a prediction model of the component information of the crude oil sample. It is equipped with.

[0135] The FT-IR data receiving unit 10 is connected to the crude oil sample component information prediction unit 20 and receives measurement data acquired from the FT-IR measuring device. The measurement data is output from the FT-IR data receiving unit 10 and input to the crude oil sample component information prediction unit 20. The FT-IR data receiving unit 10 may also be connected to the FT-IR measuring device. The FT-IR data receiving unit 10 may be equipped with input devices such as a display, keyboard, mouse, or pointer for inputting data related to the FT-IR measurement of the crude oil sample.

[0136] The crude oil sample component information prediction unit 20 is connected to the FT-IR data receiving unit 10 and consists of a storage unit 21 where each data is stored and a calculation unit 22 that calculates the component information (predicted value) of the crude oil sample using a prediction model. The calculation unit 22 can be composed of a computer that stores a computer program for executing a method for predicting the component information of the crude oil sample. Specifically, based on the information input by the FT-IR data receiving unit 10, the calculation unit 22 can predict the component information of the crude oil sample using a prediction model. The obtained crude oil sample component information is compiled into a database along with each data stored in the storage unit 21 and can be stored in the storage unit 21 as part of the crude oil assay.

[0137] The above-mentioned device may be configured as a system in which multiple units are combined. Therefore, according to another aspect of this disclosure, a system for predicting the composition information of a crude oil sample, (1) FT-IR data receiving unit that receives data of crude oil samples measured by FT-IR, (2) Crude oil sample component information prediction unit that predicts the component information of the target crude oil sample based on a prediction model of the component information of the crude oil sample. A system equipped with the above features is provided.

[0138] In this disclosure, a series of processes for predicting the composition information of a crude oil sample can be performed by hardware, software, or a combination thereof. When performing the processing by software, a program recording the processing sequence can be installed and executed in the memory of a computer incorporated into dedicated hardware, or the program can be installed and executed on a general-purpose computer capable of performing various processes. Accordingly, according to one embodiment of this disclosure, a computer program for performing a method for predicting the composition information of a crude oil sample is provided.

[0139] For example, a program can be pre-recorded on a recording medium such as a hard disk or ROM. Therefore, according to one embodiment of this disclosure, a recording medium on which a computer program is recorded is provided. In addition, a program can be temporarily or permanently stored (recorded) on a removable recording medium such as a flexible disk, CD-ROM, MO disk, DVD, magnetic disk, or semiconductor memory.

[0140] In addition to installing programs from removable storage media as described above, programs can also be downloaded from download sites and transferred wirelessly to the computer, or transferred via a wired connection over a network such as a LAN or the internet. The computer can then receive the programs transferred in this way and install them onto its internal storage media, such as a hard disk.

[0141] The method disclosed herein can be suitably implemented on a computer that stores the above-mentioned computer program in its internal memory.

[0142] Furthermore, the various processes described herein may not only be executed chronologically as described, but may also be executed in parallel or individually as needed, depending on the processing capacity of the device performing the process. Also, in this specification, a system is a logical combination of multiple devices, and is not limited to devices in each configuration being located in the same enclosure.

[0143] Furthermore, according to another embodiment of this disclosure, the following is provided: [1] A method for predicting the composition information of a target crude oil sample using a computer, Step (1): A step of preparing a computer that stores a predictive model of the component information of a crude oil sample, The above prediction model includes explanatory variables, including measurement data of crude oil samples by FT-IR, Ramping of measurement data from crude oil sample fractions obtained by FT-ICR MS allows for the determination of a target variable containing component information of crude oil samples classified by compound group. Steps generated by machine learning using training data including, Step (2): Using the computer described above, predict the component information of the target crude oil sample based on the FT-IR measurement data of the target crude oil sample. Methods that include... [2] The method according to [1], wherein the group of compounds classified by the ramping described above is further classified into at least acyclic saturated aliphatic molecules, single-core molecules, and double-core molecules. [3] The method according to [2], wherein the single-core molecule group comprises at least one selected from the group consisting of non-heterocyclic molecules, sulfur atom-containing heterocyclic molecules, and heterocyclic molecules containing heteroatoms other than sulfur atoms. [4] The method according to [2], wherein the double core molecule group comprises at least one selected from the group consisting of nonheterocyclic molecules, sulfur atom-containing heterocyclic molecules, and heterocyclic molecules containing heteroatoms other than sulfur atoms. [5] The method according to any one of [1] to [4], wherein the target variable of the above training data further includes component information of crude oil samples classified by compound group based on measurement data of crude oil sample fractions by PONA analysis. [6] The method according to any one of [1] to [5], wherein the target variable of the above training data further includes component information of crude oil samples classified by compound group based on measurement data of crude oil sample fractions by two-dimensional gas chromatography analysis. [7] The method according to any one of [1] to [6], wherein the crude oil sample fraction is selected from vacuum residue, kerosene fraction, and naphtha fraction. [8] The method according to any of [1] to [7], wherein the prediction model is a regression analysis model that predicts the component information of a crude oil sample using a regression analysis method set for each of the above compound groups from measurement data of the crude oil sample by FT-IR. [9] A method for predicting the characteristics of a target crude oil sample using a computer, A step of using a computer to estimate the characteristic information of the target crude oil sample based on the component information of the target crude oil sample predicted by the method of claim 1. Methods that include...

[10] The method according to [9], wherein the above characteristic information includes at least one selected from the group consisting of cloud point, clogging point, pour point, asphaltene content, total acid number and mercaptan sulfur content. A computer program that causes a computer to perform any of the methods described in

[11] [1] to

[10] . A recording medium on which the computer programs described in

[12]

[11] are stored. A computer that stores the computer programs described in

[13]

[11] . A device or system for predicting the composition information of crude oil samples, which stores the computer program described in

[14]

[11] . [Examples]

[0144] The present disclosure will be described below with reference to examples, but the present disclosure is not limited to these examples. Unless otherwise specified, the units and measurement methods in this disclosure shall conform to the provisions of ASTM (American Society for Testing and Materials).

[0145] Test Example 1 (1) Creation of training data Since crude oil contains over 100,000 components, creating a predictive model that forecasts all of them is not practical. Furthermore, even if such a model were possible, handling it at a refinery would be difficult. Therefore, we decided to first investigate methods for reducing the number of predicted components.

[0146] In the predictive model, the naphtha fraction of crude oil was analyzed using PONA analysis (according to the method described in JIS-K2536-2), the kerosene fraction was analyzed using two-dimensional gas chromatography (GC×GC: Schoenmkers et al., according to the method described in Journal of Chromatography A 892(2000)29), and atmospheric residue (AR: also called "AR fraction") was used to predict the components obtained by FT-ICR MS. Analysis results for each fraction were prepared for up to 55 types of crude oil samples. Figure 11 shows the relationship between API and sulfur content (mass%) of the crude oil samples used in Test Example 1.

[0147] <Predicted number of components in each fraction> (1) Naphtha fraction We decided to use PONA analysis data (5 divisions) for light naphtha (LSR) and heavy naphtha (HSR) with 3 to 13 carbon atoms each. Table 2 shows an example of the analysis results for the LSR fraction.

[0148] The data for naphtha fractions totaled 90 (2 fractions × 9 carbon number ranges × 5 PONA analysis segments), but since many components had an abundance of 0, only 50 components in total (LSR and HSR) were actually used for model creation.

[0149] [Table 1]

[0150] (2) Light oil fraction For the kerosene fraction, each component was analyzed using lamps. The data obtained from GC×GC analysis (Table 2) divided the components with 7 to 30 carbon atoms into 33 categories, resulting in a total of 792 data points. However, by consolidating these into 12 categories as shown in Table 3, we obtained 288 data points (24 carbon number range × 12 categories). Since some of these 288 components have an actual abundance of 0, we excluded these and selected 150 components for model creation.

[0151] [Table 2]

[0152] [Table 3]

[0153] (3) AR fraction The boiling point range of the components obtained from the FT-ICR MS instrument is 360°C to an unsaturation degree DBE of 50 or less (equivalent to a boiling point of 850°C). The number of components varies depending on the crude oil, ranging from 100,000 to 250,000. Therefore, it is impractical to create a model for all components, and even if a model for 100,000 to 250,000 components were created, the calculated values ​​would have almost the same meaning as random numbers. For this reason, we set the number of components in the AR fraction necessary for advanced optimization control of the crude oil distillation unit (CDU) to 300 and investigated a method for compressing the component information of the AR fraction. The ramping flow of the AR fraction is shown in Figures 4 to 9.

[0154] <Basic policy for ramp installation> To achieve advanced CDU optimization control, it is necessary to accurately predict the distillation characteristics of processed crude oil. In this technology development, the concept is to estimate the distillation characteristics within an arbitrary temperature range from the component information obtained from the prediction model and use it as input data for the CDU optimization control system, thereby ensuring that the distillation characteristics do not change significantly before and after ramp distillation.

[0155] The lamp conversion will be carried out in the following three stages. Primary hierarchy: Classified into acyclic saturated aliphatic molecules, single-core molecules, and double-core molecules. Secondary hierarchy: Single-core molecule group / double-core molecule group are classified based on the presence or absence of heterogeneity. Third level: Classified by the number of aromatic naphthenic rings and side chains.

[0156] Unlike secondary devices, CDU does not require consideration of the reactivity of various heterocompounds (S only / N only / O only / SNO complex). Therefore, in the secondary layer, the structural components of single-core and double-core molecular groups are ramped for sulfur compounds and non-sulfur compounds, which have high abundance.

[0157] <Methods for converting each structure to a lamp> Based on the basic policy, the AR fraction was ramped up to 300 components using the following method. The ramping categories were fixed and not changed for each crude oil.

[0158] Acyclic saturated aliphatic molecules (Saturate) • The acyclic saturated aliphatic molecule group consists of several compounds, including normal paraffins and isoparaffins. However, under JACD rules, these are not distinguished and are grouped by the number of carbon atoms. Since the AR fraction has a small number of carbon atoms (20-70, totaling 50), no further scaling of the component information of the acyclic saturated aliphatic molecule group was performed, and it was left as is.

[0159] Single-core molecules Following the basic principles, the single-core molecule group (1 to 11 cores) was classified into three categories at the second level: non-heterocyclic molecules, sulfur atom-containing heterocyclic molecules, and other heterocompounds (heterocyclic molecules containing heteroatoms other than sulfur atoms). Compounds with 6 to 11 cores showed small differences in physical properties such as boiling point for compounds with the same number of cores, so in the third level, they were ramped by core number (total of 18 in the three classifications). On the other hand, compounds with 1 to 5 cores showed large differences in physical properties even with the same number of cores, so in the third level, compounds with the same number of cores were first classified by the number of aromatic / naphthene rings and then separated by the number of side chains (total of 132 in the three classifications).

[0160] Double-core molecules The molecules were separated using the same method as single-core molecules. Following the basic principles, the double-core molecule group (with 2 to 20 cores) was classified into three categories at the second level: non-heterocyclic molecules, sulfur-containing heterocyclic molecules, and other heterocompounds (heterocyclic molecules containing heteroatoms other than sulfur atoms). • Compounds with 9 to 20 cores have small differences in physical properties such as boiling point for the same number of cores. In the third tier, compounds were ramped according to the number of cores (total of 33 across the three classifications). On the other hand, for compounds with 1 to 8 cores, even with the same number of cores, there were significant differences in physical properties among the compounds. Therefore, in the third tier, compounds with the same number of cores were first classified by the number of aromatic / naphthene rings, and then separated by the number of side chains (total of 67 across the three classifications).

[0161] <Methods for creating and evaluating predictive models> Based on the strategy shown in Table 4, a predictive model was created to predict the above-mentioned target variable using the regression calculation method shown in Table 5. Commercial machine learning software was used to create the model. Model validation was performed using DCV (Double Cross Validation). Furthermore, the accuracy evaluation metric for models created using various regression calculation methods was the coefficient of determination (R). 2 The analysis was performed using the method described above, and the accuracy of the regression analysis method with the highest correlation was adopted.

[0162] [Table 4]

[0163] [Table 5]

[0164] As a result, the R of each component 2 The results were favorable. In particular, for the naphtha fraction, R was observed for more than 17 component items. 2 A result of >0.7 was obtained. Furthermore, for the kerosene fraction, R was measured for more than 40 components. 2 A result of >0.7 was obtained. Furthermore, for the AR fraction, R was found for 14 or more components. 2 A result of >0.7 was obtained. These results suggest that the prediction model using the ramping method described above can predict the component information of crude oil samples with high accuracy and speed. Below, the regression analysis method used and the prediction accuracy (R) for each component are described. 2 Parts of the above are shown in Tables 6, 7, and 8.

[0165] [Table 6]

[0166] [Table 7]

[0167] [Table 8] Test Example 2 Without employing component ramping as in Test Example 1, this property prediction model uses FT-IR measurement data of crude oil samples as explanatory variables to predict the property parameters of crude oil samples (Presentation items [2B09], [2B10] at the Osaka Conference (53rd Symposium on Petroleum and Petrochemicals); https: / / confit.atlas.jp / guide / event / jpi2023f / proceedings / list 2023 JPEC Forum; extension: / / efaidnbmnnnibpcajpcglclefindmkaj / https: / / www.pecj.or.jp / wp-content / uploads / 2024 / 01 / JPEC_report_No.240101.pdf, (See https: / / www.pecj.or.jp / forum2024 / ) When this was created using machine learning, R 2 Some property parameters were <0.7 (cloud point, clogging point, pour point, asphaltene content and total acid number, mercaptan sulfur content, etc.). This is thought to be partly due to the influence of micro-components contained in the crude oil sample on its properties. Therefore, we investigated whether it is possible to improve the accuracy of predicting the properties of the crude oil sample by predicting micro-component information using the predictive models of Test Example 1 and Test Example 2.

[0168] Of the above property items, cloud point and total acid number were used as examples to predict micro-component information / abundance using the predictive model in Test Example 2.

[0169] Next, we considered creating a correlation equation from the measured values ​​of the micro-component content and the cloud point of the diesel fuel fraction or the total acid number of the lantern fuel fraction.

[0170] (Cloudy point) It is known that the cloud point may be affected by wax derived from normal paraffin among the microcomponents of crude oil. Therefore, the relationship between the amount of normal paraffin with 8 to 30 carbon atoms in the diesel fraction and the cloud point was analyzed using Excel (trademark), and the results are shown in Figure 12. In Figure 12, the white triangles represent the correlation coefficient R between normal paraffin and the cloud point. 2 The white circles indicate the number and abundance of carbon atoms in the diesel fraction. If the amount of norparaffin decreases, R 2 Although it decreases, up to 21 carbon atoms, even if the abundance decreases, R 2 The result was a high 0.80. This suggests that normal paraffins with a high number of carbon atoms are related to the cloud point even at low abundances. Further analysis revealed that R 2 The relationship with normal paraffins having 20 carbon atoms, with a correlation coefficient of 0.82, was found to have the best correlation coefficient.

[0171] Therefore, using the prediction model of Test Example 1, we predicted the content of n-paraffins with 16 to 20 carbon atoms in the crude oil sample (diesel oil fraction). Furthermore, using this predicted value, we predicted the cloud point from the correlation equation between n-paraffins and the cloud point of the diesel oil fraction. As a result, it was confirmed that using the prediction model of Test Example 1 tended to improve prediction accuracy compared to the property prediction model that does not employ ramping (a property prediction model that uses measurement data of the crude oil sample by FT-IR as an explanatory variable and directly predicts the property items of the crude oil sample). From these results, it was shown that a prediction model can be created to predict the cloud point of a crude oil sample by using the component information of n-paraffins of the crude oil sample obtained by the prediction model of Test Example 1 as an explanatory variable and the cloud point of the crude oil sample as the dependent variable.

[0172] (Total acid value) It is said that the component corresponding to the total acid value of diesel fuel has a naphthenic acid skeleton with a carboxylic acid group. Therefore, we analyzed the correlation between the total acid value and six aromatic skeletons and two other items (normal paraffin, isoparaffin) that are thought to be related to the naphthenic acid skeleton using Excel (trademark). The correlation between cyclohexane (DBE2-4 in Table 6, compounds with 8-30 carbon atoms) and the total acid value R 2 The value was high at 0.84, and it was estimated that compounds having a cyclohexane skeleton and side chains were oxidized to compounds having a naphthenic acid skeleton, corresponding to the total acid value of diesel fuel.

[0173] Next, we analyzed the relationship between the amount of cyclohexane present in the diesel fraction and the total acid value, as shown in Figure 13. In Figure 13, the white triangles represent the correlation coefficient R between cyclohexane and the cloud point. 2 The diagram shows the carbon number and abundance of cyclohexane in the diesel fraction, with white circles indicating the carbon number and abundance. For carbon numbers from 12 to 24, regardless of abundance, R 2 The result was high, around 0.80. This suggests the presence of naphthenic compounds related to the total acid value. And, R 2 The correlation equation between cyclohexane with 18 carbon atoms and the total oxidation value, with a value of =0.82, showed the best results.

[0174] Therefore, the content of carbon-18 cyclohexane in the crude oil sample (diesel oil fraction) was predicted using the prediction model of Test Example 1, and furthermore, the cloud point was predicted using the correlation equation between the cyclohexane and the cloud point of the diesel oil fraction. As a result, it was confirmed that the prediction accuracy tended to be improved when using the prediction model of Test Example 1 compared to the property prediction model that does not employ ramping (a property prediction model that uses measurement data of the crude oil sample by FT-IR as an explanatory variable and directly predicts the property items of the crude oil sample). From these results, it was shown that a prediction model can be created to predict the total acid value of a crude oil sample by using the component information of the crude oil sample (cyclohexane, etc.) obtained by the prediction model of Test Example 1 as an explanatory variable and the total acid value of the crude oil sample as the dependent variable.

[0175] (Clogging point, pour point, asphaltene content, mercaptan sulfur content) For clogging point, pour point, asphaltene content, and mercaptan sulfur content, the same method as for cloud point and total acid value was used to calculate predicted values ​​for crude oil samples using the prediction model from Test Example 1. It was confirmed that the prediction accuracy was improved compared to the property prediction model (a property prediction model that uses measurement data of crude oil samples by FT-IR as explanatory variables and directly predicts the property items of crude oil samples). In particular, for clogging point and pour point, the component information of normal paraffin can be used as an explanatory variable, similar to the cloud point. From the above, it can be seen that a prediction model can be created to predict the total acid value of a crude oil sample by using the component information of the crude oil sample obtained by the prediction model from Test Example 1 as an explanatory variable and the clogging point, pour point, asphaltene content, and mercaptan sulfur content of the crude oil sample as the dependent variables.

[0176] Based on these results, the component information obtained by the prediction model for the crude oil sample in Test Example 2 can be used as an explanatory function for predicting various characteristic information, and is considered effective for predicting the general properties and composition of crude oil and fractions, which are difficult to predict with conventional prediction models. [Explanation of Symbols]

[0177] 1. A device for predicting the composition information of crude oil samples. 10 FT-IR data receiving unit 20 Crude Oil Sample Component Information Prediction Unit 21 Memory section 22 Arithmetic section

Claims

1. A method for predicting the composition information of a target crude oil sample using a computer, Step (1): A step of preparing a computer that stores a predictive model of the component information of a crude oil sample, The aforementioned prediction model includes explanatory variables, including measurement data of crude oil samples obtained by Fourier transform infrared spectroscopy (FT-IR), Ramping of measurement data from crude oil sample fractions using a Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR MS) allows for the determination of a target variable containing component information of crude oil samples classified by compound group. Steps generated by machine learning using training data including, Step (2): Using the computer, predict the component information of the target crude oil sample based on the FT-IR measurement data of the target crude oil sample. Methods that include...

2. The method according to claim 1, wherein the group of compounds classified by ramping is classified into at least a group of acyclic saturated aliphatic molecules, a group of single-core molecules, and a group of double-core molecules.

3. The method according to claim 2, wherein the single-core molecule group includes at least one selected from the group consisting of a non-heterocyclic molecule group, a sulfur atom-containing heterocyclic molecule group, and a heterocyclic molecule group containing heteroatoms other than sulfur atoms.

4. The method according to claim 2, wherein the group of double-core molecules includes at least one selected from the group consisting of a group of non-heterocyclic molecules, a group of sulfur atom-containing heterocyclic molecules, and a group of heterocyclic molecules containing heteroatoms other than sulfur atoms.

5. The method according to claim 1, wherein the target variable of the training data further includes component information of crude oil samples classified by compound group based on measurement data of crude oil sample fractions by PONA analysis.

6. The method according to claim 1, wherein the target variable of the training data further includes component information of a crude oil sample classified by compound group based on measurement data of crude oil sample fractions by two-dimensional gas chromatography analysis.

7. The method according to claim 1, wherein the crude oil sample fraction is selected from atmospheric residue, kerosene fraction, and naphtha fraction.

8. The method according to claim 1, wherein the prediction model is a regression analysis model that predicts component information of a crude oil sample using a regression analysis method set for each compound group from measurement data of the crude oil sample by FT-IR.

9. A method for predicting the characteristics of a target crude oil sample using a computer, A step of using a computer to estimate the characteristic information of the target crude oil sample based on the component information of the target crude oil sample predicted by the method of claim 1. Methods that include...

10. The method according to claim 9, wherein the characteristic information includes at least one selected from the group consisting of cloud point, clogging point, pour point, asphaltene content, total acid value, and mercaptan sulfur content.

11. A computer program for performing the method described in claim 1.

12. A recording medium on which the computer program described in claim 11 is recorded.

13. A computer storing the computer program described in claim 11.

14. A device or system for predicting the composition information of a crude oil sample, which stores the computer program described in claim 11.