A method and device for establishing a rapid evaluation model of oil product physical properties

Abstract

The present invention discloses a method and device for establishing a rapid evaluation model of oil product physical properties. The method includes: obtaining sample data of the oil product to be modeled and dividing it into a calibration set and a validation set; the sample data of the oil product to be modeled includes spectral data, viscosity-temperature curve data of each sample, and test data corresponding to each sample; through a variety of data processing methods, preprocessing the calibration set data respectively to obtain multiple corresponding preprocessed data sets; merging the spectral data and viscosity-temperature curve data of each preprocessed data set respectively to form multiple corresponding associated data sets, and associating them with the test data sets corresponding to each sample respectively through a variety of association algorithms to construct multiple corresponding physical property analysis models; using the validation set data to verify and evaluate each physical property analysis model respectively, and screening out the optimal model as the rapid evaluation model of the physical properties of the oil product to be modeled. The model constructed by this method has high accuracy and can accurately predict the material physical property indexes of oil products.

Description

Technical Field

[0001] The present invention relates to the field of petrochemical industry, and particularly to a method and device for establishing a rapid evaluation model of oil product physical properties. Background Art

[0002] At present, when most domestic and foreign refineries analyze various raw materials and intermediate products in the production process, they mainly rely on traditional chemical analysis methods. For traditional chemical analysis methods, each index needs to be analyzed separately by relevant personnel using separate instruments or methods. Therefore, the labor cost and equipment maintenance cost are high, and it is difficult to meet the requirements of safety and environmental protection. There is a lack of effective rapid evaluation analysis methods, resulting in high work intensity and low efficiency of chemical analysis personnel.

[0003] Some petrochemical enterprises have introduced some rapid evaluation analysis technologies, such as near-infrared spectroscopy (NIR), mid-infrared spectroscopy (MIR), and nuclear magnetic resonance spectroscopy (NMR) and other analysis technologies. The detection principles and application fields of these several rapid evaluation analysis technologies are different, but the technical routes in the spectral analysis process are basically the same. Spectral analysis mainly includes two processes: (1) establishment of an analysis model; (2) analysis of the spectrum. The quality of the analysis model for each index directly determines the accuracy of the rapid evaluation analysis data. However, in the actual application process, the current analysis technologies have some common drawbacks: currently, several spectra can only achieve the analysis at the level of internal functional groups of the sample, with less information, weak correlation for some complex indexes, low model accuracy, and it is difficult for the model to predict the change trend when the physical property indexes of the material fluctuate.

[0004] The current rapid evaluation analysis technology has at least the following defects: the model accuracy is not high, and it is difficult to accurately predict the physical property indexes of the sample. Summary of the Invention

[0005] The present invention provides a method and device for establishing a rapid evaluation model of oil product physical properties to solve the technical problem that the model accuracy is not high and it is difficult to accurately predict the physical property indexes of the sample.

[0006] To solve the above technical problem, an embodiment of the present invention provides a method for establishing a rapid evaluation model of oil product physical properties, including:

[0007] Obtain the sample data of the oil product to be modeled and divide it into a calibration set and a validation set; wherein, the sample data of the oil product to be modeled includes the spectral data, viscosity-temperature curve data of each sample, and the chemical analysis data corresponding to each sample; the spectral data is the spectral data or wave spectrum data of the oil product to be modeled;

[0008] Through a variety of data processing methods, preprocess the calibration set data respectively to obtain multiple corresponding preprocessed data sets;

[0009] Merge the spectral data and viscosity-temperature curve data of each of the preprocessed data sets to form multiple corresponding associated data sets, and use a variety of association algorithms to associate with the assay data sets corresponding to each sample respectively to construct multiple corresponding physical property analysis models;

[0010] Use the validation set data to verify and evaluate each physical property analysis model respectively, and select the optimal model as the physical property quick evaluation model of the oil product to be modeled.

[0011] Based on the existing technology, the present invention creatively proposes a method for establishing a quick evaluation model by combining spectral data and viscosity-temperature curve data, further improving the accuracy of the quick evaluation model. The oil product spectrum or wave spectrum mainly reflects the group characteristics of each organic molecule in the sample, and the viscosity information at each temperature point is conducive to analyzing the physical and chemical properties of heavy oil. Therefore, combining the spectral data with the viscosity-temperature curve data, the oil product information reflected will be more detailed, rich and accurate. On this basis, the established physical property analysis model has high accuracy and can accurately predict the material physical property indexes of the oil product.

[0012] Further, the obtaining of the sample data of the oil product to be modeled and dividing it into a calibration set and a validation set is specifically as follows: Use the Kolmogorov-Smirnov algorithm to divide all samples into two categories: a calibration set and a validation set.

[0013] Further, the preprocessing of the calibration set data through a variety of data processing methods to obtain multiple corresponding preprocessed data sets is specifically as follows:

[0014] For the spectral data of each sample in the calibration set, select a variety of first data processing methods for preprocessing respectively, and obtain the preprocessed spectral data corresponding to each sample. Among them, the first data processing methods include: at least two or more combinations of no processing, mean centering processing, mean variance processing, vector normalization processing, maximum minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, baseline correction method processing;

[0015] For the viscosity-temperature curve data of each sample in the calibration set, use a preset algorithm to fit the empirical formula of the viscosity-temperature curve, and calculate the corresponding viscosity values according to a preset multiple temperature points. Then, according to a variety of second data processing methods, preprocess the multiple temperature points and the viscosity values corresponding to each temperature point respectively to obtain the preprocessed viscosity-temperature curve data corresponding to each sample; where each viscosity-temperature curve data corresponds to multiple temperature points and the viscosity values corresponding to each temperature point; the second data processing methods include: at least one or more combinations of logarithmic processing, mean processing, mean variance processing, maximum minimum normalization processing.

[0016] The quality of each physical property analysis model directly determines the accuracy of predicting the physical property indicators of materials. Pretreating the spectral data before modeling can reduce the negative impacts such as inevitable data noise and baseline drift during the spectral scanning process of the spectrum, improve the quality of the model; the viscosity value at any temperature point can be deduced through the empirical formula of the viscosity-temperature curve, making up for the defect that the structural information reflected by the viscosity data at a single temperature point is incomplete; the present invention uses a variety of data processing methods to preprocess the calibration set data, forming multiple corresponding preprocessed data sets, used to construct multiple corresponding physical property analysis models and select the best one, ensuring the quality of the model, so as to accurately predict the physical property indicators of the sample.

[0017] Further, after obtaining the preprocessed spectral data corresponding to each sample, according to a preset dimensionality reduction method, perform dimensionality reduction processing on each of the preprocessed spectral data; wherein, the dimensionality reduction method includes the simple model method or the principal component analysis method.

[0018] Spectral data are often high-dimensional data, accompanied by a lot of unnecessary information. Therefore, before formal modeling, the spectral data are first subjected to dimensionality reduction processing. Using the simple model method and the principal component analysis method can effectively reduce the number of spectral data points. On the one hand, it improves the modeling efficiency. On the other hand, it can also reduce the interference of useless information to improve the accuracy of the model, thereby improving the accuracy of predicting the physical property indicators of the sample.

[0019] Further, merge the spectral data and viscosity-temperature curve data of each preprocessed data set to form multiple corresponding associated data sets, and through a variety of association algorithms, associate them with the test data sets corresponding to each sample respectively, and construct multiple corresponding physical property analysis models. The variety of association algorithms includes at least one or a combination of multiple of partial least squares method, artificial neural network, and kernel partial least squares method.

[0020] The present invention uses a variety of association algorithms to construct multiple corresponding physical property analysis models and select the best one, excluding the possibility of inaccurate prediction results caused by the low quality of a single physical property analysis model, ensuring the accuracy of the model, and improving the accuracy of predicting the physical property indicators of the sample.

[0021] Further, use the validation set data to verify and evaluate each physical property analysis model respectively, and select the optimal model as the physical property quick evaluation model of the oil product to be modeled. Specifically:

[0022] Merge the spectral data and viscosity-temperature curve data of each sample in the validation set to form a data set to be predicted, and use each physical property analysis model to predict respectively, obtaining multiple corresponding prediction result sets;

[0023] Compare each set of prediction results with the assay data sets corresponding to each sample in the validation set, and select the model with the smallest error in the validation set as the physical property quick evaluation model of the oil product to be modeled.

[0024] The smaller the error of the validation set, the higher the accuracy of the model. Therefore, using the model with the smallest error in the validation set among all models can ensure the accuracy of predicting the material physical property indexes of the samples.

[0025] An apparatus for establishing a physical property quick evaluation model of an oil product includes: an acquisition module, a preprocessing module, a modeling module, and a screening module.

[0026] The acquisition module is used to acquire the sample data of the oil product to be modeled and divide it into a calibration set and a validation set; wherein, the sample data of the oil product to be modeled includes the spectral data, viscosity-temperature curve data of each sample, and the assay data corresponding to each sample; the spectral data is the spectral data or wave spectrum data of the oil product to be modeled;

[0027] The preprocessing module is used to preprocess the calibration set data through a variety of data processing methods to obtain multiple corresponding preprocessed data sets;

[0028] The modeling module is used to respectively merge the spectral data and viscosity-temperature curve data of each preprocessed data set to form multiple corresponding associated data sets, and associate them with the assay data sets corresponding to each sample through a variety of association algorithms to construct multiple corresponding physical property analysis models;

[0029] The screening module is used to verify and evaluate each physical property analysis model with the validation set data, and select the optimal model as the physical property quick evaluation model of the oil product to be modeled.

[0030] Among them, the preprocessing module includes: a spectral data processing unit and a viscosity-temperature curve data processing unit;

[0031] The spectral data processing unit is used to preprocess the spectral data, specifically:

[0032] For the spectral data of each sample in the calibration set, respectively select at least two or more combinations of a variety of first data processing methods for preprocessing, and obtain the preprocessed spectral data corresponding to each sample, wherein the first data processing methods include: no processing, mean centering processing, mean variance processing, vector normalization processing, maximum-minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, baseline correction method processing;

[0033] After obtaining the preprocessed spectral data corresponding to each sample, perform dimensionality reduction processing on each of the preprocessed spectral data according to a preset dimensionality reduction method; wherein, the dimensionality reduction method includes a simple model method or a principal component analysis method;

[0034] The viscosity-temperature curve data processing unit is used to preprocess the viscosity-temperature curve data, specifically:

[0035] For the viscosity-temperature curve data of each sample in the calibration set, use a preset algorithm to fit the empirical formula of the viscosity-temperature curve, and calculate the corresponding viscosity values according to a preset number of temperature points, and then, according to a variety of second data processing methods, respectively preprocess the multiple temperature points and the viscosity values corresponding to each temperature point to obtain the preprocessed viscosity-temperature curve data corresponding to each sample; wherein, each viscosity-temperature curve data corresponds to multiple temperature points and the viscosity values corresponding to each temperature point; the second data processing methods include: at least one or a combination of multiple of logarithmic processing, mean processing, mean variance processing, maximum-minimum normalization processing.

[0036] The modeling module presets a variety of correlation algorithms for modeling, wherein the variety of correlation algorithms includes: at least one or a combination of multiple of partial least squares method, artificial neural network, kernel partial least squares method.

[0037] The screening module is used to respectively verify and evaluate each physical property analysis model with the validation set data, and screen out the optimal model as the physical property quick evaluation model of the oil product to be modeled, specifically:

[0038] Combine the spectral data and viscosity-temperature curve data of each sample in the validation set to form a dataset to be predicted, and respectively use each physical property analysis model for prediction to obtain multiple corresponding prediction result sets;

[0039] Compare each prediction result set with the test dataset corresponding to each sample in the validation set, and screen out the model with the smallest error in the validation set as the physical property quick evaluation model of the oil product to be modeled.

[0040] The device provided by the present invention creatively uses a quick evaluation analysis mode combining spectral data and viscosity-temperature curve data, further improving the accuracy of the quick evaluation model. The oil product spectrum or wave spectrum mainly reflects the group characteristics of each organic molecule in the sample, and the viscosity information at each temperature point is conducive to analyzing the physical and chemical properties of heavy oil. Therefore, combining the spectral data and the viscosity-temperature curve data, the oil product information reflected will be more detailed, rich and accurate. On this basis, the physical property analysis model constructed has high accuracy and can accurately predict the physical property indexes of the oil product. Brief Description of the Drawings

[0041] Figure 1: It is a schematic flow chart of a method for establishing a rapid evaluation model of oil product physical properties provided by the present invention;

[0042] Figure 2 : It is a graph showing the change trend of the error of the validation set of the models constructed by each modeling method under different schemes in the first embodiment of the present invention;

[0043] Figure 3 : It is a graph showing the change trend of the error of the validation set of the models constructed by each modeling method under different schemes in the second embodiment of the present invention;

[0044] Figure 4 : It is a schematic structural diagram of an oil product physical property rapid evaluation model establishment device provided in the third embodiment of the present invention. Detailed implementation manners

[0045] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0046] Embodiment 1

[0047] Please refer to Figure 1 It is a schematic flow chart of a method for establishing a rapid evaluation model of oil product physical properties provided by the present invention. In this embodiment, this method is used to evaluate and analyze the asphalt components of a certain refinery. Taking the penetration as an example, the analysis index includes steps 101 to 104, and the specific methods of each step are as follows:

[0048] In this embodiment, the spectral data of the oil product sample to be modeled is collected by a mid-infrared spectrometer, the viscosity-temperature curve data of the oil product sample to be modeled is collected by a viscometer, and the viscosity data at six temperature points of 50 °C, 60 °C, 70 °C, 80 °C, 90 °C, and 100 °C is calculated according to the empirical formula of the viscosity-temperature curve fitted by a preset algorithm. The corresponding test data is the penetration test data.

[0049] Step 101: Obtain the sample data of the oil product to be modeled and divide it into a calibration set and a validation set.

[0050] Among them, the sample data of the oil product to be modeled includes the spectral data, viscosity-temperature curve data of each sample, and the test data corresponding to each sample; the spectral data is the spectral data or wave spectrum data of the oil product to be modeled;

[0051] First, obtain 100 residue oil samples for asphalt blending, with penetration values ranging from 40 to 100. Respectively obtain the mid-infrared spectrum data, viscosity-temperature curve data of each residue oil sample, and the penetration test data corresponding to each sample, and form a spectral dataset X1, a viscosity-temperature curve dataset X2, and a test dataset Y corresponding to each sample. Among them, the dimension of X1 is 100 * 4417, the dimension of X2 is 100 * 6, and the dimension of Y is 100 * 1. The mid-infrared spectrum band range is 600 - 4000 cm -1 , the number of cutting points is 4417, the number of temperature points for viscosity values is 6, namely six temperature points of 50 °C, 60 °C, 70 °C, 80 °C, 90 °C, and 100 °C, and the number of physical property indexes of the oil product for which a model needs to be established is 1, namely penetration;

[0052] Furthermore, obtain the sample data of the oil product to be modeled and divide it into a calibration set and a validation set. Specifically: Use the Kolmogorov-Smirnov algorithm to divide all samples into two categories, namely the calibration set and the validation set.

[0053] Divide the data required for modeling the 100 samples into two parts, namely the calibration set and the validation set, according to the Kolmogorov-Smirnov algorithm. The calibration set data is used to construct the model, and the validation set data is used to verify and evaluate the model. In this embodiment, the proportion of the calibration set samples is set to 80%, that is, the calibration set contains 80 samples and the validation set contains 20 samples. The calibration set data is respectively denoted as X 1c 、X 2c 、Y c， , and the validation set data is respectively denoted as X 1v 、X 2v 、Y v , where X 1c is the spectral data of the calibration set, with a dimension of 80 * 4417, X 2c is the viscosity-temperature curve data of the calibration set, with a dimension of 80 * 6, Y c is the test data corresponding to each sample in the calibration set, with a dimension of 80 * 1, X 1v is the spectral data of the validation set, with a dimension of 20 * 4417, X 2v is the viscosity-temperature curve data of the validation set, with a dimension of 20 * 6, Y v is the test data corresponding to each sample in the validation set, with a dimension of 20 * 1.

[0054] Step 102: Through a variety of data processing methods, preprocess the calibration set data respectively to obtain multiple corresponding preprocessed datasets.

[0055] For the spectral data of each sample in the calibration set, multiple first data processing methods are respectively selected for preprocessing, and the preprocessed spectral data corresponding to each sample is obtained. Among them, the first data processing methods include at least two or more combinations of: no processing, mean centering processing, mean variance processing, vector normalization processing, maximum-minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, and baseline correction method processing;

[0056] In this embodiment, the classified calibration set data X 1c is copied into 11 copies, and 11 different first data processing methods are respectively used for preprocessing to form 11 different preprocessed data sets; the first data processing methods include: no processing, mean centering processing, mean variance processing, vector normalization processing, maximum-minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, and baseline correction method processing;

[0057] Further, after obtaining the preprocessed spectral data corresponding to each sample, according to the preset dimensionality reduction method, the preprocessed spectral data is respectively subjected to dimensionality reduction processing; among them, the dimensionality reduction method includes the simple model method or the principal component analysis method;

[0058] In this embodiment, 40 principal components of the spectral data are extracted by the principal component analysis method, and the spectral dimension is reduced from 4417 to 40;

[0059] For the viscosity-temperature curve data of each sample in the calibration set, the empirical formula of the viscosity-temperature curve is fitted by using a preset algorithm, and the corresponding viscosity values are calculated according to a preset plurality of temperature points. Then, according to multiple second data processing methods, the plurality of temperature points and the viscosity values corresponding to each temperature point are respectively preprocessed to obtain the preprocessed viscosity-temperature curve data corresponding to each sample; among them, each viscosity-temperature curve data corresponds to a plurality of temperature points and the viscosity values corresponding to each temperature point; the second data processing methods include at least one or more combinations of: logarithmic processing, mean processing, mean variance processing, and maximum-minimum normalization processing;

[0060] In this embodiment, for the calibration set data X 2cThe empirical formula of the viscosity-temperature curve fitted by the preset algorithm is: log(μ) = log(μ0) + L / (T / T0 - 1), and the viscosity values corresponding to the six preset temperature points of 50°C, 60°C, 70°C, 80°C, 90°C, and 100°C are calculated according to the fitted empirical formula of the viscosity-temperature curve. One second data processing method, that is, logarithmic processing, is used for preprocessing to obtain the preprocessed viscosity-temperature curve data corresponding to each sample.

[0061] Among them, μ represents the viscosity of the oil component, μ0 represents the limiting viscosity of the oil component at infinitely high temperature, T represents the temperature, T0 represents the temperature at which the oil component solidifies into a solid (infinite viscosity), and L is a related parameter used to measure the fitting degree of the viscosity-temperature curve to the measured viscosity-temperature data.

[0062] Table 1 shows the spectral data and the processing methods of the viscosity-temperature curve data under different schemes of this embodiment.

[0063] Table 1 Processing methods of spectral data and viscosity-temperature curve data under different schemes

[0064] Scheme number Spectrum data processing method Viscosity-temperature curve data processing method Scheme 1 No processing Logarithmic processing Scheme 2 Mean centering processing Logarithmic processing Scheme 3 Mean variance processing Logarithmic processing Scheme 4 Vector normalization processing Logarithmic processing Scheme 5 Maximum-minimum normalization processing Logarithmic processing Scheme 6 Standard normal variable transformation processing Logarithmic processing Scheme 7 Multiplicative scatter correction processing Logarithmic processing Scheme 8 Savitzky-Golay convolution smoothing processing Logarithmic processing Scheme 9 First derivative method processing Logarithmic processing Scheme 10 Detrending method processing Logarithmic processing Scheme 11 Baseline correction method processing Logarithmic processing

[0065] Step 103: Merge the spectral data and the viscosity-temperature curve data of each preprocessed data set to form multiple corresponding associated data sets, and use a variety of association algorithms to associate with the test data sets corresponding to each sample respectively to construct multiple corresponding physical property analysis models.

[0066] The variety of association algorithms includes at least one or a combination of multiple of partial least squares method, artificial neural network, and kernel partial least squares method.

[0067] In this embodiment, the processed 11 sets of mid-infrared spectral data sets X 1c and the viscosity-temperature curve data sets X 2c are respectively merged to form 11 corresponding associated data sets X c , X c with a dimension of 80*46, and through three preset algorithms of partial least squares method (PLS), artificial neural network (ANN), and kernel partial least squares method (KPLS), they are respectively associated with the test data sets Y c corresponding to each sample to construct 33 penetration analysis models in total;

[0068] At the same time, the processed 11 sets of mid-infrared spectral data sets X 1c are respectively associated with the test data sets Y c corresponding to each sample through three preset algorithms of partial least squares method (PLS), artificial neural network (ANN), and kernel partial least squares method (KPLS) to establish 33 mid-infrared spectrum-penetration analysis models as reference basic models.

[0069] Step 104: Use the validation set data to verify and evaluate each physical property analysis model respectively, and select the optimal model as the physical property quick evaluation model of the oil product to be modeled.

[0070] Merge the spectral data and viscosity-temperature curve data of each sample in the validation set to form a dataset to be predicted, and use each physical property analysis model to make predictions respectively to obtain multiple corresponding prediction result sets;

[0071] Compare each prediction result set with the corresponding test dataset of each sample in the validation set, and select the model with the smallest error in the validation set as the physical property quick evaluation model of the oil product to be modeled.

[0072] In this embodiment, the validation set data X 1v , X 2v are merged to form a dataset to be predicted X v , and each penetration analysis model constructed by this method is used to make predictions respectively to obtain multiple corresponding prediction result sets, denoted as Y p ; Compare the prediction result set Y p with the corresponding test dataset Y v of each sample in the validation set, and calculate the average error of the validation set samples of 33 penetration analysis models respectively, denoted as E p1 ~E p33 ;

[0073] Meanwhile, use 33 mid-infrared spectroscopy-penetration analysis models as the basic models to make predictions respectively, and the prediction result set is denoted as Y 1p ; Compare the prediction result set Y 1p with the corresponding test dataset Y v of each sample in the validation set, and calculate the average error of the validation set samples of 33 mid-infrared spectroscopy-penetration analysis models respectively, denoted as E p34 ~E p66 , which is used as the reference for basic data;

[0074] Please refer to Figure 2 for the trend chart of the validation set error change of the models constructed by each modeling method under different schemes in this embodiment;

[0075] From Figure 2It can be seen that, compared with the models established using only mid-infrared spectra, after combining the viscosity-temperature curve data, the overall accuracy of the models established by various modeling methods under different preprocessing schemes has been greatly improved. The model established using only mid-infrared spectra has the best effect when using maximum-minimum normalization, that is, Scheme 5, and using the KPLS algorithm for modeling. The average error of the validation set samples is 3.02. After combining the viscosity-temperature curve data, the average error of the prediction results of the model established under this scheme drops to 0.95, and the accuracy is significantly improved. Therefore, the model established by the KPLS algorithm under the conditions of Scheme 5 can be selected as the optimal model for the quick evaluation model of the penetration of the oil product to be modeled.

[0076] In the above embodiments, by using the combined quick evaluation analysis method of using spectral data and viscosity-temperature curve data, the accuracy of the quick evaluation analysis results of the physical property indexes of various raw materials or intermediate products in the petrochemical field is greatly improved. After combining the spectral data or wave spectral data with the viscosity-temperature curve data, and then correlating them with each physical property index, the prediction effects of the models established by various modeling methods are far better than the models established using only spectral data. The error of the prediction results drops significantly, and the accuracy is significantly improved. Therefore, on the basis of the existing technology, introducing the viscosity-temperature curve data into the field of quick evaluation analysis of petrochemical materials can significantly improve the accuracy of the analysis models of various physical property indexes of oil products, greatly making up for the defect that the accuracy of the prediction results of the current modeling methods for complex physical properties is not high, and thus has extremely high application value.

[0077] On the basis of the existing technology, the present invention creatively proposes a method for establishing a quick evaluation model by combining spectral data and viscosity-temperature curve data, further improving the accuracy of the quick evaluation model. The oil product spectrum or wave spectrum mainly reflects the group characteristics of each organic molecule in the sample, and the viscosity information at each temperature point is beneficial to analyzing the physical and chemical properties of heavy oil. Therefore, by combining the spectral data and the viscosity-temperature curve data, the oil product information reflected will be more detailed, rich and accurate. On this basis, the physical property analysis model constructed has high accuracy and can accurately predict the physical property indexes of the oil product.

[0078] Example Two

[0079] In this example, a method for establishing a quick evaluation model of oil product physical properties provided by the present invention is used to evaluate and analyze the material of the third normal line of a refinery (straight-run diesel from the atmospheric and vacuum distillation unit). Taking the 95% distillation temperature of diesel as an example, the analysis index includes steps 201 to 204, and the specific methods of each step are as follows:

[0080] In this example, a near-infrared spectrometer is used to collect the spectral data of the oil product sample to be modeled, a viscometer is used to collect the viscosity-temperature curve data of the oil product sample to be modeled, and the viscosity data at five temperature points of 20°C, 30°C, 40°C, 50°C, and 60°C are calculated according to the empirical formula of the viscosity-temperature curve fitted by the preset algorithm. The corresponding test data is the 95% distillation temperature test data.

[0081] Step 201: Obtain the sample data of the oil product to be modeled and divide it into a calibration set and a validation set.

[0082] Among them, the sample data of the oil product to be modeled includes the spectral data, viscosity-temperature curve data of each sample, and the test data corresponding to each sample; the spectral data is the spectral data or wave spectrum data of the oil product to be modeled;

[0083] First, obtain 100 straight-run diesel samples with a 95% distillation temperature change range of approximately 340 - 380 °C. Respectively obtain the near-infrared spectral data, viscosity-temperature curve data, and 95% distillation temperature test data corresponding to each straight-run diesel sample, and form a spectral data set X1, a viscosity-temperature curve data set X2, and a test data set Y corresponding to each sample. Among them, the dimension of X1 is 100 * 4094, the dimension of X2 is 100 * 5, the dimension of Y is 100 * 1, the near-infrared spectral band range is 4000 - 12000 cm -1 , the number of cutting points is 4094, the number of temperature points for taking viscosity values is 5, namely five temperature points of 20 °C, 30 °C, 40 °C, 50 °C, and 60 °C, and the number of physical property indexes for which the oil product needs to establish a model is 1, namely the 95% distillation temperature.

[0084] Furthermore, the obtaining of the sample data of the oil product to be modeled and dividing it into a calibration set and a validation set is specifically as follows: Use the Kolmogorov-Smirnov algorithm to divide all samples into two categories: a calibration set and a validation set.

[0085] Divide the data required for modeling into a calibration set and a validation set of 100 samples according to the Kolmogorov-Smirnov algorithm. The calibration set data is used to construct the model, and the validation set data is used to verify and evaluate the model. In this embodiment, the calibration set sample ratio is set to 80%, that is, the calibration set contains 80 samples and the validation set contains 20 samples. The calibration set data is respectively denoted as X 1c 、X 2c 、Y c， , and the validation set data is respectively denoted as X 1v 、X 2v 、Y v , where X 1c is the spectral data of the calibration set, with a dimension of 80 * 4094, X 2c is the viscosity-temperature curve data of the calibration set, with a dimension of 80 * 5, Y c is the test data corresponding to each sample in the calibration set, with a dimension of 80 * 1, X 1v is the spectral data of the validation set, with a dimension of 20 * 4094, X 2v is the viscosity-temperature curve data of the validation set, with a dimension of 20 * 5, Y vThe assay data corresponding to each sample in the validation set, with a dimension of 20*1.

[0086] Step 202: Preprocess the calibration set data through various data processing methods to obtain multiple corresponding preprocessed data sets.

[0087] For the spectral data of each sample in the calibration set, select multiple first data processing methods for preprocessing respectively, and obtain the preprocessed spectral data corresponding to each sample. Among them, the first data processing methods include at least two or more combinations of no processing, mean centering processing, mean variance processing, vector normalization processing, maximum-minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, and baseline correction method processing;

[0088] In this embodiment, the classified calibration set data X 1c Is copied into 11 copies, and 11 different first data processing methods are used for preprocessing respectively to form 11 different preprocessed data sets; the first data processing methods include: no processing, mean centering processing, mean variance processing, vector normalization processing, maximum-minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, and baseline correction method processing;

[0089] Further, after obtaining the preprocessed spectral data corresponding to each sample, perform dimensionality reduction processing on each of the preprocessed spectral data according to a preset dimensionality reduction method; among them, the dimensionality reduction method includes the simple model method or the principal component analysis method.

[0090] In this embodiment, 40 principal components of the spectral data are extracted by the principal component analysis method, and the spectral dimension is reduced from 4094 to 40.

[0091] For the viscosity-temperature curve data of each sample in the calibration set, use a preset algorithm to fit the empirical formula of the viscosity-temperature curve, calculate the corresponding viscosity values according to a preset number of temperature points, and then preprocess the multiple temperature points and the viscosity values corresponding to each temperature point according to multiple second data processing methods to obtain the preprocessed viscosity-temperature curve data corresponding to each sample; where each viscosity-temperature curve data corresponds to multiple temperature points and the viscosity values corresponding to each temperature point; the second data processing methods include at least one or more combinations of logarithmic processing, mean processing, mean variance processing, and maximum-minimum normalization processing;

[0092] In this embodiment, for the calibration set data X 2cThe empirical formula of the viscosity-temperature curve fitted by the preset algorithm is: log(μ) = log(μ0) + L / (T / T0 - 1), and the viscosity values corresponding to the five preset temperature points of 20°C, 30°C, 40°C, 50°C, and 60°C are calculated according to the fitted empirical formula of the viscosity-temperature curve. One second data processing method, namely maximum-minimum normalization processing, is used for preprocessing to obtain the preprocessed viscosity-temperature curve data corresponding to each sample.

[0093] Among them, μ represents the viscosity of the oil component, μ0 represents the limiting viscosity of the oil component at infinitely high temperature, T represents the temperature, T0 represents the temperature at which the oil component solidifies into a solid (viscosity is infinitely large), and L is a related parameter used to measure the fitting degree of the viscosity-temperature curve to the measured viscosity-temperature data.

[0094] Table 2 shows the spectral data and the processing methods of the viscosity-temperature curve data under different schemes of this embodiment.

[0095] Table 2 Processing Methods of Spectral Data and Viscosity-Temperature Curve Data under Different Schemes

[0096] Scheme number Spectrum data processing method Viscosity-temperature curve data processing method Scheme 1 No processing Maximum-minimum normalization processing Scheme 2 Mean centering processing Maximum-minimum normalization processing Scheme 3 Mean variance processing Maximum-minimum normalization processing Scheme 4 Vector normalization processing Maximum-minimum normalization processing Scheme 5 Maximum-minimum normalization processing Maximum-minimum normalization processing Scheme 6 Standard normal variable transformation processing Maximum-minimum normalization processing Scheme 7 Multiplicative scatter correction processing Maximum-minimum normalization processing Scheme 8 Savitzky-Golay convolution smoothing processing Maximum-minimum normalization processing Scheme 9 First derivative method processing Maximum-minimum normalization processing Scheme 10 Detrending method processing Maximum-minimum normalization processing Scheme 11 Baseline correction method processing Maximum-minimum normalization processing

[0097] Step 203: Combine the spectral data and the viscosity-temperature curve data of each preprocessed data set to form multiple corresponding associated data sets, and use a variety of association algorithms to associate with the test data sets corresponding to each sample respectively to construct multiple corresponding physical property analysis models.

[0098] The variety of association algorithms includes at least one or a combination of multiple of partial least squares method, artificial neural network, and kernel partial least squares method.

[0099] In this embodiment, the 11 processed near-infrared spectral data sets X 1c and the viscosity-temperature curve data set X 2c are respectively combined to form 11 corresponding associated data sets X c , X c with a dimension of 80*45, and through three preset algorithms of partial least squares method (PLS), artificial neural network (ANN), and kernel partial least squares method (KPLS), they are respectively associated with the test data sets Y c corresponding to each sample to construct a total of 33 95% distillation temperature analysis models;

[0100] At the same time, the 11 processed near-infrared spectral data sets X 1c are respectively associated with the test data sets Y cAssociate them to establish 33 near-infrared spectroscopy - 95% distillation temperature analysis models as reference basic models.

[0101] Step 204: Use the validation set data to verify and evaluate each physical property analysis model respectively, and select the optimal model as the physical property quick evaluation model of the oil product to be modeled.

[0102] Merge the spectral data and viscosity-temperature curve data of each sample in the validation set to form a dataset to be predicted, and use each physical property analysis model to make predictions respectively to obtain multiple corresponding prediction result sets;

[0103] Compare each prediction result set with the corresponding chemical analysis dataset of each sample in the validation set, and select the model with the smallest error in the validation set as the physical property quick evaluation model of the oil product to be modeled.

[0104] In this embodiment, the validation set data X 1v 、X 2v are merged to form a dataset to be predicted X v , and use each 95% distillation temperature analysis model constructed by this method to make predictions respectively to obtain multiple corresponding prediction result sets, denoted as Y p ; compare the prediction result set Y p with the corresponding chemical analysis dataset Y v of each sample in the validation set, and calculate the average error of the validation set samples of 33 95% distillation temperature analysis models respectively, denoted as E p1 ~E p33 ;

[0105] Meanwhile, use 33 near-infrared spectroscopy - 95% distillation temperature analysis models as basic models to make predictions respectively, and the prediction result set is denoted as Y 1p ; compare the prediction result set Y 1p with the corresponding chemical analysis dataset Y v of each sample in the validation set, and calculate the average error of the validation set samples of 33 near-infrared spectroscopy - 95% distillation temperature analysis models respectively, denoted as E p34 ~E p66 , as reference basic data;

[0106] Please refer to Figure 3 which is the trend chart of the error change of the validation set of the models constructed by each modeling method under different schemes in this embodiment;

[0107] From Figure 3It can be seen that, compared with the models established solely using near-infrared spectra, after combining the viscosity-temperature curve data, the overall accuracy of the models established by each modeling method under different preprocessing schemes has been greatly improved. The model established solely using near-infrared spectra has the best effect when processed by the baseline correction method, i.e., Scheme 11, and the KPLS algorithm is used for modeling. The average error of the validation set samples is 1.88. After combining the viscosity-temperature curve data, the average error of the prediction results of the model established under this scheme for the validation set drops to 0.68, and the accuracy improvement is obvious. Therefore, the model established by the KPLS method under the conditions of Scheme 11 can be selected as the optimal model for the rapid evaluation model of the 95% distillation temperature of the oil product to be modeled.

[0108] Through the combined rapid evaluation analysis method of using spectral data and viscosity-temperature curve data in the above embodiments, the accuracy of the rapid evaluation analysis results of physical property indicators of various raw materials or intermediate products in the petrochemical field has been greatly improved. After combining the spectral data or wave spectral data with the viscosity-temperature curve data, and then correlating them with each physical property indicator, the prediction effect of the models established by various modeling methods is far better than that of the models established solely using spectral data. The error of the prediction results drops significantly, and the accuracy is significantly improved. Therefore, on the basis of the existing technology, introducing viscosity-temperature curve data into the field of rapid evaluation analysis of petrochemical materials can significantly improve the accuracy of the analysis models of various physical property indicators of oil products, greatly making up for the defect that the accuracy of the prediction results of complex physical properties by the current modeling methods is not high, and thus has extremely high application value.

[0109] On the basis of the existing technology, the present invention creatively proposes a method for establishing a rapid evaluation model by combining spectral data and viscosity-temperature curve data, further improving the accuracy of the rapid evaluation model. The oil product spectrum or wave spectrum mainly reflects the group characteristics of each organic molecule in the sample, and the viscosity information at each temperature point is conducive to analyzing the physical and chemical properties of heavy oil. Therefore, by combining the spectral data and viscosity-temperature curve data, the oil product information reflected will be more detailed, rich and accurate. On this basis, the physical property analysis model constructed has high accuracy and can accurately predict the physical property indicators of the oil product.

[0110] Embodiment III

[0111] Please refer to Figure 4 which is a schematic structural diagram of an oil product physical property rapid evaluation model establishment device provided by an embodiment of the present invention.

[0112] The embodiment of the present invention provides an oil product physical property rapid evaluation model establishment device, including: an acquisition module 301, a preprocessing module 302, a modeling module 303, and a screening module 304.

[0113] In this embodiment, the acquisition module 301 is used to acquire sample data of the oil product to be modeled and divide it into a calibration set and a validation set; wherein, the sample data of the oil product to be modeled includes spectral data, viscosity-temperature curve data, and assay data corresponding to each sample; the spectral data is the spectral data or wave spectrum data of the oil product to be modeled;

[0114] The acquisition module uses the Kolmogorov-Smirnov algorithm to divide all samples into two categories: a calibration set and a validation set. Among them, the calibration set data is used to build the model, and the validation set data is used to verify and evaluate the model.

[0115] The preprocessing module 302 is used to preprocess the calibration set data through a variety of data processing methods to obtain multiple corresponding preprocessed data sets; wherein, the preprocessing module includes: a spectral data processing unit and a viscosity-temperature curve data processing unit.

[0116] The spectral data processing unit is used to preprocess the spectral data, specifically:

[0117] For the spectral data of each sample in the calibration set, multiple first data processing methods are respectively selected for preprocessing, and the preprocessed spectral data corresponding to each sample is obtained. Among them, the first data processing methods include: no processing, mean centering processing, mean variance processing, vector normalization processing, maximum minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, baseline correction method processing, at least two or more combinations thereof;

[0118] After obtaining the preprocessed spectral data corresponding to each sample, according to the preset dimensionality reduction method, the preprocessed spectral data is respectively subjected to dimensionality reduction processing; wherein, the dimensionality reduction method includes the simple model method or the principal component analysis method;

[0119] The viscosity-temperature curve data processing unit is used to preprocess the viscosity-temperature curve data, specifically:

[0120] For the viscosity-temperature curve data of each sample in the calibration set, an empirical formula of the viscosity-temperature curve is fitted by using a preset algorithm, and corresponding viscosity values are calculated according to a preset number of temperature points. Then, according to a variety of second data processing methods, the multiple temperature points and the viscosity values corresponding to each temperature point are respectively preprocessed to obtain the preprocessed viscosity-temperature curve data corresponding to each sample; wherein, each viscosity-temperature curve data corresponds to multiple temperature points and the viscosity values corresponding to each temperature point; the second data processing methods include: logarithmic processing, mean processing, mean variance processing, maximum minimum normalization processing, at least one or more combinations thereof.

[0121] The modeling module 303 is used to merge the spectral data and viscosity-temperature curve data of each preprocessed data set to form multiple corresponding associated data sets, and through a variety of association algorithms, associate them with the assay data sets corresponding to each sample respectively to construct multiple corresponding physical property analysis models.

[0122] The modeling module presets a variety of association algorithms for modeling. Among them, the variety of association algorithms includes at least one or a combination of multiple of partial least squares method, artificial neural network, and kernel partial least squares method.

[0123] The screening module 304 is used to verify and evaluate each physical property analysis model with the validation set data, and screen out the optimal model as the physical property quick evaluation model of the oil product to be modeled.

[0124] The screening module is used to verify and evaluate each physical property analysis model with the validation set data, and screen out the optimal model as the physical property quick evaluation model of the oil product to be modeled. Specifically:

[0125] Merge the spectral data and viscosity-temperature curve data of each sample in the validation set to form a data set to be predicted, and use each physical property analysis model for prediction respectively to obtain multiple corresponding prediction result sets;

[0126] Compare each prediction result set with the assay data set corresponding to each sample in the validation set, and screen out the model with the smallest error in the validation set as the physical property quick evaluation model of the oil product to be modeled.

[0127] The device provided in this embodiment creatively uses a quick evaluation analysis mode that combines spectral data and viscosity-temperature curve data, further improving the accuracy of the quick evaluation model. The oil product spectrum or wave spectrum mainly reflects the group characteristics of each organic molecule in the sample, and the viscosity information at each temperature point is beneficial to the analysis of the physical and chemical properties of heavy oil. Therefore, combining the spectral data and viscosity-temperature curve data, the oil product information reflected will be more detailed, rich and accurate. On this basis, the constructed physical property analysis model has high accuracy and can accurately predict the material physical property indexes of the oil product.

[0128] The above specific embodiments have further elaborated on the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. In particular, it is pointed out that for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for establishing a rapid evaluation model of oil product physical properties, characterized in that Including: Obtain the sample data of the oil product to be modeled and divide it into a calibration set and a validation set; wherein, the sample data of the oil product to be modeled includes the spectral data, viscosity-temperature curve data of each sample, and the assay data corresponding to each sample; the spectral data is the spectral data or wave spectrum data of the oil product to be modeled; Through a variety of data processing methods, preprocess the calibration set data respectively to obtain multiple corresponding preprocessed data sets; Merge the spectral data and viscosity-temperature curve data of each preprocessed data set respectively to form multiple corresponding associated data sets, and through a variety of association algorithms, associate with the assay data sets corresponding to each sample respectively to construct multiple corresponding physical property analysis models; Use the validation set data to verify and evaluate each physical property analysis model respectively, and select the optimal model as the physical property rapid evaluation model of the oil product to be modeled; Wherein, the step of through a variety of data processing methods, preprocess the calibration set data respectively to obtain multiple corresponding preprocessed data sets is specifically: For the spectral data of each sample in the calibration set, select a variety of first data processing methods to preprocess respectively, and obtain the preprocessed spectral data corresponding to each sample, wherein the first data processing methods include: no processing, mean centering processing, mean variance processing, vector normalization processing, maximum-minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, baseline correction method processing, at least two or more combinations thereof; after obtaining the preprocessed spectral data corresponding to each sample, it further includes: according to the preset dimensionality reduction method, perform dimensionality reduction processing on each of the preprocessed spectral data respectively; wherein, the dimensionality reduction method includes the simple model method or the principal component analysis method; For the viscosity-temperature curve data of each sample in the calibration set, use a preset algorithm to fit the empirical formula of the viscosity-temperature curve, and calculate the corresponding viscosity values according to a preset number of temperature points, and then according to a variety of second data processing methods, preprocess the multiple temperature points and the viscosity values corresponding to each temperature point respectively to obtain the preprocessed viscosity-temperature curve data corresponding to each sample; wherein, each viscosity-temperature curve data corresponds to multiple temperature points and the viscosity values corresponding to each temperature point; the second data processing methods include: logarithmic processing, mean processing, mean variance processing, maximum-minimum normalization processing, at least one or more combinations thereof; Wherein, the variety of association algorithms include: partial least squares method, artificial neural network, kernel partial least squares method, at least one or more combinations thereof.

2. The method for establishing a rapid evaluation model of oil product physical properties according to claim 1, characterized in that, The step of obtain the sample data of the oil product to be modeled and divide it into a calibration set and a validation set is specifically: use the Kolmogorov-Smirnov algorithm to divide all samples into two categories: a calibration set and a validation set.

3. The method for establishing a rapid evaluation model of oil product physical properties according to claim 1, wherein, The step of use the validation set data to verify and evaluate each physical property analysis model respectively, and select the optimal model as the physical property rapid evaluation model of the oil product to be modeled is specifically: The spectral data and viscosity-temperature curve data of each sample in the validation set are combined to form a dataset to be predicted, and each physical property analysis model is used for prediction respectively to obtain multiple corresponding prediction result sets; Each prediction result set is compared with the assay dataset corresponding to each sample in the validation set respectively, and the model with the smallest error in the validation set is selected as the physical property rapid evaluation model of the oil product to be modeled.

4. An apparatus for establishing a rapid evaluation model of oil product physical properties, characterized in that, It includes: An acquisition module, a preprocessing module, a modeling module, and a screening module; The acquisition module is used to acquire the sample data of the oil product to be modeled and divide it into a calibration set and a validation set; wherein, the sample data of the oil product to be modeled includes the spectral data, viscosity-temperature curve data of each sample, and the assay data corresponding to each sample; the spectral data is the spectral data or wave spectrum data of the oil product to be modeled; The preprocessing module is used to preprocess the calibration set data respectively through a variety of data processing methods to obtain multiple corresponding preprocessed datasets; The modeling module is used to combine the spectral data and viscosity-temperature curve data of each preprocessed dataset respectively to form multiple corresponding correlation datasets, and associate them with the assay datasets corresponding to each sample respectively through a variety of correlation algorithms to construct multiple corresponding physical property analysis models; The screening module is used to verify and evaluate each physical property analysis model with the validation set data respectively, and select the optimal model as the physical property rapid evaluation model of the oil product to be modeled; Among them, the calibration set data is preprocessed respectively through a variety of data processing methods to obtain multiple corresponding preprocessed datasets, specifically: For the spectral data of each sample in the calibration set, a variety of first data processing methods are respectively selected for preprocessing, and the preprocessed spectral data corresponding to each sample is obtained. Among them, the first data processing methods include: at least two or more combinations of no processing, mean centering processing, mean variance processing, vector normalization processing, maximum-minimum normalization processing, standard normal variable transformation processing, multiplicative scatter correction processing, Savitzky-Golay convolution smoothing processing, first derivative method processing, detrending method processing, baseline correction method processing; after obtaining the preprocessed spectral data corresponding to each sample, it further includes: performing dimensionality reduction processing on each preprocessed spectral data respectively according to a preset dimensionality reduction method; wherein, the dimensionality reduction method includes the simple model method or the principal component analysis method; For the viscosity-temperature curve data of each sample in the calibration set, an empirical formula of the viscosity-temperature curve is fitted by using a preset algorithm, and the corresponding viscosity values are calculated according to a preset number of temperature points. Then, according to a variety of second data processing methods, the multiple temperature points and the viscosity values corresponding to each temperature point are preprocessed respectively to obtain the preprocessed viscosity-temperature curve data corresponding to each sample; wherein, each viscosity-temperature curve data corresponds to multiple temperature points and the viscosity values corresponding to each temperature point; the second data processing methods include: at least one or more combinations of logarithmic processing, mean processing, mean variance processing, maximum-minimum normalization processing; Among them, the multiple correlation algorithms include at least one or a combination of multiple ones among partial least squares method, artificial neural network, and kernel partial least squares method.

5. The device for establishing a quick evaluation model of oil product physical properties according to claim 4, characterized in that The screening module is used to respectively verify and evaluate each physical property analysis model with the validation set data, and screen out the optimal model as the physical property quick evaluation model of the oil product to be modeled. Specifically: Merge the spectral data and viscosity-temperature curve data of each sample in the validation set to form a dataset to be predicted, and respectively use each physical property analysis model for prediction to obtain multiple corresponding prediction result sets; Compare each prediction result set with the test dataset corresponding to each sample in the validation set respectively, and screen out the model with the smallest error in the validation set as the physical property quick evaluation model of the oil product to be modeled.

Images