Method for generating oil product molecular composition prediction network, prediction method and device

Abstract

The application discloses a generation method and a prediction method and device of an oil product molecular composition prediction network. The generation method comprises the following steps: obtaining an original data set; training a pre-constructed generator based on the original data set to obtain first network parameters of the generator; inputting macroscopic physical property data of a randomly generated to-be-tested oil product into the generator to obtain molecular composition data of the to-be-tested oil product generated by the generator; keeping the first network parameters of the generator unchanged, training a pre-constructed discriminator connected with the generator based on the molecular composition data of the known oil product in the original data set and the molecular composition data of the to-be-tested oil product generated by the generator to obtain second network parameters of the discriminator; keeping the second network parameters of the discriminator unchanged, and jointly training the trained generator and the discriminator based on the original data set to obtain an oil product molecular composition prediction network. The prediction network can predict the full molecular composition of the oil product.

Description

Technical Field

[0001] This invention relates to the field of petroleum processing technology, and in particular to a method, method and apparatus for generating and predicting oil molecular composition prediction networks. Background Technology

[0002] Molecular management was proposed relatively early, and with the increasing maturity of computer technology, obtaining the molecular composition of oil products through software prediction and simulation is convenient and labor-saving, and has become the future development trend.

[0003] Currently, in the prediction of refined oil products, due to the complex and diverse molecular composition of oil products, it is necessary to deduce hundreds of molecular compositions from a dozen or so macroscopic properties. This necessitates a certain number of constraint formulas, but these constraint formulas themselves contain inherent errors, and an excessive number of constraint formulas can affect the final prediction results. Most previous prediction methods employed a lumped approach, reducing hundreds of oil product molecules into sets of dozens or even a hundred, thereby decreasing the number of solutions to the objective function and improving prediction accuracy. However, current schemes categorize molecules through lumping rather than predicting the complete molecular composition of the oil product. Summary of the Invention

[0004] The purpose of this invention is to provide a method, a method, and an apparatus for generating a prediction network for the molecular composition of oil products, which can be used to predict the total molecular composition of oil products.

[0005] To achieve the above-mentioned technical objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for generating an oil product molecular composition prediction network, the method comprising:

[0007] Obtain the original dataset; wherein the original dataset contains multiple sets of labeled macroscopic physical property data of known oil products, and the labels of the macroscopic physical property data of the known oil products are the molecular composition data of the known oil products;

[0008] Based on the original dataset, the pre-built generator is trained to obtain the first network parameters of the generator;

[0009] The macroscopic physical property data of the randomly generated oil to be tested are input into the generator with the first network parameters to obtain the molecular composition data of the oil to be tested generated by the generator with the first network parameters.

[0010] Keeping the first network parameters of the generator unchanged, the discriminator connected to the generator is trained based on the molecular composition data of the known oil products in the original dataset and the molecular composition data of the oil products to be tested generated by the generator, so as to obtain the second network parameters of the discriminator.

[0011] Keeping the second network parameters of the discriminator unchanged, the generator and the discriminator are jointly trained based on the original dataset to obtain the oil molecular composition prediction network.

[0012] Furthermore, training the pre-built generator based on the original dataset includes:

[0013] The macroscopic physical property data of the first batch of labeled known oil products are selected from the original dataset, and the pre-built generator is trained in a supervised manner using the macroscopic physical property data of the first batch of labeled known oil products.

[0014] Further, the step of training a pre-built discriminator connected to the generator based on the molecular composition data of known oil products in the original dataset and the molecular composition data of the oil product to be tested generated by the generator includes:

[0015] Select a batch of molecular composition data of known oil products from the original dataset, and label the categories of the selected molecular composition data of known oil products and the multiple sets of molecular composition data of oil products to be tested generated by the generator.

[0016] Based on the labeled molecular composition data of the known oil and the molecular composition data of the oil to be tested, the discriminator is subjected to supervised training.

[0017] During discriminator training, the generator's first network parameters remain unchanged, while the discriminator's second network parameters are adjusted based on the error between the discriminator's predicted output and the label, until the discriminator can distinguish whether the molecular composition data input to the discriminator comes from a known oil product or the oil product to be tested.

[0018] Further, the step of jointly training the generator and the discriminator based on the original dataset includes:

[0019] A second batch of labeled macroscopic property data of known oil products is selected from the original dataset. The generator and discriminator are then jointly trained using the second batch of labeled macroscopic property data of known oil products.

[0020] Furthermore, when jointly training the trained generator and the discriminator, the second network parameters of the discriminator remain unchanged, and the first network parameters of the generator are adjusted according to the error between the molecular composition data output by the generator and the label and the error between the predicted value output by the discriminator and the label.

[0021] The loss function for the joint training is the cross-entropy loss function;

[0022] The formula for calculating the cross-entropy loss function is as follows:

[0023]

[0024] Among them, y i The true value of the content of the i-th molecule in the known oil product is input into the generator during a training session. The predicted content of the i-th molecule output by the generator in this training is denoted as n, where n is the number of predefined molecule types.

[0025] Furthermore, the generation method further includes: iteratively training the obtained oil molecule composition prediction network, including repeating the following training steps:

[0026] Based on the original dataset, the pre-built generator is trained to obtain the first network parameters of the generator;

[0027] The macroscopic physical property data of the randomly generated oil to be tested are input into the generator with the first network parameters to obtain the molecular composition data of the oil to be tested generated by the generator with the first network parameters.

[0028] Keeping the first network parameters of the generator unchanged, the discriminator connected to the generator is trained based on the molecular composition data of the known oil products in the original dataset and the molecular composition data of the oil products to be tested generated by the generator, so as to obtain the second network parameters of the discriminator.

[0029] Keeping the second network parameters of the discriminator unchanged, the generator and the discriminator are jointly trained based on the original dataset to obtain the oil molecular composition prediction network.

[0030] Furthermore, after obtaining the oil molecular composition prediction network, the generation method further includes:

[0031] Macroscopic property data of a third batch of labeled known oil products were selected from the original dataset;

[0032] For any set of macroscopic physical property data of the known oil products in the third batch with labels, input the set of macroscopic physical property data into the oil product molecular composition prediction network to obtain the molecular composition data of the known oil products generated by the oil product molecular composition prediction network; wherein, the generated molecular composition data of the known oil products includes the molecular types of the known oil products and the predicted content of each molecular type.

[0033] The various molecules in the known oil are sorted according to their predicted content, and the absolute or relative error between the predicted content and the actual content of each molecule in the known oil is calculated.

[0034] Calculate the sum of the actual contents corresponding to a set number of molecules with high predicted content in the known oil product;

[0035] If the sum of the actual contents is lower than the first threshold, then the set quantity is incremented by 1, and the process returns to the step of calculating the sum of the actual contents.

[0036] If the sum of the actual contents is not lower than the first threshold, then the average error or average relative error at this time is determined.

[0037] If the average error or average relative error does not exceed the second threshold, the oil molecular composition prediction network is deemed qualified; otherwise, it is deemed unqualified.

[0038] This invention also provides a method for predicting the molecular composition of oil products, the method comprising:

[0039] Obtain macroscopic physical property data of the oil to be tested;

[0040] The macroscopic physical property data of the oil to be tested are input into a pre-trained oil molecular composition prediction network, which is trained and generated according to the above-mentioned method for generating oil molecular composition prediction networks.

[0041] Obtain the molecular composition data of the oil to be tested generated by the oil molecular composition prediction network.

[0042] The present invention also provides an apparatus for generating an oil molecular composition prediction network, the apparatus comprising:

[0043] The acquisition module is configured to acquire a raw dataset; wherein the raw dataset contains multiple sets of labeled macroscopic physical property data of known oil products, and the labels of the macroscopic physical property data of the known oil products are the molecular composition data of the known oil products;

[0044] The first training module is configured to train a pre-built generator based on the original dataset to obtain the first network parameters of the generator.

[0045] The molecular composition acquisition module is configured to input the macroscopic physical property data of the randomly generated oil sample to be tested into a generator having the first network parameters, and obtain the molecular composition data of the oil sample to be tested generated by the generator having the first network parameters.

[0046] The second training module is configured to keep the first network parameters of the generator unchanged, and train a pre-built discriminator connected to the generator based on the molecular composition data of the known oil products in the original dataset and the molecular composition data of the oil products to be tested generated by the generator, so as to obtain the second network parameters of the discriminator.

[0047] The third training module is configured to keep the second network parameters of the discriminator unchanged, and to jointly train the trained generator and the discriminator based on the original dataset to obtain the oil molecular composition prediction network.

[0048] Furthermore, the first training module is configured as follows:

[0049] The macroscopic physical property data of the first batch of labeled known oil products are selected from the original dataset, and the pre-built generator is trained in a supervised manner using the macroscopic physical property data of the first batch of labeled known oil products.

[0050] Furthermore, the second training module is configured as follows:

[0051] Select a batch of molecular composition data of known oil products from the original dataset, and label the categories of the selected molecular composition data of known oil products and the multiple sets of molecular composition data of oil products to be tested generated by the generator.

[0052] The discriminator is trained in a supervised manner based on the molecular composition data of the known oil product after labeling and the molecular composition data of the oil product to be tested.

[0053] Furthermore, the third training module is configured as follows:

[0054] A second batch of labeled macroscopic property data of known oil products is selected from the original dataset. The generator and discriminator are then jointly trained using the second batch of labeled macroscopic property data of known oil products.

[0055] Furthermore, when jointly training the trained generator and the discriminator, the loss function for the joint training is the cross-entropy loss function;

[0056] The formula for calculating the cross-entropy loss function is as follows:

[0057]

[0058] Among them, y i The true value of the content of the i-th molecule in the known oil product is input into the generator during a training session. The predicted content of the i-th molecule output by the generator in this training is denoted as n, where n is the number of predefined molecule types.

[0059] Furthermore, the device for generating the oil molecular composition prediction network also includes: a testing module;

[0060] The test module is configured as follows:

[0061] Macroscopic property data of a third batch of labeled known oil products were selected from the original dataset;

[0062] For any set of macroscopic physical property data of the known oil products in the third batch with labels, input the set of macroscopic physical property data into the oil product molecular composition prediction network to obtain the molecular composition data of the known oil products generated by the oil product molecular composition prediction network; wherein, the generated molecular composition data of the known oil products includes the molecular types of the known oil products and the predicted content of each molecular type.

[0063] The various molecules in the known oil are sorted according to their predicted content, and the absolute or relative error between the predicted content and the actual content of each molecule in the known oil is calculated.

[0064] Calculate the sum of the actual contents corresponding to a set number of molecules with high predicted content in the known oil product;

[0065] If the sum of the actual contents is lower than the first threshold, then the set quantity is incremented by 1, and the process returns to the step of calculating the sum of the errors;

[0066] If the sum of the actual contents is not lower than the first threshold, then the average error or average relative error at this time is determined.

[0067] If the average error or average relative error does not exceed the second threshold, the oil molecular composition prediction network is deemed qualified; otherwise, it is deemed unqualified.

[0068] The present invention also provides an oil molecular composition prediction device, the device comprising:

[0069] The property acquisition module is configured to acquire macroscopic property data of the oil to be tested;

[0070] The molecular composition prediction module is configured to input the macroscopic physical property data of the oil to be tested into a pre-trained oil molecular composition prediction network, which is trained and generated according to the above-described generation method; and obtain the molecular composition data of the oil to be tested generated by the oil molecular composition prediction network.

[0071] The present invention also provides a storage medium storing a computer program for executing the steps of the above-described method for generating an oil molecular composition prediction network or the steps of the oil molecular composition prediction method.

[0072] The present invention also provides an electronic device, including a processor and a computer program running on the processor, the computer program being used to execute the steps of the above-described method for generating an oil molecular composition prediction network or the steps of the oil molecular composition prediction method.

[0073] Furthermore, it also includes a memory, on which the computer program is stored and runs on the processor.

[0074] Compared with existing technologies, the present invention has the following beneficial effects: First, based on the original dataset, a pre-constructed generator is trained to obtain the first network parameters of the generator. Then, randomly generated macroscopic property data of the oil to be tested are input into the generator with the first network parameters to obtain molecular composition data of the oil to be tested generated by the generator with the first network parameters. Then, keeping the first network parameters of the generator unchanged, based on the known molecular composition data of oils in the original dataset and the molecular composition data of the oil to be tested generated by the generator, a pre-constructed discriminator connected to the generator is trained to obtain the second network parameters of the discriminator. Then, keeping the second network parameters of the discriminator unchanged, based on the original dataset, the trained generator and discriminator are jointly trained to obtain an oil molecular composition prediction network. Using this oil molecular composition prediction network, the complete molecular composition of the oil can be predicted. Through further iterative training, the generator in the oil molecular composition prediction network can accurately generate molecular composition data. Attached Figure Description

[0075] Figure 1 A schematic flowchart illustrating the method for generating an oil molecular composition prediction network according to an embodiment of the present invention;

[0076] Figure 2 This is a schematic diagram of the generator provided in an embodiment of the present invention;

[0077] Figure 3 This is a schematic diagram of the structure of the oil molecular composition prediction network provided in an embodiment of the present invention;

[0078] Figure 4 This is a schematic diagram of the discriminator provided in an embodiment of the present invention;

[0079] Figure 5 A schematic flowchart of the method for predicting the molecular composition of oil products provided in an embodiment of the present invention;

[0080] Figure 6 A schematic diagram of a first structure of the device for generating an oil molecular composition prediction network provided in an embodiment of the present invention;

[0081] Figure 7 A schematic diagram of a second structure of the device for generating an oil molecular composition prediction network provided in an embodiment of the present invention;

[0082] Figure 8This is a schematic diagram of the structure of the oil molecular composition prediction device provided in an embodiment of the present invention;

[0083] Figure 9 This is a first structural schematic diagram of an electronic device provided in an embodiment of the present invention;

[0084] Figure 10 This is a schematic diagram of the second structure of an electronic device provided in an embodiment of the present invention.

[0085] The reference numerals in the accompanying drawings are as follows: acquisition module 11, first training module 12, molecular composition acquisition module 13, second training module 14, third training module 15, testing module 16, physical property acquisition module 21, molecular composition prediction module 22, first memory 31, first processor 32, first memory 33, first network interface 34, first internal bus 35, second memory 41, second processor 42, second memory 43, second network interface 44, second internal bus 45, generator 10, and discriminator 20. Detailed Implementation

[0086] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. When referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the present invention as detailed in the appended claims. The following embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention.

[0087] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The singular forms “a,” “the,” and “the” used in this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0088] It should be understood that although the terms first, second, third, etc., may be used in this invention to describe various information, this information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this invention, first identification information may also be referred to as second identification information, and similarly, second identification information may also be referred to as first identification information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0089] This invention utilizes the characteristics of Generative Adversarial Networks (GANs) to construct an oil molecular composition prediction network to predict the full molecular composition of oil products.

[0090] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0091] See Figure 1 This invention provides a method for generating an oil molecular composition prediction network. This method can be used in electronic devices, such as PCs or oil molecular composition prediction devices. The method includes the following steps:

[0092] S101. Obtain the original dataset;

[0093] The original dataset contains multiple sets of labeled macroscopic property data for known oil products, where the labels for these macroscopic property data are the molecular composition data of the known oil products. The original dataset can be used for supervised training.

[0094] In this embodiment of the invention, the macroscopic physical property data of the oil may include: density, cloud point, pour point, aniline point, octane number, cetane number, pour point, cold filter plugging point, flash point, vapor pressure and other physical property data. Each set of macroscopic physical property data may include data of 18 macroscopic physical properties.

[0095] The molecular composition data of petroleum products includes: the types of molecules in the petroleum product and the corresponding content of each molecule. For example, the molecular composition data of each group of petroleum products, such as gasoline, can include data on 590 molecules.

[0096] It is known that the macroscopic physical properties and molecular composition data of oil products can be obtained from existing oil product testing reports.

[0097] S102. Based on the original dataset, train the pre-built generator 10 to obtain the first network parameters of the generator 10;

[0098] In this embodiment, the pre-built generator 10 is used to generate molecular composition data based on the input macroscopic physical property data of the oil, and output the molecular composition data to the discriminator 20; the generator 10 includes multiple (e.g., 5) fully connected neural network layers, each of which is followed by a ReLU layer (or activation layer), and the structure of the generator 10 is as follows. Figure 2 As shown. In some embodiments, the generator 10 has five fully connected neural network layers from left to right, each containing 18 nodes, 64 nodes, 128 nodes, 256 nodes, and 590 nodes, respectively.

[0099] In some embodiments, when training generator 10 alone, step S102 involves training the pre-built generator 10 based on the original dataset, including:

[0100] The first batch (e.g., 30 groups) of labeled macroscopic property data of known oil products are selected from the original dataset. The pre-constructed generator 10 is then trained in a supervised manner using the first batch of labeled macroscopic property data of known oil products.

[0101] S103. Input the macroscopic physical property data of the randomly generated oil to be tested into the generator 10 with the first network parameters to obtain the molecular composition data of the oil to be tested generated by the generator 10 with the first network parameters.

[0102] S104. Keeping the first network parameters of the generator 10 unchanged, based on the molecular composition data of the known oil products in the original dataset and the molecular composition data of the oil products to be tested generated by the generator 10, the pre-built discriminator 20 connected to the generator 10 is trained to obtain the second network parameters of the discriminator 20.

[0103] In this embodiment, participants Figure 3 Generator 10 is connected to discriminator 20. The pre-built discriminator 20 is used to determine whether the molecular composition data input to it is from a known oil product or the oil product to be tested. This discriminator 20, similar to the discriminator in a conventional generative adversarial network, may include, for example, multiple (e.g., 5) fully connected neural network layers and a softmax function classifier. Each of the first four fully connected neural network layers is followed by a ReLU layer (or activation layer). The softmax function serves as the activation function for the output layer of discriminator 20, connected after the fifth fully connected neural network layer. The structure of discriminator 20 is as follows: Figure 4 As shown. In some embodiments, the discriminator 20 has five fully connected neural network layers from left to right, each containing 590 nodes, 256 nodes, 128 nodes, 64 nodes, and 1 node, respectively. The source information of the molecular composition data output by the softmax function classifier includes first identification information and second identification information. The first identification information (e.g., 0) indicates that the molecular composition data input to the discriminator 20 comes from a known oil product, and the second identification information (e.g., 1) indicates that the molecular composition data input to the discriminator 20 comes from the oil product to be tested.

[0104] In some embodiments, in step S104, based on the molecular composition data of known oil products in the original dataset and the molecular composition data of the oil product to be tested generated by the generator 10, the pre-built discriminator 20 connected to the generator 10 is trained, including:

[0105] Select a batch (e.g., 15 groups) of molecular composition data of known oil products from the original dataset, and label the categories of the selected molecular composition data of known oil products and the multiple groups (e.g., 15 groups) of molecular composition data of oil products to be tested generated by the generator 10 (e.g., the molecular composition data of known oil products is labeled as 0, and the molecular composition data of oil products to be tested is labeled as 1).

[0106] Based on the labeled molecular composition data of the known oil and the molecular composition data of the oil to be tested, the discriminator 20 is trained in a supervised manner.

[0107] In step S104 above, during the training of the discriminator 20, the first network parameter of the generator 10 remains unchanged, and the second network parameter of the discriminator 20 is adjusted according to the error between the predicted value output by the discriminator 20 and the label, so that when the discriminator 20 receives molecular composition data, it can output a value close to 0 or close to 1, that is, it can better distinguish whether the molecular composition data input to the discriminator 20 is a known oil or a test oil.

[0108] S105. Keeping the second network parameters of the discriminator 20 unchanged, the generator 10 and the discriminator 20 after training are jointly trained based on the original dataset to obtain the oil molecular composition prediction network.

[0109] In some embodiments, step S105, based on the original dataset, involves jointly training the trained generator 10 and discriminator 20, including:

[0110] A second batch (e.g., 30 groups) of labeled macroscopic property data of known oil products is selected from the original dataset. The generator 10 and the discriminator 20 are then jointly trained using the second batch of labeled macroscopic property data of known oil products.

[0111] In step S105, during joint training, the second network parameters of the discriminator 20 remain unchanged. Based on the error between the molecular composition data output by the generator 10 and the label, and the error between the predicted value output by the discriminator 10 and the label, the first network parameters of the generator 10 are adjusted so that the generator 10 can generate molecular composition data more accurately. When the molecular composition data generated by the generator 10 is sufficiently accurate, the discriminator 20 will find it difficult to distinguish whether the molecular composition data input to the discriminator 20 comes from known oil samples selected from the original dataset or from the oil sample to be tested generated by the generator 10.

[0112] In step S105, when the generator 10 and discriminator 20 are jointly trained, the loss function for the joint training is the cross-entropy loss function.

[0113] The formula for calculating the cross-entropy loss function is as follows:

[0114]

[0115] Among them, y i The true value of the content of the i-th molecule in the known oil product is input into the generator 10 during a training session. The predicted content of the i-th molecule output by the generator 10 in this training is n, where n is the number of predefined molecule types (e.g., 590 molecules).

[0116] In some embodiments, in order to ensure that the generator 10 in the oil molecular composition prediction network can accurately generate molecular composition data, the training steps in steps S102-S105 can be repeated. After several iterations, the generator 10 in the oil molecular composition prediction network can accurately generate molecular composition data.

[0117] In some embodiments, after obtaining the oil molecular composition prediction network, the method further includes:

[0118] Select a third batch (e.g., 30 groups) of macroscopic physical property data of labeled known oil products from the original dataset;

[0119] For any set of macroscopic physical property data of the known oil products in the third batch with labels, input the set of macroscopic physical property data into the oil product molecular composition prediction network to obtain the molecular composition data of the known oil products generated by the oil product molecular composition prediction network; wherein, the generated molecular composition data of the known oil products includes the molecular types of the known oil products and the predicted content of each molecular type.

[0120] The various molecules in the known oil are sorted according to their predicted content, and the absolute or relative error between the predicted content and the actual content of each molecule in the known oil is calculated.

[0121] Calculate the sum of the actual contents corresponding to a set number of molecules with high predicted content in the known oil product;

[0122] If the sum of the actual contents is lower than the first threshold, then the set quantity is incremented by 1, and the process returns to the step of calculating the sum of the actual contents.

[0123] If the sum of the actual contents is not lower than the first threshold, then the average error or average relative error at this time is determined.

[0124] If the average error or average relative error does not exceed the second threshold, the oil molecular composition prediction network is deemed qualified; otherwise, it is deemed unqualified.

[0125] For example, select 23 molecules with high predicted content, calculate the sum of the actual contents of these 23 molecules. If the sum of the actual contents is less than 97%, increment the set quantity by 1 and return to the step of calculating the sum of the actual contents. If the sum of the actual contents is not less than 97%, determine the average relative error at this time. If the average relative error does not exceed 6%, the trained oil molecule composition prediction network is deemed qualified. If the average relative error exceeds 6%, the trained oil molecule composition prediction network is deemed unqualified.

[0126] The following section uses gasoline as an example to introduce the test results of the oil molecule composition prediction network constructed using the oil molecule composition prediction network generation method provided in the embodiments of the present invention.

[0127] Multiple sets of labeled macroscopic physical property data of gasoline are selected from the original dataset as test sets. For any set of macroscopic physical property data of gasoline in the test set (including 18 macroscopic physical property data such as octane number and vapor pressure), the set of macroscopic physical property data is input into the oil molecular composition prediction network to obtain the molecular composition data of the gasoline generated by the oil molecular composition prediction network (including data of 590 molecules), as shown in Table 1. The molecular composition data generated by generator 10 includes the types of gasoline molecules and the predicted content of each molecule. The various molecules in the gasoline are sorted in descending order of predicted content, and the relative error between the predicted content and the actual content of each molecule in the gasoline is calculated.

[0128] Table 1. Molecular composition data generated by generator 10 and their relative errors with actual content.

[0129]

[0130] As can be seen from Table 1, the relative error of 1,2,3,5-tetramethylbenzene content is the largest, with a maximum relative error of 9.88%. Selecting 23 molecules with high predicted content, the sum of their actual contents is 97.864%, which is greater than the first threshold of 97%. The average relative error at this point is 4.64%, which is less than the second threshold of 6%. Therefore, it can be determined that the trained oil molecule composition prediction network is qualified. In other words, the prediction results of this oil molecule composition prediction network are basically consistent with the actual results, and it can be considered to reflect the molecular composition of gasoline and the content of each molecule relatively well.

[0131] Based on the same inventive concept, see [link to inventive concept] Figure 5 This invention also provides a method for predicting the molecular composition of oil products. This method can be used in electronic devices, such as PCs or oil molecular composition prediction devices. The method includes the following steps:

[0132] S201. Obtain the macroscopic physical property data of the oil to be tested;

[0133] S202, Input the macroscopic physical property data of the oil to be tested into a pre-trained oil molecular composition prediction network;

[0134] The oil molecule composition prediction network is generated by training the oil molecule composition prediction network generation method according to any possible implementation of the present invention.

[0135] S203. Obtain the molecular composition data of the oil to be tested generated by the oil molecular composition prediction network.

[0136] In this embodiment of the invention, in the oil molecular composition prediction network obtained after training, the discriminator 20 has difficulty distinguishing whether the molecular composition data input to the discriminator 20 comes from a known oil product selected from the original dataset or the oil product to be tested generated by the generator 10. This enables the generator 10 to generate molecular composition data more accurately.

[0137] It should be noted that the oil molecular composition prediction network provided in this embodiment of the invention can be used to predict the molecular composition of gasoline, as well as the molecular composition of heavy oil and other oil products.

[0138] Based on the same inventive concept, see [link to inventive concept] Figure 6 The present invention also provides a device for generating an oil molecular composition prediction network, the device comprising: an acquisition module 11, a first training module 12, a molecular composition acquisition module 13, a second training module 14, and a third training module 15.

[0139] The acquisition module 11 is configured to acquire a raw dataset; wherein the raw dataset contains multiple sets of labeled macroscopic physical property data of known oil products, and the labels of the macroscopic physical property data of the known oil products are the molecular composition data of the known oil products;

[0140] The first training module 12 is configured to train the pre-built generator 10 based on the original dataset to obtain the first network parameters of the generator 10.

[0141] The pre-built generator 10 is used to generate molecular composition data from the input macroscopic physical property data of the oil, and outputs the molecular composition data to the discriminator 20; the generator 10 includes multiple fully connected neural network layers, and each fully connected neural network layer is connected to an activation layer.

[0142] The molecular composition acquisition module 13 is configured to input the macroscopic physical property data of the randomly generated oil to be tested into the generator 10 having the first network parameters, and obtain the molecular composition data of the oil to be tested generated by the generator 10 having the first network parameters.

[0143] The second training module 14 is configured to keep the first network parameters of the generator 10 unchanged, and train the pre-built discriminator 20 connected to the generator 10 based on the molecular composition data of the known oil products in the original dataset and the molecular composition data of the oil products to be tested generated by the generator 10, so as to obtain the second network parameters of the discriminator 20.

[0144] The pre-built discriminator 20 is used to determine whether the molecular composition data input to the discriminator 20 is from a known oil product or the oil product to be tested. The discriminator 20, similar to the discriminator of a conventional generative adversarial network, may include n fully connected neural network layers, n-1 activation layers and a softmax function classifier. Each of the first n-1 fully connected neural network layers is followed by an activation layer. The softmax function is used as the activation function of the output layer of the discriminator 20 and is connected after the nth fully connected neural network layer.

[0145] The third training module 15 is configured to keep the second network parameters of the discriminator 20 unchanged, and to jointly train the trained generator 10 and the discriminator 20 based on the original dataset to obtain the oil molecular composition prediction network.

[0146] In some embodiments, the first training module 12 is configured as follows:

[0147] The macroscopic physical property data of the first batch of labeled known oil products are selected from the original dataset, and the pre-constructed generator 10 is trained in a supervised manner using the macroscopic physical property data of the first batch of labeled known oil products.

[0148] In some embodiments, the second training module 14 is configured as follows:

[0149] Select a batch of molecular composition data of known oil products from the original dataset, and label the categories of the selected molecular composition data of known oil products and the multiple sets of molecular composition data of oil products to be tested generated by the generator 10.

[0150] Based on the labeled molecular composition data of the known oil and the molecular composition data of the oil to be tested, the discriminator 20 is trained in a supervised manner.

[0151] In some embodiments, the third training module 15 is configured as follows:

[0152] A second batch of labeled macroscopic property data of known oil products is selected from the original dataset. The generator 10 and the discriminator 20 are then jointly trained using the second batch of labeled macroscopic property data of known oil products.

[0153] In some embodiments, when the generator 10 and the discriminator 20 are jointly trained, the loss function for the joint training is the cross-entropy loss function;

[0154] The formula for calculating the cross-entropy loss function is as follows:

[0155]

[0156] Among them, y i The true value of the content of the i-th molecule in the known oil product is input into the generator 10 during a training session. The predicted content of the i-th molecule output by the generator 10 in this training is denoted as n, where n is the number of predefined molecule types.

[0157] In some embodiments, such as Figure 7 As shown, the device for generating the above-mentioned oil molecular composition prediction network also includes a test module 16.

[0158] Test module 16 is configured as follows:

[0159] Macroscopic property data of a third batch of labeled known oil products were selected from the original dataset;

[0160] For any set of macroscopic physical property data of the known oil products in the third batch with labels, input the set of macroscopic physical property data into the oil product molecular composition prediction network to obtain the molecular composition data of the known oil products generated by the oil product molecular composition prediction network; wherein, the generated molecular composition data of the known oil products includes the molecular types of the known oil products and the predicted content of each molecular type.

[0161] The various molecules in the known oil are sorted according to their predicted content, and the absolute or relative error between the predicted content and the actual content of each molecule in the known oil is calculated.

[0162] Calculate the sum of the actual contents corresponding to a set number of molecules with high predicted content in the known oil product;

[0163] If the sum of the actual contents is lower than the first threshold (e.g., the sum of relative errors < 97%), then the set quantity is incremented by 1, and the process returns to the step of calculating the sum of the actual contents.

[0164] If the sum of the actual contents is not lower than the first threshold, then the average error or average relative error at this time is determined.

[0165] If the average error or average relative error does not exceed the second threshold (e.g., average relative error ≤ 6%), then the oil molecular composition prediction network is deemed qualified; otherwise, it is deemed unqualified.

[0166] Based on the same inventive concept, see [link to inventive concept] Figure 8 This invention also provides an oil molecular composition prediction device, which includes:

[0167] The property acquisition module 21 is configured to acquire macroscopic property data of the oil to be tested;

[0168] The molecular composition prediction module 22 is configured to input the macroscopic physical property data of the oil to be tested into a pre-trained oil molecular composition prediction network to obtain the molecular composition data of the oil to be tested generated by the oil molecular composition prediction network.

[0169] The oil molecule composition prediction network is generated by training the oil molecule composition prediction network generation method according to any possible implementation of the present invention.

[0170] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0171] For the apparatus embodiments, since they basically correspond to the method embodiments, relevant details are referred to in the description of the method embodiments. The apparatus embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0172] Based on the same inventive concept, embodiments of the present invention also provide a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method for generating an oil molecule composition prediction network in any of the above possible implementations.

[0173] Optionally, the storage medium may be a non-transitory computer-readable storage medium, such as a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0174] Based on the same inventive concept, embodiments of the present invention also provide a computer program product, including a computer program, which, when executed by a processor, implements the steps of the method for generating an oil molecule composition prediction network in any of the above possible implementations.

[0175] Based on the same inventive concept, see [link to inventive concept] Figure 9This invention also provides an electronic device, including a first memory 31 (e.g., non-volatile memory), a first processor 32, and a computer program stored in the first memory 31 and executable on the first processor 32. When the first processor 32 executes the computer program, it implements the steps of the method for generating an oil molecular composition prediction network in any of the above possible implementations. This electronic device can be, for example, a PC or an oil molecular composition prediction device.

[0176] like Figure 9 As shown, the electronic device may also include: a first memory 33, a first network interface 34, and a first internal bus 35; wherein the first memory 31, the first processor 32, the first memory 33, and the first network interface 34 communicate with each other through the first internal bus 35. In addition to these components, other hardware may also be included, which will not be described in detail here.

[0177] It should be noted that the above-mentioned oil molecular composition prediction network generation device can be implemented by software. As a logical device, it is formed by the first processor 32 of the electronic device in which it is located reading the computer program instructions stored in the non-volatile memory into the first memory 33 for execution.

[0178] Based on the same inventive concept, embodiments of the present invention also provide a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the oil molecular composition prediction method in any of the above possible implementations.

[0179] Based on the same inventive concept, embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the oil molecular composition prediction method in any of the above possible implementations.

[0180] Based on the same inventive concept, see [link to inventive concept] Figure 10 The present invention also provides another electronic device, including a second memory 41 (e.g., non-volatile memory), a second processor 42, and a computer program stored in the second memory 41 and executable on the second processor 42. When the second processor 42 executes the program, it implements the steps of the oil molecular composition prediction method in any of the above possible implementations. This electronic device may be, for example, a PC or an oil molecular composition prediction device.

[0181] like Figure 10As shown, the electronic device may also include: a second memory 43, a second network interface 44, and a second internal bus 45; wherein the second memory 41, the second processor 42, the second memory 43, and the second network interface 44 communicate with each other through the second internal bus 45. In addition to these components, other hardware may also be included, which will not be described in detail here.

[0182] It should be noted that the above-mentioned oil molecular composition prediction device can be implemented by software. As a device in a logical sense, it is formed by the second processor 42 of the electronic device in which it is located reading the computer program instructions stored in the non-volatile memory into the second memory 43 for execution.

[0183] It should be noted that the above-mentioned method for generating the oil molecular composition prediction network and the method for predicting oil molecular composition can be executed in the same electronic device. For example, the training of the oil molecular composition prediction network and the prediction of oil molecular composition can both be executed in the oil molecular composition prediction device. Of course, they can also be executed in different electronic devices. For example, the oil molecular composition prediction network can be trained in one electronic device, and the trained oil molecular composition prediction network can be used when performing molecular composition prediction in another electronic device.

[0184] The embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, the program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0185] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0186] Suitable computers for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0187] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. Processors and memory may be supplemented by or incorporated into dedicated logic circuitry.

[0188] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0189] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0190] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0191] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A method for generating an oil molecular composition prediction network, characterized in that, The generating method comprises: obtaining an original data set; wherein the original data set comprises a plurality of groups of macroscopic property data of known oil products with labels, the labels of the macroscopic property data of the known oil products being molecular composition data of the known oil products; based on the original data set, a pre-constructed generator (10) is trained to obtain first network parameters of the generator (10); inputting randomly generated macroscopic property data of a to-be-tested oil product into the generator (10) with the first network parameters to obtain molecular composition data of the to-be-tested oil product generated by the generator (10) with the first network parameters; keeping the first network parameters of the generator (10) unchanged, based on the molecular composition data of the known oil products in the original data set and the molecular composition data of the to-be-tested oil products generated by the generator (10), a pre-constructed discriminator (20) connected to the generator (10) is trained to obtain second network parameters of the discriminator (20); keeping the second network parameters of the discriminator (20) unchanged, based on the original data set, the trained generator (10) and the discriminator (20) are jointly trained to obtain an oil product molecular composition prediction network; selecting a third batch of macroscopic property data of known oil products with labels from the original data set; for any one group of the third batch of macroscopic property data of known oil products with labels, inputting the group of macroscopic property data into the oil product molecular composition prediction network to obtain molecular composition data of the known oil product generated by the oil product molecular composition prediction network; wherein the generated molecular composition data of the known oil product comprises molecular species of the known oil product and a predicted content corresponding to each molecular species; sorting various molecules in the known oil product according to the predicted content, and calculating absolute errors or relative errors of the predicted content corresponding to each molecular species in the known oil product and the actual content; calculating a sum of the actual content corresponding to a set number of molecular species with high predicted content in the known oil product; if the sum of the actual content is lower than a first threshold, then adding 1 to the set number and returning to the step of calculating the sum of the actual content; if the sum of the actual content is not lower than the first threshold, then determining an average error or an average relative error at this time; if the average error or the average relative error does not exceed a second threshold, then determining that the oil product molecular composition prediction network is qualified, otherwise, it is unqualified.

2. The method of claim 1, wherein: The training of the pre-constructed generator (10) based on the original data set comprises: selecting a first batch of macroscopic property data of known oil products with labels from the original data set, and performing supervised training on the pre-constructed generator (10) by using the first batch of macroscopic property data of known oil products with labels.

3. The method of claim 1, wherein: The training of the pre-constructed discriminator (20) connected to the generator (10) based on the molecular composition data of the known oil products in the original data set and the molecular composition data of the to-be-tested oil products generated by the generator (10) comprises: Select a batch of known oil product molecular composition data from the original data set, and mark the categories of the selected known oil product molecular composition data and the multiple groups of molecular composition data of the to-be-tested oil products generated by the generator (10); Based on the marked known oil product molecular composition data and the to-be-tested oil product molecular composition data, the discriminator (20) is supervised trained; Wherein, during the training of the discriminator (20), the first network parameter of the generator (10) remains unchanged, and the second network parameter of the discriminator (20) is adjusted according to the error between the predicted value output by the discriminator (20) and the label, until the discriminator (20) can distinguish whether the molecular composition data input into the discriminator (20) is from a known oil product or a to-be-tested oil product.

4. The method of claim 1, wherein: The joint training of the trained generator (10) and the discriminator (20) based on the original data set comprises: Selecting a second batch of labeled known oil product macroscopic physical property data from the original data set, and using the second batch of labeled known oil product macroscopic physical property data to jointly train the trained generator (10) and the discriminator (20).

5. The oil product molecular composition prediction network generation method according to claim 4, characterized in that: During the joint training of the trained generator (10) and the discriminator (20), the second network parameter of the discriminator (20) remains unchanged, and the first network parameter of the generator (10) is adjusted according to the error between the molecular composition data output by the generator (10) and the label, and the error between the predicted value output by the discriminator (20) and the label; The loss function of the joint training is a cross-entropy loss function; and the calculation formula of the cross-entropy loss function is: wherein y i is the true value of the content of the i-th molecule in the known oil product input to the generator (10) in a training, is the predicted value of the content of the i-th molecule output by the generator (10) in the training, and n is the number of predefined molecule types.

6. The method of claim 1, wherein: The generation method further comprises: iteratively training the obtained oil product molecular composition prediction network, comprising repeating the following training steps: Training the pre-constructed generator (10) based on the original data set to obtain the first network parameter of the generator (10); Inputting the randomly generated macroscopic physical property data of the to-be-tested oil product into the generator (10) with the first network parameter to obtain the molecular composition data of the to-be-tested oil product generated by the generator (10) with the first network parameter; Keeping the first network parameter of the generator (10) unchanged, training the pre-constructed discriminator (20) connected with the generator (10) based on the known oil product molecular composition data in the original data set and the to-be-tested oil product molecular composition data generated by the generator (10) to obtain the second network parameter of the discriminator (20); Keeping the second network parameter of the discriminator (20) unchanged, jointly training the trained generator (10) and the discriminator (20) based on the original data set to obtain an oil product molecular composition prediction network.

7. A method of predicting the molecular composition of an oil product, characterized in that, The method comprises: Obtaining macroscopic physical property data of a to-be-tested oil product; inputting macroscopic physical property data of the to-be-tested oil product into a pre-trained oil product molecular composition prediction network, the oil product molecular composition prediction network being trained and generated according to the method for generating an oil product molecular composition prediction network in any one of claims 1-6; obtaining molecular composition data of the to-be-tested oil product generated by the oil product molecular composition prediction network.

8. An apparatus for generating a predictive network for the molecular composition of oil products, characterized in that, The device comprises: an acquisition module (11) configured to acquire an original data set; wherein the original data set contains a plurality of sets of macroscopic physical property data of known oil products with labels, the labels of the macroscopic physical property data of the known oil products being molecular composition data of the known oil products; a first training module (12) configured to train a pre-constructed generator (10) based on the original data set, to obtain first network parameters of the generator (10); a molecular composition acquisition module (13) configured to input randomly generated macroscopic physical property data of a to-be-tested oil product into the generator (10) with the first network parameters, to obtain molecular composition data of the to-be-tested oil product generated by the generator (10) with the first network parameters; a second training module (14) configured to keep the first network parameters of the generator (10) unchanged, to train a pre-constructed discriminator (20) connected to the generator (10) based on the molecular composition data of the known oil products in the original data set and the molecular composition data of the to-be-tested oil products generated by the generator (10), to obtain second network parameters of the discriminator (20); a third training module (15) configured to keep the second network parameters of the discriminator (20) unchanged, to jointly train the trained generator (10) and the discriminator (20) based on the original data set, to obtain the molecular composition prediction network; a test module (16) configured to: select macroscopic physical property data of a third batch of known oil products with labels from the original data set; for any one of the macroscopic physical property data of the third batch of known oil products with labels, input the set of macroscopic physical property data into the oil product molecular composition prediction network, to obtain molecular composition data of the known oil product generated by the oil product molecular composition prediction network; wherein the generated molecular composition data of the known oil product includes molecular species of the known oil product and a predicted content corresponding to each molecular species; sort the various molecules in the known oil product according to the predicted content, and calculate absolute errors or relative errors of the predicted content and the actual content corresponding to each molecular species in the known oil product; calculate a sum of the actual content corresponding to a set number of molecular species with high predicted content in the known oil product; if the sum of the actual content is lower than a first threshold value, add 1 to the set number, and return to the step of calculating the sum of the actual content; if the sum of the actual content is not lower than the first threshold value, determine the average error or the average relative error at this time; if the average error or the average relative error does not exceed a second threshold value, determine that the oil product molecular composition prediction network is qualified, otherwise, it is unqualified.

9. The apparatus of claim 8, wherein: The first training module (12) is configured to: The macroscopic physical property data of a first batch of labeled known oil products are selected from the original data set, and the pre-constructed generator (10) is supervised trained by using the macroscopic physical property data of the first batch of labeled known oil products.

10. The apparatus for generating an oil molecular composition prediction network according to claim 8, wherein: The second training module (14) is configured to: The molecular composition data of a batch of known oil products are selected from the original data set, and the categories of the selected molecular composition data of the known oil products and the molecular composition data of a plurality of to-be-tested oil products generated by the generator (10) are labeled; and the discriminator (20) is supervised trained based on the labeled molecular composition data of the known oil products and the molecular composition data of the to-be-tested oil products.

11. The apparatus for generating an oil molecular composition prediction network of claim 8, wherein: The third training module (15) is configured to: The macroscopic physical property data of a second batch of labeled known oil products are selected from the original data set, and the trained generator (10) and the discriminator (20) are jointly trained by using the macroscopic physical property data of the second batch of labeled known oil products.

12. The apparatus of claim 11, wherein: When the trained generator (10) and the discriminator (20) are jointly trained, the loss function of the joint training is a cross-entropy loss function; and the calculation formula of the cross-entropy loss function is: wherein y i is the true value of the content of the i-th molecule in the known oil product input to the generator (10) in a training, is the predicted value of the content of the i-th molecule output by the generator (10) in the training, and n is the number of predefined molecule types.

13. An oil product molecular composition prediction device, characterized by comprising: The device comprises: The physical property acquisition module (21) is configured to acquire the macroscopic physical property data of the to-be-tested oil product; The molecular composition prediction module (22) is configured to input the macroscopic physical property data of the to-be-tested oil product into a pre-trained oil product molecular composition prediction network, the oil product molecular composition prediction network is trained and generated according to the generation method of the oil product molecular composition prediction network in any one of claims 1-6; and the molecular composition data of the to-be-tested oil product generated by the oil product molecular composition prediction network is obtained.

14. A storage medium, characterized by The storage medium has stored thereon a computer program, the computer program being used to execute the steps of the generation method of the oil product molecular composition prediction network in any one of claims 1-6 or the steps of the oil product molecular composition prediction method in claim 7.

15. An electronic device, comprising: The device comprises a processor and a computer program running on the processor, the computer program being used to execute the steps of the generation method of the oil product molecular composition prediction network in any one of claims 1-6 or the steps of the oil product molecular composition prediction method in claim 7.

16. The electronic device of claim 15, wherein: The device further comprises a memory, and the computer program is stored on the memory and runs on the processor.

Images