Oil blending quality prediction method and system based on physical information neural network

Abstract

The present invention discloses a method and system for predicting the quality of oil blending based on a physical information neural network, and relates to the field of chemical control technology. The system discloses a blending parameter prediction module, a quality prediction tracing module, and a raw material process screening module. The present invention realizes the deep integration of physical constraints and data-driven, and can directly and accurately predict various indicators and comprehensive quality of the blended oil products based on the process parameters of the blending raw materials. When it is predicted that there are problems with the quality of the blended oil products, the blending process tracing step is started, and the influence of various process parameters in reference raw materials similar to the blending raw materials on the quality is deeply analyzed according to the comprehensive difference screening range, and then the process parameter screening order is customized to ensure that the key process parameters affecting the quality are screened first, and the screening efficiency and accuracy of the process parameters of the blending raw materials are guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of chemical control technology, and more particularly to an oil blending quality prediction method and system based on a physical information neural network. Background Art

[0002] In the field of oil refining, oil blending is a crucial process. It aims to produce oil products that meet various quality standards by mixing raw materials of different properties according to specific proportions and processes to meet diverse market demands. For example, the quality requirements of oil products for different purposes such as automotive gasoline, diesel and aviation kerosene are different.

[0003] Traditional oil blending quality control methods rely primarily on empirical formulas and linear models derived from extensive experimental data. These methods have numerous limitations in practical application:

[0004] First, empirical formulas are often derived based on specific operating conditions and a limited range of raw materials, lacking the ability to comprehensively and accurately describe complex oil blending systems. With the continuous advancement of refining processes and the increasing diversification of raw material sources, the interactions between the various components involved in oil blending have become increasingly complex, exhibiting highly nonlinear characteristics. Traditional empirical formulas struggle to accurately describe these complex relationships, resulting in significant deviations in the prediction of various quality indicators of blended oils (such as octane number, viscosity, flash point, pour point, and distillation range), making them unable to meet the requirements of modern high-precision oil quality control.

[0005] Secondly, relying solely on data-driven linear models, while they can extract certain patterns from large amounts of historical data, ignores the inherent physical mechanisms and constraints of the oil blending process. The various quality indicators in the oil blending process do not exist in isolation; they influence each other and are governed by physical laws such as mixing rules and thermodynamic principles. For example, the effect of mixing different components on oil viscosity follows specific mixing rules (linear mixing rules are applicable in some simple cases, while more realistic models such as the Grunberg-Nissan nonlinear mixing rule are required for complex systems). Linear models struggle to incorporate these physical constraints, resulting in predictions that may violate basic physical common sense, significantly reducing their reliability in practical applications.

[0006] Furthermore, when predicted oil blending quality issues arise, traditional methods lack systematic and efficient methods for tracing the root causes and troubleshooting the blending raw material process parameters. Typically, this involves checking each process parameter individually or relying on operator experience to identify potential influencing factors. This approach is not only time-consuming and labor-intensive, but also makes it difficult to accurately identify key influencing factors, making it easy to miss important clues. This leads to an inability to optimize and adjust the blending process in a timely and effective manner, thus impacting the overall quality and production efficiency of the oil blending process.

[0007] In today's chemical industry, which is increasingly pursuing high-quality, high-efficiency production and refined management, there is an urgent need for a method and system that can deeply integrate the physical mechanisms of the oil blending process with data intelligence. This method can not only accurately predict the various indicators and overall quality of the blended oil, but also quickly and accurately trace the blending process when quality problems arise, and screen out the key factors affecting process parameters, thereby achieving effective control of oil blending quality and targeted optimization of process parameters, thereby improving the stability and reliability of the entire oil blending process. Summary of the Invention

[0008] In view of the shortcomings of the existing technology, the purpose of the present invention is to provide an oil blending quality prediction method and system based on physical information neural network.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] An oil blending quality prediction system based on physical information neural networks, including a blending parameter prediction module, a quality prediction and traceability module, and a raw material process screening module;

[0011] The blending parameter prediction module is used to build a physical information neural network model of multiple blending quality indicators and determine the prediction parameter group of multiple blending quality indicators based on various process parameters of the blending raw materials;

[0012] The quality prediction and tracing module determines the predicted quality index of each blending quality indicator based on the prediction parameter group of multiple blending quality indicators, and then determines the comprehensive quality index of the blending raw materials. Based on the comparison result of the comprehensive quality index of the blending raw materials and the comprehensive quality threshold index, it determines whether to start the blending process tracing step;

[0013] The raw material process screening module generates a comprehensive difference screening range after starting the blending process traceability step, obtains all blending raw materials before the current time of the system and the corresponding comprehensive quality index, marks the blending raw materials whose comprehensive quality index is within the comprehensive difference screening range as reference raw materials, and then generates a process parameter screening sequence, and screens the process parameters of the blending raw materials in sequence according to the process parameter screening sequence.

[0014] Furthermore, based on the various process parameters of the blending raw materials, a prediction parameter group of multiple blending quality indicators is determined: the various process parameters of the blending raw materials are obtained, the process parameters of the blending raw materials are input into the physical information neural network model of multiple blending quality indicators, and the prediction parameter group of multiple blending quality indicators is output.

[0015] Furthermore, the prediction quality index of each harmonic quality indicator is determined based on the prediction parameter group of multiple harmonic quality indicators: the prediction parameters of each harmonic quality indicator in the prediction parameter group of multiple harmonic quality indicators are obtained, the prediction quality quantification model corresponding to each harmonic quality indicator is obtained, the prediction parameters of the harmonic quality indicator are input into the corresponding prediction quality quantification model respectively, and the prediction quality index of each harmonic quality indicator is output.

[0016] Furthermore, the comprehensive quality index of the blending raw materials is determined by the following process: obtaining the predicted quality index of each blending quality indicator, comparing the predicted quality index of each blending quality indicator pairwise, calculating the absolute difference between the predicted quality indexes of the two compared blending quality indicators, calculating the predicted quality fluctuation index, summing up the average of all predicted quality fluctuation indices, calculating the average predicted quality fluctuation index, setting the predicted quality threshold index, and when the predicted quality index of the blending quality indicator is less than the predicted quality threshold index, increasing the number of indicator depressions by one, and determining the comprehensive quality index of the blending raw materials based on the average predicted quality fluctuation index and the number of indicator depressions.

[0017] Furthermore, based on the comparison result of the quality comprehensive index of the blending raw materials and the quality comprehensive threshold index, it is determined whether to start the blending process tracing step: the quality comprehensive threshold index is set, and when the quality comprehensive index of the blending raw materials is greater than or equal to the quality comprehensive threshold index, the blending process tracing step is started, and the quality comprehensive difference index corresponding to the blending raw materials is simultaneously determined.

[0018] Furthermore, the comprehensive difference screening range is generated in the following process: obtain the upper difference index and the lower difference index, calculate the sum of the quality comprehensive difference index and the upper difference index to obtain the quality comprehensive upper index, calculate the difference between the quality comprehensive difference index and the lower difference index to obtain the quality comprehensive lower index, and generate the comprehensive difference screening range based on the quality comprehensive lower index and the quality comprehensive upper index.

[0019] Furthermore, the process parameter screening sequence generates the following process: determining various process parameters for blending raw materials, obtaining the quality optimization impact index of various process parameters, and sorting the quality optimization impact index of various process parameters in order, thereby generating the process parameter screening sequence.

[0020] Furthermore, the quality optimization influence index of a process parameter is obtained through the following process: selecting a process parameter, determining all reference raw materials, determining the parameter influence index of all reference raw materials for the process parameter, calculating the sum of all parameter influence indices, and obtaining the quality optimization influence index of the process parameter.

[0021] Furthermore, the parameter influence index of a reference raw material for the process parameter is determined by the following process: obtaining the process parameter of the reference raw material, randomly adjusting the process parameter s times, determining the quality comprehensive index of the reference raw material after each adjustment, comparing the quality comprehensive indexes of the reference raw material after each adjustment pairwise, calculating the absolute difference between the two compared quality comprehensive indices to obtain the adjustment swing index, summing up and averaging all the adjustment swing indices to obtain the average adjustment swing index, comparing the quality comprehensive index of the reference raw material after each adjustment with the quality comprehensive index of the reference raw material before adjustment, when the quality comprehensive index of the reference raw material after adjustment is less than the quality comprehensive index of the reference raw material before adjustment, increasing the quality optimization quantity by one, and determining the parameter influence index of the reference raw material for the process parameter based on the average adjustment swing index and the quality optimization quantity.

[0022] Furthermore, the oil blending quality prediction method based on physical information neural network has the following steps:

[0023] Step 1: Construct a physical information neural network model of multiple harmonic quality indicators;

[0024] Step 2: Based on various process parameters of blending raw materials, determine the prediction parameter group of multiple blending quality indicators;

[0025] Step 3: Determine the predicted quality index of each blending quality index based on the prediction parameter group of multiple blending quality indicators, and then determine the comprehensive quality index of the blending raw materials;

[0026] Step 4: Based on the comparison result of the quality comprehensive index of the blending raw materials and the quality comprehensive threshold index, determine whether to start the blending process traceability step;

[0027] Step 5: After starting the blending process traceability step, generate a comprehensive difference screening range;

[0028] Step 6: Generate a process parameter screening sequence, and screen the process parameters of the blending raw materials in sequence according to the process parameter screening sequence.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] The system of the present invention realizes the deep integration of physical constraints and data-driven through the blending parameter prediction module, the quality prediction and tracing module and the raw material process screening module. It can directly and accurately predict various indicators and comprehensive quality of the blended oil products according to the process parameters of the blending raw materials. When it is predicted that there are problems with the quality of the blended oil products, the blending process tracing step is started, and the influence of various process parameters in the reference raw materials similar to the blending raw materials on the quality is deeply analyzed according to the comprehensive difference screening range, and then the process parameter screening order is customized to ensure that the key process parameters affecting the quality are screened first, and the screening efficiency and accuracy of the process parameters of the blending raw materials are guaranteed. The method of the present invention realizes the deep integration of physical mechanism and data intelligence, accurately predicts the blending quality and optimizes the process parameters in a targeted manner, significantly improving the quality controllability of oil blending. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a schematic diagram showing the principles of the system of the present invention;

[0032] Figure 2 This is a principle block diagram of the raw material process screening module of the present invention;

[0033] Figure 3 Flowchart for determining the predicted quality index for reconciling quality indicators. DETAILED DESCRIPTION

[0034] Example 1: Reference Figures 1 to 3 , an oil blending quality prediction system based on physical information neural network, including blending parameter prediction module, quality prediction and traceability module, and raw material process screening module.

[0035] The blending parameter prediction module constructs a physical information neural network model of multiple blending quality indicators (multiple blending quality indicators include octane number, viscosity, flash point, pour point, and distillation range, all of which need to be collected after blending. These indicators jointly determine the performance and use of the finished oil), obtains various process parameters of the blending raw materials (types of process parameters include component ratio, temperature, and pressure, etc.), inputs the process parameters of the blending raw materials into the physical information neural network model of multiple blending quality indicators, and outputs a prediction parameter group of multiple blending quality indicators (the prediction parameter group of multiple blending quality indicators includes prediction parameters for octane number, viscosity, flash point, pour point, and distillation range).

[0036] Physical information neural network model of multiple harmonic quality indicators: Construct a neural network model and embed physical constraints of octane number into the neural network model: 1. Linear mixing rule constraints: ; : component mass fraction, : Component octane number; Physical meaning: Assume that the octane number is the linear weighted sum of each component; 2. Nonlinear synergistic effect constraints: ; : Synergistic effect correction term (such as synergistic improvement of MTBE and aromatics) through historical data fitting; 3. Boundary constraints: ; Ensure that the blended octane number does not exceed the octane number range of the components; Physical constraints on viscosity embedded in the neural network model: 1. Linear volume weighted rule: ; : Component density, suitable for approximate mixing with low viscosity differences; 2. Grunberg-Nissan nonlinear mixing rule: ; : component mole fraction; : Interaction parameters between components (fitted by experiments); 3. Temperature dependence constraints: ; Embed the Andrade equation to describe the effect of temperature on viscosity; embed the physical constraints of flash point into the neural network model: 1. Phase equilibrium vapor pressure constraint: ; Calculate the saturation temperature using the Antoine equation ,in, ; 2. Reid vapor pressure (RVP) association constraints: ; Establish a thermodynamic correlation between RVP and flash point to ensure the consistency of vapor pressure and flash point; embed physical constraints of the freezing point into the neural network model: 1. Linear summation of n-alkane content: ; : The carbon number of n-alkanes in the component, : Contribution coefficient of carbon number to pour point; 2. Wax crystal precipitation boundary constraint: The cold filter point is usually 3-5°C higher than the pour point and serves as an empirical boundary constraint. Physical constraints for embedding the neural network model into the distillation range: 1. Boundary constraints between the initial boiling point (IBP) and the final boiling point (FBP): ; The distillation range after blending does not exceed the minimum initial distillation point and maximum final distillation point of the component distillation range; 2. 10% / 50% / 90% evaporation temperature summation rule: ; : Correction terms for volatility differences between components (calculated by UNIFAC model), these rules are used as part of the loss function to ensure that the model's prediction results conform to physical laws. The loss function consists of two parts: data fitting loss and physical constraint loss. The total loss is obtained by weighted summation. The process parameters of i groups of blending raw materials are collected, and the process parameters of blending raw materials are used as training data for the neural network model. A prediction parameter group of multiple blending quality indicators is assigned to each training data. The training data is divided into training set, verification set and validation set according to the set ratio of 5:2:1. The neural network training is performed on the training set, verification set and validation set. After the training is completed, a physical information neural network model of multiple blending quality indicators is constructed.

[0037] The quality prediction and traceability module determines the predicted quality index of each blending quality indicator based on the prediction parameter group of multiple blending quality indicators, and then generates the quality comprehensive index of the blending raw material, sets the quality comprehensive threshold index (the quality comprehensive threshold index is a preset index used for comparison with the quality comprehensive index), and when the quality comprehensive index of the blending raw material is greater than or equal to the quality comprehensive threshold index, starts the blending process tracing step (when the quality comprehensive index of the blending raw material is less than the quality comprehensive threshold index, the blending process tracing step is not started), and simultaneously determines the quality comprehensive difference index corresponding to the blending raw material (the quality comprehensive difference index is the difference between the quality comprehensive index and the quality comprehensive threshold index).

[0038] Determine the prediction quality index of each harmonic quality indicator based on the prediction parameter group of multiple harmonic quality indicators: obtain the prediction parameters of each harmonic quality indicator in the prediction parameter group of multiple harmonic quality indicators, obtain the prediction quality quantification model corresponding to each harmonic quality indicator, input the prediction parameters of the harmonic quality indicator into the corresponding prediction quality quantification model respectively, and output the prediction quality index of each harmonic quality indicator.

[0039] The comprehensive quality index of the blending raw materials is generated in the following process: obtain the predicted quality index of each blending quality index, compare the predicted quality index of each blending quality index with each other, calculate the absolute difference between the predicted quality indexes of the two compared blending quality indexes, calculate the predicted quality fluctuation index, calculate the sum and average of all the predicted quality fluctuation indexes, calculate the average predicted quality fluctuation index Kdzbp, set the predicted quality threshold index (the predicted quality threshold index is a preset index used for comparison with the predicted quality index), when the predicted quality index of the blending quality index is less than the predicted quality threshold index, increase the index depression times by one (when the predicted quality index of the blending quality index is greater than or equal to the predicted quality threshold index, no processing is done), mark the index depression times as Setgv, and pass Calculate the comprehensive quality index of the blending raw material , where vg1 is the first coefficient, vg2 is the second coefficient, the value of vg1 is 1.01, and the value of vg2 is 0.84.

[0040] Each blending quality indicator corresponds to a predictive quality quantification model, such as octane number and viscosity, each corresponding to a predictive quality quantification model. All predictive quality quantification models are constructed based on deep learning models. This embodiment uses octane number and viscosity as examples to disclose the steps for constructing predictive quality quantification models for octane number and viscosity.

[0041] Octane number prediction quality quantification model: Build a deep learning model, obtain j octane number prediction parameters, train the deep learning model with the octane number prediction parameters, assign a prediction quality index to each octane number prediction parameter. The prediction quality index ranges from 1.5 to 5.0. The larger the prediction quality index, the more the octane number prediction parameter meets the quality standard. The j octane number prediction parameters are divided into training set, validation set, and test set in a ratio of 70%:15%:15%. The training set, validation set, and test set are trained. After training is completed, a prediction quality quantification model for octane number is constructed.

[0042] Prediction quality quantification model for viscosity: Construct a deep learning model, obtain j viscosity prediction parameters, train the deep learning model with the viscosity prediction parameters, assign a prediction quality index to each viscosity prediction parameter. The prediction quality index ranges from 1.5 to 5.0. The larger the prediction quality index, the more the viscosity prediction parameter meets the quality standard. Divide the j viscosity prediction parameters into training set, validation set, and test set in a ratio of 70%:15%:15%. Train the training set, validation set, and test set. After training is completed, a prediction quality quantification model for viscosity is constructed.

[0043] The raw material process screening module, when the blending process traceability step is started, generates a comprehensive difference screening range, obtains all blending raw materials before the current time of the system and the corresponding comprehensive quality index, marks the blending raw materials with comprehensive quality index within the comprehensive difference screening range as reference raw materials, and then generates a process parameter screening sequence, and screens the process parameters of the blending raw materials in sequence according to the process parameter screening sequence (for example: the process parameter screening sequence of the blending raw materials is 1, component ratio; 2, pressure; 3, temperature, then the component ratio of the blending raw materials is first screened, and the component ratio of the blending raw materials can be adjusted, and the adjusted component ratio is input into the physical information neural network model of multiple blending quality indicators. According to the output results, it is analyzed whether the component ratio of the blending raw materials needs to be adjusted. After the screening of the process parameter ranked first is completed, the process parameter ranked next is screened, and so on).

[0044] Comprehensive difference screening range, generation process: obtain the upper difference index and the lower difference index (the upper difference index is greater than the lower difference index, and both the upper difference index and the lower difference index are preset indexes), calculate the sum of the quality comprehensive difference index and the upper difference index to obtain the quality comprehensive upper index, calculate the difference between the quality comprehensive difference index and the lower difference index to obtain the quality comprehensive lower index, and generate the comprehensive difference screening range based on the quality comprehensive lower index and the quality comprehensive upper index.

[0045] Process parameter screening sequence, generation process: determine the various process parameters for blending raw materials, obtain the quality optimization impact index of various process parameters, sort the quality optimization impact index of various process parameters in order, and then generate the process parameter screening sequence.

[0046] The quality optimization influence index of a process parameter is obtained by: selecting a process parameter, determining all reference raw materials, determining the parameter influence index of all reference raw materials for the process parameter, calculating the sum of all parameter influence indices, and calculating the quality optimization influence index of the process parameter.

[0047] The parameter influence index of a reference raw material for the process parameter is determined by the following process: obtaining the process parameter of the reference raw material (such as obtaining the component ratio of the reference raw material), randomly adjusting the process parameter s times (such as randomly adjusting the component ratio of the reference raw material s times), determining the quality comprehensive index of the reference raw material after each adjustment, comparing the quality comprehensive index of the reference raw material after each adjustment, calculating the absolute difference between the two compared quality comprehensive indexes, and obtaining the adjustment swing index, summing up all the adjustment swing indices and calculating the average adjustment swing index Taeds, comparing the quality comprehensive index of the reference raw material after each adjustment with the quality comprehensive index of the reference raw material before adjustment, when the quality comprehensive index of the reference raw material after adjustment is less than the quality comprehensive index of the reference raw material before adjustment, increasing the quality optimization quantity by one (when the quality comprehensive index of the reference raw material after adjustment is greater than or equal to the quality comprehensive index of the reference raw material before adjustment, no processing is performed), marking the quality optimization quantity as FskL, through Calculate the parameter influence index of the reference raw material for this process parameter , where vg3 is the third coefficient, vg4 is the fourth coefficient, the value of vg3 is 2.14, and the value of vg4 is 1.17.

[0048] Through the blending parameter prediction module, quality prediction and traceability module, and raw material process screening module, a deep integration of physical constraints and data-driven is achieved. The various indicators and comprehensive quality of the blended oil products can be directly and accurately predicted based on the process parameters of the blending raw materials. When it is predicted that there are quality problems with the blended oil products, the blending process traceability step is started. According to the comprehensive difference screening range, the impact of various process parameters in the reference raw materials similar to the blending raw materials on the quality is deeply analyzed, and then the process parameter screening order is customized to ensure that the key process parameters affecting the quality are screened first, and to ensure the efficiency and accuracy of the screening of the process parameters of the blending raw materials.

[0049] Example 2: A method for predicting oil blending quality based on a physical information neural network, the steps are as follows:

[0050] Step 1: Construct a physical information neural network model of multiple harmonic quality indicators;

[0051] Step 2: Based on various process parameters of blending raw materials, determine the prediction parameter group of multiple blending quality indicators;

[0052] Step 3: Determine the predicted quality index of each blending quality index based on the prediction parameter group of multiple blending quality indicators, and then determine the comprehensive quality index of the blending raw materials;

[0053] Step 4: Based on the comparison result of the quality comprehensive index of the blending raw materials and the quality comprehensive threshold index, determine whether to start the blending process traceability step;

[0054] Step 5: After starting the blending process traceability step, generate a comprehensive difference screening range;

[0055] Step 6: Generate a process parameter screening sequence, and screen the process parameters of the blending raw materials in sequence according to the process parameter screening sequence.

[0056] The above method realizes the deep integration of physical mechanism and data intelligence, accurately predicts the blending quality and optimizes the process parameters in a targeted manner, significantly improving the quality controllability of oil blending.

[0057] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0058] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0059] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0060] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0061] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0062] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0063] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0064] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. The oil blending quality prediction system based on physical information neural network is characterized by: Including blending parameter prediction module, quality prediction and traceability module, and raw material process screening module; The blending parameter prediction module is used to build a physical information neural network model of multiple blending quality indicators and determine the prediction parameter group of multiple blending quality indicators based on various process parameters of the blending raw materials; The quality prediction and traceability module determines the predicted quality index of each blending quality indicator based on the prediction parameter group of multiple blending quality indicators, compares the predicted quality index of each blending quality indicator pairwise, calculates the absolute difference between the predicted quality indexes of the two compared blending quality indicators, calculates the predicted quality fluctuation index, calculates the average of all the predicted quality fluctuation indexes, and sets the predicted quality threshold index. When the predicted quality index of the blending quality indicator is less than the predicted quality threshold index, the number of indicator depressions is increased once. The quality comprehensive index of the blending raw material is determined based on the average predicted quality fluctuation index and the number of indicator depressions. Based on the comparison result of the quality comprehensive index of the blending raw material and the quality comprehensive threshold index, it is determined whether to start the blending process traceability step; The raw material process screening module generates a comprehensive difference screening range after starting the blending process traceability step, obtains all blending raw materials before the current time of the system and the corresponding comprehensive quality index, marks the blending raw materials whose comprehensive quality index is within the comprehensive difference screening range as reference raw materials, and then generates a process parameter screening sequence, and screens the process parameters of the blending raw materials in sequence according to the process parameter screening sequence.

2. The oil blending quality prediction system based on physical information neural network according to claim 1 is characterized in that: Based on various process parameters of the blending raw materials, a prediction parameter group of multiple blending quality indicators is determined: various process parameters of the blending raw materials are obtained, the process parameters of the blending raw materials are input into the physical information neural network model of multiple blending quality indicators, and the prediction parameter group of multiple blending quality indicators is output.

3. The oil blending quality prediction system based on physical information neural network according to claim 1 is characterized in that: Determine the prediction quality index of each harmonic quality indicator based on the prediction parameter group of multiple harmonic quality indicators: obtain the prediction parameters of each harmonic quality indicator in the prediction parameter group of multiple harmonic quality indicators, obtain the prediction quality quantification model corresponding to each harmonic quality indicator, input the prediction parameters of the harmonic quality indicator into the corresponding prediction quality quantification model respectively, and output the prediction quality index of each harmonic quality indicator.

4. The oil blending quality prediction system based on physical information neural network according to claim 1 is characterized in that: Based on the comparison result of the quality comprehensive index of the blending raw materials and the quality comprehensive threshold index, determine whether to start the blending process tracing step: set the quality comprehensive threshold index, and when the quality comprehensive index of the blending raw materials is greater than or equal to the quality comprehensive threshold index, start the blending process tracing step and simultaneously determine the quality comprehensive difference index corresponding to the blending raw materials.

5. The oil blending quality prediction system based on physical information neural network according to claim 1 is characterized in that: Comprehensive difference screening range, generation process: obtain the upper difference index and the lower difference index, calculate the sum of the quality comprehensive difference index and the upper difference index to obtain the quality comprehensive upper index, calculate the difference between the quality comprehensive difference index and the lower difference index to obtain the quality comprehensive lower index, and generate the comprehensive difference screening range based on the quality comprehensive lower index and the quality comprehensive upper index.

6. The oil blending quality prediction system based on physical information neural network according to claim 1 is characterized in that: Process parameter screening sequence, generation process: determine the various process parameters for blending raw materials, obtain the quality optimization impact index of various process parameters, sort the quality optimization impact index of various process parameters in order, and then generate the process parameter screening sequence.

7. The oil blending quality prediction system based on physical information neural network according to claim 6 is characterized in that: The quality optimization influence index of a process parameter is obtained by: selecting a process parameter, determining all reference raw materials, determining the parameter influence index of all reference raw materials for the process parameter, calculating the sum of all parameter influence indices, and calculating the quality optimization influence index of the process parameter.

8. The oil blending quality prediction system based on physical information neural network according to claim 7 is characterized in that: The parameter influence index of a reference raw material for the process parameter is determined by the following process: obtaining the process parameter of the reference raw material, randomly adjusting the process parameter s times, determining the quality comprehensive index of the reference raw material after each adjustment, comparing the quality comprehensive indexes of the reference raw material after each adjustment pairwise, calculating the absolute difference between the two compared quality comprehensive indices to obtain the adjustment swing index, summing up the average of all the adjustment swing indices to obtain the average adjustment swing index, comparing the quality comprehensive index of the reference raw material after each adjustment with the quality comprehensive index of the reference raw material before adjustment, when the quality comprehensive index of the reference raw material after adjustment is less than the quality comprehensive index of the reference raw material before adjustment, increasing the quality optimization quantity by one, and determining the parameter influence index of the reference raw material for the process parameter based on the average adjustment swing index and the quality optimization quantity.

9. A method for predicting oil blending quality based on a physical information neural network, applied to an oil blending quality prediction system based on a physical information neural network according to any one of claims 1 to 8, characterized in that: Here are the steps: Step 1: Construct a physical information neural network model of multiple harmonic quality indicators; Step 2: Based on various process parameters of blending raw materials, determine the prediction parameter group of multiple blending quality indicators; Step 3: Determine the predicted quality index of each blending quality index based on the prediction parameter group of multiple blending quality indicators, and then determine the comprehensive quality index of the blending raw materials; Step 4: Based on the comparison result of the quality comprehensive index of the blending raw materials and the quality comprehensive threshold index, determine whether to start the blending process traceability step; Step 5: After starting the blending process traceability step, generate a comprehensive difference screening range; Step 6: Generate a process parameter screening sequence, and screen the process parameters of the blending raw materials in sequence according to the process parameter screening sequence.

Images