Oil product blending component oil proportioning optimization control method, system, equipment and medium based on soft measurement

Abstract

The application discloses an oil blending component oil proportioning optimization control method, system, device and medium based on soft measurement. The application predicts the quality attribute of the finished oil in real time through a soft measurement module, obtains an attribute prediction value, determines a target gain with the original proportioning data, and corrects the model gain of a multivariable controller online by using the target gain, generates adjusted target proportioning data by the multivariable controller, and issues a formula control instruction to a control device, so that the component oil proportioning is accurately adjusted. When the component oil formula or attribute changes, the application can automatically and timely adjust the controller gain, effectively overcome the nonlinearity of the blending process, improve the control precision and system robustness, thereby guaranteeing the stability of the finished oil quality and avoiding the waste of high-value component oil. By introducing the soft measurement technology, the method can replace the traditional online analyzer to a certain extent, reduce the dependence on expensive instruments, and improve the economy and maintenance convenience of the system.

Description

Technical Field

[0001] This application relates to the field of petrochemical process control technology, and in particular to a method, system, equipment and medium for optimizing the oil ratio control of oil blending components based on soft measurement. Background Technology

[0002] Pipeline blending is a key process in modern oil refineries for producing refined petroleum products. Its core lies in optimizing the proportions of each component oil using a multivariable controller to ensure that the blended refined petroleum product meets target quality attributes, including octane number, distillation range, and vapor pressure. The performance of the multivariable controller is highly dependent on the accuracy of its internal control model, and the gain of the control model is a crucial parameter, reflecting the degree to which changes in the formulation affect the quality attributes.

[0003] However, oil blending is a typical nonlinear process. For example, there is a complex nonlinear relationship between octane number and component oil ratio (manifested as a blending effect). Traditional multivariable controllers use a fixed-gain linear model. When the operating point moves due to changes in formulation or feedstock properties, the fixed-gain linear model cannot accurately describe the dynamic characteristics of the process, leading to decreased control accuracy and causing fluctuations in the quality of finished oil products or waste of high-value component oils.

[0004] Therefore, how to improve the control precision of oil blending, ensure the stability of finished oil quality, and avoid the waste of high-value component oils are problems that need to be solved by those skilled in the art. Summary of the Invention

[0005] This application provides a method, system, equipment, and medium for optimizing the proportion of oil components in oil blending based on soft measurement, so as to improve the control accuracy of oil blending, ensure the stability of finished oil quality, and avoid the waste of high-value component oils.

[0006] In a first aspect, this application provides a method for optimizing and controlling the oil ratio of oil blending components based on soft sensing, including:

[0007] Obtain the quality attribute values ​​and original proportion data of each component oil;

[0008] The quality attribute prediction value of the finished oil is predicted by using the soft measurement module, the quality attribute values ​​of each component oil and the original proportion data, and the target gain is determined based on the quality attribute prediction value and the original proportion data.

[0009] The model gain of the original multivariable controller is adjusted using the target gain to obtain the adjusted target multivariable controller.

[0010] The formula adjustment value is obtained through the target multivariate controller and the quality attribute deviation; wherein, the quality attribute deviation is the deviation between the target value of the quality attribute and the predicted value of the quality attribute.

[0011] Based on the formula adjustment value and the original proportion data, determine the target proportion data;

[0012] The target ratio data is used to generate a formula control command, which is then sent to the control equipment to adjust the component oil ratio in the oil blending process.

[0013] Optionally, using a soft measurement module and the quality attribute values ​​of each component oil and the original blending data, the predicted quality attribute values ​​of the finished oil are calculated, and the target gain is determined based on the predicted quality attribute values ​​and the original blending data, including:

[0014] The quality attribute values ​​of each component oil and the original proportion data are input into the soft measurement module; wherein, the soft measurement module includes a nonlinear parameterized blending model; the nonlinear parameterized blending model is a data-driven model or a semi-mechanistic model;

[0015] The nonlinear parameterized blending model is used to predict the quality attributes of the finished oil, and the target gain is determined based on the predicted quality attributes and the original blending data.

[0016] Optionally, the target gain can be determined based on the predicted values ​​of quality attributes and the original proportioning data, including:

[0017] Calculate the partial derivatives of the predicted quality attribute values ​​and the original proportion data, and use these partial derivatives as the target gain. The formula for calculating the target gain is as follows: ;in, For target gain, For partial differential operators, For predicted values ​​of quality attributes, These are the original proportions.

[0018] Optionally, predicting the quality attribute prediction values ​​of the refined oil product using the nonlinear parameterized blending model includes:

[0019] Using the aforementioned nonlinear parameterized blending model, the quality attribute values ​​of each component oil, and the original proportioning data, the predicted quality attribute values ​​of the finished oil are calculated.

[0020] The nonlinear parameterized harmonic model is as follows:

[0021] ;

[0022] in, For predicted values ​​of quality attributes, Let be the weighting parameter for the i-th component oil. This represents the original proportioning data for the i-th component oil. Let i be the quality attribute value of the i-th component oil. The bias parameter is used to characterize the harmonic nonlinearity.

[0023] Optionally, the oil-to-liquid ratio control method further includes:

[0024] Obtain the actual quality attribute values ​​of the refined oil;

[0025] Based on the actual quality attribute values ​​of the finished oil and the quality attribute values ​​of each component oil, the weight parameters and bias parameters of the nonlinear parameterized blending model are updated.

[0026] Optionally, if the original multivariate controller is a model predictive controller, then adjusting the model gain of the original multivariate controller using the target gain to obtain the adjusted target multivariate controller includes:

[0027] The steady-state gain matrix of the model predictive controller is dynamically updated using the target gain, and the updated model predictive controller is used as the target multivariate controller.

[0028] Optionally, obtain the quality attribute values ​​and original proportion data of each component oil, including:

[0029] The quality attribute values ​​of each component oil are measured using different types of sensors; each component oil includes at least one quality attribute value.

[0030] Use the target ratio data from the previous moment as the original ratio data.

[0031] Secondly, this application provides a soft-sensor-based oil blending component ratio optimization control system, comprising:

[0032] The soft measurement module is used to acquire the quality attribute values ​​and original proportion data of each component oil, and predict the quality attribute values ​​of the finished oil based on the quality attribute values ​​and original proportion data of each component oil, and determine the target gain based on the quality attribute prediction values ​​and original proportion data.

[0033] A multivariable controller is used to adjust the model gain through the target gain and obtain the formula adjustment value based on the input quality attribute deviation; wherein the quality attribute deviation is the deviation between the target value and the predicted value of the quality attribute; the target ratio data is determined based on the formula adjustment value and the original ratio data; the formula control command is generated using the target ratio data and sent to the control device to adjust the component oil ratio in the oil blending process.

[0034] Thirdly, this application provides an electronic device, comprising:

[0035] Memory, used to store computer programs;

[0036] A processor is used to execute the computer program to implement the steps of the above-described method for optimizing the oil ratio of oil blending components.

[0037] Fourthly, this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for optimizing the oil ratio of blending components.

[0038] Compared with the prior art, the technical solutions provided in this application have the following advantages: This application provides a method, system, device, and medium for optimizing the proportion of component oils in oil blending based on soft sensing. In this application, when the component oil formula or properties change, this application can use a soft sensing module to predict the quality attribute value of the finished oil based on the quality attribute value and the original proportion data, and determine the target gain based on the predicted quality attribute value and the original proportion data; then, the model gain of the multivariable controller is adjusted based on the target gain, so that the adjusted target proportion data can be obtained through the adjusted target multivariable controller, and a formula control command is generated and sent to the control device to adjust the component oil proportions in the oil blending process.

[0039] As can be seen, this application can determine the target gain based on the predicted value of the quality attribute and the original proportioning data when the formulation or properties of the component oil change, and adjust the model gain of the multivariable controller. In this way, the model gain of the multivariable controller can be corrected online, effectively overcoming the nonlinearity of the blending process, significantly improving the control accuracy and robustness of oil blending, ensuring the stability of the finished oil quality, and avoiding the waste of high-value component oils. Attached Figure Description

[0040] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0043] Figure 1 A flowchart of a soft-sensor-based oil blending component ratio optimization control method is provided for embodiments of this application;

[0044] Figure 2 A schematic diagram of the online blending variable gain multivariable formulation control system for oil products provided in this application embodiment;

[0045] Figure 3 A schematic diagram of the neural network structure of the soft measurement harmonic model provided in the embodiments of this application;

[0046] Figure 4 This is a structural diagram of the online blending variable gain multivariable formulation control system for oil products provided in the embodiments of this application;

[0047] Figure 5 This is a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0048] This specific embodiment is merely an explanation of this application and is not intended to limit it. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of this application.

[0049] It should be noted that, in the optional embodiments of this application, the data related to object information, when applied to specific products or technologies, requires the permission or consent of the object. Furthermore, the collection, use, and processing of this data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. In other words, if the embodiments of this application involve data related to an object, it must be obtained with the object's authorization and consent, the authorization and consent of relevant departments, and in accordance with the relevant laws, regulations, and standards of the country and region. If the embodiments involve personal information, the acquisition of all personal information requires the individual's consent. If sensitive information is involved, the separate consent of the information subject is required. The embodiments also need to be implemented with the object's authorization and consent.

[0050] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0051] Furthermore, the term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship. To better understand and illustrate the solutions of the embodiments of this application, some technical terms involved in the embodiments of this application are briefly explained below.

[0052] Oil blending: The technique of mixing different component oils to produce qualified finished oil products.

[0053] Pipeline blending: This method involves pumping the component oils and additives into pipelines in different proportions, mixing them to achieve a homogeneous state, and then storing them in finished product tanks or shipping them directly out of the factory.

[0054] Component oil: Component oil is a semi-finished gasoline product produced by oil refineries. It does not contain oxygenated compounds and is mainly used to blend different grades of ethanol gasoline, such as catalytic gasoline, reformed gasoline, and alkylated oil.

[0055] Key quality attributes of component oils include: octane number, distillation range, olefin content, vapor pressure, aromatic content, oxygen content, benzene content, density, cetane number, flash point, naphtha yield, and residue yield, etc. Among them, octane number (RON) is the only scale for gasoline's anti-knock ability; distillation range (e.g., dry point) is used to describe the evaporation performance of the oil; olefin content, aromatic content, benzene content, and oxygen content are environmental and health indicators; vapor pressure affects the evaporation performance of gasoline; density affects the volumetric calorific value of the engine and the mass measurement of the injection system; cetane number is the core indicator of diesel ignition and combustion performance; flash point is a key indicator for the safe storage and transportation of oil products; naphtha yield / residue yield is an indicator for measuring the production efficiency and economy of a certain processing unit (such as catalytic cracking, atmospheric and vacuum distillation) in a refinery.

[0056] Multivariate control: An advanced control strategy that can systematically handle multiple coupled controlled and manipulated variables and perform synergistic optimization through model prediction; here, the controlled variables are multiple quality attributes of refined oil products, and the manipulated variables are the proportions of multiple component oils.

[0057] In traditional control, the gain of the control model of a multivariable controller is usually a fixed value. However, the oil blending process has significant nonlinear characteristics. Attributes such as octane number, distillation range, and saturated vapor pressure have a non-simple linear relationship with the formulation. Therefore, using a linear control model with fixed gain leads to increased control error, making it difficult to meet the requirements of high-precision blending. Thus, there is an urgent need in this field for an optimized control method that can adapt to process nonlinearity and correct the controller model in real time to achieve more accurate and economical blending production.

[0058] Therefore, this application proposes a soft-sensor-based method, system, equipment, and medium for optimizing the proportion of oil components in oil blending. This application uses soft-sensor technology to calculate and update the model gain of a multivariable controller in real time, enabling the controller to adapt to the nonlinear changes in the blending process, thereby significantly improving control accuracy and system robustness, enhancing the control accuracy of oil blending, ensuring the stability of finished oil quality, and avoiding the waste of high-value component oils.

[0059] See Figure 1 The flowchart below illustrates a method for optimizing and controlling the oil blending ratio of oil components based on soft sensing, as provided in this application embodiment. The method includes:

[0060] S101. Obtain the quality attribute values ​​and original proportion data of each component oil;

[0061] In this application, the control method is implemented through a control system, which includes a soft sensor module and a multivariable controller. In traditional control, the model gain of the multivariable controller is typically a fixed value. However, the oil blending process exhibits significant nonlinear characteristics (e.g., the relationship between properties such as octane number, distillation range, and saturated vapor pressure and the formulation is not simple linear). Using a linear model with a fixed gain leads to increased control error, making it difficult to meet the requirements of high-precision blending. To address this issue, this application constructs a dynamic, adaptive intelligent control system to solve the problem of insufficient control accuracy caused by process nonlinearity in traditional blending control. This system predicts the finished product properties in real time through soft sensors and calculates the key control model gain accordingly, thereby achieving online optimization of the multivariable controller.

[0062] In this step, the soft measurement module can acquire and integrate key characteristic parameters of the oil by monitoring its state in real time, such as Prozi1 (key characteristic parameter 1 of the i-th oil), Prozi2 (key characteristic parameter 2 of the i-th oil), and so on up to Prozoj (key characteristic parameter j of the o-th oil). These parameters cover multiple important indicators mentioned above, including RON, aromatic content, oxygen content, benzene content, density, cetane number, flash point, dry point, naphtha yield, and residue yield. This application refers to the specific values ​​of the key characteristic parameters as quality attribute values, and the monitored oil includes each component oil and the actual finished oil.

[0063] In another embodiment of this application, obtaining the quality attribute values ​​of each component oil and the original proportion data includes: measuring the quality attribute values ​​of each component oil using different types of sensors; wherein each component oil includes at least one quality attribute value; and using the target proportion data from the previous moment as the original proportion data.

[0064] In this embodiment, different types of sensors can be used to measure the quality attribute values ​​of each component oil. The soft measurement module integrates the data from multiple sensors to generate the quality attribute values ​​of each component oil. Furthermore, the soft measurement module and multiple sensors can form an integrated analysis platform with powerful data analysis capabilities, capable of rapidly processing large amounts of real-time data. By monitoring the state of the oil in real time and utilizing data fusion technology, the soft measurement module improves the accuracy and reliability of the measurement results to cope with dynamic attribute changes in complex environments. The soft measurement module also has a fault self-diagnosis function, automatically switching to a backup sensor when a sensor fails, ensuring continuous system operation and data acquisition integrity. Specifically, when a sensor fails, the system can automatically switch to a backup sensor to ensure continuous system operation and data acquisition integrity. For example, if a sensor fails, the soft measurement module will immediately detect the fault signal, automatically switch to the backup sensor, and simultaneously issue an alarm to notify maintenance personnel for repair.

[0065] In this step, the raw proportion data obtained is the target proportion data output by the multivariable controller at the previous moment. This application uses the target proportion data from the previous moment as the raw proportion data input to the soft measurement module at the current moment.

[0066] S102. Using the soft measurement module and the quality attribute values ​​of each component oil and the original proportion data, predict the quality attribute value of the finished oil, and determine the target gain based on the quality attribute prediction value and the original proportion data.

[0067] In this embodiment, the soft measurement module employs data mining and machine learning algorithms to analyze the current quality attribute values ​​of each component oil and the original blending data, predicting the quality attribute values ​​of the finished oil. This method of predicting the quality attribute values ​​of the finished oil through the soft measurement module enables real-time identification of fluctuations in quality attribute values, providing accurate quality attribute predictions based on these changes. This allows for the determination of the target gain based on the predicted quality attribute values ​​and the original blending data, enabling timely adjustments to the model gain of the multivariate controller to obtain more accurate target blending data. For example, when market demand changes or raw material supply fluctuates, the soft measurement module can adjust the target gain to generate blending data that meets new production needs.

[0068] In another embodiment of this application, the quality attribute prediction value of the finished oil is predicted using a soft measurement module and the quality attribute values ​​and original proportion data of each component oil, and a target gain is determined based on the quality attribute prediction value and the original proportion data. This includes: inputting the quality attribute values ​​and original proportion data of each component oil into the soft measurement module; wherein the soft measurement module includes a nonlinear parameterized blending model; the nonlinear parameterized blending model is a data-driven model or a semi-mechanistic model; the quality attribute prediction value of the finished oil is predicted using the nonlinear parameterized blending model, and a target gain is determined based on the quality attribute prediction value and the original proportion data.

[0069] In this application, after receiving the current raw blending data and the quality attribute values ​​of each component oil, the soft measurement module, based on the received data, uses its internal nonlinear parametric blending model to predict the quality attribute prediction values ​​of one or more quality attributes of the finished oil in real time. Based on this nonlinear parametric blending model, the soft measurement module calculates the target gain in real time. In another embodiment of this application, the process of determining the target gain based on the quality attribute prediction values ​​and the raw blending data includes: calculating the partial derivatives of the quality attribute prediction values ​​and the raw blending data, using the partial derivatives as the target gain, and the formula for calculating the target gain is: ;in, For target gain, For partial differential operators, For predicted values ​​of quality attributes, These are the original proportions.

[0070] Specifically, the target gain is the partial derivative of the predicted value of the refined oil quality attributes with respect to the original blending data, for example: (Octane number) / (Reformed gasoline ratio) indicates the marginal impact of fine-tuning the reformed gasoline blending ratio on the finished product's octane number.

[0071] Furthermore, the nonlinear parameterized blending model in this application can be a neural network model or a semi-mechanistic model. If it is a neural network model, the model takes the original blending data and the quality attribute values ​​of each component oil as input, calculates through hidden layers, and outputs predicted values ​​of the quality attributes of the finished oil; the target gain is obtained by taking the partial derivative of the input original blending data using the backpropagation algorithm of the neural network. If the nonlinear parameterized blending model is a semi-mechanistic model, the structure of the semi-mechanistic model is constructed based on the empirical mechanism formula of oil blending, and its model parameters are determined by fitting process data; the target gain is obtained by analytically differentiating or numerically differencing the predicted value of the quality attributes output by the semi-mechanistic model with respect to the input original blending data.

[0072] In another embodiment of this application, predicting the quality attribute prediction value of the refined oil using the nonlinear parameterized blending model includes: predicting the quality attribute prediction value of the refined oil using the nonlinear parameterized blending model, the quality attribute values ​​of each component oil, and the original proportioning data; the nonlinear parameterized blending model is:

[0073] ;

[0074] in, For predicted values ​​of quality attributes, Let be the weighting parameter for the i-th component oil. This represents the original proportioning data for the i-th component oil. Let i be the quality attribute value of the i-th component oil. The bias parameter is used to characterize the harmonic nonlinearity.

[0075] In this application, The predicted values ​​for different quality attributes of refined oil products are given here. Only the octane number is used as an example. Specifically, the semi-mechanistic model predicts the octane number of refined oil products based on the following nonlinear blending formula:

[0076] .

[0077] in, For the predicted octane number of refined oil products, For the first The proportions of the component oils, For the first Octane number of the component oil, These are the weighting coefficients. Empirical parameters characterizing harmonic nonlinearity; model parameters and It was obtained through regression analysis of historical harmonized data.

[0078] S103. Adjust the model gain of the original multivariable controller by using the target gain to obtain the adjusted target multivariable controller;

[0079] Specifically, this application uses a dynamically calculated target gain in real time, which is directly fed back to a multivariable controller to correct the model gain of its internal model. This model gain characterizes the sensitivity of the refined oil properties to changes in the blending formula. The multivariable controller then uses this gain, along with the quality attribute deviation between the predicted and target values ​​of the refined oil properties, to dynamically and accurately adjust the control inputs.

[0080] In this application, the final target proportion data is determined by a multivariate controller. Since this application adjusts the model gain of the multivariate controller, the multivariate controller before model gain adjustment is referred to as the original multivariate controller, and the multivariate controller after model gain adjustment is also referred to as the original multivariate controller. The original multivariate controller and the original multivariate controller can be two multivariate controllers with different model gains, or they can be the same multivariate controller with different model gains; no specific limitation is made here.

[0081] In another embodiment of this application, if the original multivariate controller is a model predictive controller, then adjusting the model gain of the original multivariate controller through the target gain to obtain the adjusted target multivariate controller includes: dynamically updating the steady-state gain matrix of the model predictive controller through the target gain, and using the updated model predictive controller as the target multivariate controller.

[0082] Specifically, the multivariate controller in this application is a model predictive controller. Its internal model employs a step response model or a state-space model, and the steady-state gain matrix of the model is updated online using real-time updated model gains. This multivariate controller can utilize a combination of fuzzy control and genetic algorithms. This combination enables the system to simulate biological decision-making processes, making optimal decisions under uncertain conditions, thereby improving the flexibility and adaptability of the blending process. For example, when the properties of oil fluctuate or the external environment changes, the multivariate controller can quickly adjust the control strategy through the synergistic effect of fuzzy control and genetic algorithms, generating the target blending data for the current moment to adapt to the new situation.

[0083] S104. Obtain the formula adjustment value through the target multivariate controller and the quality attribute deviation; wherein, the quality attribute deviation is the deviation between the target value of the quality attribute and the predicted value of the quality attribute;

[0084] S105. Determine the target proportion data based on the formula adjustment value and the original proportion data;

[0085] In this application, after the multivariate controller obtains the target value of the quality attribute of the finished oil, the predicted value of the quality attribute in real time by the soft measurement module, and the target gain calculated in real time, the multivariate controller will use the real-time target gain to update the model gain of its internal model so as to perform optimization calculation based on the deviation between the target value of the quality attribute and the predicted value of the quality attribute, and after obtaining the formula adjustment value, generate the final target ratio data.

[0086] S106. Generate formula control instructions using target ratio data and send the formula control instructions to the control equipment to adjust the component oil ratio in the oil blending process.

[0087] In this application, the multivariate controller can calculate and optimize the blending amount of oil products in real time to generate target blending data. For example, during the oil blending process, the multivariate controller can automatically adjust the proportions of different components based on the real-time monitored quality attribute prediction values ​​to ensure that the characteristics of various oil products are accurately controlled.

[0088] Furthermore, the multivariable controller in this application is not only connected to the soft measurement module, but also to the control equipment. After generating the target proportion data, the multivariable controller generates new formula control instructions based on the target proportion data and sends the new formula control instructions to the control equipment to control the blending process.

[0089] In summary, compared with the prior art, this application has at least the following significant advantages:

[0090] High-precision control: By updating the model gain in real time, the multivariable controller can adapt to the nonlinearity of the process, greatly reducing the control error caused by model mismatch and significantly improving the control accuracy of key quality attributes.

[0091] Strong robustness: The real-time prediction function of soft measurement overcomes the time lag of offline analysis. Combined with the self-updating capability of the model, the system has a stronger ability to adapt to disturbances such as fluctuations in component oil properties and changes in equipment characteristics.

[0092] Significant economic benefits: More precise control ensures the stability of product quality, avoids over-quality, and enables "edge optimization," thereby reducing the consumption of high-value component oils and saving production costs.

[0093] High level of automation and intelligence: It realizes full automation from prediction and gain correction to optimization control, reducing human intervention and is a key technology for realizing intelligent blending plant.

[0094] In another embodiment of this application, the oil-to-liquid ratio control method further includes:

[0095] Obtain the actual quality attribute values ​​of the refined oil; based on the actual quality attribute values ​​of the refined oil and the quality attribute values ​​of each component oil, update the weight parameters and bias parameters of the nonlinear parameterized blending model.

[0096] In this application, to ensure the prediction accuracy of the nonlinear parameterized blending model within the soft sensing module, the parameters of the nonlinear parameterized blending model can be updated. Specifically, the parameters (weight parameters and bias parameters) of the nonlinear parameterized blending model can be periodically or triggered using offline laboratory analysis data and / or quality attribute values ​​of refined oil measured by online analyzers. This allows the model to adapt to changes in raw materials and maintain its prediction accuracy. Parameter updates can be implemented using recursive least squares, gradient descent, or their variants. For example, as the system runtime increases, the weight parameters and bias parameters of the nonlinear parameterized blending model are periodically updated using the quality attribute values ​​of refined oil to improve the accuracy of predicting oil properties.

[0097] It should be noted that, in order to achieve wider application and management, this system also has a variety of communication interfaces, including but not limited to wired and wireless communication protocols, to facilitate the connection with cloud platforms or other industrial information systems, so as to realize real-time data uploading, remote monitoring and comprehensive management, and realize full-process visualization and intelligent decision support for oil blending.

[0098] In summary, this application, by updating the weight and bias parameters of the nonlinear parameterized blending model, enables the model to adapt to changes in the external environment, improving the system's stability and robustness, and effectively addressing the uncertainties and complexities in oil blending. For example, when the composition of the oil changes or production conditions alter, the weight and bias parameters are automatically adjusted based on real-time feedback of the finished oil's quality attributes to ensure stable system operation. In oil blending, facing various uncertainties and complexities, this approach ensures the model maintains a consistently good operating state.

[0099] For a clear explanation of this plan, please refer to [link / reference]. Figure 2 The above is a schematic diagram of a formula control system provided in an embodiment of this application, which is combined with... Figure 2 This paper describes the complete process of a multivariable optimization control method for oil blending with integrated soft sensing. As shown in the figure, this system integrates a soft sensing module and a multivariable controller, and forms a closed-loop optimization through dynamic gain. (See also...) Figure 3 Figure 1 is a schematic diagram of the neural network structure of the soft measurement harmonic model provided in this application embodiment. Each processing layer will be described below:

[0100] 1. Input Layer: Used for data acquisition and formula initialization. Its input data consists of two types of key real-time data received by the system: Quality attribute values ​​of each component oil: Key attribute indicators of each component oil involved in blending (such as catalytic gasoline, reformed gasoline, alkylate oil, etc.), such as octane number, distillation range, olefin content, vapor pressure, etc. Target values ​​of finished product quality attributes: The desired quality standard of the final finished product. Initial blending data: Based on the target and the initial model, the multivariate controller calculates an initial blending formula, i.e., the recommended blending ratio of each component oil.

[0101] 2. Core Processing Layer: Employing machine learning models such as neural networks, based on the input data, the layer passes through hidden layers and output layers, outputting predicted quality attribute values ​​for key properties of refined oil products, as shown in the diagram. .

[0102] 3. Online learning and parameter updates: such as Figure 2 As shown in the parameter update section, the weights and biases of this neural network are calibrated periodically or triggered using offline laboratory analysis data or higher-precision online analyzer data. This ensures that the soft measurement model can adapt to changes in raw materials and equipment aging, maintaining prediction accuracy over the long term.

[0103] 4. Variable Gain Calculation: The soft measurement module includes a dedicated gain calculation unit. This allows the model gain to be dynamic, changing dynamically with the current formulation point and component oil properties, rather than being a fixed value. It accurately reflects the sensitivity of the control variable to the controlled variable near the current operating point.

[0104] See Figure 2 The multivariable controller's inputs include the following: Attribute deviation: the difference between the predicted value of the finished product quality attribute from the soft sensor and the target value of the quality attribute; Variable gain: the latest target gain transmitted in real time from the soft sensor module; Online optimization signal: from the upper-level optimization system, which issues optimization instructions after considering economic objectives. Based on these instructions, the target value of the quality attribute is fine-tuned so that the adjusted target value and the predicted value of the quality attribute can be used to calculate the formula adjustment value and regenerate the target proportion data under the current target value of the quality attribute. In this way, the target value of the quality attribute can be adjusted in real time according to the optimization instructions, improving the flexibility of system control.

[0105] The decision-making process of the multivariable controller is as follows: using the built-in control model and the accurate process sensitivity information provided by the target gain, the optimal formula adjustment value ΔR is calculated to make the attribute deviation approach zero as quickly and smoothly as possible, while satisfying constraints such as cost.

[0106] Output data: The original proportion data R is added to the formula adjustment value ΔR, and the final real-time formula control command is output and sent to the flow controller on site for execution.

[0107] Furthermore, the entire process implemented in this application through the soft measurement module and multivariable controller forms a tight closed loop. After the formulation is executed, the new component oil quality attribute values ​​and the new original proportioning data are input into the soft measurement module to start a new cycle of prediction-gain calculation-optimization adjustment, thereby achieving continuous and precise control.

[0108] See Figure 4 This is a structural diagram of the online blending variable gain multivariable formulation control system for oil products provided in this application embodiment, based on... Figure 4 The complete process of the method is explained as follows:

[0109] Step 1: Input the oil characteristic characterization data into the system.

[0110] The system receives two sets of key inputs: a) real-time attribute data of each component oil, which are the quality attribute values ​​of each component oil in the above embodiments, and can come from online analyzers or laboratory data; b) target quality attribute values ​​of the finished oil. Data a is input to the soft measurement module, and data b is input to the multivariate controller.

[0111] Step 2: Soft measurement prediction and variable gain calculation, and send the results to the multivariable controller.

[0112] The soft sensing module receives the raw blending data (R) and the quality attribute values ​​of each component oil. Its internal blending model (e.g., a neural network or semi-mechanistic model) predicts the finished oil's quality attribute value (Y) in real time. The module's gain calculation unit calculates the model gain (g) under the current operating conditions in real time. This gain is the partial derivative of the finished oil attribute with respect to the formula, i.e. .

[0113] Step 3: The online optimization module enables online updates of model parameters.

[0114] To ensure long-term accuracy, the parameter update unit of the soft measurement module uses actual refined oil quality attribute values ​​measured in laboratory analysis and other actual measurements to periodically or trigger the calibration of the blending model parameters (such as the weights and biases of the neural network).

[0115] Step 4: Optimization control of the multivariable controller.

[0116] The multivariable controller receives three key signals: the target value of the quality attributes of the refined oil, the predicted value of the quality attributes of the refined oil (the two are compared to form a deviation), and the real-time model gain (g). The controller uses this real-time gain (g) to update the steady-state gain matrix of its internal model, and performs optimization calculations based on the updated model to solve for the optimal formula adjustment value (ΔR) that minimizes the quality deviation and meets the economic objective, thereby generating a new formula control command.

[0117] Step S5: Closed-loop execution. New recipe control commands are issued to actuators such as flow control valves. The entire process repeats continuously, forming a dynamic closed-loop optimized control loop.

[0118] The above method constitutes a complete closed-loop optimization control process: recipe input → real-time prediction by soft sensors → dynamic gain calculation → multivariable controller optimization adjustment. Through real-time attribute prediction by soft sensors and online correction of variable gain, the nonlinearity of the tuning process is effectively overcome, significantly improving the accuracy and robustness of the control.

[0119] Here, taking gasoline octane blending as an example, and combining the above process, a specific embodiment of this application will be described in detail:

[0120] Scenario: The goal is to blend gasoline with an octane rating of 95, mainly using two components: catalytic gasoline (octane rating 92) and reformed gasoline (octane rating 98).

[0121] Initial state:

[0122] The system is set to a target octane rating of 95.

[0123] The initial formula is: 60% catalytic gasoline and 40% reformed gasoline.

[0124] The soft measurement model predicts the current octane number to be 94.5, with a bias of -0.5.

[0125] Variable gain comes into play:

[0126] The soft measurement module calculates the current gain in real time. Assuming the gain g is calculated to be 0.8 under the current formulation, this means that for every 1% increase in the reformed gasoline proportion, the octane rating increases by approximately 0.8 units. After receiving the deviation and gain, the multivariate controller calculates that the reformed gasoline proportion needs to be increased by approximately 0.625% (i.e., 0.5 / 0.8). The new formulation is adjusted to: 59.375% catalytic gasoline and 40.625% reformed gasoline.

[0127] Nonlinear Response (Value of Variable Gain): Assuming that due to the nonlinearity of the blending process, the octane boost effect weakens when the reformed gasoline proportion exceeds 41%. At the new formulation point (reformed gasoline proportion ~40.625%), the soft-sensor module recalculates the gain and finds that g has decreased from 0.8 to 0.6. This accurately reflects the change in process sensitivity. If the predicted octane number is still below the target (e.g., 94.8), the controller will use a new, smaller gain g=0.6 to calculate the adjustment. This avoids over-control due to model mismatch, allowing the system to smoothly and accurately approach the target value and effectively suppressing overshoot.

[0128] This application offers the following advantages: High control precision: The variable gain mechanism ensures the controller model always matches the instantaneous characteristics of the process, significantly reducing control errors caused by nonlinearity. Strong robustness: It can adapt to disturbances such as fluctuations in raw material component properties and changes in catalyst activity. Stable product quality: It reduces the fluctuation range of finished oil properties and improves product quality consistency. Good economic benefits: More precise control can reduce the excessive addition of high-value components, achieving "edge-limit" optimization and reducing production costs.

[0129] The experimental data examples of this application (based on typical industrial cases) are as follows: After implementing this scheme on the gasoline blending unit of a refinery, it was compared with an advanced control scheme using a fixed gain model. The results are shown in Table 1:

[0130] Table 1

[0131]

[0132] Note: The above data are illustrative and the actual effect may vary depending on the specific device, but the trend is consistent.

[0133] Conclusion: This solution organically integrates the real-time predictive capabilities of soft sensors with the optimization capabilities of multivariable controllers through dynamic gain, forming an integrated intelligent system encompassing perception, decision-making, and execution. It not only solves the core nonlinear challenges of oil blending but also achieves a significant leap in control accuracy and economic efficiency, making it a key technology for realizing refined production.

[0134] The following describes the oil blending component oil ratio optimization control system provided in the embodiments of this application. The control system described below can be referred to in correspondence with the control method described above.

[0135] This application discloses an oil blending component ratio optimization control system based on soft measurement, specifically including:

[0136] The soft measurement module is used to acquire the quality attribute values ​​and original proportion data of each component oil, and predict the quality attribute values ​​of the finished oil based on the quality attribute values ​​and original proportion data of each component oil, and determine the target gain based on the quality attribute prediction values ​​and original proportion data.

[0137] A multivariable controller is used to adjust the model gain through the target gain and obtain the formula adjustment value based on the input quality attribute deviation; wherein the quality attribute deviation is the deviation between the target value and the predicted value of the quality attribute; the target ratio data is determined based on the formula adjustment value and the original ratio data; the formula control command is generated using the target ratio data and sent to the control device to adjust the component oil ratio in the oil blending process.

[0138] The soft measurement module in this application is the core of the invention for achieving variable gain, and it comprises three sub-units:

[0139] Blending Model Unit: Used to predict the quality attribute values ​​of the finished oil based on the input quality attribute values ​​of each component oil and the original proportion data, through the nonlinear parameterized blending model in the soft measurement module; the nonlinear parameterized blending model is a data-driven model or a semi-mechanistic model;

[0140] Specifically, the blending model unit is used to: measure the quality attribute values ​​of each component oil through different types of sensors; wherein each component oil includes at least one quality attribute value; and use the target proportion data from the previous moment as the original proportion data.

[0141] Specifically, the blending model unit is used to establish a nonlinear mapping relationship from the formulation (R) to the finished product attributes (Y). This nonlinear parameterized blending model can be a data-driven model, such as an Artificial Neural Network (ANN) or a Support Vector Machine (SVM). Taking a neural network as an example, its input is the proportion and attributes of each component oil, which undergoes nonlinear transformation through hidden layers to output predicted values ​​of the finished product attributes. This model excels at learning complex nonlinear relationships from historical data. The nonlinear parameterized blending model can also be a semi-mechanistic model, incorporating prior knowledge of the blending process.

[0142] Specifically, the blending model unit is used to: predict the quality attribute values ​​of the finished oil by using the nonlinear parameterized blending model, the quality attribute values ​​of each component oil and the original proportion data.

[0143] The nonlinear parameterized harmonic model is as follows:

[0144] ;

[0145] in, For predicted values ​​of quality attributes, Let be the weighting parameter for the i-th component oil. This represents the original proportioning data for the i-th component oil. Let i be the quality attribute value of the i-th component oil. The bias parameter is used to characterize the harmonic nonlinearity.

[0146] Gain Calculation Unit: Used to determine the target gain based on the predicted quality attribute values ​​and the original proportioning data; that is, to calculate the partial derivatives of the predicted quality attribute values ​​and the original proportioning data, and use the partial derivatives as the target gain. The formula for calculating the target gain is: ;in, For target gain, For partial differential operators, For predicted values ​​of quality attributes, These are the original proportions.

[0147] Specifically, the gain calculation unit is responsible for calculating the variable gain in real time. For neural network models, the gain can be automatically obtained through efficient backpropagation algorithms. For semi-mechanistic models, the gain can be calculated through analytical differentiation or numerical difference methods.

[0148] Parameter update unit: used to obtain the actual quality attribute values ​​of the refined oil, and update the weight parameters and bias parameters of the nonlinear parameterized blending model based on the actual quality attribute values ​​of the refined oil and the quality attribute values ​​of each component oil.

[0149] Specifically, the parameter update unit can use algorithms such as recursive least squares and gradient descent to update the parameters of the blending model online using the latest actual measurement data, so that the model can adapt to changes in raw materials and maintain prediction accuracy.

[0150] In another embodiment of this application, if the multivariate controller is a model predictive controller, the multivariate controller is specifically used to dynamically update the steady-state gain matrix of the model predictive controller through the target gain, and use the updated model predictive controller as the target multivariate controller.

[0151] Specifically, the multivariable controller in this application is preferably a model predictive controller (MPC). Internally, it typically employs a step response or state-space model. The key improvement of this application lies in the fact that the steady-state gain matrix in the MPC model is no longer fixed, but dynamically updated by the real-time gain (g) from the soft-sensor module. This ensures that the MPC's predictive model always reflects the true sensitivity of the tuning process at the current operating point, thereby making more accurate optimization decisions.

[0152] Figure 5 A structural diagram of an electronic device provided in an embodiment of the present invention, such as... Figure 5 As shown, it includes:

[0153] Memory 20 is used to store computer programs;

[0154] The processor 21 is used to execute a computer program to implement the steps of the component oil ratio optimization control method as described in the above embodiments.

[0155] The electronic devices provided in this embodiment may include, but are not limited to, smartphones, tablets, laptops, or desktop computers.

[0156] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one hardware form selected from Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an Artificial Intelligence (AI) processor, which handles computational operations related to machine learning.

[0157] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 20 is used to store at least the following computer program 201, which, after being loaded and executed by the processor 21, is capable of implementing the relevant steps of the component oil ratio optimization control method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, and the storage method may be temporary storage or permanent storage. The operating system 202 may include Windows, Unix, Linux, etc.

[0158] In some embodiments, the electronic device may further include a display screen 22, an input / output interface 23, a communication interface 24, a power supply 25, and a communication bus 26.

[0159] Those skilled in the art will understand that Figure 5 The structures shown do not constitute a limitation on electronic devices and may include more or fewer components than those shown.

[0160] In another exemplary embodiment, a computer storage medium is also provided, wherein the program instructions, when executed by a processor, implement the steps of the data deduplication method described in any of the above method embodiments.

[0161] It is understood that if the component oil ratio optimization control method in the above embodiments is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the current technology, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), electrically erasable programmable ROM, register, hard disk, removable disk, CD-ROM, magnetic disk, or optical disk, and other media capable of storing program code.

[0162] The various embodiments described in this specification are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” used herein may also mean the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a specific order described or illustrated, unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0163] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0164] The above are only some embodiments of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for optimizing and controlling the oil ratio of oil blending components based on soft sensing, characterized in that, include: Obtain the quality attribute values ​​and original proportion data of each component oil; The quality attribute prediction value of the finished oil is predicted by using the soft measurement module, the quality attribute values ​​of each component oil and the original proportion data, and the target gain is determined based on the quality attribute prediction value and the original proportion data. The model gain of the original multivariable controller is adjusted using the target gain to obtain the adjusted target multivariable controller. The formula adjustment value is obtained through the target multivariate controller and the quality attribute deviation; wherein, the quality attribute deviation is the deviation between the target value of the quality attribute and the predicted value of the quality attribute. Based on the formula adjustment value and the original proportion data, determine the target proportion data; The target ratio data is used to generate a formula control command, which is then sent to the control equipment to adjust the component oil ratio in the oil blending process.

2. The method for optimizing and controlling the oil ratio of blending components according to claim 1, characterized in that, Using a soft measurement module and the quality attribute values ​​of each component oil, along with the original blending data, the predicted quality attribute values ​​of the finished oil are calculated. Based on these predicted quality attribute values ​​and the original blending data, the target gain is determined, including: The quality attribute values ​​of each component oil and the original proportion data are input into the soft measurement module; wherein, the soft measurement module includes a nonlinear parameterized blending model; the nonlinear parameterized blending model is a data-driven model or a semi-mechanistic model; The nonlinear parameterized blending model is used to predict the quality attributes of the finished oil, and the target gain is determined based on the predicted quality attributes and the original blending data.

3. The method for optimizing and controlling the oil ratio of blending components according to claim 2, characterized in that, The target gain is determined based on the predicted values ​​of quality attributes and the original proportioning data, including: Calculate the partial derivatives of the predicted quality attribute values ​​and the original proportion data, and use these partial derivatives as the target gain. The formula for calculating the target gain is as follows: ;in, For target gain, For partial differential operators, For predicted values ​​of quality attributes, These are the original proportions.

4. The method for optimizing and controlling the oil ratio of oil blending components according to claim 2, characterized in that, The prediction of the quality attributes of refined oil products using the nonlinear parameterized blending model includes: Using the aforementioned nonlinear parameterized blending model, the quality attribute values ​​of each component oil, and the original proportioning data, the predicted quality attribute values ​​of the finished oil are calculated. The nonlinear parameterized harmonic model is as follows: ； in, For predicted values ​​of quality attributes, Let be the weighting parameter for the i-th component oil. This represents the original proportioning data for the i-th component oil. Let i be the quality attribute value of the i-th component oil. The bias parameter is used to characterize the harmonic nonlinearity.

5. The method for optimizing and controlling the oil ratio of blending components according to claim 4, characterized in that, The oil-to-liquid ratio control method further includes: Obtain the actual quality attribute values ​​of the refined oil; Based on the actual quality attribute values ​​of the finished oil and the quality attribute values ​​of each component oil, the weight parameters and bias parameters of the nonlinear parameterized blending model are updated.

6. The method for optimizing and controlling the oil ratio of oil blending components according to claim 1, characterized in that, If the original multivariate controller is a model predictive controller, then adjusting the model gain of the original multivariate controller using the target gain to obtain the adjusted target multivariate controller includes: The steady-state gain matrix of the model predictive controller is dynamically updated using the target gain, and the updated model predictive controller is used as the target multivariate controller.

7. The method for optimizing and controlling the proportion of oil components in oil blending according to any one of claims 1 to 6, characterized in that, Obtain the quality attribute values ​​and original proportions of each component oil, including: The quality attribute values ​​of each component oil are measured using different types of sensors; each component oil includes at least one quality attribute value. Use the target ratio data from the previous moment as the original ratio data.

8. A soft-sensor-based oil blending component ratio optimization control system, characterized in that, include: The soft measurement module is used to acquire the quality attribute values ​​and original proportion data of each component oil, and predict the quality attribute values ​​of the finished oil based on the quality attribute values ​​and original proportion data of each component oil, and determine the target gain based on the quality attribute prediction values ​​and original proportion data. A multivariate controller is used to adjust the model gain through the target gain and obtain the formula adjustment value based on the input quality attribute bias; wherein, the quality attribute bias is the deviation between the target value of the quality attribute and the predicted value of the quality attribute; and the target proportion data is determined based on the formula adjustment value and the original proportion data. The target ratio data is used to generate a formula control command, which is then sent to the control equipment to adjust the component oil ratio in the oil blending process.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the oil blending component oil ratio optimization control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the oil blending component oil ratio optimization control method as described in any one of claims 1 to 7.

Images