A multi-objective blending optimization method for refined oil products in refining and chemical enterprises

Abstract

The present invention discloses a multi-objective blending optimization method for refined oil products of a refining enterprise, and specifically relates to the technical field of target optimization. The method comprises the following steps: constructing an original physical property feature set by identifying nonlinear response sensitive variables in the blending component ratio; constructing an oil product performance prediction model and a blending component interaction influence model based on the original physical property feature set; utilizing the oil product performance prediction model, analyzing the influence of nonlinear coupling of simultaneous changes in the ratios of multiple blending components on the variation trend of key performance indicators of refined oil products; utilizing the blending component interaction influence model, analyzing the influence of changes in the ratios of single or individual blending components on the stability of key performance indicators of refined oil products; constructing a constraint equation based on the variation trend characteristics of key performance and fluctuation sensitivity indicators, generating a blending solution set, performing target performance achievement evaluation and global optimal solution screening, obtaining an optimal blending solution, improving the blending prediction accuracy, and realizing stable optimization control of refined oil product quality.

Description

Technical Field

[0001] The present invention relates to the technical field of target optimization, and more particularly to a multi-target blending optimization method for refined oil products in a refining enterprise. Background Art

[0002] In the production process of refined oil products in refining and chemical enterprises, the blending process, as a key link to achieve product performance standards and cost control, usually relies on the combined prediction and optimization of the physical and chemical properties of different blending components.

[0003] Existing technologies mostly use linear superposition assumptions or static rules to deal with the relationship between blending components, without considering the nonlinear coupling effects that may exist among the blending components under actual process conditions. This leads to deviations between predicted results and actual performance, which in turn causes inaccurate blending plans, significant fluctuations in finished oil performance, and increased difficulty in production control. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a multi-objective blending optimization method for refined oil products of a refinery enterprise to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A multi-objective blending optimization method for refined oil products in a refinery comprises the following steps:

[0007] S1: Obtain historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components, identify nonlinear response sensitive variables in the blending component ratio, and construct the original physical property feature set;

[0008] S2: Based on the original physical property feature set, the oil performance prediction model and the blending component interaction model are constructed respectively;

[0009] S3: Based on the oil performance prediction model, analyze the impact of nonlinear coupling of simultaneous changes in the proportions of multiple blending components on the changing trends of key performance indicators of refined oil products;

[0010] S4: Based on the component interaction model, analyze the impact of changes in the proportion of single or individual blending components on the stability of key performance indicators of finished oil products;

[0011] S5: Based on the trend characteristics of key performance changes and fluctuation sensitivity indicators, construct constraint equations for multi-objective blending optimization and generate a set of blending solutions;

[0012] S6: Based on the blending solution set, perform target performance achievement evaluation and global optimal solution screening to obtain the optimal blending solution to guide production.

[0013] In a preferred embodiment, S1 is specifically:

[0014] Obtain historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components;

[0015] Pre-process the acquired historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components;

[0016] Based on the pre-processed historical blending formula data, the actual quality test data of the finished oil, and the physical and chemical property index data of the blending components, the nonlinear response sensitive variables in the blending component ratio are determined through sensitivity analysis;

[0017] The original physical property feature set is constructed based on the pre-processed historical blending formula data, the actual quality inspection data of the finished oil, the physicochemical property index data of the blending components and the determined nonlinear response sensitive variables.

[0018] In a preferred embodiment, S2 is specifically:

[0019] The model inputs are historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components, and nonlinear response sensitive variables;

[0020] Use neural network algorithm to establish oil performance prediction model;

[0021] The support vector regression algorithm was used to establish the interaction model of blending components.

[0022] In a preferred embodiment, S3 is specifically:

[0023] The historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components and nonlinear response sensitive variables are used as inputs of the oil performance prediction model;

[0024] The oil product performance prediction model is used to calculate and obtain the nonlinear coupling relationship of multiple blending component ratios that change simultaneously and predict the key performance indicators of finished oil products.

[0025] Based on the prediction results of the key performance indicators of refined oil products, the nonlinear coupling relationship between the simultaneous changes in the proportions of multiple blending components and the changing rules of the key performance indicators of refined oil products are analyzed to obtain the changing trend characteristics of the key performance indicators of refined oil products.

[0026] In a preferred embodiment, S4 is specifically:

[0027] The historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components and nonlinear response sensitive variables are used as inputs of the blending component interaction model;

[0028] The interaction model of blending components is used to calculate and obtain the prediction results of the stability of key performance indicators of finished oil products due to the change of the proportion of single or individual blending components;

[0029] Based on the prediction results of the stability of the key performance indicators of refined oil products, the relationship between the changes in the proportion of single or individual blending components and the stability of the key performance indicators of refined oil products is analyzed to obtain the fluctuation sensitivity index of the key performance indicators of refined oil products.

[0030] In a preferred embodiment, S5 is specifically:

[0031] Determine the key performance constraints in the multi-objective blending optimization process based on the changing trend characteristics of the key performance indicators of refined oil products;

[0032] Determine the component ratio constraints in the multi-objective blending optimization process based on the fluctuation sensitivity index of the key performance indicators of the refined oil products;

[0033] Based on the performance index constraints and component ratio constraints, the constraint equations for multi-objective optimization are established;

[0034] The constraint equations are solved by the particle swarm optimization algorithm to generate multiple blending solutions that meet the key performance indicator constraints and component ratio constraints, forming a blending solution set.

[0035] In a preferred embodiment, S6 is specifically:

[0036] Calculate the target performance achievement of the key performance indicators of the refined oil corresponding to each blending solution in the blending solution set;

[0037] Comprehensively evaluate all blending solutions in the blending solution set based on the target performance achievement of key performance indicators of refined oil products;

[0038] Based on the comprehensive evaluation results, the blending solution set is screened to obtain the optimal blending solution with the highest comprehensive evaluation of key performance indicators.

[0039] The technical effects and advantages of the multi-objective blending optimization method for refined oil products of a refinery enterprise of the present invention are as follows:

[0040] By introducing the identification of nonlinear response sensitive variables, the nonlinear impact factors of blending component changes on finished oil performance are effectively revealed, and the expression integrity of data features is improved; by separately constructing oil performance prediction models and blending component interaction models, the performance trends and stability of key performance indicators of finished oil can be accurately modeled and separated and identified in different dimensions; by analyzing the changing trends of key performance indicators of finished oil under nonlinear coupling with simultaneous changes in the proportions of multiple blending components, the prediction accuracy is enhanced; by capturing the impact of changes in the proportion of single or individual components on the stability of key performance indicators, the robustness of the optimization scheme to process disturbances is enhanced; constructing constraint equations for multi-objective blending optimization, generating a set of blending solutions, and improving the feasibility and practicality of blending solutions; through performance achievement and comprehensive evaluation, the optimal blending scheme is screened to achieve stable control of finished oil quality and improvement of blending efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The present invention is a schematic diagram of a multi-objective blending optimization method for refined oil products of a refinery. DETAILED DESCRIPTION

[0042] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0043] Example 1

[0044] Figure 1 The present invention provides a multi-objective blending optimization method for refined oil products in a refinery, which comprises the following steps:

[0045] S1: Obtain historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components, identify nonlinear response sensitive variables in the blending component ratio, and construct the original physical property feature set;

[0046] S2: Based on the original physical property feature set, the oil performance prediction model and the blending component interaction model are constructed respectively;

[0047] S3: Based on the oil performance prediction model, analyze the impact of nonlinear coupling of simultaneous changes in the proportions of multiple blending components on the changing trends of key performance indicators of refined oil products;

[0048] S4: Based on the component interaction model, analyze the impact of changes in the proportion of single or individual blending components on the stability of key performance indicators of finished oil products;

[0049] S5: Based on the trend characteristics of key performance changes and fluctuation sensitivity indicators, construct constraint equations for multi-objective blending optimization and generate a set of blending solutions;

[0050] S6: Based on the blending solution set, perform target performance achievement evaluation and global optimal solution screening to obtain the optimal blending solution to guide production.

[0051] S1: Obtain historical blending formula data, actual finished oil quality test data, and physical and chemical property index data of blending components, identify nonlinear response sensitive variables in the blending component ratio, and construct the original physical property feature set, including:

[0052] Obtain historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components;

[0053] Historical blending recipe data refers to the mass ratio data of each blending component actually used and recorded by a refinery during past production cycles. For example, in the production of No. 92 gasoline, a refinery uses blending components such as naphtha, reformate, isomerized gasoline, and MTBE. The recorded ratios for each blending operation, expressed in weight percentage, constitute historical blending recipe data. Actual quality test data for finished oil products refers to performance indicators determined through sampling and testing of each batch of finished oil after blending. For example, gasoline product performance indicators include octane number, density, vapor pressure, sulfur content, and distillation range. These test results are recorded using measuring instruments, and each batch corresponds to unique test data. Blending component physicochemical property data refers to the inherent physical and chemical properties of each blending component, including but not limited to density, sulfur content, olefin content, aromatics content, and distillation range. Physicochemical property data are measured using experimental methods or instruments. For example, density is measured using a standard densitometer, and sulfur content is determined using fluorescence or combustion methods.

[0054] Pre-process the acquired historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components;

[0055] Preprocessing includes data cleaning, data correction, and data standardization. Data cleaning involves checking historical blending formula data, actual finished oil quality test data, and the physicochemical property data of blending components, removing abnormal data points that contain recording errors or deviate from the acceptable range. Data correction involves checking the logical consistency of actual finished oil quality test data with historical blending formula data. For example, this involves determining the logical matching relationship between the recorded blending component ratio data and the actual measured finished oil performance indicators. If the performance indicators and blending component ratios do not conform to the known production rules of the refinery, re-verification and correction are performed. Data standardization involves converting the physicochemical property data of blending components according to a unified standard or scale to make the indicators of different components comparable. Specifically, the value of each physicochemical property indicator is subtracted from the minimum value of the corresponding physicochemical property indicator in the entire data set, and then divided by the difference between the maximum and minimum values ​​of the physicochemical property indicator in the entire data set to obtain the standardized indicator data. After data preprocessing, the historical blending formula data, the actual quality inspection data of finished oil products, and the physical and chemical properties index data of the blending components meet the requirements of data quality and availability.

[0056] Based on the pre-processed historical blending formula data, the actual quality test data of the finished oil, and the physical and chemical property index data of the blending components, the nonlinear response sensitive variables in the blending component ratio are determined through sensitivity analysis;

[0057] During the blending process, the ratio of each blending component in historical blending recipe data was used as the analysis object, and each key performance indicator recorded in the actual quality inspection data of the finished oil product was used as the response variable. A quantitative relationship was established between changes in the blending component ratio and the key performance indicator. Specifically, variance analysis was used to calculate the significance of the changes in the key performance indicator caused by changes in the proportion of each blending component. The significance calculation was based on the ratio of the between-group difference to the within-group difference in the change in the key performance indicator caused by the change in the proportion of each blending component. If the ratio exceeded the set significance threshold, the change in the blending component ratio was determined to be sensitive to the key performance indicator.

[0058] For example, when the proportion of reformed oil gradually increases from a lower proportion to a specific threshold in the historical blending formula data, it is found that the corresponding finished oil octane number shows an obvious nonlinear saturation trend, that is, the rate at which the octane number decreases significantly with the increase of the reformed oil proportion and tends to be constant. This specific threshold is the nonlinear response sensitive threshold of the reformed oil, that is, the proportion of the reformed oil is marked as a nonlinear response sensitive variable.

[0059] Based on the pre-processed historical blending formula data, the actual quality test data of the finished oil, the physicochemical property index data of the blending components, and the determined nonlinear response sensitive variables, the original physical property feature set is constructed;

[0060] The original physical property feature set represents a data set formed after data preprocessing and sensitive variable identification, integrating historical blending formula data, actual finished oil quality test data, physicochemical property index data of blending components, and nonlinear response sensitive variables. The original physical property feature set represents all the basic data characteristics required for the finished oil blending process and is organized in a data table format. For example, each data sample clearly lists the blending formula component ratios, physicochemical property indexes, corresponding actual test performance indicators, and identified nonlinear response sensitive variables.

[0061] S2: Based on the original physical property feature set, an oil performance prediction model and a blending component interaction model are constructed, including:

[0062] The model inputs are historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components, and nonlinear response sensitive variables;

[0063] Use neural network algorithm to establish oil performance prediction model;

[0064] The oil performance prediction model is a nonlinear prediction model constructed using a multi-layered neural network algorithm. It is used to predict the actual key performance indicators (KPIs) of refined oil products under specific blending conditions. The neural network algorithm is a multi-layer feedforward neural network structure consisting of an input layer, multiple hidden layers, and an output layer. The neural network algorithm is trained using a backpropagation algorithm. Historical blending recipe data and corresponding actual refined oil quality test data are used as training samples. The error between the predicted output and the actual performance indicators is calculated, and an error function is used to guide the iterative adjustment and optimization of the connection weights in the network. The error function is calculated by calculating the difference between each KPI predicted by the neural network and the actual performance indicator, squaring the differences, summing them, and averaging them to obtain the error function value. The training process is repeated until the error function value falls below the set error limit, completing the establishment and parameter optimization of the oil performance prediction model. After training, the oil performance prediction model is capable of predicting new recipe data.

[0065] The support vector regression algorithm was used to establish the interaction model of blending components;

[0066] The blending component interaction model is represented as a nonlinear regression analysis model established using the support vector regression algorithm. It is used to analyze and quantify the sensitivity of changes in the blending component ratios to the stability of key performance indicators of refined oil products. The support vector regression algorithm is a nonlinear regression method based on statistical learning theory. The algorithm uses historical blending recipe data, the physicochemical properties of the blending components, and nonlinear response-sensitive variables as input variables, and the fluctuation sensitivity indicators of key performance indicators from actual refined oil quality testing data as output variables. The support vector regression algorithm establishes a nonlinear quantitative functional relationship between changes in the blending component ratios and the stability of key performance indicators. The support vector regression algorithm uses a kernel function to nonlinearly map the input data, consisting of historical blending recipe data, the physicochemical properties of the blending components, and the nonlinear response-sensitive variables, from a low-dimensional space to a high-dimensional feature space. In this high-dimensional feature space, a nonlinear regression functional relationship is constructed between changes in the blending component ratios and the fluctuation sensitivity indicators of key performance indicators of refined oil products. The kernel function uses a radial basis kernel function. The calculation method for the radial basis kernel function is as follows: first, the square of the difference between each two input data samples is calculated, and then an exponential function transformation is performed with the negative value of the square of the difference as the exponent. This results in the kernel function value approaching zero as the difference between each two data samples increases, i.e., the kernel function value decreases. After completing the data mapping, the support vector regression algorithm establishes a nonlinear regression function relationship in the high-dimensional feature space. This nonlinear regression function relationship is established by solving an optimization problem to determine the model parameters in the regression function. The goal of the optimization problem is to accurately fit the key performance indicator's volatility sensitivity index while reducing the complexity of the nonlinear regression function. The regression function's fitting accuracy for the key performance indicator's volatility sensitivity index is evaluated by calculating the sum of squared errors between the volatility sensitivity index predicted by the support vector regression algorithm and the actual volatility sensitivity index. The complexity of the regression function is reduced by introducing a regularization term in the optimization problem that constrains the numerical value of the function parameters, thereby constraining the size of the function model parameters and reducing the complexity of the function model. The support vector regression algorithm uses a sequential minimum optimization algorithm to solve optimization problems. The sequential minimum optimization algorithm selects two data samples each time to optimize and update parameters. The optimization update steps are repeated step by step, and the objective function value of the optimization problem is calculated and recorded after each update. When the numerical change in the optimization problem objective function corresponding to two consecutive optimization update iterations falls below a set minimum threshold, the optimization solution is determined to have reached a convergence state, and the optimization parameter update process is terminated. Through the above complete support vector regression algorithm process, a blending component interaction model is established, which expresses the nonlinear functional relationship between the changes in the ratio of a single or individual blending component and the sensitivity to fluctuations in the key performance indicators of the refined oil product.

[0067] S3: Based on the oil performance prediction model, analyze the impact of nonlinear coupling of simultaneous changes in the proportions of multiple blending components on the changing trends of key performance indicators of refined oil products, including:

[0068] The historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components and nonlinear response sensitive variables are used as inputs of the oil performance prediction model;

[0069] The oil product performance prediction model is used to calculate and obtain the nonlinear coupling relationship of multiple blending component ratios that change simultaneously and predict the key performance indicators of finished oil products.

[0070] The historical blending formula data, physicochemical property indicators of the blending components, and nonlinear response-sensitive variables for each row in the data table are input to the input layer of the oil performance prediction model. Each neuron in the input layer corresponds to an input variable in the data table. Each input layer neuron performs a weighted summation of the input values ​​using the corresponding connection weights and transmits the weighted summation to the next layer. In the hidden layer, each processing unit receives the weighted summation from the input layer and performs calculations using a nonlinear activation function. The nonlinear activation function uses either the hyperbolic tangent function or the logistic function. The hyperbolic tangent function is calculated by first subtracting the exponential function value of the input value from the exponential function value of the negative input value, and then dividing it by the sum of the exponential function value of the input value and the exponential function value of the negative input value. The logistic function is calculated by first calculating the exponential function value of the negative input value plus one, and then dividing the result by one to obtain the output value. The hidden layer passes the data through these nonlinear calculations to the output layer. The output layer generates predictions for the key performance indicators (KPIs) of refined oil products. Each neuron in the output layer corresponds to a KPI for refined oil products, including predicted values ​​for octane number, density, vapor pressure, sulfur content, and distillation range. The calculation method for each neuron in the output layer is to multiply the data transmitted from the previous hidden layer by the weights connecting the neurons in the output layer, and then sum all these products to obtain the predicted KPI value. Through these calculation steps, the oil performance prediction model predicts the KPIs of refined oil products under specific formulations based on the nonlinear coupling relationship of multiple blending component ratios that vary simultaneously.

[0071] Based on the prediction results of the key performance indicators of refined oil products, the nonlinear coupling relationship between the simultaneous changes in the proportions of multiple blending components and the changing patterns of the key performance indicators of refined oil products is analyzed to obtain the changing trend characteristics of the key performance indicators of refined oil products;

[0072] Based on a series of prediction results generated by an oil product performance prediction model, this study analyzes the nonlinear trends and patterns in the key performance indicators of refined oil products when the blending ratios of multiple blending components are simultaneously varied within a specific range. The analysis method is trend analysis. Trend analysis involves analyzing the regular trends in the predicted values ​​of key performance indicators as the blending ratios gradually change. For example, the nonlinear saturation trend of the octane number indicator as the reformate ratio gradually increases is analyzed. This trend indicates that the octane number increases linearly and rapidly at low blending ratios, but the increase slows or even flattens out after reaching a specific blending ratio. Through this analysis method, the patterns between the changes in the blending ratios of all involved blending components in the prediction results and the changes in key performance indicators are analyzed and summarized item by item, thereby obtaining the trend characteristics of the key performance indicators of refined oil products. The trend characteristics of the key performance indicators of refined oil products are represented as a series of feature descriptions, including a description of the nonlinear pattern of each key performance indicator's change with each blending component ratio change. For example, the nonlinear saturation trend characteristic of octane number as the proportion of reformate increases.

[0073] S4: Based on the component interaction model, analyze the impact of changes in the proportion of single or individual blending components on the stability of key performance indicators of finished oil products, including:

[0074] The historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components and nonlinear response sensitive variables are used as inputs of the blending component interaction model;

[0075] The interaction model of blending components is used to calculate and obtain the prediction results of the stability of key performance indicators of finished oil products due to the change of the proportion of single or individual blending components;

[0076] The historical blending formula data, physicochemical property data of each blending component, and nonlinear response sensitive variables for each row in the data table are input into the blending component interaction model. This model calculates a nonlinear regression function in a high-dimensional feature space to predict the stability of the key performance indicators of the refined oil product under changes in the proportions of individual blending components. The prediction results are expressed as a fluctuation sensitivity index for the key performance indicator, indicating the degree to which changes in the proportions of individual blending components affect the stability of the key performance indicator of the refined oil product.

[0077] Based on the prediction results of the stability of the key performance indicators of refined oil products, the relationship between the changes in the proportion of single or individual blending components and the stability of the key performance indicators of refined oil products is analyzed to obtain the fluctuation sensitivity index of the key performance indicators of refined oil products;

[0078] The volatility sensitivity index represents the sensitivity of a refined oil product's key performance indicator to fluctuations when the ratio of a single or individual blending component changes within a specific range. The specific analysis process involves analyzing and comparing the predicted volatility sensitivity index values ​​one by one to assess the impact of changes in the ratio of a single or individual blending component on each key performance indicator. The specific analysis method is the sensitivity value comparison method. This method compares the volatility sensitivity index values ​​output by the model under conditions of changes in the ratio of a single or individual blending component to determine the degree of impact of changes in the ratio of a single or individual blending component on the stability of the key performance indicator. For example, the analysis determined that when the naphtha ratio changes within a specific range, the volatility sensitivity value for the octane number indicator is small, indicating that the impact of changes in the naphtha ratio on the stability of the octane number indicator is low. However, when the methyl tertiary butyl ether ratio changes within a specific range, the volatility sensitivity value for the vapor pressure indicator is large, indicating that changes in the methyl tertiary butyl ether ratio have a high impact on the stability of the vapor pressure indicator.

[0079] S5: Based on the trend characteristics and fluctuation sensitivity indicators of key performance indicators, a multi-objective blending optimization constraint equation is constructed to generate a blending solution set, including:

[0080] Determine the key performance constraints in the multi-objective blending optimization process based on the changing trend characteristics of the key performance indicators of refined oil products;

[0081] Key performance constraints represent the target performance range or trend limits allowed for each key performance indicator during the blending process, ensuring that the final blended oil product meets expected performance standards. For example, if the oil performance prediction model and trend analysis indicate that the octane number exhibits a nonlinear saturation trend after reaching a specific reformate ratio, meaning that increasing the reformate ratio beyond that specific ratio has limited octane improvement, then the specific ratio will be used as the upper limit of the performance constraint for the reformate ratio. Through this analytical approach, corresponding performance constraints are determined for each key performance indicator, generating a set of key performance constraints.

[0082] Determine the component ratio constraints in the multi-objective blending optimization process based on the fluctuation sensitivity index of the key performance indicators of the refined oil products;

[0083] Component ratio constraints represent specific constraints on the variation range of the ratios of individual or multiple blending components during multi-objective blending optimization. They are used to ensure the stability of key performance indicators of the refined oil product remains within acceptable limits. Specifically, the component ratio constraints are determined based on the fluctuation sensitivity index values ​​of each blending component. For example, considering the high sensitivity of changes in the methyl tertiary butyl ether (MTBE) ratio to the stability of the vapor pressure index, if the fluctuation sensitivity analysis indicates that significant fluctuations in the vapor pressure index occur when the MTBE ratio exceeds a specific upper limit or falls below a specific lower limit, these upper and lower limits will be used as the component ratio constraint range for MTBE. Through this analytical method, the ratio ranges of individual or multiple blending components are determined, forming a set of component ratio constraints.

[0084] Based on the performance index constraints and component ratio constraints, the constraint equations for multi-objective optimization are established;

[0085] The constraint equation for multi-objective optimization is expressed as a mathematical programming model, clarifying the relationship between specific objective functions and constraints. The objective function is defined as optimizing multiple key performance indicators (KPIs) during the blending process. For example, the objective function requires simultaneously achieving optimization targets for octane number, density, and vapor pressure. The optimization objective function is calculated as the weighted sum of the squared differences between each KPI and the target performance, with the weights explicitly indicating the importance of each KPI. For example, if a refinery prioritizes octane number over vapor pressure, the squared difference in octane number will account for a larger proportion of the optimization objective function. The constraints in the constraint equation include KPI constraints and component ratio constraints, expressed as mathematical inequalities or equations. For example, the octane number must be greater than or equal to a specific target value, and the MTBE ratio must be within specific upper and lower limits. This method forms a multi-objective optimization constraint equation system.

[0086] The constraint equations are solved using the particle swarm optimization algorithm to generate multiple blending solutions that meet the key performance indicator constraints and component ratio constraints, forming a blending solution set;

[0087] Multiple particles are initialized and randomly distributed throughout the search space. Each particle represents a specific blending solution, including a blending component ratio. Particles are evaluated according to a defined optimization objective function, and the optimization search is performed by repeatedly updating their positions. The updated position of each particle is expressed as follows: using the current particle's optimal position and the optimal position of all particles as references, the distance and direction between the current particle and the optimal position are calculated. After multiplying the distance and direction by a random weight, the current particle's position is adjusted to achieve the updated position. The updated position represents the new blending solution. The update is repeated until the optimization convergence condition is reached. The optimization convergence condition is expressed as the optimization objective function repeatedly changing below a set threshold. Through this optimization process, multiple specific blending solutions that meet the key performance indicator constraints and component ratio constraints are obtained.

[0088] The multiple blending solutions that satisfy the constraints, calculated by the particle swarm optimization algorithm, are collected and organized to form a blending solution set. This blending solution set is explicitly represented as a dataset, where each solution represents a blending component ratio and the corresponding predicted key performance indicator value.

[0089] S6: Based on the blending solution set, perform target performance achievement evaluation and global optimal solution screening to obtain the optimal blending solution to guide production, including:

[0090] Calculate the target performance achievement of the key performance indicators of the refined oil corresponding to each blending solution in the blending solution set;

[0091] The target performance achievement degree represents the degree of consistency or deviation between the predicted key performance indicators (KPIs) for each blending solution and the predetermined target performance indicators. The calculation process is as follows: For each blending solution in the blending solution set, the corresponding predicted KPI values ​​are extracted, including the predicted octane number, density, vapor pressure, sulfur content, and distillation range. The predicted KPI values ​​are compared with the target values ​​for each KPI predetermined by the refinery, and the achievement degree of each KPI is calculated. The target performance achievement degree is calculated by first determining the absolute value of the difference between the predicted KPI and the target KPI. Using the target KPI as a benchmark, the ratio of the absolute value of the difference to the target KPI is calculated by dividing the absolute value of the difference by the target KPI. A smaller ratio indicates a closer match between the predicted KPI and the target, i.e., a higher achievement degree; a larger ratio indicates a greater deviation between the predicted KPI and the target, i.e., a lower achievement degree. The ratio is calculated for each KPI to obtain the target performance achievement degree for each KPI. The target performance achievement degrees of all KPIs are then weighted and summed. The weight represents the importance a company places on each performance indicator during the production process. For example, if a refinery predetermines that octane number is given the highest weight, followed by vapor pressure, then the octane number's target performance achievement will be the largest factor in the weighted summation. This calculation yields the target performance achievement for each blending solution.

[0092] Comprehensively evaluate all blending solutions in the blending solution set based on the target performance achievement of key performance indicators of refined oil products;

[0093] Comprehensive evaluation involves comprehensively analyzing the target performance achievement of each KPI for each blending solution, thereby obtaining a comprehensive evaluation score for the blending solution. This score measures the overall achievement of the target performance across all KPIs. The comprehensive evaluation method is a weighted summation approach. The calculation process is as follows: First, a weight is assigned to each KPI. The weight represents the relative importance of the finished oil performance indicator in the actual production and application of the refinery. For example, if octane number is the most important KPI for gasoline production, the highest weight is assigned to octane number. For KPIs such as vapor pressure, density, or sulfur content, weights are assigned based on production requirements and market demand. The target performance achievement of each KPI for each blending solution is multiplied by the corresponding weight. These products are then summed to obtain a comprehensive evaluation score for the blending solution. A higher comprehensive evaluation score indicates that the blending solution more effectively meets the requirements of the multi-objective blending optimization. Through this method, comprehensive evaluation scores are calculated for all blending solutions, forming a set of comprehensive evaluation scores.

[0094] Based on the comprehensive evaluation results, the blending solution set is screened to obtain the optimal blending solution with the highest comprehensive evaluation of key performance indicators;

[0095] The screening process is as follows: sort the comprehensive evaluation score sets of all blending schemes; the sorting method is: sort from high to low according to the size of the comprehensive evaluation scores, thereby generating an ordered set of blending schemes; select the blending scheme that is at the front of the blending scheme set after sorting, that is, the blending scheme with the highest comprehensive evaluation score as the optimal blending scheme. The optimal blending scheme is represented by the blending ratio scheme that has the highest target performance achievement of key performance indicators and is closest to the company's expected goals. For example, after screening, if the octane number, density, and vapor pressure of a specific blending scheme are closest to the target performance indicators, and the comprehensive evaluation score of the specific blending scheme is the highest, then the specific blending scheme is determined to be the optimal blending scheme for actual production guidance of the refining enterprise, including the ratio of each blending component and the predicted values ​​of each key performance indicator predicted by the corresponding ratio.

[0096] The above embodiments can be implemented in whole or in part via 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 comprises one or more computer instructions or computer programs. When loaded or executed on a computer, the processes or functions described in the embodiments of this application are fully or partially performed. 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. 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 means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. The available medium can be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0097] Those skilled in the art will appreciate that the modules 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.

[0098] 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 modules described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0099] 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 modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules 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 modules, which can be electrical, mechanical or other forms.

[0100] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of this embodiment according to actual needs.

[0101] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0102] If the functions are implemented in the form of software function modules 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, server, or 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.

[0103] 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.

[0104] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multi-objective blending optimization method for refined oil products in a refinery, characterized in that: The steps include: S1: Obtain historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components, identify nonlinear response sensitive variables in the blending component ratio, and construct the original physical property feature set; S2: Based on the original physical property feature set, the oil performance prediction model and the blending component interaction model are constructed respectively; S3: Based on the oil performance prediction model, analyze the impact of nonlinear coupling of simultaneous changes in the proportions of multiple blending components on the changing trends of key performance indicators of refined oil products; The historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components and nonlinear response sensitive variables are used as inputs of the oil performance prediction model; The oil product performance prediction model is used to calculate and obtain the nonlinear coupling relationship of multiple blending component ratios that change simultaneously and predict the key performance indicators of finished oil products. Based on the prediction results of the key performance indicators of refined oil products, the nonlinear coupling relationship between the simultaneous changes in the proportions of multiple blending components and the changing patterns of the key performance indicators of refined oil products is analyzed to obtain the changing trend characteristics of the key performance indicators of refined oil products; S4: Based on the component interaction model, analyze the impact of changes in the proportion of single or individual blending components on the stability of key performance indicators of finished oil products; The historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components and nonlinear response sensitive variables are used as inputs of the blending component interaction model; The interaction model of blending components is used to calculate and obtain the prediction results of the stability of key performance indicators of finished oil products due to the change of the proportion of single or individual blending components; Based on the prediction results of the stability of the key performance indicators of refined oil products, the relationship between the changes in the proportion of single or individual blending components and the stability of the key performance indicators of refined oil products is analyzed to obtain the fluctuation sensitivity index of the key performance indicators of refined oil products; S5: Based on the trend characteristics of key performance changes and fluctuation sensitivity indicators, construct constraint equations for multi-objective blending optimization and generate a set of blending solutions; S6: Based on the blending solution set, perform target performance achievement evaluation and global optimal solution screening to obtain the optimal blending solution to guide production.

2. The multi-objective blending optimization method for refined oil products of a refinery according to claim 1, characterized in that: S1, specifically: Obtain historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components; Pre-process the acquired historical blending formula data, actual quality test data of finished oil products, and physical and chemical property index data of blending components; Based on the pre-processed historical blending formula data, the actual quality test data of the finished oil, and the physical and chemical property index data of the blending components, the nonlinear response sensitive variables in the blending component ratio are determined through sensitivity analysis; The original physical property feature set is constructed based on the pre-processed historical blending formula data, the actual quality inspection data of the finished oil, the physicochemical property index data of the blending components and the determined nonlinear response sensitive variables.

3. The multi-objective blending optimization method for refined oil products of a refinery according to claim 2, characterized in that: S2, specifically: The model inputs are historical blending formula data, actual quality test data of finished oil, physicochemical property index data of blending components, and nonlinear response sensitive variables; Use neural network algorithm to establish oil performance prediction model; The support vector regression algorithm was used to establish the interaction model of blending components.

4. The multi-objective blending optimization method for refined oil products of a refinery according to claim 3, characterized in that: S5, specifically: Determine the key performance constraints in the multi-objective blending optimization process based on the changing trend characteristics of the key performance indicators of refined oil products; Determine the component ratio constraints in the multi-objective blending optimization process based on the fluctuation sensitivity index of the key performance indicators of the refined oil products; Based on the performance index constraints and component ratio constraints, the constraint equations for multi-objective optimization are established; The constraint equations are solved by the particle swarm optimization algorithm to generate multiple blending solutions that meet the key performance indicator constraints and component ratio constraints, forming a blending solution set.

5. The multi-objective blending optimization method for refined oil products of a refinery according to claim 4, characterized in that: S6, specifically: Calculate the target performance achievement of the key performance indicators of the refined oil corresponding to each blending solution in the blending solution set; Comprehensively evaluate all blending solutions in the blending solution set based on the target performance achievement of key performance indicators of refined oil products; Based on the comprehensive evaluation results, the blending solution set is screened to obtain the optimal blending solution with the highest comprehensive evaluation of key performance indicators.

Images