Optimization methods, equipment, electronic devices and storage media for residue hydrotreating units

Abstract

This invention provides an optimization method, apparatus, electronic equipment, and storage medium for a residue hydrotreating unit. The method includes: performing cluster analysis on the current operating data of the residue hydrotreating unit based on actual operating data and generated operating data to obtain similar operating data, and determining the range of operating parameters based on the similar operating data; performing global optimization within the range of operating parameters based on an objective function and the constraint relationship between the operating parameters and product information to obtain optimized operating parameters, wherein the objective function is determined by the operating parameters and the corresponding product information; and optimizing the operating parameters of the residue hydrotreating unit based on the optimized operating parameters. The method, apparatus, electronic equipment, and storage medium provided by this invention can ensure that the optimization result approximates the theoretical optimal value, improve optimization efficiency and effect, and enhance the overall economic benefits of the unit.

Description

Technical Field

[0001] This invention relates to the field of petroleum refining technology, and in particular to an optimization method, apparatus, electronic equipment and storage medium for a residue hydrotreating unit. Background Technology

[0002] With the continuous development of artificial intelligence and big data in recent years, and the constant iteration and updating of technology, the petrochemical industry, as one of the pillar industries of energy, is also attempting to integrate with new information technologies and develop towards intelligentization. Moreover, due to the continuous deterioration of crude oil quality and the increasing demand for middle distillate oils, residue hydrocracking has become one of the most important secondary processing units. At the same time, the production process of residue hydrocracking is complex and the feedstocks are variable, offering significant room for optimization. Therefore, establishing a unit optimization method with high computational accuracy, fast response speed, and strong practicality is key to improving the production efficiency and profitability of current residue hydrocracking units.

[0003] Currently, the optimization method for residue hydrotreating units is based on matching the historical operating conditions of the unit, which cannot provide the optimal result and has its limitations. Summary of the Invention

[0004] This invention provides an optimization method, apparatus, electronic equipment, and storage medium for a residue hydrotreating unit, which addresses the shortcomings of poor optimization performance in existing residue hydrotreating units and improves the optimization effect of the residue hydrotreating unit.

[0005] This invention provides an optimization method for a residue hydrotreating unit, comprising:

[0006] Based on actual operating data and generated operating data, cluster analysis is performed on the current operating data of the residue hydrotreating unit to obtain similar operating data that are similar to the current operating data, and the range of operating parameters is determined based on the similar operating data.

[0007] Based on the objective function and the constraint relationship between the operating parameters and the product information, global optimization is performed within the range of the operating parameters to obtain optimized operating parameters. The objective function is determined by the operating parameters and the corresponding product information.

[0008] Based on the optimized operating parameters, the operating parameters of the residue hydrotreating unit are optimized.

[0009] According to the optimization method of the residue hydrotreating unit provided by the present invention, the constraint relationship between the operating parameters and product information is determined based on the product information prediction model;

[0010] The product information prediction model is trained based on actual sample operation parameters, actual product information corresponding to the actual sample operation parameters, sample operation parameters, and predicted product information corresponding to the generated sample operation parameters. The predicted product information is predicted based on the residue oil hydrogenation mechanism model.

[0011] According to the optimization method of a residue hydrotreating unit provided by the present invention, the product information prediction model is a long short-term memory neural network.

[0012] According to the optimization method of a residue hydrotreating unit provided by the present invention, the step of determining the range of operating parameters based on the similar operating condition data includes:

[0013] Based on the product information corresponding to the current working condition data, candidate working condition data that is superior to the product information is determined from the similar working condition data.

[0014] Based on the candidate operating condition data, the range of the operating parameters is determined.

[0015] According to the optimization method of a residue hydrotreating unit provided by the present invention, the step of performing cluster analysis on the current operating data of the residue hydrotreating unit based on actual operating data and generated operating data to obtain similar operating data that are similar to the current operating data includes:

[0016] Based on actual operating data and generated operating data, including feed flow rate, feed composition, feed properties, and operating parameters, a hierarchical clustering tree is constructed.

[0017] Based on the hierarchical clustering tree, cluster analysis is performed on the current operating condition data to obtain similar operating condition data that are similar to the current operating condition data.

[0018] According to the optimization method of a residue hydrotreating unit provided by the present invention, the operating parameters are obtained by performing correlation analysis between various parameters in actual operating data and generated operating data and product information, and obtaining parameters with high correlation to product information.

[0019] The present invention also provides an optimization device for a residue hydrotreating unit, comprising:

[0020] The clustering analysis unit is used to perform clustering analysis on the current operating data of the residue hydrotreating unit based on actual operating data and generated operating data, to obtain similar operating data that are similar to the current operating data, and to determine the range of operating parameters based on the similar operating data.

[0021] A global optimization unit is used to perform global optimization within the range of the operation parameters based on the objective function and the constraint relationship between the operation parameters and the product information, to obtain optimized operation parameters. The objective function is determined by the operation parameters and the corresponding product information.

[0022] The device optimization unit is used to optimize the operating parameters of the residue hydrotreating unit based on the optimized operating parameters.

[0023] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the optimization method of any of the above-described residue hydrotreating apparatuses.

[0024] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the optimization method of the residue hydrotreating apparatus as described above.

[0025] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the optimization method of any of the above-described residue hydrotreating apparatuses.

[0026] The present invention provides an optimization method, apparatus, electronic equipment, and storage medium for a residue hydrotreating unit. By performing cluster analysis on the current operating data of the residue hydrotreating unit based on actual and generated operating data, the range of operating parameters is obtained. Then, global optimization is performed by combining the constraint relationship between operating parameters and product information. This ensures that the optimization result approximates the theoretical optimal value, improving the reliability and accuracy of the optimization scheme and increasing computational efficiency. Based on this, the operating parameters of the residue hydrotreating unit are optimized, thereby improving optimization efficiency and effectiveness, significantly saving labor costs, and enhancing the unit's operating efficiency and overall economic benefits. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0028] Figure 1 This is one of the flowcharts illustrating the optimized method for the residue hydrotreating unit provided by the present invention;

[0029] Figure 2 This is a schematic diagram of the hierarchical clustering tree provided by the present invention;

[0030] Figure 3 This is a schematic diagram of the product information prediction model provided by the present invention;

[0031] Figure 4 This is a schematic diagram of the forget gate structure in the product information prediction model provided by the present invention;

[0032] Figure 5 This is a schematic diagram of the memory gate structure in the product information prediction model provided by the present invention;

[0033] Figure 6 This is a schematic diagram of the training process of the product information prediction model provided by the present invention;

[0034] Figure 7 This is the second schematic diagram of the optimized method for the residue hydrotreating unit provided by the present invention;

[0035] Figure 8 This is a schematic diagram of the structure of the optimized device for the residue hydrotreating unit provided by the present invention;

[0036] Figure 9 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0038] For highly complex residue oil reaction systems, a lumped approach is generally used for reaction kinetic modeling. This involves treating various individual compounds as virtual components based on similar kinetic properties, thus dividing the residue oil feed into several virtual components and constructing a lumped kinetic model based on the reactions. However, the lumped kinetic model composed of virtual residue oil components differs from the actual residue oil feed composition, failing to reflect changes in composition within the lumped model. It cannot accurately describe changes in feedstock and the reaction transformations of individual components within the lumped model, leading to inaccurate predictions of product yields and properties. Molecular-level mechanistic models are more complex than lumped kinetic models, reflecting more detailed information on the properties of individual product components. They have a wider range of applicability to feedstocks but also require higher-level analytical instruments, representing the future trend in refinery modeling. However, due to their complexity, high computational accuracy, high computational resource consumption, and long computation time, they are difficult to use in optimization calculations and cannot be used for real-time on-site optimization.

[0039] Current optimization methods for residue hydrotreating units are based on matching historical operating conditions, which cannot provide optimal results for units not previously operated, thus having limitations. To address this, this invention provides an optimization method for residue hydrotreating units. Figure 1 This is one of the flowcharts illustrating the optimized method for the residue hydrotreating unit provided by the present invention, such as... Figure 1 As shown, the method includes:

[0040] Step 110: Based on actual operating condition data and generated operating condition data, perform cluster analysis on the current operating condition data of the residue hydrotreating unit to obtain similar operating condition data that are similar to the current operating condition data, and determine the range of operating parameters based on the similar operating condition data.

[0041] Step 120: Based on the objective function and the constraint relationship between the operation parameters and the product information, perform global optimization within the range of the operation parameters to obtain the optimized operation parameters. The objective function is determined by the operation parameters and the corresponding product information.

[0042] Step 130: Optimize the operating parameters of the residue hydrotreating unit based on optimized operating parameters.

[0043] Specifically, current operating condition data refers to data related to the current operating conditions of the residue hydrotreating unit, such as feed information like composition and content, and operating parameters like reactor bed temperature, bed pressure, and liquid hourly space velocity. Actual operating condition data refers to the operating condition data from the actual operation of the residue hydrotreating unit, while generated operating condition data refers to the operating condition data generated mechanistically. Product information can be product yield or product properties. Here, product yield can be, for example, the yield of each component of the product, such as gasoline, diesel, kerosene, or tail oil; however, this embodiment of the invention does not specifically limit this.

[0044] Cluster analysis can be performed based on the feed information and operating parameter information of the current operating data of the residue hydrotreating unit. Similar operating data, i.e., similar operating data, can be identified from the actual and generated operating data. The upper and lower limits of the operating parameters in the similar operating data can then be extracted to determine the range of operating parameters. Here, the cluster analysis can be implemented using density clustering, hierarchical clustering, model clustering, etc., and this embodiment of the invention does not specifically limit the method used.

[0045] Subsequently, based on the objective function characterizing the overall efficiency of the residue hydrotreating unit and the constraints between operating parameters and product information, global optimization can be performed within the range of the operating parameters to obtain a set of operating parameters that maximizes the overall efficiency of the residue hydrotreating unit; these are the optimized operating parameters. The objective function can be determined based on the operating parameters, corresponding product information, and the prices of each component of the product. Based on these optimized operating parameters, the operating parameters that need optimization in the residue hydrotreating unit can then be further optimized.

[0046] Here, the global optimization algorithm can be one or more of the following methods: Pattern Search, Improved Adaptive Differential Evolution (JADE), Composite Differential Evolution Algorithm (CoDE), Simplex Nelder-Mead, and Particle Swarm Optimization (PSO). This embodiment of the invention does not impose specific limitations on this method.

[0047] It should be noted that by first identifying similar operating condition data to the current operating condition data, and then determining the range of operating parameters based on this similar operating condition data, the reliability of the optimized operating parameters can be guaranteed, the accuracy of the optimization scheme can be improved, and the constraint relationship between operating parameters and product information can be established. Furthermore, since some operating parameters within the range are not parameters actually used in the residue hydrotreating unit, the constraint relationship between operating parameters and product information is applied during the global optimization process. This constraint relationship can reflect the operating law of the residue hydrotreating unit from a mechanistic perspective, thereby further ensuring the reliability of the optimized operating parameters.

[0048] Furthermore, in practical applications, the operating parameters of residue hydrotreating units do not change frequently, and the upper and lower limits of each operating parameter are relatively narrow. Therefore, existing matching and optimization methods based on historical operating conditions of residue hydrotreating units cannot provide optimal operating conditions that the unit has not operated before. In contrast, this invention generates operating condition data under different feed conditions and operating conditions based on the applicable scope of the residue hydrotreating mechanism model. This generated operating condition data is then combined with actual operating condition data and the generated operating condition data to filter similar operating condition data. Global optimization is then performed within the range of operating parameters determined by these operating condition data to obtain the globally optimal solution, i.e., the optimized operating parameters. This ensures that the optimization result approximates the theoretical optimal value, which can then be applied to online optimization. Moreover, by searching for the globally optimal solution within the narrowed range, the consumption of computational resources is significantly reduced, and computational efficiency is improved.

[0049] The method provided in this invention performs cluster analysis on the current operating data of the residue hydrotreating unit based on actual and generated operating data to obtain the range of operating parameters. Then, it performs global optimization by combining the constraint relationship between operating parameters and product information. This ensures that the optimization result approximates the theoretical optimal value, improving the reliability and accuracy of the optimization scheme and increasing computational efficiency. Based on this, the operating parameters of the residue hydrotreating unit are optimized, thereby improving optimization efficiency and effect, significantly saving labor costs, and improving the operating efficiency and overall economic benefits of the unit.

[0050] Based on any of the above embodiments, the constraint relationship between the operating parameters and the product information is determined based on the product information prediction model;

[0051] The product information prediction model is trained based on actual sample operation parameters, actual product information corresponding to actual sample operation parameters, generated sample operation parameters, and predicted product information corresponding to generated sample operation parameters. The predicted product information is predicted based on the residue oil hydrogenation mechanism model.

[0052] Specifically, considering that lumped dynamics models cannot accurately predict product yields and require modification of the prediction model for different residue hydrotreating units, resulting in poor model scalability; while molecular-level mechanistic models have long computation times and cannot be applied to real-time optimization, this invention addresses this issue by employing a neural network model from artificial intelligence algorithms to construct a product information prediction model. This model is then used to predict product information for the input operating parameters, obtaining the product information corresponding to the operating parameters output by the product information prediction model. This allows the determination of the constraint relationship between the operating parameters and the product information, thereby improving the computational efficiency of the residue hydrotreating unit prediction model while ensuring the accuracy of product yield and property predictions.

[0053] Based on this, considering that training the product information prediction model using only actual operating parameters may lead to overfitting due to the lack of device mechanism information, while the residue hydrotreating mechanism model can follow the operating rules of the residue hydrotreating unit in principle, the predicted product yield may have slight deviations, but its trend will definitely conform to reality, this embodiment of the invention combines actual operating parameters and the residue hydrotreating mechanism model to train the product information prediction model, thereby making the prediction results of the product information prediction model more in line with reality and further improving the accuracy of the model's product yield and physical property prediction.

[0054] The specific training process can be as follows: First, collect actual operating parameters as actual operating parameters, and the actual product information corresponding to the actual operating parameters; according to the applicable scope of the residue hydrotreating mechanism model, simulate and generate production operating parameters under different feed and operating conditions, and then input the generated operating parameters into the residue hydrotreating mechanism model to obtain the predicted product information corresponding to the generated operating parameters; based on this, apply the actual operating parameters and the actual product information corresponding to the actual operating parameters to generate operating parameters and the predicted product information corresponding to the generated operating parameters, and train the initial model to obtain the product information prediction model. Here, the initial model can be a single neural network model or a combination of multiple neural network models, and this embodiment of the invention does not specifically limit it.

[0055] The method provided in this invention trains a product information prediction model by combining actual operating parameters and a residue hydrotreating mechanism model, enabling real-time prediction of the yield of each component under current operating conditions. This improves the accuracy and efficiency of product information prediction for the residue hydrotreating unit. The product information prediction model is then used to determine optimized operating parameters, and these parameters are further optimized to provide a rapid optimized operating plan for the residue hydrotreating unit. This improves optimization efficiency and effectiveness, enabling online optimization of the residue hydrotreating unit.

[0056] Based on any of the above embodiments, step 110, determining the range of operating parameters based on similar operating condition data, includes:

[0057] Based on the product information corresponding to the current working condition data, candidate working condition data that are superior to the product information are determined from similar working condition data;

[0058] Based on candidate operating condition data, determine the range of operating parameters.

[0059] Specifically, after determining similar operating condition data, based on the product information corresponding to the current operating condition data and the product information corresponding to each similar operating condition data, operating condition data with better product information than the current operating condition data can be further selected from the similar operating condition data as candidate operating condition data. Then, the range of operating parameters is determined based on the candidate operating condition data for subsequent optimization of operating parameters.

[0060] It should be noted that first identifying similar operating condition data that are similar to the current operating condition data, and then filtering candidate operating condition data with product information that is better than the current operating condition data, can further narrow down the range of operating parameters, thereby further improving the search efficiency of the global optimization solution and enhancing the optimization efficiency of the residue hydrotreating prediction model.

[0061] Based on any of the above embodiments, in step 110, based on actual operating condition data and generated operating condition data, cluster analysis is performed on the current operating condition data of the residue hydrotreating unit to obtain similar operating condition data that are similar to the current operating condition data, including:

[0062] Based on actual operating data and generated operating data, including feed flow rate, feed composition, feed properties, and operating parameters, a hierarchical clustering tree is constructed.

[0063] Based on hierarchical clustering trees, cluster analysis is performed on the current operating condition data to obtain similar operating condition data that are similar to the current operating condition data.

[0064] Feed properties can include, for example, density, distillation temperature, viscosity, sulfur content, metal content, residual carbon content, and the compositional characteristics of the four components. Operating parameters can include, for example, feed temperature, hydrogen injection temperature, bed temperature of each reactor, bed pressure, liquid hourly space velocity, and hydrogen-to-oil ratio.

[0065] Specifically, Figure 2 This is a schematic diagram of the hierarchical clustering tree provided by the present invention, as shown below. Figure 2 As shown, firstly, it is possible to generate operating condition data based on actual operating condition data and... Figure 2 Based on the characteristics of the feed flow rate in the residual oil hydrotreating operating data, different operating condition data are clustered into N categories. Within these N categories, based on the feed flow rate, they are further clustered into M categories according to the characteristics of the feed composition. Then, based on the N×M categories formed in the above steps, they are further clustered into P categories according to the feed properties. Finally, based on the N×M×P categories formed in the above steps, and according to key operating parameters (i.e.... Figure 2 The operating conditions (as described in the text) are further clustered into K categories, thereby establishing a hierarchical clustering tree with a nested structure of operating conditions. Optionally, a hierarchical clustering algorithm is used for clustering.

[0066] Based on this, the feed information and operating parameter information in the current operating condition data are input into the clustering tree at this level, which can complete the clustering analysis of the current operating condition data, obtain operating condition data similar to the current operating condition data as similar operating condition data, and select candidate operating condition data with a better product yield than the current operating condition data.

[0067] Based on any of the above embodiments, the operating parameters are obtained by performing correlation analysis between each parameter in the actual working condition data and the generated working condition data and the product information, and the parameters are highly correlated with the product information.

[0068] Specifically, a correlation analysis is performed on the parameters and product information in the actual operating condition data stored in the real-time updated operating database and the production operating condition data stored in the residue hydrotreating mechanism model database. Parameters with high correlation to product information are obtained, which are key variables that play an important role in predicting product yield and physical properties. These key variables are determined as input variables of the product information prediction model, and the output variables are the yield and physical properties of each component of the product from the residue hydrotreating unit.

[0069] Data from the real-time updated operational database and the residue hydrotreating mechanism model database are used as training datasets and input into the initial model for training, resulting in a trained product information prediction model. During global optimization, this product information prediction model is applied to predict product information for the input operational parameters, obtaining the product information corresponding to the operational parameters output by the product information prediction model. This allows the determination of the constraint relationship between the operational parameters and the product information.

[0070] Here, correlation analysis can be performed using methods such as principal component analysis and factor analysis, and the embodiments of the present invention do not specifically limit this.

[0071] Based on any of the above embodiments, the product information prediction model is a long short-term memory neural network.

[0072] Specifically, considering that the operation of a residue hydrotreating unit is a dynamic process with varying states at different times, some historical information can influence the current operation of the unit and therefore requires long-term memory; while other information can be ignored. For example, the residual carbon properties of a certain feedstock can cause changes in the catalyst properties and affect the current operating state of the unit; information on the residual carbon properties of such feedstocks needs to be stored long-term. Long Short-Term Memory (LSTM) neural networks can effectively transmit and express information over long time sequences without ignoring relevant information from a long period. Therefore, this embodiment of the invention uses LSTM to construct a product information prediction model, thereby further improving the accuracy of product information prediction.

[0073] It should be noted that long short-term memory neural networks retain important information for a long time, and the stored information is dynamically adjusted according to the input. By applying long short-term memory neural networks with the function of remembering historical data, a product information prediction model can be established to predict the yield of each component after hydrogenation of residue oil under the current working conditions in real time, which can further improve the accuracy and efficiency of product yield prediction.

[0074] Furthermore, the product information prediction model can control whether the current state needs to be derived from the historical state, i.e., the unit state c of the previous time step, by updating the parameters of the forget gate and the memory gate. t-1How much information is forgotten, and from the current candidate state? The amount of new information received is used to control the current cell state c by updating the parameters of the output gate. t How much information needs to be output to the current external state H? t ;

[0075] Figure 3 This is a schematic diagram of the product information prediction model provided by the present invention, as shown below. Figure 3 As shown, with neuron 2 (i.e. Figure 3 Taking Neuron2 as an example, this neuron can adjust its behavior based on the external state H from the previous time step. t-1 and the current input X t The external state H at the current moment is calculated. t Here, the function tanh is the learnable sigmoid function in the neural network model, which can map variables to the interval [0, 1]. The calculation formula is:

[0076] The function σ is also the learnable sigmoid function in the neural network model, and its calculation formula is:

[0077] Figure 4 This is a schematic diagram of the forget gate structure in the product information prediction model provided by this invention, as shown below. Figure 4 As shown, the formula for calculating the forget gate parameters is:

[0078] f t =σ(W f ·[H t-1 ,X t ]+b f )

[0079] Among them, f t W is the output vector of the forget gate. f Let b be the weight matrix. f For the bias term, [H t-1 ,X t The symbol ] represents concatenating two vectors into a longer vector.

[0080] The memory gate is a control unit used to determine whether the input at time t is retained in the cell state. The tanh function is used to extract the effective information from the vector, while the σ function controls how much of the effective information at the current time is retained in the cell state. Figure 5 This is a schematic diagram of the memory gate structure in the product information prediction model provided by the present invention, as shown below. Figure 5 As shown, the formula for calculating the memory gate parameters is:

[0081] it =σ(W i ·[H t-1 ,X t ]+b i )

[0082]

[0083] Among them, i t W is the output vector of the σ layer. i W c Let b be the weight matrix. i b c For bias terms, This represents the candidate state at the current moment. Therefore, different σ-function outputs will lead to different information being remembered and forgotten. In this way, the product information prediction model retains important information for a long time, and the remembered information is dynamically adjusted with the input.

[0084] The formula for calculating the current cell state (i.e., internal state) is:

[0085]

[0086] Among them, c t This represents the current state of the cell.

[0087] The formula for calculating the output gate parameters is:

[0088] o t =σ(W o ·[H t-1 ,X t ]+b o )

[0089] Among them, o t W is the output vector of the output gate. o Let b be the weight matrix of the output gate. o This is the bias term for the output gate.

[0090] The formula for calculating the external state at the current moment is:

[0091]

[0092] Among them, H t This represents the external state at the current moment. Based on this, according to H... t The output of the product information prediction model is calculated, which is the product information.

[0093] Based on any of the above embodiments, the present invention provides an optimization method for a residue hydrotreating unit based on a combination of reaction mechanism model and artificial intelligence algorithm. Taking the simulation and optimization of a residue hydrotreating unit in a refining and chemical enterprise as an example, the method includes a training phase of product information prediction model and an optimization phase of residue hydrotreating unit.

[0094] Figure 6 This is a schematic diagram of the training process of the product information prediction model provided by the present invention, as shown below. Figure 6 As shown, the method includes:

[0095] S1. First, construct a mechanism model for residue hydrotreating and use this model for calculation. By simulating different residue hydrotreating feed compositions and feed flow rates, as well as the yields and properties of each component of the reaction products under different operating conditions, establish a database of residue hydrotreating mechanism models.

[0096] (1) Based on the applicable scope of the residual oil hydrotreating mechanism model and historical operating condition data, the reasonable upper and lower limits of each operating parameter shall be specified. The applicable scope here may be determined from papers, books, research and other documents.

[0097] (2) Set a reasonable value step size according to the upper and lower limits of each operating parameter in the residual oil hydrogenation mechanism model, take a certain number of values ​​for each operating parameter within its upper and lower limits, and arrange and combine the values ​​taken for each operating parameter with different feed composition, flow rate and other feed information to obtain different combinations of model input parameters, that is, obtain the generated operating condition data.

[0098] (3) Input the generation condition data obtained in step (2) into the residue hydrotreating mechanism model to obtain the predicted product yield and predicted product properties corresponding to the generation condition data output by the residue hydrotreating mechanism model.

[0099] (4) Combine the generated operating conditions data obtained in steps (2) and (3) with the corresponding predicted product yield and predicted product properties into a set of data and save it in the residue hydrotreating mechanism model database.

[0100] Optionally, a residue hydrotreating mechanism model can be constructed based on a traditional lumped kinetic model. The residue hydrotreating mechanism model includes a residue hydrotreating reactor model, a fractionation model, and a kinetic parameter correction model.

[0101] S2. Collect and gather data from the DCS (Distributed Control System), MES (Manufacturing Execution System), and LIMS (Laboratory Information Management System) of the residue hydrotreating unit, and perform data preprocessing such as outlier handling, smoothing and noise reduction, steady-state analysis, and normalization to form a real-time updated operating database.

[0102] The operational database for residue hydrotreating units may include:

[0103] Residue hydrotreating feedstock: composition, density, viscosity, sulfur content, metal content, carbon residue, distillation temperature, and properties of the four components;

[0104] Catalysts: Types of catalysts, carbon content, properties, etc.;

[0105] Operating parameters of the residue hydrotreating unit: temperature of each bed in the reactor, reactor pressure and pressure drop, hydrogen-to-oil ratio, liquid hourly space velocity, etc.

[0106] S3. First, preprocess the data in the real-time updated operating database and the residue hydrotreating mechanism model database to form a training dataset. Use the training dataset to train the product information prediction model for residue hydrotreating. Figure 7 This is the second schematic diagram of the optimized method for the residue hydrotreating device provided by the present invention.

[0107] S4. Use hierarchical clustering algorithm to perform cluster analysis on the current working condition data, and combine the predicted product yield obtained in step S4 to screen out working conditions that are better than the corresponding product information of the current working condition, and obtain the range of operating parameters.

[0108] (1) Based on the characteristics of the feed flow rate in the residual oil hydrotreating mechanism model database and the operation database, hierarchical clustering algorithm is used to cluster them into N categories;

[0109] (2) Under the N categories clustered by feed flow rate, according to the characteristics of feed composition, hierarchical clustering algorithm is used to cluster into M categories;

[0110] (3) Under the N×M categories formed by steps (1) and (2), according to the feeding properties, hierarchical clustering algorithm is used to cluster them into P categories;

[0111] (4) Under the N×M×P categories formed by the above steps, based on key operating parameters, a hierarchical clustering algorithm is used to cluster them into K categories, thereby establishing a hierarchical clustering tree based on the nested structure of operating conditions using DCS data, MES data, and LIMS analysis data; for example Figure 2 As shown, optionally, the ranges of N, M, P and K are all ∈{(0,15]|Z};

[0112] (5) Input the feed information and operation parameter information in the current working condition data into the hierarchical clustering algorithm to obtain similar working condition data that are similar to the current working condition data, and filter out candidate working condition data that are better than the product information corresponding to the current working condition data. Extract the upper and lower limits of the operation parameters of these candidate working condition data to obtain the range of their operation parameters.

[0113] The loss function of the hierarchical clustering algorithm is defined as follows:

[0114]

[0115] That is, the sum of squared errors of each sample from its cluster center. Where x i Let c represent the i-th sample. i It is x i The cluster to which it belongs J represents the centroid of the cluster, and J is the total number of samples. Hierarchical clustering is an unsupervised learning algorithm that iteratively finds a way to partition k clusters to minimize the loss function of the clustering algorithm.

[0116] S5. Apply the trained product information prediction model to predict the product information of the input operation parameters, and obtain the product information corresponding to the operation parameters output by the product information prediction model. Thus, the constraint relationship between the operation parameters and the product information can be determined.

[0117] Based on the set objective function, the range obtained in step S4 and the constraint relationship between the operating parameters and product information are globally optimized to obtain optimized operating parameters. For example, if the optimization objective is to produce more diesel components, then the optimized operating parameters are the operating parameters corresponding to the highest diesel component yield within the range. Based on these optimized operating parameters, the operating parameters of the residue hydrotreating unit are then optimized to obtain the optimized results, namely, the optimized product yield and optimized product properties. Optionally, the global optimization algorithm can employ a pattern search method for optimization calculation.

[0118] This invention not only provides a rapid optimized operating scheme for a residue hydrotreating unit but also yields the theoretically optimal optimization results, which can be used for online optimization. This invention applies a neural network algorithm with historical data memory capabilities, namely a Long Short-Term Memory (LSTM) neural network, to predict the yields of each component after residue hydrotreating under current operating conditions in real time. Furthermore, this invention sets up several operating condition data points by taking values ​​within the upper and lower limits of various operating parameters and using a residue hydrotreating mechanism model to predict the yield results. This establishes a complete description of the operating status and yields of each component of the residue hydrotreating unit under different feed changes and operating conditions, forming a residue hydrotreating mechanism model database. A hierarchical clustering algorithm is used to hierarchically decompose the operating condition information in the residue hydrotreating mechanism model database and the operating database, analyzing information such as the feed and key operating parameters under current operating conditions. Combined with the real-time predicted yields of each component, the range of operating parameters is selected for global optimization. By obtaining the upper and lower limits of the operating parameters within a narrowed range and searching for the globally optimal solution within that range, computational resources are greatly saved, computational efficiency is improved, and near-theoretical optimal values ​​are provided.

[0119] The optimized apparatus of the residue hydrotreating device provided by the present invention is described below. The optimized apparatus of the residue hydrotreating device described below and the optimized method of the residue hydrotreating device described above can be referred to in correspondence with each other.

[0120] Based on any of the above embodiments, the present invention provides an optimized device for a residue hydrotreating unit based on a combination of reaction mechanism model and artificial intelligence algorithm. Figure 8 This is a schematic diagram of the optimized device of the residue hydrotreating unit provided by the present invention, as shown below. Figure 8 As shown, the device includes:

[0121] Cluster analysis unit 810 is used to perform cluster analysis on the current operating data of the residue hydrotreating unit based on actual operating data and generated operating data, to obtain similar operating data that are similar to the current operating data, and to determine the range of operating parameters based on the similar operating data.

[0122] The global optimization unit 820 is used to perform global optimization within the range of the operation parameters based on the objective function and the constraint relationship between the operation parameters and the product information, so as to obtain the optimized operation parameters. The objective function is determined by the operation parameters and the corresponding product information.

[0123] The device optimization unit 830 is used to optimize the operating parameters of the residue hydrotreating unit based on optimized operating parameters.

[0124] The apparatus provided in this invention trains a product information prediction model by combining actual operating parameters and a residue hydrotreating mechanism model, enabling real-time prediction of the yield of each component under current operating conditions. This improves the accuracy and efficiency of product information prediction for the residue hydrotreating unit. The product information prediction model is then used to determine optimized operating parameters, and based on this, the operating parameters of the residue hydrotreating unit are optimized. This allows for the rapid provision of an optimized operating scheme for the residue hydrotreating unit, improving optimization efficiency and effectiveness, and enabling online optimization of the residue hydrotreating unit.

[0125] Based on any of the above embodiments, the constraint relationship between the operating parameters and the product information is determined based on the product information prediction model;

[0126] The product information prediction model is trained based on actual sample operation parameters, actual product information corresponding to actual sample operation parameters, generated sample operation parameters, and predicted product information corresponding to generated sample operation parameters. The predicted product information is predicted based on the residue oil hydrogenation mechanism model.

[0127] Based on any of the above embodiments, the product information prediction model is a long short-term memory neural network.

[0128] Based on any of the above embodiments, determining the range of operating parameters based on similar operating condition data includes:

[0129] Based on the product information corresponding to the current working condition data, candidate working condition data that are superior to the product information are determined from similar working condition data;

[0130] Based on candidate operating condition data, determine the range of operating parameters.

[0131] Based on any of the above embodiments, cluster analysis is performed on the current operating data of the residue hydrotreating unit based on actual operating data and generated operating data to obtain similar operating data that are similar to the current operating data, including:

[0132] Based on actual operating data and generated operating data, including feed flow rate, feed composition, feed properties, and operating parameters, a hierarchical clustering tree is constructed.

[0133] Based on hierarchical clustering trees, cluster analysis is performed on the current operating condition data to obtain similar operating condition data that are similar to the current operating condition data.

[0134] Based on any of the above embodiments, the operating parameters are obtained by performing correlation analysis between each parameter in the actual working condition data and the generated working condition data and the product information, and the parameters are highly correlated with the product information.

[0135] Figure 9 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 9As shown, the electronic device may include a processor 910, a communication interface 920, a memory 930, and a communication bus 940, wherein the processor 910, the communication interface 920, and the memory 930 communicate with each other via the communication bus 940. The processor 910 can call logical instructions in the memory 930 to execute an optimization method for the residue hydrotreating unit. This method includes: performing cluster analysis on the current operating condition data of the residue hydrotreating unit based on actual operating condition data and generated operating condition data to obtain similar operating condition data, and determining the range of operating parameters based on the similar operating condition data; performing global optimization within the range of operating parameters based on an objective function and the constraint relationship between the operating parameters and product information to obtain optimized operating parameters, wherein the objective function is determined based on the operating parameters and the corresponding product information; and optimizing the operating parameters of the residue hydrotreating unit based on the optimized operating parameters.

[0136] Furthermore, the logical instructions in the aforementioned memory 930 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a 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 includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0137] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the optimization method for the residue hydrotreating unit provided by the above methods. The method includes: performing cluster analysis on the current operating condition data of the residue hydrotreating unit based on actual operating condition data and generated operating condition data to obtain similar operating condition data that is similar to the current operating condition data, and determining the range of operating parameters based on the similar operating condition data; performing global optimization within the range of operating parameters based on an objective function and the constraint relationship between the operating parameters and product information to obtain optimized operating parameters, wherein the objective function is determined based on the operating parameters and the corresponding product information; and optimizing the operating parameters of the residue hydrotreating unit based on the optimized operating parameters.

[0138] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements an optimization method for the residue hydrotreating unit provided by the methods described above. This method includes: performing cluster analysis on the current operating condition data of the residue hydrotreating unit based on actual operating condition data and generated operating condition data to obtain similar operating condition data that is similar to the current operating condition data; determining a range of operating parameters based on the similar operating condition data; performing global optimization within the range of the operating parameters based on an objective function and the constraint relationship between the operating parameters and product information to obtain optimized operating parameters, wherein the objective function is determined based on the operating parameters and the corresponding product information; and optimizing the operating parameters of the residue hydrotreating unit based on the optimized operating parameters.

[0139] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0140] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An optimization method for a residue hydrotreating unit, characterized in that, include: Based on actual operating data and generated operating data, cluster analysis is performed on the current operating data of the residue hydrotreating unit to obtain similar operating data that are similar to the current operating data, and the range of operating parameters is determined based on the similar operating data. Based on the objective function and the constraint relationship between the operating parameters and the product information, global optimization is performed within the range of the operating parameters to obtain optimized operating parameters. The objective function is determined by the operating parameters and the corresponding product information. Based on the optimized operating parameters, the operating parameters of the residue hydrotreating unit are optimized; The constraint relationship between the operating parameters and product information is determined based on a product information prediction model. The product information prediction model is trained based on actual sample operating parameters, actual product information corresponding to the actual sample operating parameters, sample operating parameters, and predicted product information corresponding to the generated sample operating parameters. The predicted product information is predicted based on a residue oil hydrogenation mechanism model.

2. The optimization method for the residue hydrotreating unit according to claim 1, characterized in that, The product information prediction model is a long short-term memory neural network.

3. The optimization method for the residue hydrotreating unit according to claim 1, characterized in that, The determination of the range of operating parameters based on the similar operating condition data includes: Based on the product information corresponding to the current working condition data, candidate working condition data that is superior to the product information is determined from the similar working condition data; Based on the candidate operating condition data, the range of the operating parameters is determined.

4. The optimization method for the residue hydrotreating unit according to claim 1, characterized in that, The process involves clustering the current operating data of the residue hydrotreating unit based on actual operating data and generated operating data to obtain similar operating data that are similar to the current operating data, including: Based on actual operating data and generated operating data, including feed flow rate, feed composition, feed properties, and operating parameters, a hierarchical clustering tree is constructed. Based on the hierarchical clustering tree, cluster analysis is performed on the current operating condition data to obtain similar operating condition data that are similar to the current operating condition data.

5. The method for optimizing a residue hydrotreating unit according to any one of claims 1 to 4, characterized in that, The operating parameters are obtained by performing a correlation analysis between various parameters in the actual operating data and the generated operating data and the product information, and are highly correlated with the product information.

6. An optimized device for a residue hydrotreating unit, characterized in that, include: The clustering analysis unit is used to perform clustering analysis on the current operating data of the residue hydrotreating unit based on actual operating data and generated operating data, to obtain similar operating data that are similar to the current operating data, and to determine the range of operating parameters based on the similar operating data. A global optimization unit is used to perform global optimization within the range of the operation parameters based on the objective function and the constraint relationship between the operation parameters and the product information, to obtain optimized operation parameters. The objective function is determined by the operation parameters and the corresponding product information. The device optimization unit is used to optimize the operating parameters of the residue hydrotreating unit based on the optimized operating parameters. The constraint relationship between the operating parameters and product information is determined based on a product information prediction model. The product information prediction model is trained based on actual sample operating parameters, actual product information corresponding to the actual sample operating parameters, sample operating parameters, and predicted product information corresponding to the generated sample operating parameters. The predicted product information is predicted based on a residue oil hydrogenation mechanism model.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the optimization method for the residue hydrotreating apparatus as described in any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the optimization method for the residue hydrotreating apparatus as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the optimization method for the residue hydrotreating apparatus as described in any one of claims 1 to 5.

Images