Aromatic hydrocarbon combined device collaborative optimization control method

Abstract

The application discloses a kind of aromatic hydrocarbon combined device synergic optimization control method, belong to petroleum chemical industry process optimization technical field.The application is aimed at the problems of the existing aromatic hydrocarbon combined device single unit independent optimization, such as the working condition fragmentation, the deficiency of global optimality, the low efficiency of mechanism model optimization, the long-period optimization effect attenuation, by collecting hydrogen reforming-aromatic extraction combined device whole-process basic data, the calibration mechanism model of two units is respectively constructed and connected to form whole-process integrated model, and the direct coupling and energy cascade coupling relationship of material are quantitatively characterized;Based on integrated model, generate multi-condition simulation data, train big data agent model that can quickly map the correlation between working condition and index;With the maximum total aromatic yield and the minimum comprehensive net energy consumption of the whole device as the double synergistic target, a multi-objective optimization algorithm embedded with cross-unit parameter linkage rules is used to carry out global collaborative optimization, and a pareto optimal operating parameter solution set is obtained.

Description

Technical Field

[0001] This invention belongs to the field of petrochemical process optimization technology, specifically a collaborative optimization control method for an aromatic hydrocarbon complex. Background Technology

[0002] Aromatics are essential basic organic feedstocks in the petrochemical industry. Light aromatics such as benzene, toluene, and xylene are widely used in synthetic materials, fine chemicals, and other industries. Their production capacity and operational level are important indicators of the development level of the petrochemical industry. Aromatics complexes typically consist of a hydroreforming unit and an aromatics extraction and separation unit connected in series. The hydroreforming unit uses naphtha as feedstock and, under the action of a catalyst, produces aromatic-rich reformate through dehydrogenation, cyclization, and isomerization reactions, while also producing high-purity hydrogen as a byproduct. The aromatics extraction unit uses solvents such as sulfolane and separates benzene, toluene, and xylene products that meet purity requirements through extraction distillation, solvent recovery, and aromatics rectification processes. These two units have a close material and energy coupling relationship, and their coordinated operation level directly affects the unit's product yield, quality stability, and overall energy consumption.

[0003] In current production practice, the optimization of aromatics complexes mostly adopts a single-unit independent optimization model: the hydroreforming unit focuses on adjusting parameters such as reaction temperature, pressure, space velocity, and hydrogen-to-oil ratio, with the core objective of improving aromatics yield and hydrogen production; the aromatics extraction unit focuses on optimizing parameters such as solvent ratio, reflux ratio, and operating temperature and pressure of each tower, with the core objective of ensuring product purity and improving aromatics recovery rate. This type of technical solution has certain effectiveness in optimizing the operating conditions of a single unit, but it still has the following limitations: First, the coupling and linkage effects between units were not fully considered. The component distribution and distillation range characteristics of the reformed oil directly determine the feeding difficulty of the extraction unit, and the separation requirements of the extraction unit also constrain the operating range of the reforming unit. The optimal operating conditions obtained by optimizing a single unit independently are often only local optima at the overall process level. This can easily lead to situations where the yield of reformed aromatics increases but the energy consumption of extraction and separation increases significantly, and the overall benefit does not reach the optimal level. It is impossible to achieve global synergistic optimization of the total aromatics yield and the energy consumption of the entire unit.

[0004] Secondly, rigorous mechanistic models have low computational efficiency and are difficult to adapt to dynamic operating conditions. Although mechanistic models based on reaction kinetics and phase equilibrium have high steady-state calculation accuracy, single-condition simulations are computationally intensive and time-consuming. When faced with dynamic changes such as fluctuations in feedstock properties and adjustments in production load, they cannot quickly output suitable operating parameters. On-site adjustments rely heavily on the experience of operators, resulting in problems such as adjustment lag and poor accuracy consistency.

[0005] Third, there is a lack of a continuous iterative full-cycle optimization mechanism. Most optimization schemes are offline calibration-based phased optimizations. After being put into use, no normalized model calibration and parameter update mechanism has been established. As slow time-varying factors such as catalyst activity decay, tray efficiency decline, and equipment scaling accumulate, the model prediction bias gradually increases, and the optimization effect gradually weakens with operating time. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a collaborative optimization control method for aromatic hydrocarbon complexes, which solves the problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides a collaborative optimization control method for an aromatic hydrocarbon complex, comprising the following steps: S1. Collect basic data of the entire process of the hydroreforming-aromatics extraction combined unit, including feedstock properties, operating parameters of each unit, structural parameters of key equipment, reaction kinetics, and equilibrium parameters of the separated phases. S2. To meet the feed requirements of the aromatics extraction unit, a calibration mechanism model for the hydroreforming reaction unit is constructed to predict the distribution of aromatic components, non-aromatic content and by-product hydrogen in the reformed oil under different reaction conditions, providing suitable feed prediction input for downstream applications. S3. Match the fluctuation range of the feed composition of the reforming unit, construct a calibration mechanism model for the aromatic extraction and separation unit, predict the purity, recovery rate and unit energy consumption of aromatic products under different feed composition, solvent and distillation conditions, and realize dynamic response to changes in upstream feed. S4. Connect the two unit mechanism models according to the actual process flow, construct an integrated model covering the entire process of hydroreforming, reforming oil fractionation, aromatic extraction, solvent recovery, and aromatic distillation, quantitatively characterize the direct coupling relationship between the two units' materials and the energy gradient coupling relationship, and eliminate the working condition fragmentation bias of independent modeling of single units. S5. Based on the integrated model, multi-condition simulation is carried out to generate a full-condition dataset. A big data proxy model is trained to quickly map the relationship between raw material composition, cross-unit operating parameters and the yield / purity / energy consumption of the whole process, supporting efficient collaborative optimization. S6. With the dual objectives of maximizing the total aromatics yield of the entire process and minimizing the overall net energy consumption of the entire unit as the two collaborative objectives, and with constraints such as product quality specifications, equipment safety boundaries, and cross-unit operating condition matching, a multi-objective optimization algorithm with embedded cross-unit parameter linkage rules is used to perform global collaborative optimization to obtain the Pareto optimal operating parameter solution set. S7. Select the operating parameters with the best overall benefits from the Pareto optimal solution set, complete the on-site debugging and verification using an upstream and downstream progressive strategy, evaluate the optimization effect based on actual production data and iteratively calibrate the integrated model, and form a full-cycle optimization closed loop with the matching cross-unit feedforward-feedback control logic.

[0008] Preferably, in step S2, the mechanism model of the hydroreforming reaction unit adopts a 13-lumped kinetic framework adapted to the requirements of the extracted feed fraction: the reactants are divided into 13 lumped groups according to carbon number ranges, namely, C5-C6, C7-C8, and C9+ n-alkanes, isoalkanes, cycloalkanes, benzene, toluene, xylene, and C9+ aromatics. The lumped group division granularity is perfectly matched with the feed component identification accuracy of the aromatic extraction unit. When calibrating the model, the historical data of stable plant operation are used as the basis, and the prediction error weights of benzene, toluene, xylene, and key non-aromatic components are weighted to ensure that the relative error of the reformate fraction feed composition prediction output by the model is ≤2%.

[0009] Preferably, in step S3, the mechanism model of the aromatic extraction unit calibrates the binary interaction phase equilibrium parameters for the sulfolane solvent system, and corrects the mass transfer efficiency of each tray and the solvent-aromatic liquid phase mass transfer coefficient by combining the actual calibration data of the device; the calibration process covers the fluctuation range of the feed composition output under all operating conditions of the reforming unit, and focuses on calibrating the corresponding mapping relationship between the solvent ratio, reflux ratio and aromatic recovery rate and product purity, so as to ensure that the relative error of the extraction unit in predicting the fluctuation of the feed composition is ≤3%, and can accurately respond to the changes in the output composition of the upstream reforming unit.

[0010] Preferably, in step S4, the integrated model quantitatively represents the coupling relationship through unified material and energy balance throughout the entire process: At the material coupling level, the C7-C8 aromatic enriched fraction produced from the side stream of the reforming oil distillation tower is directly used as the feed for the aromatic extraction unit, without simplifying and merging intermediate components, thus achieving unbiased transfer of all components from the reforming oil to the extraction feed. At the energy coupling level, a cross-unit waste heat cascade matching logic is established, using the high-level waste heat of the reforming reaction products as the preheating heat source for the reforming oil fractionation tower feed, and the mid-level waste heat at the top of the solvent recovery tower as the preheating heat source for the benzene and toluene tower feeds. The net energy consumption of the entire process is calculated according to the following formula: ; in The overall net energy consumption of the entire unit, This represents the total input energy consumption when each unit operates independently. For the i-th cross-unit waste heat load, To correspond to the actual utilization efficiency of waste heat exchange, the value range is 95%~97%; through unified calculation, omissions and duplications in single-unit energy consumption accounting are eliminated, and the energy coupling benefits of the whole process are truly reflected.

[0011] Preferably, in step S5, the multi-condition simulation uses the Latin hypercube sampling method to generate samples. The sampling variables cover nine key parameters, including feedstock composition, reforming reaction temperature / pressure / space velocity / hydrogen-to-oil ratio, extraction solvent ratio / reflux ratio, and reflux ratio / recovery rate of each distillation column. The sampling process incorporates cross-unit parameter linkage constraints, and the sampling range is 10% outside the upper and lower limits of the actual operation of the unit, generating a total of 2000-3000 sets of valid operating condition data. The big data proxy model is constructed using the gradient boosting tree algorithm. After training on the full operating condition dataset, it provides the prediction determination coefficients for total aromatics yield, product purity, and overall net energy consumption. The calculation time for a single working condition is reduced by more than 99% compared to the mechanism model, supporting large-scale global collaborative optimization.

[0012] Preferably, in step S6, the quantification expression of the dual collaborative optimization objective is: The total aromatics yield of the entire process, i.e., the proportion of the total mass of qualified benzene, toluene, and xylene products to the mass of the reformate feedstock: ; in These represent the output flow rates of benzene, toluene, and xylene products, respectively. To correspond to the quality and purity of the product, The total feed flow rate for reforming feedstock; The overall net energy consumption per unit of raw material for the entire unit, i.e. the converted value of the net energy consumption of the entire process under the unit raw material processing volume; The constraints include: hard constraints on product quality (benzene purity ≥ 99.9%, toluene purity ≥ 99.5%, xylene purity ≥ 99.2%), hard constraints on equipment safety (temperature, pressure, and operating parameter ranges for each piece of equipment), cross-unit operating condition matching constraints (deviation between reforming feed distillation range and extraction feed requirements ≤ 3℃), and soft constraints on operational stability (optimized parameters deviating from current operating values ​​by no more than 15%).

[0013] Preferably, the multi-objective optimization algorithm in step S6 is an improved NSGA-II algorithm, with the core improvement being the introduction of a cross-unit parameter cooperative mutation rule based on overall process sensitivity: when the mutation operation adjusts the reforming reaction temperature... Simultaneously adjust the solvent ratio of the extraction distillation column. The linkage relationship is calculated using the following formula: ; in This is the adjustment amount for the variation in reforming reaction temperature. To assess the sensitivity of the overall aromatic yield obtained from the surrogate model to the current operating conditions of the reforming reaction temperature, The adjustment coefficient is set to 0.2-0.5. This rule ensures real-time matching of upstream and downstream operating conditions during the optimization process, avoids infeasible solutions caused by single-unit parameter adjustments, and improves the efficiency of global optimization and the engineering feasibility of the solution. After non-dominated sorting and crowding screening, the algorithm outputs a Pareto optimal solution set.

[0014] Preferably, in step S7, the optimal target parameters are selected from the Pareto solution set using the hourly comprehensive economic benefit increment function, and the calculation formula is as follows: ; in To optimize the incremental hourly economic benefits of the proposed solution relative to the current baseline operating conditions, This represents the average selling price of mixed aromatic hydrocarbon products. To optimize the overall aromatics yield improvement of the entire process, The selling price of by-product hydrogen, To optimize the change in the output flow rate of by-product hydrogen, Standard energy unit price, To optimize the change in net energy consumption per unit of raw material; select The solution with the largest value is used as the final target operation parameter, taking into account the global optimum of product revenue, by-product revenue and energy consumption cost.

[0015] Preferably, the on-site commissioning in step S7 adopts a coordinated and progressive strategy between upstream and downstream: the operating parameters of the hydroreforming unit are adjusted first, with each adjustment not exceeding 20% ​​of the total deviation between the target value and the current value. After the reforming unit is running stably, the deviation between the reformed oil components and the model prediction value is ≤±2%, and the extraction feed requirements are met, the operating parameters of the reformed oil fractionation tower, the aromatics extraction unit, and the aromatics distillation tower are adjusted sequentially. After each adjustment, the unit is run stably for 4-8 hours before proceeding to the next step to avoid significant fluctuations and mismatches in the upstream and downstream operating conditions. After commissioning, the unit is run stably for 72 hours to conduct an effect evaluation. If any of the key indicators, such as aromatics yield, product purity, or comprehensive net energy consumption, has an error of more than 5% relative to the model prediction value, the actual operating data is added to the full-condition dataset, and the integrated mechanism model and the big data proxy model are calibrated simultaneously to complete the coordinated iteration of the entire process model.

[0016] Preferably, after the optimized solution is implemented, it is equipped with cross-unit feedforward-feedback collaborative control logic to form a full-cycle optimization closed loop: Feedforward control: Real-time monitoring of feedstock density, boiling point, and group composition using an online near-infrared analyzer. When the feedstock density deviation is ≥0.02 g / cm³... 3 If the final boiling point deviation is ≥5℃, it is determined to be a significant fluctuation in the raw material. Based on the big data proxy model, the feedforward linkage adjustment of the operating parameters of the reforming unit and the extraction unit is calculated synchronously to adapt to changes in the raw material in advance and ensure that the upstream and downstream operating conditions are always matched. Feedback adjustment: The actual operating values ​​of the purity of aromatic products and the comprehensive net energy consumption of the whole process are collected every 2 hours. After comparing with the target values, the incremental PID algorithm is used to fine-tune the operating parameters. The single fine-tuning amplitude does not exceed 10% of the feedforward adjustment amount to eliminate the influence of model bias and unknown disturbances. Regular iteration: Every 30 consecutive days of operation, the entire process model is recalibrated and global collaborative optimization is carried out based on the accumulated operating data, the optimal operating parameter library is updated, and the slow time-varying characteristics related to catalyst activity decay and equipment scaling are adapted.

[0017] This invention provides a synergistic optimization control method for an aromatic hydrocarbon complex, which has the following beneficial effects: 1. This invention constructs a full-process integrated model to quantitatively characterize the material and energy coupling relationship between the hydroreforming and aromatics extraction units, breaking the limitations of isolated operating conditions in single-unit independent modeling. It can carry out collaborative optimization across the entire process, achieving coordinated matching between aromatics yield and unit energy consumption. At the same time, based on the mechanism simulation dataset, a big data proxy model is constructed, which significantly improves the efficiency of operating condition calculation and optimization while ensuring the prediction accuracy of key indicators. It can quickly respond to dynamic scenarios such as raw material fluctuations and load adjustments.

[0018] 2. A multi-objective optimization algorithm with embedded cross-unit parameter linkage rules is adopted to simultaneously match upstream and downstream operation parameters during the optimization process, thereby improving the engineering feasibility of the optimal solution. The output Pareto optimal solution set can be flexibly adapted to product demand, energy prices and raw material conditions at different stages, and the incremental comprehensive economic benefits are used as the basis for scheme selection. Product revenue, by-product revenue and energy consumption costs are considered simultaneously, which is more in line with actual production and operation goals.

[0019] 3. With a supporting upstream and downstream progressive debugging strategy and cross-unit feedforward-feedback closed-loop control logic, the adjustment range of control parameters is controlled during the implementation of the optimized scheme, reducing the risk of large fluctuations in operating conditions and ensuring a smooth transition in production; by regularly carrying out model calibration and parameter updates, it can adapt to slow time-varying characteristics such as catalyst activity decay and equipment scaling, and maintain the long-term stability of the optimization effect. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the collaborative optimization control method for the aromatic hydrocarbon complex of the present invention; Figure 2 This is a block diagram illustrating the principle of the synergistic optimization system for the aromatic hydrocarbon complex of the present invention. Figure 3 This is a comparison chart of the core indicators under different optimization modes of the present invention; Figure 4 This is a comparison chart of the core operating indicators of different optimization modes of the present invention; Figure 5This is a diagram illustrating the frontier of the Pareto optimal solution set of this invention; Figure 6 This is a comparison chart of the changes in aromatic yield during long-term operation of this invention; Figure 7 This is a comparison chart of the operational stability of the present invention under fluctuating raw material conditions. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment uses a typical hydroreforming-aromatics extraction combined unit in the petrochemical field as an application carrier. The hydroreforming unit is a 1.2 million tons / year continuous reforming unit, employing a platinum-based reforming catalyst and using straight-run naphtha as the main feedstock, with the option to blend in some hydrocracked naphtha. The aromatics extraction unit uses a sulfolane extraction distillation process, with a designed processing capacity of 800,000 tons / year, producing benzene, toluene, and mixed xylenes. Those skilled in the art can make adaptive adjustments based on the technical solution of this invention according to different unit scales, feedstock types, and solvent systems; all such adjustments should be covered within the scope of protection of this invention.

[0022] like Figure 1 As shown, this method follows a hierarchical progression logic: data acquisition—unit mechanism modeling—full-process integration—surrogate model training—multi-objective optimization—on-site implementation and closed-loop control. It sequentially completes seven steps, S1 to S7, achieving full-process control from basic data input to optimized closed-loop output from top to bottom, clearly demonstrating the logical connections and data transmission relationships between each step. For example... Figure 2 As shown, the system corresponding to this method is divided into six layers from top to bottom: a data input layer, a hydroreforming unit model layer, an aromatics extraction unit model layer, a big data proxy model layer, a multi-objective collaborative optimization layer, and a closed-loop control layer. The two unit models are interconnected through material coupling and energy coupling. The proxy model receives the simulation output of the mechanism model and provides rapid computational support for upper-level optimization. Finally, the closed-loop control layer is applied to the device operation, fully presenting the hierarchical structure and coupling logic of the system. This method is adaptable to various industrial scenarios such as steady-state production, raw material fluctuations, and long-cycle operation. It can be implemented through the following methods: The synergistic optimization control method for aromatic hydrocarbon complexes described in this invention is adaptable to various industrial scenarios, including steady-state production, feedstock fluctuations, and long-cycle operation. It is implemented through the following methods: Example 1: Full-process collaborative optimization under steady-state production conditions: This embodiment targets the normal production conditions of stable raw materials and stable load, and conducts offline full-process collaborative optimization to replace the traditional single-unit independent optimization mode. The specific implementation process is as follows: Step 1: Complete basic data collection: The data acquisition device collected historical operational data and basic design data from 30 consecutive days of stable operation, covering four core data categories: ① Raw material properties: Straight-run naphtha density 0.721 g / cm³ 3 The distillation range is 78℃~179℃, and the group composition consists of 65.2% alkanes, 22.8% cycloalkanes, and 12.0% aromatics. ② Operating parameters: average reaction temperature of the reforming reactor 498℃, reaction pressure 1.5MPa, mass hourly space velocity 1.8h. -1 Hydrogen-to-oil ratio 5.2; solvent ratio 4.2 and reflux ratio 1.0 for the extraction distillation column; top / bottom temperature, reflux ratio, and recovery rate for the benzene, toluene, and xylene columns; ③ Equipment parameters: inner diameter of reforming reactor and catalyst loading, number of trays and feed location of each distillation column, heat exchange area of ​​main heat exchangers, etc. ④ Basic parameters: Kinetic parameters of the reforming reaction and binary phase equilibrium data of the sulfolane-aromatic system obtained through laboratory calibration.

[0023] ⑤ Step 2: Construction of the calibration mechanism model for the hydrogenation reforming unit: Based on chemical process simulation software, a reforming reaction model was constructed using a 13-lumped kinetic framework adapted to the requirements of the extracted feed fractions. The reactants were divided into 13 lumped groups according to carbon number ranges: C5-C6, C7-C8, and C9+ n-alkanes, isoalkanes, cycloalkanes, benzene, toluene, xylene, and C9+ aromatics. The lumped group division granularity was perfectly matched with the feed component detection accuracy of the aromatics extraction unit.

[0024] Based on the steady-state operating data of the unit, the kinetic parameters are calibrated using the weighted least squares method. During the calibration process, the prediction error weights of benzene, toluene, xylene and key non-aromatic components are increased. The final model output has a relative error of ≤2% in predicting the composition of reformate fractionation feed. It can accurately predict the distribution of aromatic components, the content of non-aromatic components and the production of by-product hydrogen under different reaction conditions.

[0025] Step 3: Construction of the calibration mechanism model for the aromatics extraction unit: A full-process separation model of extraction distillation column, solvent recovery column, benzene column, toluene column and xylene column was constructed in the same simulation platform. The phase equilibrium behavior of sulfolane solvent system was described by NRTL activity coefficient model. Binary interaction parameters were calibrated by combining the device calibration data, and the tray mass transfer efficiency and solvent-aromatic liquid phase mass transfer coefficient of each column were corrected.

[0026] The calibration process covers the fluctuation range of feed components output under all operating conditions of the reforming unit, focusing on fitting the correspondence between solvent ratio, reflux ratio and aromatic recovery rate and product purity. Finally, the relative error of the extraction unit in predicting the fluctuation of feed components is ≤3%, which can accurately respond to the changes in the output composition of the upstream reforming unit.

[0027] Step 4: Building a full-process integrated model: By connecting the reforming reaction model and the extraction separation model according to the actual process flow, a full-process integrated model covering "hydrogenation reforming—reform oil fractionation—aromatic extraction—solvent recovery—aromatic distillation" is constructed. The coupling relationship is quantitatively characterized by unified material balance and energy balance. Material coupling: The C7-C8 aromatic enriched fraction produced from the side stream of the reforming oil distillation tower is directly used as the feed for the aromatic extraction unit, without simplifying and merging intermediate components, so as to achieve unbiased transfer of all components from the reforming product to the extraction feed. Energy Coupling: A cross-unit waste heat cascade matching logic is established, using the high-grade waste heat of the reforming reaction products as the preheating heat source for the reforming oil fractionation tower feed, and using the mid-grade waste heat at the top of the solvent recovery tower as the preheating heat source for the benzene and toluene tower feeds. The net energy consumption of the entire process is calculated using the following formula: ; In the formula, The total net energy consumption of the entire unit is the total amount of energy that the unit actually needs to input from the outside after deducting the waste heat recovered across units. The total input energy consumption when each unit of hydroreforming and aromatics extraction operates independently and there is no waste heat sharing. The heat load for the i-th cross-unit waste heat recovery; To ensure the actual utilization efficiency of the waste heat exchange process, the heat transfer loss of the heat exchanger, heat dissipation of the pipeline, and bypass loss are taken into account. To account for the number of waste heat recovery items across units, this embodiment includes two main waste heat recovery processes. This accounting method differs from the traditional statistical model that independently adds up the energy consumption of a single unit. By unifying the calculation, it eliminates omissions and duplications in the energy consumption accounting of a single unit, and truly reflects the energy-saving benefits brought about by the energy coupling of the entire process.

[0028] The key parameters for cross-unit waste heat recovery in this embodiment are shown in the table below: Table 1 Key Parameters for Cross-Unit Waste Heat Recovery: ; Step 5: Training the Big Data Agent Model Multi-condition simulations were conducted based on an integrated model. The Latin hypercube sampling method was used to generate samples. The sampling variables covered nine key parameters, including feedstock composition, reforming reaction temperature / pressure / space velocity / hydrogen-to-oil ratio, extraction solvent ratio / reflux ratio, and reflux ratio / recovery rate of each distillation column. The sampling process incorporated cross-unit parameter linkage constraints. The sampling range was 10% outside the upper and lower limits of the actual operation of the unit, generating a total of 2,500 sets of valid operating condition data.

[0029] A gradient boosting tree algorithm was used to train a big data proxy model. After training, the model's prediction coefficients for total aromatics yield, product purity, and overall net energy consumption were calculated. The calculation time for a single working condition is reduced by more than 99% compared to the mechanism model, and it can support large-scale global collaborative optimization.

[0030] Step 6: Multi-objective collaborative optimization solution: The dual optimization objectives are to maximize the total aromatics yield and minimize the overall net energy consumption of the entire unit. The total aromatics yield is calculated using the following formula: ; This formula uses the quality of pure aromatic hydrocarbon products as the accounting benchmark, incorporating product purity into the total yield statistics. This avoids inflated quality figures caused by calculating solely based on product flow rate, and conforms to the common statistical standards for aromatic hydrocarbon product yields in the petrochemical industry. In the formula... These represent the output flow rates of benzene, toluene, and xylene, respectively. To correspond to the quality and purity of the product, To ensure the total feed flow rate of reforming feedstock, all flow parameters are calculated using mass flow rates, and the accounting basis remains consistent.

[0031] The constraints are set as follows: hard constraints on product quality (benzene purity ≥ 99.9%, toluene purity ≥ 99.5%, xylene purity ≥ 99.2%); hard constraints on equipment safety (temperature, pressure, and operating parameter ranges for each piece of equipment); cross-unit operating condition matching constraints (deviation between reforming feed range and extraction feed requirement ≤ 3℃); and soft constraints on operational stability (optimized parameters deviating from current operating values ​​by no more than 15%). The cross-unit operating condition matching constraints are used to ensure that the reforming feed range matches the extraction unit's feed design range, avoiding infeasible solutions where the reforming range exceeds the range, causing the extraction unit to malfunction.

[0032] An improved NSGA-II algorithm with embedded cross-unit parameter linkage rules is used for global optimization. The core improvement is the introduction of cross-unit parameter cooperative mutation rules based on full-process sensitivity: when the mutation operation adjusts the reforming reaction temperature... Simultaneously adjust the solvent ratio of the extraction distillation column. The linkage relationship is calculated using the following formula: ; In the formula, This is the adjustment amount for the variation in the reforming reaction temperature; The sensitivity of the total aromatic yield of the entire process to the current operating condition of the reforming reaction temperature, calculated based on the big data proxy model, reflects the magnitude of the change in the total aromatic yield of the entire process caused by a unit temperature change. It is calculated by the finite difference method at the current operating point. The coordination adjustment coefficient is used to control the linkage amplitude of the solvent ratio and avoid excessive adjustment that could lead to operating condition mismatch. The value is determined by on-site calibration based on the coupling strength between the upstream and downstream of the device. In this embodiment, the value is 0.35.

[0033] The physical logic behind this linkage is as follows: when the reforming reaction temperature increases, the aromatic content in the reformed oil increases while the proportion of non-aromatic hydrocarbons decreases. This reduces the separation difficulty in the extraction unit, allowing for a slight reduction in the solvent ratio to meet product purity requirements. Therefore, the solvent ratio adjustment is negatively correlated with the temperature adjustment. This rule ensures real-time matching of upstream and downstream operating conditions during the optimization process, avoiding infeasible solutions arising from single-unit parameter adjustments, and improving optimization efficiency and the engineering feasibility of the solution.

[0034] After non-dominated sorting and crowding filtering, the algorithm outputs a Pareto optimal solution set containing 30 solutions. The parameters and indices of typical solutions are shown in the table below: Table 2. Comparison of Typical Schemes for Pareto Optimal Solution Sets: ; like Figure 5 As shown, the above typical schemes are evenly distributed along the Pareto front, showing a trade-off relationship in which the total yield of aromatics gradually increases with the increase of comprehensive energy consumption; among them, scheme 2 is located in the range of optimal comprehensive benefits, taking into account both yield improvement and energy consumption control, and is the final preferred scheme, which intuitively reflects the solution set distribution characteristics and scheme selection logic of multi-objective optimization.

[0035] Step 7: Optimal Solution Selection and On-site Implementation The optimal objective parameters are selected from the Pareto solution set using the hourly comprehensive economic benefit increment function. The calculation formula is as follows: ; This formula uses the current steady-state operating conditions of the unit as a benchmark to calculate the hourly economic benefit increment brought about by the optimization scheme. It covers three parts: the increased revenue from aromatics production, the revenue from changes in by-product hydrogen, and the changes in energy consumption costs, comprehensively reflecting the overall economic benefits of the optimization scheme; where To optimize the incremental hourly economic benefits of the proposed solution relative to the current baseline operating conditions, This represents the average selling price of mixed aromatic hydrocarbon products. To optimize the overall aromatics yield improvement of the entire process, The selling price of by-product hydrogen, To optimize the change in the output flow rate of by-product hydrogen, Standard energy unit price, To optimize the change in net energy consumption per unit of raw materials, all price parameters use the factory's internally calculated, tax-exclusive benchmark price to ensure that the calculation results are consistent with actual business accounting standards.

[0036] Select The solution with the largest value is taken as the final implementation plan. The corresponding optimized reforming reaction temperature is adjusted to 503℃, hydrogen-to-oil ratio to 5.0, and extraction solvent ratio to 4.0, taking into account product benefits, by-product benefits, and energy consumption costs.

[0037] The on-site commissioning adopts a coordinated and progressive strategy between upstream and downstream: the operating parameters of the hydroreforming unit are adjusted first, and the adjustment range of each parameter does not exceed 20% of the total deviation between the target value and the current value. After the reforming unit is running stably, the deviation between the reformed oil components and the model prediction value is ≤±2%, and the extraction feed requirements are met, the operating parameters of the reformed oil fractionation tower, the aromatics extraction unit, and the aromatics distillation tower are adjusted in sequence. After each adjustment, the unit is run stably for 6 hours before proceeding to the next step.

[0038] After debugging, the system was continuously and stably run for 72 hours to evaluate its effectiveness. The core indicators of the system under normal operating conditions, single-unit independent optimization conditions, and the collaborative optimization conditions of this invention were compared. The results are shown in the table below: Table 3 Comparison of core operating indicators under different optimization modes: ; like Figure 3 , Figure 4 As shown, the comparison trend of total aromatic yield and net energy consumption per unit of raw material under the three operating conditions is presented intuitively: the total aromatic yield reaches the highest value under the collaborative optimization mode of this invention, while the net energy consumption per unit of raw material is reduced to the lowest value. Compared with conventional operation and single-unit independent optimization, there is a significant improvement, and the collaborative optimization effect of the two indicators is significant.

[0039] As shown in Table 3, compared with the single-unit independent optimization mode, the whole-process collaborative optimization of the present invention can further improve the total yield of aromatics by 0.5 percentage points, reduce the comprehensive net energy consumption per unit of raw material by 2.8%, and significantly improve the overall economic benefits. Moreover, the purity of each product meets the quality standards, which verifies the effectiveness of the optimization scheme.

[0040] Example 2, Feedforward-feedback coordinated control under fluctuating raw material properties: This embodiment addresses dynamic operating conditions such as adjustments to the device's feeding structure and fluctuations in raw material properties. It applies the feedforward-feedback collaborative control logic of this invention to ensure stable production and product quality. The specific scenario and implementation process are as follows: Application scenarios: The feedstock was changed from 100% straight-run naphtha to a blend of 20% hydrocracked naphtha. An online near-infrared analyzer detected a real-time increase in feedstock density of 0.026 g / cm³. 3 A 7°C increase in the final boiling point is considered a significant fluctuation in the feedstock. Traditional single-unit lag adjustment methods are prone to problems such as abrupt changes in the composition of the extraction feedstock, a decrease in aromatic hydrocarbon recovery, and excessive fluctuations in product purity.

[0041] Implementation process: Feedforward linkage adjustment: After raw material fluctuation detection triggers, the feedforward linkage adjustment amount of the operating parameters of the reforming unit and the extraction unit is synchronously calculated based on the big data proxy model: the reforming reaction temperature is increased by 1.8℃ and the hydrogen-to-oil ratio is increased by 0.3, and the solvent ratio of the extraction distillation column is simultaneously increased by 0.22 and the reflux ratio is increased by 0.08, adapting in advance to changes in raw material composition and ensuring that the upstream and downstream operating conditions are always matched. The parameters are synchronously sent to the DCS system to achieve synchronous adjustment of parameters throughout the entire process, avoiding mismatch between upstream and downstream operating conditions caused by adjustments of a single unit.

[0042] Feedback fine-tuning: The actual operating values ​​of the purity of aromatic products and the overall net energy consumption of the entire process are collected every 2 hours. After comparing with the target values, an incremental PID algorithm is used to fine-tune the operating parameters in a coordinated manner. The single fine-tuning range does not exceed 10% of the feedforward adjustment. In this operating condition, the solvent ratio is fine-tuned by 0.02 in the feedback loop to further correct model bias and eliminate the influence of unknown disturbances.

[0043] like Figure 7 As shown, under fluctuating raw material properties, the feedforward-feedback control mode of this invention significantly reduces the fluctuation range of aromatic yield and product purity compared to the traditional lag adjustment mode. During raw material switching, the purity of benzene, toluene, and xylene products consistently meets quality requirements, and the fluctuation range of the total aromatic yield is controlled within 0.15%. Compared to the traditional single-unit lag adjustment mode, the fluctuation range of product purity is reduced by 62%, and the yield loss is reduced by 0.21 percentage points, effectively improving the operational stability under fluctuating raw material conditions.

[0044] Implementation results: During the raw material switching process, the purity of benzene, toluene, and xylene products always met the quality requirements, and the fluctuation range of the total aromatics yield was controlled within 0.15%. Compared with the traditional single-unit lag adjustment mode, the fluctuation range of product purity was reduced by 62%, and the yield loss was reduced by 0.21 percentage points, effectively improving the operational stability under raw material fluctuation conditions.

[0045] Example 3: Model Iteration and Continuous Optimization under Long-Term Operation: This embodiment addresses the slow time-varying characteristics of catalyst activity decay and tray scaling during long-term operation of the equipment. It applies the periodic iteration mechanism of this invention to maintain the long-term stability of the optimization effect. The specific scenario and implementation process are as follows: Application scenarios: During continuous operation of the unit, the activity of the reforming catalyst slowly declines month by month, and the mass transfer efficiency of the extraction tower trays decreases due to slight scaling. If a one-time offline optimization scheme is adopted, the optimization effect will gradually weaken with operating time, and the aromatic yield may decline significantly after 3 months.

[0046] Implementation process: Regular data accumulation: During the operation of the device, the entire process operation data is continuously collected and compiled into a monthly steady-state operation dataset every 30 days.

[0047] Model Collaborative Calibration: The actual operating data of the current month is added to the full-condition dataset, and the integrated mechanism model and big data proxy model are calibrated simultaneously: the reforming model corrects the catalyst activity coefficient to adapt to the activity decay; the extraction model corrects the tray mass transfer efficiency to adapt to changes in equipment performance.

[0048] Global collaborative optimization update: Based on the calibrated model, a new full-process collaborative optimization is carried out to update the optimal operating parameter library. For example, in the third month of operation, the catalyst activity decreased by about 4% compared to the initial value. After optimization, the reforming reaction temperature was increased by 3°C compared to the initial optimal value, and the extraction solvent ratio was adjusted by 0.15 to maintain a stable aromatic yield within the equipment safety boundary.

[0049] Iterative cycle: Iterate continuously according to the above cycle to adapt to the slow time-varying characteristics of the device throughout its entire operating cycle.

[0050] like Figure 6 As shown, under the periodic iterative optimization mode, the decline in total aromatics yield over a 6-month operating cycle is significantly less than that of the traditional one-time optimization mode. In the traditional mode, the yield continuously declines with operating time, while this invention effectively maintains yield stability through monthly iterative calibration. After 6 months of continuous operation, the total aromatics yield only decreased from the initial optimized 38.7% to 38.5%, a decline of 0.2 percentage points. Compared to the traditional optimization mode without periodic iteration, the yield decline is reduced by 0.4 percentage points, and the overall energy consumption increase is reduced by 3.1%, effectively extending the effective period of the optimization scheme and ensuring long-term operational benefits.

[0051] Implementation results: After six months of continuous operation, the total yield of aromatics decreased from 38.7% to 38.5% after initial optimization, a decrease of only 0.2 percentage points. Compared with the traditional optimization mode without regular iteration, the yield decrease was reduced by 0.4 percentage points, and the increase in comprehensive energy consumption was reduced by 3.1%, effectively extending the effective period of the optimization scheme and ensuring long-term operational benefits.

[0052] In summary, the method described in this invention is adaptable to various production scenarios in aromatics complexes, including steady-state optimization, dynamic fluctuations, and long-cycle operation. Through a technical approach combining mechanistic modeling and data-driven methods, along with offline optimization and online control, it achieves deep synergistic optimization of the hydroreforming and aromatics extraction units. This improves operational efficiency and resilience to disturbances while ensuring product quality and production safety. The application scenarios of this invention are not limited to the above embodiments; it can be extended to reforming-extraction complexes of different scales and process routes, as well as other petrochemical complexes involving material and energy coupling.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for coordinated optimization control of an aromatic hydrocarbon complex, characterized in that, Includes the following steps: S1. Collect basic data for the entire process of the hydroreforming-aromatics extraction combined unit; S2. To meet the feed requirements of the aromatics extraction unit, a calibration mechanism model for the hydroreforming reaction unit is constructed to predict the distribution of aromatic components, non-aromatic content and by-product hydrogen in the reformed oil under different reaction conditions, providing suitable feed prediction input for downstream applications. S3. Match the fluctuation range of the feed composition of the reforming unit, construct a calibration mechanism model for the aromatic extraction and separation unit, predict the purity, recovery rate and unit energy consumption of aromatic products under different feed composition, solvent and distillation conditions, and realize dynamic response to changes in upstream feed. S4. Connect the two unit mechanism models according to the actual process flow, construct an integrated model covering the entire process of hydroreforming, reforming oil fractionation, aromatic extraction, solvent recovery, and aromatic distillation, quantitatively characterize the direct coupling relationship between the two units' materials and the energy gradient coupling relationship, and eliminate the working condition fragmentation bias of independent modeling of single units. S5. Based on the integrated model, multi-condition simulation is carried out to generate a full-condition dataset. A big data proxy model is trained to quickly map the relationship between raw material composition, cross-unit operating parameters and the yield / purity / energy consumption of the whole process, supporting efficient collaborative optimization. S6. With the dual objectives of maximizing the total aromatics yield of the entire process and minimizing the overall net energy consumption of the entire unit as the two collaborative objectives, and with constraints such as product quality specifications, equipment safety boundaries, and cross-unit operating condition matching, a multi-objective optimization algorithm with embedded cross-unit parameter linkage rules is used to perform global collaborative optimization to obtain the Pareto optimal operating parameter solution set. S7. Select the operating parameters with the best overall benefits from the Pareto optimal solution set, complete the on-site debugging and verification using an upstream and downstream progressive strategy, evaluate the optimization effect based on actual production data and iteratively calibrate the integrated model, and form a full-cycle optimization closed loop with the matching cross-unit feedforward-feedback control logic.

2. The synergistic optimization control method for an aromatic hydrocarbon complex according to claim 1, characterized in that, In step S2, the mechanism model of the hydroreforming reaction unit adopts a 13-lumped kinetic framework adapted to the requirements of the feed fraction extraction: the reactants are divided into 13 lumped groups according to carbon number ranges, namely, C5-C6, C7-C8, and C9+ n-alkanes, isoalkanes, cycloalkanes, benzene, toluene, xylene, and C9+ aromatics. The lumped group division granularity is perfectly matched with the feed component identification accuracy of the aromatic extraction unit. During model calibration, historical data of stable plant operation are used as the basis, and the prediction error weights of benzene, toluene, xylene, and key non-aromatic components are weighted to ensure that the relative error of the reformate fraction feed composition prediction output by the model is ≤2%.

3. The synergistic optimization control method for an aromatic hydrocarbon complex according to claim 2, characterized in that, In step S3, the mechanism model of the aromatic extraction unit is used to calibrate the binary interaction phase equilibrium parameters for the sulfolane solvent system, and the mass transfer efficiency of each tray and the solvent-aromatic liquid phase mass transfer coefficient are corrected by combining the actual calibration data of the device. The calibration process covers the fluctuation range of feed components output under all operating conditions of the reforming unit. It focuses on calibrating the corresponding mapping relationship between solvent ratio, reflux ratio and aromatic recovery rate and product purity, ensuring that the relative error of the extraction unit in predicting the fluctuation of feed components is ≤3%, and can accurately respond to changes in the output composition of the upstream reforming unit.

4. The synergistic optimization control method for an aromatic hydrocarbon complex according to claim 3, characterized in that, In step S4, the integrated model quantitatively represents the coupling relationship through unified material balance and energy balance throughout the entire process: At the material coupling level, the C7-C8 aromatic enriched fraction produced from the side stream of the reforming oil distillation tower is directly used as the feed for the aromatic extraction unit, without simplifying and merging intermediate components, thus achieving unbiased transfer of all components from the reforming oil to the extraction feed. At the energy coupling level, a cross-unit waste heat cascade matching logic is established, using the high-level waste heat of the reforming reaction products as the preheating heat source for the reforming oil fractionation tower feed, and the mid-level waste heat at the top of the solvent recovery tower as the preheating heat source for the benzene and toluene tower feeds. The net energy consumption of the entire process is calculated according to the following formula: ； in The overall net energy consumption of the entire unit, This represents the total input energy consumption when each unit operates independently. For the i-th cross-unit waste heat load, To correspond to the actual utilization efficiency of waste heat exchange, the value range is 95%~97%; through unified calculation, omissions and duplications in single-unit energy consumption accounting are eliminated, and the energy coupling benefits of the whole process are truly reflected.

5. The method for coordinated optimization control of an aromatic hydrocarbon complex according to claim 4, characterized in that, In step S5, the multi-condition simulation uses the Latin hypercube sampling method to generate samples. The sampling variables cover nine key parameters, including feedstock composition, reforming reaction temperature / pressure / space velocity / hydrogen-to-oil ratio, extraction solvent ratio / reflux ratio, and reflux ratio / recovery rate of each distillation column. The sampling process incorporates cross-unit parameter linkage constraints, and the sampling range is 10% outside the upper and lower limits of the actual operation of the unit, generating a total of 2000-3000 sets of valid operating condition data. The big data proxy model is constructed using the gradient boosting tree algorithm. After training on the full-condition dataset, the prediction determination coefficients for total aromatics yield, product purity, and overall net energy consumption are obtained. The calculation time for a single working condition is reduced by more than 99% compared to the mechanism model, supporting large-scale global collaborative optimization.

6. The method for coordinated optimization control of an aromatic hydrocarbon complex according to claim 5, characterized in that, In step S6, the quantification expression of the dual collaborative optimization objective is: The total aromatics yield of the entire process, i.e., the proportion of the total mass of qualified benzene, toluene, and xylene products to the mass of the reformate feedstock: ； in These represent the output flow rates of benzene, toluene, and xylene products, respectively. To correspond to the quality and purity of the product, The total feed flow rate for reforming feedstock; The overall net energy consumption per unit of raw material for the entire unit, i.e. the converted value of the net energy consumption of the entire process under the unit raw material processing volume; The constraints include: hard constraints on product quality (benzene purity ≥ 99.9%, toluene purity ≥ 99.5%, xylene purity ≥ 99.2%), hard constraints on equipment safety (temperature, pressure, and operating parameter ranges for each piece of equipment), cross-unit operating condition matching constraints (deviation between reforming feed distillation range and extraction feed requirements ≤ 3℃), and soft constraints on operational stability (optimized parameters deviating from current operating values ​​by no more than 15%).

7. The method for coordinated optimization control of an aromatics complex according to claim 6, characterized in that, The multi-objective optimization algorithm in step S6 is an improved NSGA-II algorithm. The core improvement is the introduction of a cross-unit parameter cooperative mutation rule based on the overall process sensitivity: when the mutation operation adjusts the reforming reaction temperature... Simultaneously adjust the solvent ratio of the extraction distillation column. The linkage relationship is calculated using the following formula: ； in This is the adjustment amount for the variation in reforming reaction temperature. To assess the sensitivity of the overall aromatic yield obtained from the surrogate model to the current operating conditions of the reforming reaction temperature, The adjustment coefficient is set to 0.2-0.

5. This rule ensures real-time matching of upstream and downstream operating conditions during the optimization process, avoids infeasible solutions caused by single-unit parameter adjustments, and improves the efficiency of global optimization and the engineering feasibility of the solution. After non-dominated sorting and crowding screening, the algorithm outputs a Pareto optimal solution set.

8. The method for coordinated optimization control of an aromatic hydrocarbon complex according to claim 7, characterized in that, In step S7, the optimal target parameters are selected from the Pareto solution set using the hourly comprehensive economic benefit increment function, and the calculation formula is as follows: ； in To optimize the incremental hourly economic benefits of the proposed solution relative to the current baseline operating conditions, This represents the average selling price of mixed aromatic hydrocarbon products. To optimize the overall aromatics yield improvement of the entire process, The selling price of by-product hydrogen, To optimize the change in the output flow rate of by-product hydrogen, Standard energy unit price, This represents the change in net energy consumption per unit of raw material after optimization. Select The solution with the largest value is used as the final target operation parameter, taking into account the global optimum of product revenue, by-product revenue and energy consumption cost.

9. The method for coordinated optimization control of an aromatic hydrocarbon complex according to claim 8, characterized in that, The on-site commissioning in step S7 adopts a progressive strategy of upstream and downstream collaboration: the operating parameters of the hydroreforming unit are adjusted first, with each adjustment not exceeding 20% ​​of the total deviation between the target value and the current value. After the reforming unit is running stably, the deviation between the reformed oil components and the model prediction value is ≤±2%, and the extraction feed requirements are met, the operating parameters of the reformed oil fractionation tower, aromatics extraction unit, and aromatics distillation tower are adjusted sequentially. After each adjustment, the unit is run stably for 4-8 hours before proceeding to the next step to avoid large fluctuations and mismatches in upstream and downstream operating conditions. After commissioning, the unit is run stably for 72 hours to conduct an effect evaluation. If any of the key indicators, such as aromatics yield, product purity, or comprehensive net energy consumption, has an error of more than 5% relative to the model prediction value, the actual operating data is added to the full-condition dataset. The integrated mechanism model and the big data proxy model are calibrated simultaneously to complete the collaborative iteration of the entire process model.

10. The method for coordinated optimization control of an aromatic hydrocarbon complex according to claim 9, characterized in that, After the optimization scheme is implemented, it is equipped with cross-unit feedforward-feedback collaborative control logic to form a full-cycle optimization closed loop: Feedforward control: Real-time monitoring of feedstock density, boiling point, and group composition using an online near-infrared analyzer. When the feedstock density deviation is ≥0.02 g / cm³... 3 If the final boiling point deviation is ≥5℃, it is determined to be a significant fluctuation in the raw material. Based on the big data proxy model, the feedforward linkage adjustment of the operating parameters of the reforming unit and the extraction unit is calculated synchronously to adapt to changes in the raw material in advance and ensure that the upstream and downstream operating conditions are always matched. Feedback adjustment: The actual operating values ​​of the purity of aromatic products and the comprehensive net energy consumption of the whole process are collected every 2 hours. After comparing with the target values, the incremental PID algorithm is used to fine-tune the operating parameters. The single fine-tuning amplitude does not exceed 10% of the feedforward adjustment amount to eliminate the influence of model bias and unknown disturbances. Regular iteration: Every 30 consecutive days of operation, the entire process model is recalibrated and global collaborative optimization is carried out based on the accumulated operating data, the optimal operating parameter library is updated, and the slow time-varying characteristics related to catalyst activity decay and equipment scaling are adapted.

Images