An optimized method and system for separating p-xylene

Abstract

This invention provides an optimized method and system for separating para-xylene, relating to the field of xylene separation technology. The method involves constructing a simulated moving bed mechanism model; building a simulated moving bed surrogate model based on the sampling interval of each decision variable and the simulated moving bed mechanism model; establishing optimization conditions; optimizing the simulated moving bed surrogate model using a particle swarm optimization algorithm based on the optimization conditions to determine the optimal set of decision variables; and controlling the simulated moving bed based on the optimal set of decision variables. This invention, by constructing a simulated moving bed surrogate model, can directly simulate the para-xylene yield, para-xylene purity, and cycle energy consumption under different sets of decision variables, thereby determining the optimal set of decision variables to control the simulated moving bed, which can improve the accuracy and efficiency of optimizing the para-xylene separation process.

Description

Technical Field

[0001] This invention relates to the field of xylene separation technology, and in particular to an optimized method and system for separating para-xylene. Background Technology

[0002] p-Xylene is an important organic chemical raw material used in the production of common materials such as polyesters, fibers, films, and plastics. Industrially, p-xylene is typically produced using a mixed xylene separation process. Mixed xylenes include multiple components such as ethylbenzene (EB), p-xylene (PX), m-xylene (MX), o-xylene (OX), and p-diethylbenzene (PDEB). Except for their melting points, these components have very similar physical properties, making conventional distillation methods difficult to apply to their separation. Currently, the main industrial methods for separating mixed xylenes include cryogenic crystallization, complexation extraction, and adsorption separation. Among these, adsorption separation technology based on a continuous countercurrent simulated moving bed is currently the mainstream process technology for producing p-xylene.

[0003] Simulated moving bed chromatography evolved from moving bed chromatography, retaining the high separation efficiency of fixed bed chromatography while incorporating the countercurrent liquid-solid flow characteristics of moving bed chromatography. It boasts advantages such as high interphase mass transfer driving force, strong separation capacity, high adsorbent utilization, and low solvent consumption, and is widely used in petrochemical, food, fine chemical, and biopharmaceutical industries. Simulated moving bed chromatography, for example... Figure 1 and Figure 2 As shown, the simulated moving bed has its beds connected end to end and connected to the material pipeline through a rotary valve. The feed and discharge positions are switched periodically to achieve continuous countercurrent contact between the liquid phase and the adsorption phase. Taking advantage of the different affinities of the adsorbent to the four xylene isomers, p-diethylbenzene is preferentially adsorbed. Then, the p-diethylbenzene on the adsorbent is desorbed by the desorbent, and then high-purity p-diethylbenzene is obtained by distillation.

[0004] Adsorption separation is the core unit in the production of p-diethylbenzene. Due to the high coupling degree, computational complexity, and slow solution of the mechanistic model of the simulated moving bed, and the fact that the traditional optimization process repeatedly calls the mechanistic model, the computational efficiency is low, which is not conducive to the research of optimization algorithms and the implementation of optimization schemes. In contrast, the surrogate model, with its flexibility, simplicity, and speed, is increasingly becoming an important means of process research. Summary of the Invention

[0005] The purpose of this invention is to provide an optimized method and system for separating para-xylene, which can improve the accuracy and efficiency of optimizing the para-xylene separation process.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] An optimized method for separating p-xylene includes:

[0008] Based on the structural parameters of the simulated moving bed and the operating parameters of the p-xylene adsorption separation process, a simulated moving bed mechanism model is constructed; the simulated moving bed mechanism model is used to describe the component concentration of each bed layer of the simulated moving bed;

[0009] Obtain the sampling intervals for multiple decision variables;

[0010] Based on the sampling interval of each decision variable and the simulated moving bed mechanism model, a simulated moving bed surrogate model is constructed; the simulated moving bed surrogate model is used to simulate the para-xylene yield, para-xylene purity and cycle energy consumption under different decision variables;

[0011] Optimization conditions are established; the optimization conditions include: maximizing the yield of p-xylene, minimizing the cycle energy consumption, and ensuring that the purity of p-xylene is greater than the purity threshold.

[0012] Based on the aforementioned optimization conditions, the simulated moving bed proxy model is optimized using the particle swarm optimization algorithm to determine the optimal set of decision variables.

[0013] The simulated moving bed is controlled based on the optimal set of decision variables.

[0014] Optionally, after constructing the simulated moving bed mechanism model based on the structural parameters of the simulated moving bed and the operating parameters of the p-xylene adsorption separation process, the method further includes:

[0015] The instantaneous concentration values ​​of each component in each bed layer of the simulated moving bed are obtained; the components include ethylbenzene, p-xylene, m-xylene, o-xylene, and p-diethylbenzene;

[0016] Standard curves were constructed to simulate the concentration variation of each component in a moving bed, with the bed number as the x-axis and the instantaneous concentration value as the y-axis.

[0017] The simulated concentration values ​​of each component in each bed layer of the simulated moving bed are obtained; the simulated concentration values ​​are obtained by the simulated moving bed mechanism model; the acquisition time corresponding to the simulated concentration values ​​is the same as the acquisition time corresponding to the instantaneous concentration values.

[0018] Using the bed number as the x-axis and the simulated concentration value as the y-axis, simulated curves of the concentration change of each component in the simulated moving bed as a function of the bed layer were constructed.

[0019] By adjusting the structural parameters of the simulated moving bed and the operating parameters of the xylene adsorption separation process, the error between the simulated curve of the concentration of the same component changing with the bed and the standard curve of the concentration changing with the bed is within the preset error range, thus obtaining the corrected simulated moving bed mechanism model.

[0020] Optionally, the decision variables include: desorption zone reflux ratio, purification zone reflux ratio, adsorption zone reflux ratio, and buffer zone reflux ratio.

[0021] Optionally, the sampling interval of the decision variable is determined by finding the device process of the simulated moving bed.

[0022] Optionally, the construction of the simulated moving bed surrogate model based on the sampling interval of each decision variable and the simulated moving bed mechanism model includes:

[0023] The decision variables are sampled using the Latin hypercube sampling method to obtain multiple decision variable groups; each decision variable group includes a desorption zone reflux ratio sample value, a purification zone reflux ratio sample value, an adsorption zone reflux ratio sample value, and a buffer zone reflux ratio sample value.

[0024] The multiple decision variable groups are respectively input into the simulated moving bed mechanism model to obtain the process simulation data corresponding to each decision variable group;

[0025] Based on the process simulation data, calculate the p-xylene yield, p-xylene purity, and cycle energy consumption for each decision variable group.

[0026] Using decision variables as independent variables and p-xylene yield, p-xylene purity, and cycle energy consumption as dependent variables, regression models for p-xylene yield, p-xylene purity, and cycle energy consumption were constructed.

[0027] The para-xylene yield regression model, para-xylene purity regression model, and cycle energy consumption regression model were determined to be simulated moving bed surrogate models.

[0028] Optionally, the yield of p-xylene is:

[0029] The purity of the p-xylene is:

[0030] The cycle energy consumption is:

[0031] Where REX represents the yield of p-xylene; This represents the mass fraction of p-xylene in the extract. Q represents the mass fraction of p-xylene in the feed. E Q is the flow rate of the extracted fluid; F PUX represents the feed flow rate; PUX represents the purity of p-xylene. This represents the mass fraction of o-xylene in the extract. This represents the mass fraction of intermediate xylene in the extract. ρ is the mass fraction of ethylbenzene in the extract; CEC is the cycle energy consumption; ρ is the feed density; g is the acceleration due to gravity. To simulate the average flow rate of the circulating pump in a moving bed within a single cycle; Q k To simulate the instantaneous flow rate of the kth bed layer of the moving bed; H is the head; η1 is the efficiency of the circulating pump; η2 is the efficiency of the motor.

[0032] Optionally, based on the optimization conditions, the simulated moving bed surrogate model is optimized using a particle swarm optimization algorithm to determine the optimal set of decision variables, including:

[0033] Let the number of iterations n equal 1;

[0034] An initial particle swarm is constructed as the particle swarm for the first iteration; the number of particles in the initial particle swarm is equal to the number of decision variable groups; the positions of the particles in the initial particle swarm are used to represent the reflux ratio of the desorption zone, the reflux ratio of the purification zone, the reflux ratio of the adsorption zone, and the reflux ratio of the buffer zone.

[0035] Input the position of the particle swarm at the nth iteration into the simulated moving bed surrogate model to obtain the paraxylene yield and cycle energy consumption of each particle, which are used as the fitness value of the particle.

[0036] Crowding is calculated and fast non-dominated sorting is performed on each particle based on its fitness value;

[0037] A predetermined number of particles are designated as the elite particle set.

[0038] Determine any particle in the particle swarm at the nth iteration as the current particle;

[0039] Get any two particles from the elite particle set as the current elite particles;

[0040] Calculate the angles between the current particle and the two current elite particles respectively;

[0041] The current elite particle corresponding to the smallest included angle is determined as the winning elite particle.

[0042] Adjust the speed and position of the current particle based on the winning elite particle;

[0043] By iterating through all particles in the particle swarm at the nth iteration, we determine whether the iteration number n has reached the iteration number threshold and obtain the result.

[0044] If the judgment result is negative, then the value of the iteration number n is increased by 1, and the process returns to the step "Input the position of the particle swarm at the nth iteration into the simulated moving bed surrogate model to obtain the p-xylene yield and cycle energy consumption of each particle".

[0045] If the judgment result is yes, then the decision variable group corresponding to the particles in the particle swarm that meet the optimization conditions in the nth iteration is determined as the optimal decision variable group.

[0046] An optimized system for separating p-xylene includes:

[0047] The simulated moving bed mechanism model construction module is used to construct a simulated moving bed mechanism model based on the structural parameters of the simulated moving bed and the operating parameters of the p-xylene adsorption separation process; the simulated moving bed mechanism model is used to describe the component concentration of each bed layer of the simulated moving bed;

[0048] The sampling interval acquisition module is used to acquire the sampling intervals of multiple decision variables;

[0049] The simulated moving bed surrogate model construction module is used to construct a simulated moving bed surrogate model based on the sampling interval of each decision variable and the simulated moving bed mechanism model; the simulated moving bed surrogate model is used to simulate the para-xylene yield, para-xylene purity and cycle energy consumption under different decision variables;

[0050] An optimization condition construction module is used to construct optimization conditions; the optimization conditions include: maximum p-xylene yield, minimum cycle energy consumption, and p-xylene purity greater than a purity threshold.

[0051] The optimal decision variable set determination module is used to optimize the simulated moving bed surrogate model based on the optimization conditions using the particle swarm optimization algorithm to determine the optimal decision variable set.

[0052] A simulated moving bed control module is used to control the simulated moving bed based on the optimal decision variable set.

[0053] Optionally, the system further includes:

[0054] The instantaneous concentration value acquisition module is used to acquire the instantaneous concentration value of each component in each bed layer of the simulated moving bed; the components include ethylbenzene, p-xylene, m-xylene, o-xylene, and p-diethylbenzene;

[0055] The standard curve construction module is used to construct standard curves for the concentration of each component in a moving bed as a function of the bed layer, with the bed layer number as the x-axis and the instantaneous concentration value as the y-axis.

[0056] The concentration simulation value acquisition module is used to acquire the concentration simulation value of each component in each bed layer of the simulated moving bed; the concentration simulation value is obtained by the simulated moving bed mechanism model; the acquisition time corresponding to the concentration simulation value is the same as the acquisition time corresponding to the instantaneous concentration value;

[0057] The simulation curve construction module is used to construct simulation curves of the concentration of each component in the simulated moving bed as a function of the bed layer, with the bed layer number as the horizontal axis and the simulated concentration value as the vertical axis.

[0058] The simulated moving bed mechanism model calibration module is used to adjust the structural parameters of the simulated moving bed and the operating parameters of the xylene adsorption separation process, so that the error between the simulated curve of the concentration of the same component changing with the bed and the standard curve of the concentration changing with the bed is within the preset error range, thus obtaining the calibrated simulated moving bed mechanism model.

[0059] Optionally, the decision variables include: desorption zone reflux ratio, purification zone reflux ratio, adsorption zone reflux ratio, and buffer zone reflux ratio.

[0060] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0061] This invention provides an optimized method and system for separating para-xylene. Based on the structural parameters of a simulated moving bed and the operating parameters of the para-xylene adsorption separation process, a simulated moving bed mechanism model is constructed. This model describes the component concentration in each bed layer of the simulated moving bed. Sampling intervals for multiple decision variables are obtained. Based on the sampling intervals for each decision variable and the simulated moving bed mechanism model, a simulated moving bed surrogate model is constructed. This surrogate model simulates the para-xylene yield, para-xylene purity, and cycle energy consumption under different decision variables. Optimization conditions are established, including maximizing the para-xylene yield, minimizing cycle energy consumption, and ensuring para-xylene purity exceeds a purity threshold. Based on these optimization conditions, a particle swarm optimization algorithm is used to optimize the simulated moving bed surrogate model, determining the optimal set of decision variables. The simulated moving bed is then controlled based on this optimal set of decision variables. By constructing the simulated moving bed surrogate model, the para-xylene yield, para-xylene purity, and cycle energy consumption under different sets of decision variables can be directly simulated, thereby determining the optimal set of decision variables to control the simulated moving bed. This improves the accuracy and efficiency of optimizing the para-xylene separation process. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0063] Figure 1 This is a schematic diagram of a simulated moving bed structure in the prior art;

[0064] Figure 2 This is a schematic diagram illustrating the working principle of a simulated moving bed in existing technology.

[0065] Figure 3 This is a flowchart of the optimized method for separating p-xylene in an embodiment of the present invention;

[0066] Figure 4 This is a standard curve of concentration variation with bed layer in an embodiment of the present invention;

[0067] Figure 5 This is a simulated curve of concentration variation with bed layer in an embodiment of the present invention;

[0068] Figure 6 This is a schematic diagram of the NSGA-II algorithm solving the test function ZDT2 in an embodiment of the present invention;

[0069] Figure 7 This is a schematic diagram of the CMOPSO algorithm solving the test function ZDT2 in an embodiment of the present invention;

[0070] Figure 8 This is a schematic diagram of the NSGA-II algorithm solving the test function ZDT6 in an embodiment of the present invention;

[0071] Figure 9 This is a schematic diagram of the CMOPSO algorithm solving the test function ZDT6 in an embodiment of the present invention;

[0072] Figure 10 This is a comparison chart showing the effects of the CMOPSO algorithm and the NSGA-II algorithm on optimizing the operation of a simulated moving bed in this embodiment of the invention. Detailed Implementation

[0073] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0074] The purpose of this invention is to provide an optimized method and system for separating para-xylene, which can improve the accuracy and efficiency of optimizing the para-xylene separation process.

[0075] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0076] Example

[0077] like Figure 3 As shown, the present invention provides an optimized method for separating p-xylene, comprising:

[0078] Step 101: Based on the structural parameters of the simulated moving bed and the operating parameters of the p-xylene adsorption separation process, construct a simulated moving bed mechanism model; the simulated moving bed mechanism model is used to describe the component concentration of each bed layer in the simulated moving bed.

[0079] Step 102: Obtain the sampling intervals for multiple decision variables; the decision variables include: desorption zone reflux ratio, purification zone reflux ratio, adsorption zone reflux ratio, and buffer zone reflux ratio. The sampling intervals for the decision variables are determined by examining the process of the simulated moving bed device.

[0080] Step 103: Based on the sampling interval of each decision variable and the simulated moving bed mechanism model, construct a simulated moving bed surrogate model. The simulated moving bed surrogate model is used to simulate the p-xylene yield, p-xylene purity, and cycle energy consumption under different decision variables. Combining the unit process manual and on-site commissioning experience, determine the decision variables and their sampling intervals: desorption zone reflux ratio: 0.218–0.284; purification zone reflux ratio: 0.120–0.170; adsorption zone reflux ratio: 0.388–0.418; buffer zone reflux ratio: -0.076–-0.010.

[0081] Step 103 includes:

[0082] Step 1031: Sample the decision variables using the Latin hypercube sampling method to obtain multiple decision variable groups; each decision variable group includes a desorption zone reflux ratio sample value, a purification zone reflux ratio sample value, an adsorption zone reflux ratio sample value, and a buffer zone reflux ratio sample value; perform sampling according to the Latin hypercube sampling method, divide the optimization interval of each decision variable into 50 equal segments, and randomly generate a point in each segment. Randomly select one point of each decision variable in each sampling, and the vector formed is the sampling point, generating a total of 50 sampling points.

[0083] Step 1032: Input multiple decision variable groups into the simulated moving bed mechanism model to obtain the process simulation data corresponding to each decision variable group;

[0084] Step 1033: Based on the process simulation data, calculate the p-xylene yield, p-xylene purity, and cycle energy consumption for each decision variable group.

[0085] The yield of p-xylene was:

[0086] The purity of p-xylene is:

[0087] The cycle energy consumption is:

[0088] Where REX represents the yield of p-xylene; This represents the mass fraction of p-xylene in the extract. Q represents the mass fraction of p-xylene in the feed. E Q is the flow rate of the extracted fluid; F PUX represents the feed flow rate; PUX represents the purity of p-xylene. This represents the mass fraction of o-xylene in the extract. This represents the mass fraction of intermediate xylene in the extract. ρ is the mass fraction of ethylbenzene in the extract; CEC is the cycle energy consumption; ρ is the feed density; g is the acceleration due to gravity. To simulate the average flow rate of the circulating pump in a moving bed within a single cycle; Q k To simulate the instantaneous flow rate of the kth bed layer of the moving bed; H is the head; η1 is the efficiency of the circulating pump; η2 is the efficiency of the motor.

[0089] Step 1034: Using the decision variable as the independent variable and the para-xylene yield, para-xylene purity, and cycle energy consumption as the dependent variables, construct regression models for para-xylene yield, para-xylene purity, and cycle energy consumption, respectively; based on the input-output relationship of the mechanism model, construct a polynomial regression surrogate model in Matlab, and use the polynomial to replace the complex adsorption mechanism to complete the calculation of the above performance indicators from the reflux ratio of each zone of the simulated moving bed.

[0090] Step 1035: Determine that the para-xylene yield regression model, para-xylene purity regression model, and cycle energy consumption regression model are simulated moving bed proxy models.

[0091] Step 104: Construct optimization conditions; optimization conditions include: maximum p-xylene yield, minimum cycle energy consumption, and p-xylene purity greater than the purity threshold.

[0092] Step 105: Based on the optimization conditions, the particle swarm optimization algorithm is used to optimize the simulated moving bed proxy model to determine the optimal set of decision variables. Determining the optimization strategy: The optimization of a simulated moving bed can generally be divided into design optimization and operational optimization. Design optimization studies how to design the device for a specific production process to reduce system complexity and construction costs. Operational optimization studies how to determine the optimal operating point for a pre-designed device to improve production efficiency and reduce production costs. Given the complexity of the simulated moving bed device process design, this invention only optimizes the operational conditions, assuming that the bed length, diameter, and number of the simulated moving bed are fixed. The reflux ratio of each zone is used as the decision variable, maximizing PX yield and minimizing cycle energy consumption are used as optimization objectives, and PX purity is used as a constraint.

[0093] Step 105 includes:

[0094] Step 1051: Set the iteration number n to equal 1.

[0095] Step 1052: Construct an initial particle swarm as the particle swarm for the first iteration; the number of particles in the initial particle swarm is equal to the number of decision variable groups; the positions of the particles in the initial particle swarm are used to represent the reflux ratio of the desorption zone, the reflux ratio of the purification zone, the reflux ratio of the adsorption zone, and the reflux ratio of the buffer zone.

[0096] Step 1053: Input the position of the particle swarm at the nth iteration into the simulated moving bed surrogate model to obtain the paraxylene yield and cycle energy consumption of each particle, which are used as the fitness value of the particle.

[0097] Step 1054: Calculate the crowding of particles and perform fast non-dominated sorting based on the fitness value of each particle.

[0098] Step 1055: Determine the preset number of particles as the elite particle set.

[0099] Step 1056: Determine any particle in the particle swarm at the nth iteration as the current particle.

[0100] Step 1057: Obtain any two particles from the elite particle set as the current elite particles.

[0101] Step 1058: Calculate the angles between the current particle and the two current elite particles respectively.

[0102] Step 1059: Determine the current elite particle corresponding to the smallest included angle as the winning elite particle.

[0103] Step 10510: Adjust the speed and position of the current particle based on the winning elite particle.

[0104] Step 10511: Traverse all particles in the particle swarm at the nth iteration, determine whether the iteration number n has reached the iteration number threshold, and obtain the determination result; if the determination result is no, proceed to step 10512; if the determination result is yes, proceed to step 10513.

[0105] Step 10512: Increment the iteration number n by 1 and return to step 1053.

[0106] Step 10513: Determine the set of decision variables corresponding to the particles in the particle swarm that meet the optimization conditions in the nth iteration as the optimal set of decision variables.

[0107] For the requirements of this implementation case, the optimization objective is defined as: finding a set of input-to-return ratios that, while satisfying purity constraints, maximizes the PX yield and minimizes cycle energy consumption. This problem is a constrained multi-objective optimization problem, and its mathematical description is as follows:

[0108] max REX = f1(x)

[0109] min CEC=f2(x)

[0110] st PUX=g(x)≥99.7%

[0111] x=[r1,r2,r3,r4]∈D

[0112] Where r1, r2, r3, r4 are the reflux ratios of each zone, D is the optimization interval of each reflux ratio, and f1(x), f2(x), g(x) are the polynomial functions corresponding to each performance index in the proxy model.

[0113] The algorithm was selected using test functions ZDT2 and ZDT6. Both test functions ZDT2 and ZDT6 have two objective functions, and the position and shape of the optimal frontier are known, which can be used to test the performance of the algorithm.

[0114] Both test function ZDT2 and test function ZDT6 are two-objective optimization problems, constructed according to the following process:

[0115] min f1(x)

[0116]

[0117] Where x = (x1, x2, ..., x i ,…,x n ) T Let x be an n-dimensional decision vector. i Let be the value of the i-th decision variable.

[0118] The test function ZDT2 has 30 decision variables, defined as follows:

[0119] f1(x)=x1

[0120]

[0121] st 0≤x i ≤1 i=1,……,30

[0122] The test function ZDT6 has 10 decision variables, defined as follows:

[0123] f1(x)=1-exp(-4x1)sin 6 (6πx1)

[0124]

[0125] st 0≤x i ≤1 i=1,……,10

[0126] Experiments show that the competition-based multi-objective particle swarm optimization algorithm (CMOPSO) has significant advantages in solving this type of optimization problem, such as... Figures 6-9 As shown, the optimal solution set found by the CMOPSO algorithm on the test function is very close to the optimal front, outperforming the classic NSGA-II algorithm in both convergence and diversity, making it suitable for the optimization problem studied in this invention. The competition-based multi-objective particle swarm optimization algorithm (CMOPSO) originates from the competitive swarm optimizer. Compared with other particle swarm algorithms, its main difference lies in that the particle swarm update process is driven by the current internal competitive winners, rather than by historically excellent particles. Therefore, there is no external archive set, reducing computational cost. Experiments show that the competition-based learning strategy achieves a better balance between convergence and diversity.

[0127] The main process of applying this algorithm in the proxy model is as follows:

[0128] (1) Initialization. Set the particle swarm size to 50 and the maximum number of iterations to 100. Particle p i It has two attributes: position and velocity. Position x i =[r1,r2,r3,r4] is composed of the recirculation ratios of each zone, with velocity v i =[v1,v2,v3,v4] is used to update the positions of particles. An initial particle swarm P = {p1,p2,…,p4} is generated in the decision space. 50}, and randomly assign the position and velocity of each particle.

[0129] (2) Competition Phase. The positions of each particle are input into the surrogate model, and the corresponding PX yield and cycle energy consumption are calculated as the fitness values ​​of the particles. Based on the fitness values, the particles are crowding-out and fast non-dominated sorting is performed. The top 10 particles from the sorting results are selected as the elite particle set Q = {q1,q2,…,q}. 10 For each particle p i Two elite particles q are randomly selected from Q. a q b Conduct a competition: In the target space consisting of f1(x) and f2(x), calculate p respectively. i With q a q b The angle between the particles, with the smaller angle, is used as p. i winner

[0130] (3) Learning Phase. According to the learning strategy, each particle learns from its current winner: utilizing the winner... Location information adjustment p i The speed, that is Where R1, R2 ∈ [0, 1] are random values, and then p is updated. i The position, i.e., x i ′=x i +v i Once all particle positions have been updated, the algorithm terminates if it has reached the maximum number of iterations; otherwise, it returns to step 2 to continue with the next generation.

[0131] Step 106: Control the simulated moving bed based on the optimal decision variable set.

[0132] The present invention provides an optimized method for separating p-xylene, which further includes the following steps after step 101 and before step 102:

[0133] Step 107: Obtain the instantaneous concentration values ​​of each component in each bed layer of the simulated moving bed; the components include ethylbenzene, p-xylene, m-xylene, o-xylene, and p-diethylbenzene.

[0134] Step 108: Construct standard curves for the concentration of each component in the simulated moving bed as a function of the bed layer, with the bed layer number as the x-axis and the instantaneous concentration value as the y-axis.

[0135] Step 107: Obtain the simulated concentration values ​​of each component in each bed layer of the simulated moving bed; the simulated concentration values ​​are obtained by the simulated moving bed mechanism model; the acquisition time corresponding to the simulated concentration values ​​is the same as the acquisition time corresponding to the instantaneous concentration values.

[0136] Step 1010: Using the bed number as the x-axis and the simulated concentration value as the y-axis, construct simulation curves for the concentration of each component in the simulated moving bed as a function of the bed layer.

[0137] Step 1011: Adjust the structural parameters of the simulated moving bed and the operating parameters of the xylene adsorption separation process so that the error between the simulated curve of the concentration of the same component changing with the bed and the standard curve of the concentration changing with the bed is within the preset error range, and obtain the corrected simulated moving bed mechanism model.

[0138] The simulated moving bed mechanism model was built based on Aspen Chromatography simulation software. First, the adsorption and separation process of the simulated moving bed was constructed. According to... Figure 1 The partitioning strategy shown illustrates the overall structure of the mechanistic model built in Aspen Chromatography simulation software, including inlet and outlet pipelines, pipeline pumps, rotary valves, and a simulated moving bed. The results are as follows... Figure 2As shown, zone I is the desorption zone, zone II is the purification zone, zone III is the adsorption zone, zone IV is the buffer zone, D and Desorbent both represent desorbents, E and Extract both represent extracts, F and Feed both represent feedstocks, R and Raffinate both represent raffinate, and G1, G2, G3, and G4 are four different flushing solutions. Then, the parameters of the simulated moving bed are configured. The parameters required for modeling include design parameters such as bed length, bed inner diameter, and adsorbent packing density, as well as operating parameters such as step time and reflux ratio in each zone. Parameter details are shown in Table 1. Finally, the adsorption kinetic equations of the simulated moving bed are corrected. Based on the Langmuir adsorption mechanism, this invention relates to five components (four C8 aromatics and a desorbent) and four kinetic parameters (Peckley number, mass transfer coefficient, and two Langmuir adsorption equilibrium constants), therefore a total of 20 kinetic parameters need to be checked. Since the entire unit operates stably during actual industrial operation, the correction process uses the most recent set of actual values ​​for the concentration distribution of the simulated moving bed components.

[0139] Using the optimization tools built into the Aspen Chromatography simulation software, the aforementioned kinetic parameters were corrected to minimize the deviation between the simulated and actual values ​​of component concentrations in each bed layer of the simulated moving bed. The actual and simulated values ​​were then plotted as corresponding curves, such as... Figure 4 and Figure 5 As shown, the horizontal axis represents the bed layer, and the vertical axis represents the concentration. Since the simulated moving bed in industry operates intermittently, the actual values ​​are instantaneous values ​​at a certain sampling moment. However, the mechanistic model treats this dynamic process as a steady-state process. Therefore, there is still a certain deviation between the two sets of curves. However, the simulated concentration results at the extraction port and the residual port are basically consistent with the actual values, indicating that the mechanistic model can accurately describe the industrial adsorption separation process of this device.

[0140] Table 1. Parameters involved in the simulated moving bed mechanism model.

[0141]

[0142] Under the same population size and number of iterations, the CMOPSO algorithm and the NSGA-II algorithm show the following performance in optimizing simulated moving bed operations: Figure 10 As shown. From Figure 10The termination of the particle swarm optimization (PSO) iteration reveals that maximizing the p-xylene yield and minimizing the cycle energy consumption are conflicting optimization objectives, with no unique optimal solution. This is because decreasing the cycle energy consumption also reduces the average flow rate of the simulated moving bed, leading to reduced relative contact between the PX in the feed liquid and the adsorbent. Consequently, some excess PX remains unadsorbed and is lost in the raffinate, resulting in a decrease in PX yield. Both the CMOPSO and NSGA-II algorithms yielded Pareto optimal fronts with similar trends when addressing this optimization problem. However, the majority of optimal solutions obtained by the NSGA-II algorithm were dominated by the optimal individuals from the CMOPSO algorithm, and the optimal fronts of the CMOPSO algorithm were more widely and uniformly distributed. The results show that the CMOPSO algorithm outperforms the NSGA-II algorithm in both convergence and diversity, demonstrating a significant advantage in optimizing simulated moving bed adsorption separation processes. The optimization results obtained by the CMOPSO algorithm are shown in Table 2. Operators can flexibly select ideal operating conditions from the optimal solution set according to production needs. The model and optimization method established based on the above steps can be applied to optimize the operating conditions of industrial simulated moving bed adsorption separation processes, improving the operational efficiency of the equipment.

[0143] Table 2. Schematic diagram of CMOPSO algorithm optimization results

[0144]

[0145]

[0146]

[0147] Furthermore, the present invention also provides an optimized system for separating p-xylene, comprising:

[0148] The simulated moving bed mechanism model construction module is used to construct a simulated moving bed mechanism model based on the structural parameters of the simulated moving bed and the operating parameters of the p-xylene adsorption separation process; the simulated moving bed mechanism model is used to describe the component concentration of each bed layer of the simulated moving bed.

[0149] The sampling interval acquisition module is used to acquire the sampling intervals for multiple decision variables. These decision variables include: desorption zone reflux ratio, purification zone reflux ratio, adsorption zone reflux ratio, and buffer zone reflux ratio.

[0150] The simulated moving bed surrogate model construction module is used to construct a simulated moving bed surrogate model based on the sampling interval of each decision variable and the simulated moving bed mechanism model. The simulated moving bed surrogate model is used to simulate the yield, purity and cycle energy consumption of para-xylene under different decision variables.

[0151] The optimization condition construction module is used to construct optimization conditions. The optimization conditions include: maximizing the yield of p-xylene, minimizing the cycle energy consumption, and ensuring that the purity of p-xylene is greater than the purity threshold.

[0152] The optimal decision variable set determination module is used to optimize the simulated moving bed surrogate model based on optimization conditions using the particle swarm optimization algorithm to determine the optimal decision variable set.

[0153] The simulated moving bed control module is used to control the simulated moving bed based on the optimal set of decision variables.

[0154] The instantaneous concentration value acquisition module is used to acquire the instantaneous concentration values ​​of each component in each bed layer of the simulated moving bed; the components include ethylbenzene, p-xylene, m-xylene, o-xylene, and p-diethylbenzene.

[0155] The standard curve construction module is used to construct standard curves for simulating the concentration changes of each component in a moving bed, with the bed number as the x-axis and the instantaneous concentration value as the y-axis.

[0156] The concentration simulation value acquisition module is used to acquire the concentration simulation value of each component in each bed layer of the simulated moving bed; the concentration simulation value is obtained by the simulated moving bed mechanism model; the acquisition time corresponding to the concentration simulation value is the same as the acquisition time corresponding to the instantaneous concentration value.

[0157] The simulation curve construction module is used to construct simulation curves of the concentration of each component in a moving bed as a function of the bed layer, with the bed layer number as the x-axis and the simulated concentration value as the y-axis.

[0158] The simulated moving bed mechanism model calibration module is used to adjust the structural parameters of the simulated moving bed and the operating parameters of the xylene adsorption separation process, so that the error between the simulated curve of the concentration of the same component changing with the bed and the standard curve of the concentration changing with the bed is within the preset error range, thus obtaining the calibrated simulated moving bed mechanism model.

[0159] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0160] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An optimized method for separating p-xylene, characterized in that, include: Based on the structural parameters of the simulated moving bed and the operating parameters of the p-xylene adsorption separation process, a simulated moving bed mechanism model is constructed; the simulated moving bed mechanism model is used to describe the component concentration of each bed layer in the simulated moving bed. Obtain the sampling intervals for multiple decision variables; Based on the sampling interval of each decision variable and the simulated moving bed mechanism model, a simulated moving bed surrogate model is constructed; the simulated moving bed surrogate model is used to simulate the p-xylene yield, p-xylene purity and cycle energy consumption under different decision variables; Construct optimization conditions; The optimization conditions include: maximizing the yield of p-xylene, minimizing the cycle energy consumption, and ensuring that the purity of p-xylene is greater than the purity threshold. Based on the aforementioned optimization conditions, the simulated moving bed proxy model is optimized using the particle swarm optimization algorithm to determine the optimal set of decision variables. The simulated moving bed is controlled based on the optimal set of decision variables; The construction of the simulated moving bed surrogate model based on the sampling interval of each decision variable and the simulated moving bed mechanism model includes: The decision variables are sampled using the Latin hypercube sampling method to obtain multiple decision variable groups; each decision variable group includes a desorption zone reflux ratio sample value, a purification zone reflux ratio sample value, an adsorption zone reflux ratio sample value, and a buffer zone reflux ratio sample value. The multiple decision variable groups are respectively input into the simulated moving bed mechanism model to obtain the process simulation data corresponding to each decision variable group; Based on the process simulation data, the p-xylene yield, p-xylene purity, and cycle energy consumption for each decision variable group are calculated respectively; the p-xylene yield is: The purity of p-xylene is: The cycle energy consumption is: ;in, The yield of p-xylene; This represents the mass fraction of p-xylene in the extract. This represents the mass fraction of p-xylene in the feed. This refers to the flow rate of the extracted fluid. This refers to the feed flow rate; For the purity of p-xylene, This represents the mass fraction of o-xylene in the extract. This represents the mass fraction of intermediate xylene in the extract. This represents the mass fraction of ethylbenzene in the extract. For cycle energy consumption, This refers to the feed density; It is the acceleration due to gravity; To simulate the average flow rate of the circulating pump in a moving bed within a single cycle; , To simulate the instantaneous flow rate of the kth bed layer of the moving bed; For Yang Cheng; For circulating pump efficiency; For motor efficiency; Using decision variables as independent variables and p-xylene yield, p-xylene purity, and cycle energy consumption as dependent variables, regression models for p-xylene yield, p-xylene purity, and cycle energy consumption were constructed. The para-xylene yield regression model, para-xylene purity regression model, and cycle energy consumption regression model were determined to be simulated moving bed surrogate models. Based on the aforementioned optimization conditions, the simulated moving bed surrogate model is optimized using the particle swarm optimization algorithm to determine the optimal set of decision variables, including: Let the number of iterations n equal 1; An initial particle swarm is constructed as the particle swarm for the first iteration; the number of particles in the initial particle swarm is equal to the number of decision variable groups; the positions of the particles in the initial particle swarm are used to represent the reflux ratio of the desorption zone, the reflux ratio of the purification zone, the reflux ratio of the adsorption zone, and the reflux ratio of the buffer zone; Input the position of the particle swarm at the nth iteration into the simulated moving bed surrogate model to obtain the paraxylene yield and cycle energy consumption of each particle, which are used as the fitness value of the particle. Crowding is calculated and fast non-dominated sorting is performed on each particle based on its fitness value; A predetermined number of particles are designated as the elite particle set. Determine any particle in the particle swarm at the nth iteration as the current particle; Get any two particles from the elite particle set as the current elite particles; Calculate the angles between the current particle and the two current elite particles respectively; The current elite particle corresponding to the smallest included angle is determined as the winning elite particle. Adjust the speed and position of the current particle based on the winning elite particle; By iterating through all particles in the particle swarm at the nth iteration, we determine whether the iteration number n has reached the iteration number threshold and obtain the result. If the judgment result is negative, then the value of the iteration number n is increased by 1, and the process returns to the step "Input the position of the particle swarm at the nth iteration into the simulated moving bed surrogate model to obtain the p-xylene yield and cycle energy consumption of each particle". If the judgment result is yes, then the decision variable group corresponding to the particles in the particle swarm that meet the optimization conditions in the nth iteration is determined as the optimal decision variable group.

2. The optimized method for separating p-xylene according to claim 1, characterized in that, After constructing the simulated moving bed mechanism model based on the structural parameters of the simulated moving bed and the operating parameters of the p-xylene adsorption separation process, the following is also included: The instantaneous concentration values ​​of each component in each bed layer of the simulated moving bed are obtained; the components include ethylbenzene, p-xylene, m-xylene, o-xylene, and p-diethylbenzene; Standard curves were constructed to simulate the concentration variation of each component in a moving bed, with the bed number as the x-axis and the instantaneous concentration value as the y-axis. The simulated concentration values ​​of each component in each bed layer of the simulated moving bed are obtained; the simulated concentration values ​​are obtained by the simulated moving bed mechanism model; the acquisition time corresponding to the simulated concentration values ​​is the same as the acquisition time corresponding to the instantaneous concentration values. Using the bed number as the x-axis and the simulated concentration value as the y-axis, simulated curves of the concentration change of each component in the simulated moving bed as a function of the bed layer were constructed. By adjusting the structural parameters of the simulated moving bed and the operating parameters of the xylene adsorption separation process, the error between the simulated curve of the concentration of the same component changing with the bed and the standard curve of the concentration changing with the bed is within the preset error range, thus obtaining the corrected simulated moving bed mechanism model.

3. The optimized method for separating p-xylene according to claim 1, characterized in that, The decision variables include: desorption zone reflux ratio, purification zone reflux ratio, adsorption zone reflux ratio, and buffer zone reflux ratio.

4. The optimized method for separating p-xylene according to claim 1, characterized in that, The sampling interval of the decision variables is determined by searching the device process of the simulated moving bed.

5. An optimized system for separating p-xylene, characterized in that, include: The simulated moving bed mechanism model construction module is used to construct a simulated moving bed mechanism model based on the structural parameters of the simulated moving bed and the operating parameters of the p-xylene adsorption separation process; the simulated moving bed mechanism model is used to describe the component concentration of each bed layer of the simulated moving bed; The sampling interval acquisition module is used to acquire the sampling intervals of multiple decision variables; A simulated moving bed surrogate model construction module is used to construct a simulated moving bed surrogate model based on the sampling interval of each decision variable and the simulated moving bed mechanism model. The simulated moving bed surrogate model is used to simulate the p-xylene yield, p-xylene purity, and cycle energy consumption under different decision variables. The simulated moving bed surrogate model construction module is also used to sample the decision variables using the Latin hypercube sampling method to obtain multiple decision variable groups. Each decision variable group includes a desorption zone reflux ratio sampling value, a purification zone reflux ratio sampling value, an adsorption zone reflux ratio sampling value, and a buffer zone reflux ratio sampling value. The multiple decision variable groups are input into the simulated moving bed mechanism model to obtain process simulation data corresponding to each decision variable group. Based on the process simulation data, the p-xylene yield, p-xylene purity, and cycle energy consumption corresponding to each decision variable group are calculated. The p-xylene yield is: The purity of p-xylene is: The cycle energy consumption is: ;in, The yield of p-xylene; This represents the mass fraction of p-xylene in the extract. This represents the mass fraction of p-xylene in the feed. This refers to the flow rate of the extracted fluid. This refers to the feed flow rate; For the purity of p-xylene, This represents the mass fraction of o-xylene in the extract. This represents the mass fraction of intermediate xylene in the extract. This represents the mass fraction of ethylbenzene in the extract. For cycle energy consumption, This refers to the feed density; It is the acceleration due to gravity; To simulate the average flow rate of the circulating pump in a moving bed within a single cycle; , To simulate the instantaneous flow rate of the kth bed layer of the moving bed; For Yang Cheng; For circulating pump efficiency; For motor efficiency; using decision variables as independent variables and paraxylene yield, paraxylene purity, and cycle energy consumption as dependent variables, respectively, a paraxylene yield regression model, a paraxylene purity regression model, and a cycle energy consumption regression model are constructed; the paraxylene yield regression model, paraxylene purity regression model, and cycle energy consumption regression model are determined to be simulated moving bed proxy models; An optimization condition construction module is used to construct optimization conditions; the optimization conditions include: maximum p-xylene yield, minimum cycle energy consumption, and p-xylene purity greater than a purity threshold. The optimal decision variable set determination module is used to optimize the simulated moving bed surrogate model based on the optimization conditions using a particle swarm optimization algorithm to determine the optimal decision variable set. This module further sets the iteration number n to 1; constructs an initial particle swarm as the particle swarm for the first iteration; the number of particles in the initial particle swarm is equal to the number of decision variable sets; the positions of the particles in the initial particle swarm represent the reflux ratios in the desorption zone, purification zone, adsorption zone, and buffer zone; inputs the positions of the particle swarm corresponding to the nth iteration into the simulated moving bed surrogate model to obtain the paraxylene yield and cycle energy consumption of each particle, which are used as the particle's fitness value; performs crowding calculation and fast non-dominated sorting on the particles based on their fitness values; determines a preset number of particles as the elite particle set; and determines the optimal set for the nth iteration. Any particle in the particle swarm at the nth iteration is taken as the current particle; any two particles in the elite particle set are taken as the current elite particles; the angle between the current particle and the two current elite particles is calculated respectively; the current elite particle corresponding to the smallest angle is determined as the winning elite particle; the speed and position of the current particle are adjusted according to the winning elite particle; all particles in the particle swarm at the nth iteration are traversed, and it is determined whether the iteration number n has reached the iteration number threshold, and the judgment result is obtained; if the judgment result is not, the value of the iteration number n is increased by 1, and the process returns to the step "input the position corresponding to the particle swarm at the nth iteration into the simulated moving bed surrogate model to obtain the paraxylene yield and cycle energy consumption of each particle"; if the judgment result is yes, the decision variable group corresponding to the particle in the particle swarm at the nth iteration that meets the optimization conditions is determined as the optimal decision variable group. A simulated moving bed control module is used to control the simulated moving bed based on the optimal decision variable set.

6. The optimized system for separating p-xylene according to claim 5, characterized in that, The system also includes: The instantaneous concentration value acquisition module is used to acquire the instantaneous concentration value of each component in each bed layer of the simulated moving bed; the components include ethylbenzene, p-xylene, m-xylene, o-xylene, and p-diethylbenzene; The standard curve construction module is used to construct standard curves for the concentration of each component in a moving bed as a function of the bed layer, with the bed layer number as the x-axis and the instantaneous concentration value as the y-axis. The concentration simulation value acquisition module is used to acquire the concentration simulation value of each component in each bed layer of the simulated moving bed; the concentration simulation value is obtained by the simulated moving bed mechanism model; the acquisition time corresponding to the concentration simulation value is the same as the acquisition time corresponding to the instantaneous concentration value; The simulation curve construction module is used to construct simulation curves of the concentration of each component in the simulated moving bed as a function of the bed layer, with the bed layer number as the horizontal axis and the simulated concentration value as the vertical axis. The simulated moving bed mechanism model calibration module is used to adjust the structural parameters of the simulated moving bed and the operating parameters of the xylene adsorption separation process, so that the error between the simulated curve of the concentration of the same component changing with the bed and the standard curve of the concentration changing with the bed is within the preset error range, thus obtaining the calibrated simulated moving bed mechanism model.

7. The optimized system for separating p-xylene according to claim 5, characterized in that, The decision variables include: desorption zone reflux ratio, purification zone reflux ratio, adsorption zone reflux ratio, and buffer zone reflux ratio.

Images