A method to improve the efficiency of parameter estimation of continuous renormalization model

By dividing the lumped model according to the reactant structure and kinetic characteristics and simplifying the calculation of fast reaction components, the problem of low estimation efficiency of continuous reform model parameters is solved, and a significant improvement in the lumped model parameter estimation speed and production optimization are achieved.

CN114819462BActive Publication Date: 2024-11-26NANJING RICHISLAND INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210082355.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-01-24
Publication Date
2024-11-26
Estimated Expiration
2042-01-24

AI Technical Summary

Technical Problem

The parameter estimation efficiency of continuous reforming model is low. Traditional algorithms are difficult to calculate the mutual conversion and reaction rate gap between complex reactants, resulting in poor efficiency. Intelligent optimization algorithms are prone to falling into local optimization and difficult to improve solution efficiency.

Method used

The lumped model is divided according to the molecular structure and kinetic characteristics of the reactants, and the calculation of fast reaction components is simplified, and the calculation of fast reaction components is approximate to steady-state components. The objective function is used to solve the reaction rate correction factor, and the lumped model is simplified to improve the efficiency of parameter estimation.

Benefits of technology

It significantly improves the estimation speed of lumped model parameters, reduces the calculation accuracy loss, improves the efficiency of model parameter setting, and helps enterprises optimize production and improve economic benefits.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114819462B_ABST
    Figure CN114819462B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for improving the efficiency of continuous reforming model parameter estimation. The method screens the fast-reacting components in the continuous reforming process and uses them as steady-state components to simplify the calculation of the lumped model and then perform parameter estimation. Compared with the method of directly estimating the parameters of the original model using an optimization algorithm, this method can simplify the calculation and significantly improve the efficiency of model parameter estimation, thereby helping enterprises to adjust the model more timely and effectively, which is of great value to stabilizing and optimizing continuous reforming production and improving the economic benefits of enterprises.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the production field of refining enterprises, and in particular to a method for establishing a mechanism model based on process knowledge, and specifically to a method for improving the efficiency of continuous reforming model parameter estimation. Background Art

[0002] Continuous reforming is the mainstream process of catalytic reforming in refining enterprises. It is mainly used for the production of high-octane gasoline, light aromatics and hydrogen, and is crucial to the economic benefits of the entire enterprise. The process simulation of the reforming process can find the weak links in the production process, help enterprises optimize production operations and improve efficiency.

[0003] Establishing an efficient and accurate model is an important part of the process simulation of continuous reforming. At present, the parameter estimation of the continuous reforming model is generally solved by the traditional algorithm relying on the gradient. However, due to the extremely complex continuous reforming process, there is a strong coupling between the mutual transformation of reactants, and there is a large difference in the reaction rate between different reactions, which makes it difficult for the traditional algorithm to solve by relying on the gradient. Traditional algorithms often need to call the lumped model nearly 10,000 times to obtain satisfactory results for parameter estimation, and the parameter estimation efficiency is low. In response to the above problems, many studies have focused on the optimization of parameter estimation algorithms, such as using algorithms such as the bounded Fourier transform (BFGS) and the Marquardt method to optimize the gradient calculation. However, as the lumped division continues to be refined and the lumped model becomes more complex, the gradient is still difficult to calculate. Therefore, intelligent optimization algorithms such as evolutionary algorithms and swarm intelligence algorithms are used to speed up the parameter estimation speed. Although intelligent algorithms do not rely on gradients and have fast calculation speeds, they often focus on improving the performance of the algorithm itself, and are prone to problems such as falling into local optimality and poor convergence. In fact, from the analysis of the production process, it is found that in the continuous reforming model, some reactions are relatively fast, while some reactions are relatively slow. For fast-reacting components, they can actually be approximated as steady-state components, which greatly simplifies the kinetic model and greatly improves the solution efficiency without reducing the accuracy.

[0004] In short, the continuous reforming process urgently needs a calculation method for a simplified lumped model so as to improve the efficiency of model parameter estimation and thereby guide enterprises to optimize production and improve efficiency. Summary of the invention

[0005] In view of the above problems, the present invention discloses a method for improving the parameter estimation efficiency of the continuous reforming model, establishing a lumped model according to the molecular structure of the continuous reforming reactants and the reaction kinetics characteristics, and then simplifying the calculation of the fast reaction components in the lumped model to improve the parameter setting efficiency of the lumped model. The method comprises the following steps:

[0006] 1) According to the molecular structure and kinetic characteristics of the reactants in the continuous reforming process, the lumps are divided and an initial lumped model including I lumps, J reactions and K reactors is established:

[0007]

[0008] Where Y i,k represents the actual molar flow rate of the lumped i in reactor k, kmol / h, k = 1, 2, ..., K; x k is the coordinate of the reactant flow direction relative to the outer diameter of the reactor k, m, the reactants flow radially from the outer diameter to the inner diameter of the reactor; T k represents the actual temperature at the outlet of reactor k, K; where Y ik0 is the initial molar flow rate of lumped i at the inlet of reactor k; A k is the cross-sectional area of ​​the reactant flow in reactor k, m 2 , for radial reactors A = 2πxH; V c,k is the catalyst loading in reactor k, m 3 ;LHSV k is the volume space velocity, 1 / h, which is the ratio of the volume flow rate of the feed in reactor k to the catalyst loading volume; r k + is the total generation rate of all reactions in reactor k, kmol / (kg catalyst h), r k - is the total consumption rate of all reactions in reactor k; C pik is the lumped heat capacity of the i-th reactor in reactor k, kJ / (mol·K); △H j,k is the enthalpy change of the jth reaction in reactor k, kJ / mol; T k0 is the reactor k inlet temperature;

[0009] Calculate the reaction rate using the following formula:

[0010]

[0011] Where λ is the forward reaction rate, eq is the reaction equilibrium constant, for an irreversible reaction λ eq =∞;Y - is the molar concentration of the reaction consumables; Y + is the molar concentration of the reaction product; λ0 is the reaction frequency factor, h -1 ·Mpa -θ ; λ' is the reaction rate correction factor; E is the reaction activation energy, kJ / mol; R is the gas constant, 8.314 J / (mol·K); T is the temperature in the reactor, K; P h is the partial pressure of reactant hydrogen, MPa; θ is the pressure index; is the average activity of the catalyst in the reactor; S is the catalyst activity factor; F is the device factor; ψ is the adsorption term of the catalyst active center.

[0012] 2) The model Components that meet the following conditions are classified as fast-reacting components:

[0013]

[0014] In the formula and are the generation and consumption rates of the fast-reacting components at position x in reactor k, r + ≈r - , δ is the threshold;

[0015] 3) Divide the lumped component Y into two sets, the fast response component and non-rapid reacting components

[0016] 4) The fast-reacting component is approximated as the steady-state component, and its rate of change becomes Solve to get the molar concentration calculation formula, Y fast =g(Y,x);

[0017] 5) Use the fast-reacting component molar concentration calculation formula g(Y,x) to simplify the lumped model This gives a simplified lumped model:

[0018]

[0019] 6) The parameters to be estimated are reduced to J reaction rate correction factors λ j ',j=1,2,...,J;

[0020] 7) Use the following objective function to solve the model The J reaction rate correction factors λ' are:

[0021]

[0022] Where Y i,k is the actual molar flow rate of the ith lumped volume at the outlet of the kth reactor, is the model's predicted value for the ith lumped molar flow rate at the outlet of the kth reactor, T k is the actual temperature at the outlet of the kth reactor, is the model's predicted value for the outlet temperature of the kth reactor.

[0023] Beneficial effects:

[0024] The present invention discloses a method for improving the efficiency of continuous reforming model parameter estimation. The method aims at the problem of poor efficiency of continuous reforming lumped model parameter estimation. By screening fast-reacting components in the continuous reforming process and simplifying the lumped model calculation, the lumped model parameter estimation efficiency is significantly improved. The method can quickly set the model parameters, thereby helping enterprises to adjust the model more timely and effectively, which is of great value to stabilizing and optimizing continuous reforming production and improving the economic benefits of enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A flow chart of a method for improving the efficiency of continuous reformulation model parameter estimation according to the present invention;

[0026] Figure 2 It is a curve diagram of P11 changes before and after the model is simplified in the implementation method of the present invention;

[0027] Figure 3 It is a curve diagram of P12 changes before and after the model is simplified in the implementation method of the present invention;

[0028] Figure 4 It is a curve diagram of N11 variation before and after the model is simplified in the implementation method of the present invention;

[0029] Figure 5 It is a curve diagram of N12 variation before and after the model is simplified in the implementation method of the present invention;

[0030] Figure 6 A11 change curve diagram before and after model simplification in the implementation method of the present invention;

[0031] Figure 7 A12 variation curve diagram before and after model simplification in the implementation method of the present invention;

[0032] Figure 8 It is the change curve of the parameter estimation objective function before and after the model is simplified in the implementation method of the present invention. DETAILED DESCRIPTION

[0033] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, and the implementation effect of the method in improving the efficiency of continuous reforming parameter estimation is described by a specific operation flow. This implementation case is implemented based on the technical solution of the present invention, but the protection scope of the present invention is not limited to the following embodiments.

[0034] The continuous reforming unit of a certain refinery contains 4 reactors, K = 4. The lumped model structure in each reactor is the same, but the initial value of the lumped model and the reactor parameters are different. To avoid repetition, the present invention takes the first reactor as an example. According to the molecular structure of the reactants and the reaction kinetics, a lumped model containing 30 lumps, I = 30, and 40 reactions is established, J = 40.

[0035] Figure 1 This is a flow chart of a method for improving the parameter estimation efficiency of the continuous reformulation model of the present invention. First, the threshold of the fast response judgment condition is set to δ = 10 -6 , analyzing the reactants in the lumped model, it is found that the lumped carbon 11-cycloalkanes (N11) and carbon 12-cycloalkanes (N12) meet the fast reaction conditions. The chemical reaction equations related to N11 and N12 are:

[0036]

[0037] Where P11 is C11 alkane, P12 is C12 alkane, A11 is C11 aromatic hydrocarbon, A12 is C12 aromatic hydrocarbon. The net generation rates of N11 and N12 are:

[0038]

[0039]

[0040] Where Y P11 ,Y N11 ,Y A11 are the molar flow rates of components P11, N11 and A11 respectively; Y P12 ,Y N12 ,Y A12 are the molar flow rates of components P12, N12 and A12 respectively; λ1,λ eq1 They are Reaction rate and reaction equilibrium constant; λ2,λ eq2 They are Reaction rate and reaction equilibrium constant; λ3,λ eq3 They are Reaction rate and reaction equilibrium constant; λ4,λ eq4 They are reaction rate and reaction equilibrium constant.

[0041] N11 and N12 are classified as fast-reacting components and assumed to be steady-state components. The concentration calculation formulas of N11 and N12 are obtained as follows:

[0042]

[0043]

[0044] The concentration calculation formula of N11 and N12 is brought into the original model to simplify the lumped model calculation. The change curves of P11, P12, N11, N12, A11 and A12 are as follows: Figure 2-Figure 7As shown in the figure, the change curves of P11 and P12 after the simplified lumped model calculation are exactly the same as before, while the change curves of N11, N12, A11 and A12 have the same trend but with certain deviations. This shows that the simplified model still maintains the chemical kinetic characteristics of the original lumped components, which is consistent with the actual changes of chemical components in the reaction. Moreover, for the improvement of the model parameter estimation efficiency after the simplified model, a small loss of calculation accuracy is completely acceptable.

[0045] The 40 reaction correction factors λ were calculated using the following objective function: j ':

[0046]

[0047] Comparing the parameter estimation efficiency of the original model and the corrected model, the objective function change curve is as follows Figure 8 As shown, the parameter estimation indicators are shown in Table 1:

[0048] Table 1 Parameter estimation indicators before and after the simplified lumped model

[0049]

[0050] from Figure 8 As shown in Table 1, after simplifying the lumped model, the speed of lumped model parameter estimation is significantly improved, about 10 times faster than the original lumped model. Moreover, by simplifying the lumped model, the time scale gap between lumped components is reduced without losing the calculation accuracy of the model, making the lumped model easier to fit, the convergence of the lumped model objective function becomes faster, and the efficiency of model parameter estimation is significantly improved.

Claims

1. A method for improving the efficiency of continuous renormalization model parameter estimation, characterized in that By identifying the fast-reacting component in the continuous reforming model, simplifying the calculation of the component and improving the efficiency of model parameter estimation, the following steps are included: 1) According to the molecular structure and kinetic characteristics of the reactants in the continuous reforming process, the lumps are divided and an initial lumped model including I lumps, J reactions and K reactors is established: Where Y i,k represents the actual molar flow rate of the lumped i in reactor k, i = 1, 2, ..., I, k = 1, 2, ..., K; x k T is the coordinate of the reactant flow direction relative to the outer diameter of the reactor k, and the reactants flow radially from the outer diameter to the inner diameter of the reactor; k represents the actual temperature at the outlet of reactor k; 2) The model Components that meet the following conditions are classified as fast-reacting components: In the formula and are the generation and consumption rates of the fast-reacting components at position x in reactor k, r + ≈r - , δ is the threshold; 3) Divide the lumped component Y into two sets, the fast response component and non-rapid reacting components 4) The fast-reacting component is approximated as the steady-state component, and its rate of change becomes Solve the formula for calculating its molar concentration, Y fast =g(Y,x); 5) Use the fast-reacting component molar concentration calculation formula g(Y,x) to simplify the lumped model Calculate and get the simplified lumped model: 6) The parameters to be estimated are reduced to J reaction rate correction factors λ′ j ,j=1,2,...,J; 7) Use the following objective function to solve the model The J reaction rate correction factors λ′ j : Where Y i,k is the actual molar flow rate of the ith lumped volume at the outlet of the kth reactor, is the model's predicted value for the ith lumped molar flow rate at the outlet of the kth reactor, T k is the actual temperature at the outlet of the kth reactor, is the model's predicted value for the outlet temperature of the kth reactor.

2. A method for improving the efficiency of continuous retraining model parameter estimation according to claim 1, characterized in that Create a lumped model consisting of I lumps, J reactions, and K reactors: Where Y ik0 is the initial molar flow rate of lumped i at the inlet of reactor k; A k is the cross-sectional area of ​​the reactant flow in reactor k, and for radial reactors it is A = 2πxH; V c,k is the catalyst loading in reactor k; LHSV k is the volume space velocity, which is the ratio of the volume flow rate of the feed in reactor k to the catalyst loading volume; r k + is the total generation rate of all reactions in reactor k, r k - is the total consumption rate of all reactions in reactor k; C pik is the lumped heat capacity of the reactor k; △H j,k is the enthalpy change of the jth reaction in reactor k; r j,k is the reaction rate of the jth reaction in reactor k; T k0 is the reactor k inlet temperature.

3. A method for improving the efficiency of continuous retraining model parameter estimation according to claim 1, characterized in that The reaction rate r is calculated using the reaction rate correction factor λ': Where λ is the forward reaction rate, eq is the reaction equilibrium constant, for an irreversible reaction λ eq =∞;Y - is the molar concentration of the reaction consumables; Y + is the molar concentration of the reaction product; λ0 is the reaction frequency factor; λ' is the reaction rate correction factor; E is the reaction activation energy; R is the gas constant; T is the temperature in the reactor; P h is the partial pressure of reactant hydrogen; θ is the pressure index; is the average activity of the catalyst in the reactor; S is the catalyst activity factor; F is the device factor; ψ is the adsorption term of the catalyst active center.