CN108710779A - Optimal modeling method for FCC reaction regeneration process of micro-charge interaction P system in membrane - Google Patents

Abstract

The invention discloses an optimal modeling method for the FCC reaction regeneration process of the internal micro-charge mutual force P system in the membrane, aiming at the optimal modeling of the fluid catalytic cracking (Fluid Catalytic Cracking, FCC) reaction-regeneration process in the oil refining process, fast and high-precision Condition forecast. It includes the following steps: 1) Obtain the sampling data of the process through on-site operations or experiments, determine the approximate structure of each sub-model of the input and output, and use the sum of squared errors between the model estimated output and the actual output as the minimization objective function; After Ca ²⁺ , Na ⁺ , Cl ^‑ plasma, inspired by the interaction between ions in the new intracellular environment, a specific high-efficiency optimization algorithm is abstracted; 3) Setting the operating parameters of the algorithm; 4) By minimizing the objective function, The algorithm is used to estimate the unknown parameters in the reaction-regeneration model, obtain the optimal parameters and form a mathematical model. The modeling method of the invention has the characteristics of anti-premature, high precision of optimization and fast convergence, and is also suitable for modeling other complex chemical reaction processes.

Description

An Optimal Modeling Method for FCC Reaction Regeneration Process of Intramembrane Micro-charge Mutual Force P System

technical field

The present invention relates to an optimization modeling method, in particular to a technology for optimally establishing a prediction model in the fluid catalytic cracking (FCC) reaction-regeneration process aimed at lightening heavy crude oil in petroleum enterprises.

Background technique

Energy is the source of social progress. The popularity of automobiles has accelerated the pace of urban development. Petrochemical energy, especially gasoline and diesel for vehicles, is the main energy source for automobiles. Only a very small amount of gasoline and diesel can be obtained by direct distillation of crude oil in a mixed state (the yield of straight-run light oil is about 10% to 40%), mainly because the main component of crude oil is a mixture of heavy oils. According to statistics, at present, 78.1% of gasoline in my country comes from the fluidized catalytic cracking (FCC) process. Distillate oil (or heavy distillate oil mixed with vacuum oil extraction) contacts with the catalyst, and undergoes cracking reaction to produce a series of high-value products such as light diesel oil, gasoline, and combustible gas. This process greatly increases the conversion rate from petroleum to light oil products. Conversion rate. Considering the contradiction between the increasingly depleted supply of crude oil resources in the world and the increasing demand for light product oil resources that social development must face, heavy oil fluid catalytic cracking will be an important direction for the development of catalytic cracking in the world in the future ¹ .

The basic components of a commonly used fluid catalytic cracking unit (Fluid Catalytic Cracking Unit, FCCU) mainly include: reaction-regenerator, lifting reactor, main fractionator, absorption stripper, main fan and wet air compressor. As one of the most important parts, the reaction-regenerator plays a vital role in achieving advanced process control and maximizing benefits in the FCC process. Reaction temperature, catalyst circulation rate, steam flow, primary air volume, among other factors within the reaction-regenerator determine product quality and revenue distribution. However, the reaction-regenerator is a multi-parameter, nonlinear and multi-variable tightly coupled complex system. In order to improve its design, establish an automatic control system, analyze the reliable and safe operation of the control system, and establish a dynamic system that can reflect the law of its chemical reaction The meaning of the learning model is crucial ² .

For modeling methods, it can be roughly divided into two categories: mechanism modeling and experimental modeling. Usually people tend to model the mechanism, thinking that such a model has a basic theory as a guarantee, and the physical meaning is clear. For more complex systems, many simplifications are required before the mechanism model can be established. Experimental modeling seems to be a last resort, but it also has strong vitality in today's greatly improved data processing capabilities. The composition of the process raw materials processed by the fluid catalytic cracking process is complex, the temperature gradient in the physical pipeline of the reactor regenerator changes greatly, and the calorific value of each point in the fluid changes dynamically, which brings great difficulties to the mechanism modeling of the process . Aiming at various input variables, output variables and their matrix couplings that affect the fluid catalytic cracking process, the response curve and other appearance characteristics of the multiple-input multiple-output (MIMO) system can be obtained by using the step perturbation test method. According to the response curve to give its scientific model structure, using the specific sampling data in the curve, the efficient optimization algorithm is used to make global optimal estimation of the unknown parameters in the model structure. This system identification method obtains the mathematical model of the research object through actual verification , the model that has been trained can predict the output of a period of time in the future under the same working conditions (or minor adjustments) with better accuracy.

Therefore, it is an effective and feasible scientific modeling problem to make reasonable assumptions on the model structure on the basis of field data sampling and transform the parameter estimation of the FCC process model into an optimization problem. However, for the actual engineering optimization problems of the oil cracking process in the fluidized catalytic cracking unit, which are characterized by complexity, constraints, nonlinearity, and multiple local minimum points, the traditional optimization method is easy to fall into the local optimal value or even fail to obtain the optimal value. The value of figure of merit has made the intelligent optimization method inspired by biology get people's attention. A cell is the cornerstone of a biological system, and at the same time it is an exquisite "machine" with a complex structure that has evolved over a long period of time. Its internal intricate behaviors are all self-regulated in an effective manner. In the past, computing science did not fully consider cells as a source of inspiration for computing models, and membrane computing was created to meet this challenge. In 1998, Gheorghe Pǎun, an academician of the European Academy of Sciences, proposed the concept of membrane computing. The initial letter of Pǎun is "P", so membrane computing is also called P system. Biochemical reactions in biological cell membranes or material exchange between cell membranes are regarded as a computational process, and even material exchange between cells can be regarded as information exchange between computational units; however, existing scientists seem to only consider It prevents the transfer of nutrients and metabolites by the cell membrane, but in fact, calcium, magnesium, sodium, carbon and other substances that enter the cell membrane are mostly in the ion state, such as Ca ²⁺ , Mg ²⁺ , Na ⁺ , In the framework of the cell membrane, each ion entering the membrane from outside the membrane will have a synergistic effect on the existing charged particles in the membrane. According to the nature of the mutual repulsion of the same kind of charges and the mutual attraction of different kinds of charges, it is possible to understand the deeper level of the cell. The mechanism and internal connection of all life activities in the body.

According to the conversion of intracellular substances, energy transfer and transport functions of biological cells, and the interaction between charged ions entering the cell membrane and the original various ions in the membrane, the method of the present invention abstracts a kind of biological energy inspired by the mutual force of charged ions in the membrane. P system optimization algorithm and corresponding rules can be used to solve complex nonlinear optimization problems ^; 2982 fluid catalytic cracking reaction process and equipment, considering the highly nonlinear, coupling and complexity of the actual process, the proposed P system optimization modeling method inspired by the micro-charge interaction in the membrane is used to solve the problem of fluid catalytic cracking In the parameter estimation of the cracking (FCC) reaction-regeneration process model, a relatively ideal effect has been achieved.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art and provide a P system FCC reaction-regeneration process modeling method inspired by charged ions in the membrane.

A kind of P system FCC process modeling method inspired by charged ions in the membrane of the present invention comprises the following steps:

1) Through on-site operations or experiments, the sampling data (including feed preheating temperature MV ₁ , circulating oil flow MV ₂ , residual oil flow MV ₃ , feed Oil flow MV ₄ , crude oil to riser flow MV ₅ , riser outlet temperature MV ₆ ; actual process temperature CV ₁ of No. 1 regenerator, actual process temperature CV ₂ of No. 2 regenerator, reaction heat of riser CV ₃ );

2) Based on the characteristics of the input and output data, combined with public experience, select the appropriate transfer function matrix model and determine the model parameter set and parameter search range to be determined, and sample the model estimated output and actual output of the FCC reactor-regenerator The sum of squared errors of the data is used as the minimization objective function;

3) Based on the influence and variation of charged ions entering the biological cell membrane on other charged ions and the transport mechanism of the cell membrane, an optimal modeling method for fluid catalytic cracking process inspired by the interaction of charged ions in the membrane is abstracted; Inspired by the influence of other charged ions, the following rules for the optimization method of the micro-charge mutual force P system in the membrane are proposed: selection rules, adaptive variation rules, micro-charge mutual force rules, exchange rules, and termination rules;

4) Set the initial environment and parameters of the P-system optimization algorithm inspired by the mutual force of charged ions in the membrane: the number of layers of the P-system nested membrane structure m=5, the number of object sets in each membrane layer n=10, loop The cycle number G=500, the adaptive mutation probability is a _p =0.01, b _p =0.29, c _p =15/G, G ₀ =G/2, interaction factor K _d =0.2, size of acceptable error ε=1×10 ^-4 and termination rule, where g is the current running algebra;

5) Extract the transfer function z transformation of each input MVi to output CV _j , as follows:

6) Run the P system optimization algorithm inspired by the interaction of charged ions in the membrane to estimate the unknown parameters in the fluid catalytic cracking process model, because the process has 6 input MV _i and 3 output CV _j ; i=1,2 , ..., 6; j=1, 2, 3; after selecting a certain sampling period, for each sub-process MV _i →CV _j , as long as the 36 parameters are determined to be less than 36), and at the same time collect The timed sampling of the control variables and the measured data table are reserved. At the same time, a parameterized predictive output model is formed;

7) The sum of squared errors (SSE) between the predicted output and the measured output under the parameter model is used as the objective function, as follows:

Where f _j is the SSE of the jth control variable, ns is the number of sampling points used when estimating each group of CV _j parameters; CV _{j, k} is the predicted output of the kth model; is the true process sampled value in step 6 for the same given operation corresponding to CV _j,k .

8) Run the proposed P system optimization algorithm inspired by the mutual force of charged ions in the membrane, and obtain the estimated values of the unknown parameters in the fluid catalytic cracking process model that best fit the actual output data by minimizing the objective function. Values are substituted into the fluid catalytic cracking process model to form a mathematical model of the fluid catalytic cracking process;

The membrane topology structure of the membrane calculation optimization algorithm of the optimal modeling method for the FCC reaction regeneration process of the micro-charge mutual force P system in the membrane is a nested membrane structure, except for the surface membrane and the innermost membrane, other membranes Both contain its adjacent inner membrane and are contained by its outer membrane at the same time, as shown in Figure 1;

The optimal modeling method for the FCC reaction regeneration process of the micro-charge mutual force P system in the membrane estimates the unknown parameters in the FCC process model, and obtains the FCC process model by minimizing the objective function. Unknown parameter estimates, the estimated values are substituted into the FCC process model, and the steps to form a mathematical model of the FCC process are as follows:

1) Set the initial environment of the P system optimization method and the search range of 36 (maximum 36) parameters to be estimated in each key output variable model group of the FCC model, and randomly generate an object set;

2) The FCC reaction taken from the reaction regenerator is at the feed preheating temperature MV ₁ , the preset flow rate of circulating oil MV ₂ , the preset flow rate of remaining oil MV ₃ , the preset flow rate of feedstock oil MV ₄ , the feedstock oil The preset flow rate MV ₅ supplied to the riser, the actual temperature value CV ₁ of the No. 1 regenerator at the preset temperature MV ₆ of the riser outlet, the actual temperature CV ₂ of the No. 2 regenerator, and the reaction heat CV ₃ of the riser The sum of the squares of the error between the predicted value of the model of each variable and the sampling value of the corresponding three output variables under the six operating variables of the actual FCCU process is used as the objective function;

3) The P-system optimization algorithm with a nested membrane structure is calculated from the innermost membrane, and all objects in the inner area of the membrane except the surface membrane are sequentially executed by the selection rule, the exchange rule, the adaptive variation rule, and the micro-charge mutual force rule operate;

4) If all objects in each membrane area are updated, then enter step 5), otherwise return to step 3);

5) Objects in the surface layer execute adaptive mutation rules and selection rule operations; if the termination rules are met, then go to step 6); otherwise, part of the optimal solutions, that is, part of the candidate objects, is sent to the innermost layer of the membrane through the communication rules In, continue step 3), complete the next-generation optimization search;

6) When the P system optimization algorithm inspired by the mutual force of charged ions in the membrane runs to the termination criterion of the algorithm, the optimal value obtained is used as the estimated value of the unknown parameters of the fluid catalytic cracking process model, and the estimated value is substituted into the fluid catalytic cracking process model In, a mathematical model of the fluid catalytic cracking process is formed.

The process flow of the optimal modeling method for the FCC reaction regeneration process of the micro-charge mutual force P system in the membrane is that the nested membrane structure starts from the innermost membrane, and the individual sets in it are updated and communicated according to the rules. The selection rules, exchange rules, adaptive mutation rules, micro-charge interaction rules, and termination rules inspired by the recombination and variation of intracellular matter and energy in biological cells and the interaction of charged ions in the membrane are:

1) Selection rules

According to the description of the physical characteristics of the cell membrane, the cell membrane is a jujube cake model of lecithin bilayer and protein mosaic. The lecithin bilayer has fluidity and can selectively allow some substances needed by life to be transported through active transport, cooperative transport and free transport. In and out of the cell membrane; the selection rule is defined as an object with a small objective function value as a candidate communication object; all objects are a combination of feasible solutions in the search space, the type of current microcharge and its objective function value;

2) Communication rules

The communication rules are expressed as: if the current membrane is not the outermost membrane, the current membrane will send some of its candidate objects to its adjacent outer membrane; if the current membrane is the outermost membrane, judge whether the algorithm satisfies the termination rule, and if , then output the best individual in the outermost membrane directly; otherwise, send the best individual in the outermost membrane to the innermost membrane to replace the equivalent amount in the innermost membrane worst individual. The scale of communication = ceil (the number of individuals in the current membrane * communication probability), and the ceil() function is rounded up.

3) Adaptive mutation rules

In the P-system optimization algorithm inspired by the mutual force of charged ions in the membrane, individuals can generate new individuals through the local mutation of monomers, which guides the micro-charge mutual force optimization algorithm to jump out of the local minimum, and proposes the following sudden mutation rules:

p = rand(m, n) ζ = rand v = 1, 2, ..., nl = 1, 2, ..., m

In formula (8), p is a randomly generated m×n matrix, ζ is a random number between [0, 1], m is the set number of layers of the nested membrane structure, and n is the set object in each layer of membrane number, p _m respectively represent the probability of sudden mutation; the parameters in formula (9) are set as follows: a _p =0.01, b _p =0.29, c _p =15/G, G ₀ =G/2, G is micro The maximum operating algebra set by the charge mutual force P system optimization algorithm, g is the current operating algebra;

4) Micro-charge mutual force rule

From the point of view of cell biology, the cell membrane can recognize various types of ionic substances and actively transport these ions in a targeted manner (the continuous transfer from low concentration to high concentration of some ions, which is different from self-diffusion). For example, the actual measurement proves that there are huge differences in the concentrations of Na ⁺ , K ⁺ , Ca ²⁺ , and Cl ^- ions inside and outside the cell membrane. Taking potassium ion K ⁺ and sodium ion Na+ as examples, the cell membrane is moving from the inside to the membrane of K ⁺ After the end of the pumping process of Na ⁺ from the inside of the membrane to the outside of the membrane, it is bound to cause a slight change in the environment of other ions outside the membrane. There is a certain interaction between these charged ions and the existing charged particles inside and outside the membrane. According to the concept that the same kind of charges repel each other and different kinds of charges attract each other, the micro-charge mutual force rule is expressed as:

a. Randomly assign a charge "+" or "-" to each object at the beginning of the algorithm, L=1.

b. When all the individuals in the current membrane have evolved, assuming that the number of excellent individuals obtained through the communication rules is C (C>=1), then for these C excellent individuals, sort them in order according to their respective fitness values, and rank the Lth excellent individual integrated into its outer membrane;

c. After the individuals in the outer membrane are sorted by fitness value, keep the original K=round(n*P _ch ) individuals (generally, the value of P _ch will eventually make the number of K less, for example, 3) , starting from k=K+1 individuals: calculate the Euclidean space distance between the current K+1th individual and the individual entering the current membrane If the polarity of the current charge of the current K+1th individual is the same as that of the charge entering the current membrane, according to the principle of mutual repulsion of the same charges, the current K+1th individual is carried out according to the variable degree of freedom scan, j=1;

d. If x _{c1, j} < x _{K+1, j} , then otherwise

e. If the polarity of the current charge of the current K+1th individual is different from the polarity of the charge entering the current membrane, according to the principle of mutual attraction of different charges, proceed as follows:

f. If x _{c1, j} < x _{K+1, j} , then otherwise

g. If the other dimensions of the current individual have not completed the traversal, then j=j+1, return to step d; otherwise, the charged attribute of the current K+1 individual is reversed (this strategy prevents the better initial candidate solution from being replaced by the constant current optimal solution is always repelling or always attracting aggregation), and k=k+1 at the same time; if k exceeds the maximum index of the number of individuals in a single layer, go directly to the next step, otherwise return to step c; until the K+1th individual to the nth individual have been updated;

h.L=L+1, and delete the individual with the worst fitness value in the membrane (to compensate for the increase in the number of individuals caused by the entry of the L individual into the membrane, to ensure that the number of individuals in the single-layer membrane is constant). If L exceeds the maximum index of C outstanding individuals, go directly to the next step, otherwise return to step b;

i. end.

It refers to: 1) When two comparison individuals have the same charge, and the repulsive force under the distance between them makes a certain variable of the object to be updated execute During operation, it is possible to make the updated object exceed the search boundary of the current variable, which needs to be detected in practice, and adopts out-of-bounds equidistant rebound (possibly multiple rebounds) to ensure that the updated individual variables in each dimension do not cross the boundary; 2 ) algorithm allows, for each Both evaluate the fitness value of new individuals, and the elite retention strategy ensures that only valid updates are retained.

The algorithm under the rule of micro-charge mutual force will redistribute the individual sets randomly in the entire search space according to the charging properties between objects and the principle of charge mutual force;

5) Termination rules

The termination rule is expressed as the algorithm reaches the maximum running algebra or satisfies the following formula:

In the formula, f ^* and f ^best respectively denote the currently found optimal solution and the global optimal solution of the optimization problem, and ε=1×10 ^-4 is an acceptable error.

The invention simulates the recombination and variation of intracellular substances and energy in biological cells, and the impact of charged particles entering the cell membrane on other charged ions, and proposes that the same kind of charges between charged ions and other ions in the membrane repel each other, and different kinds of charges attract each other , the attenuation characteristics of the force between charged ions with the space distance, etc., extract the corresponding mechanism and propose a new P system optimization algorithm inspired by it, in which the mutual repulsion effect of the same charge will disperse the repulsion force of a large number of individuals gathered around the optimal individual , which increases the diversity of the population in the optimization process, and the mutual attraction effect of heterogeneous charges uses the optimal individual to guide other individuals. Through elite retention and inheritance of excellent experience, the algorithm has both good global search ability and Better local search ability, better algorithm convergence speed and precision.

Description of drawings

Figure 1 is a schematic diagram of a P system with a nested membrane structure;

Fig. 2 is the schematic diagram of the reaction of the fluidized catalytic cracking process;

Figure 3 FCC reaction-regeneration input-output relationship matrix

Fig.4 The temperature response diagram of No. 1 regenerator under the step disturbance of the manipulated variable MV6

Fig.5 Temperature response graph of No. 1 regenerator under common step disturbance

Figure 6 Comparison of temperature modeling results of No. 1 regenerator by different methods

Detailed ways

The optimal modeling method for the FCC reaction-regeneration process of the micro-charge mutual force P system in the membrane is used to accurately model the same or similar process, including the following steps:

1) After the fluid catalytic cracking reaction process is in a steady state, the 6 operating variables of the fluid catalytic cracking process are given a disturbance of 5% to 10% by field operation (or the original FCC process is in a certain state after the static steady state A new setting is given to each of its operating variables at a time), and the temperature CV ₁ of the No. 1 reactor, the temperature CV ₂ of the No. 2 reactor, and the reaction heat CV ₃ of the riser are respectively regularly sampled and measured, and 3 batches are obtained. sample data.

2) Draw the mixed time-domain response curve of each output variable with the time axis as the horizontal axis by the method of drawing points, compare the observed curve with the known general mathematical model structure, and reconfirm the structure of the model and the parameters with optimization in the model Number and parameter definition domain.

3) For the input sampling data of MV ₁ ~ MV ₆ of the fluid catalytic cracking process with the same output variable, the square of the error between the estimated output variable response of the fluid catalytic cracking process model and the actual sampling output data of the fluid catalytic cracking process and as the objective function;

4) According to the recombination and variation of intracellular substances and energy in biological cells and the impact mechanism of charged ions entering the cell membrane on other charged ions, abstract the micro-charge interaction P system optimization algorithm in the membrane; affected by the functions of various organelles in the cell membrane Inspired by this, the rules of the new P system optimization method are proposed as follows: selection rules, exchange rules, self-adaptive mutation rules, micro-charge mutual force rules, and termination rules;

5) Set the initial environment and parameters of the micro-charge mutual force P system optimization algorithm: topology structure, the evolution algebra G of the P system optimization algorithm operation = 500, and the probability of sudden mutation is AC probability p _c = 0.2, mutual force factor K _d = 0.2, p _a = 0.8, size of acceptable error ε = 1×10 ^-4 and termination rules of autophagic membrane calculation method, where a _p = 0.001, b _p =0.099, c _p =15/G, G ₀ =G/2, g is the current running algebra;

6) Run the proposed P system optimization algorithm inspired by the micro-charge interaction in the membrane to estimate 36 unknown parameters in the model of the temperature CV ₁ input to the No. 1 regenerator in the fluid catalytic cracking reaction-regeneration process: Among them, The dynamic data of the actual response process comes from on-site experimental measurements. After the original process has entered the steady state S ₁ , step disturbances are applied to the six input variables at the same time from the specified time t ₁ = 0, and the series of temperature values of the FCC process are recorded at the same time. Sampling The period is consistent, and all data sampling is synchronized; by minimizing the objective function, the estimated value of the unknown parameters in the FCC process model is obtained, and the estimated value is substituted into the FCC process model; various inputs of the FCC process are formed to the mathematical model of the first output variable;

7) Change the output variable, and form mathematical models of other output variables in the same way; comprehensively obtain the overall model with multiple inputs and multiple outputs.

Implementation example

The 1.4 million tons of heavy oil FCC reaction-regeneration unit of a refinery is shown in Figure 2. It can be seen from the figure that the feed oil is first mixed with the remaining oil, and then mixed with the circulating oil. After adding the regenerated catalyst, it enters the bottom of the riser and starts to circulate. In the riser, under the action of the catalyst, the crude oil is cracked into a dense phase mixture such as hydrocarbon vapor and carbon. At the same time, the unremoved sulfur, nitrogen, carbon, nickel, vanadium and other elements or impurities in the crude oil cause the catalyst to be poisoned and deactivated. In the separation tower, dilute-phase gases such as hydrocarbons rise and are sent to fractionation devices to obtain gasoline, diesel, gas and other products, while spent catalysts containing a large amount of coke settle to the bottom of the separation tower by gravity and flow out to 1# regenerator. With the introduction of pressurized air, a large amount of coke in the dense phase fluid will be continuously ignited and burned to generate flue gas such as CO and CO ₂ , and the remaining spent catalyst containing a small amount of coke will be discharged from the bottom of the 1# regenerator , is transported to the 2# regenerator again, where a small amount of coke is completely burned and becomes gas to escape, and then the regenerated catalyst is obtained. The heat generated by combustion during catalyst regeneration provides the required temperature environment for the cracking reaction in the riser. The regenerated catalyst is mixed with residual oil and circulating oil and then enters the riser to enter the next cycle of circulation.

In Fig. 2, the combustion heat of 1# regenerator and 2# regenerator must be kept strictly within a relatively small fluctuation range, which requires the actual process to be relatively stable. Otherwise, insufficient coke combustion will not give enough temperature to the pyrolysis reaction in the riser, and too intense coke combustion may disrupt the normal cracking process and even cause a fire accident. From the analysis, it can be seen that the factors affecting the cracking process are: the actual temperature of the 1# regenerator, the actual temperature of the 2# regenerator and the reaction heat of the riser. The controllable factors are: feed preheating temperature, circulating oil flow, residual oil flow, feed flow, crude oil supply flow to riser and riser outlet oil temperature. For the whole process, define the former as system control variables, expressed by CV ₁ , CV ₂ and CV ₃ ; define the latter as system operation variables, and use MV ₁ , MV ₂ , MV ₃ , MV ₄ , MV ₅ and MV ₆ , The input and output diagram of the whole system can be shown in the following matrix diagram as shown in Figure 3.

The commonly used variation ranges for selecting the system control variables and operating variables in the reaction regenerator are as follows:

Table 1 FCC reaction-operated variables and control variables in the regenerator

Based on the above input and output diagrams of the fluidized catalytic process, combined with the actual reaction process, the system is analyzed first, and then the process model is established in three steps. Analysis system: 1) The bed temperature of 1# regenerator and 2# regenerator must be strictly kept within a very small range in the actual site; 2) It is suggested that the online gas composition and concentration measurement is not reliable, and 1# regeneration The concentration of CO ₂ /CO in the regenerator and the concentration of O ₂ in the 2# regenerator are all in the dynamic process, and the equal volume of gas is sampled regularly and measured offline accurately;

3) The heat of reaction in the riser is directly related to the cracking degree of the feed oil, and the dynamic cracking degree cannot be expressed in the static dilute phase components, so the reaction heat of the riser is directly used to soft measure the actual riser The degree of cleavage in . For each test, the selected step size should be as reasonable as possible in order to ensure that the response of the process output can be clearly observed, and to obtain the process model by identifying the entire step test data set for a parametric model.

first step:

a) When the entire system is in a steady state, the set value is normalized (take MV ₆ as an example: since MV ₆ is the set value of the outlet temperature of the riser, and its range is 507-517°C, the normalized Afterwards, its normal range is 0.98~1), when the MV ₆ set value is increased by 3.33% of the current set value and the rest of the MV _k remain unchanged, observe the actual temperature response curve of the No. 1 regenerator, and the MV ₆ → Under the input step disturbance of CV ₁ , the recorded value of the temperature CV ₁ of No. 1 regenerator, the sampling period is 1min, and the whole process is recorded for a total time of 90min. The whole data is shown in Table 2:

Table 2 Recording table of temperature detection data of No. 1 regenerator under riser temperature step disturbance

The record table forms a graph according to the time axis, as shown in Figure 4.

According to the disturbance response results of CV ₁ , the process of MV ₆ → CV ₁ can be roughly regarded as a second-order system with damping, so it can be assumed that the model structure of MV ₆ → CV ₁ is as follows:

In view of timing sampling, the model structure of transforming the above formula into a discrete form through z-transformation is as follows:

Since the step disturbance of MV ₆ is positively rising, and the response of CV ₁ is decaying and falling, C ₆₁ is negative.

b) Repeat the operation of sub-step a) in the first step, but change the operating variable from MV ₆ to MV ₅ , and obtain new MV _{i, 5 model structures of 1-5} →CV ₁ through monitoring 5 times;

c) Repeat the operation of sub-steps a) and b) in the first step, but change the control variable from CV ₁ to CV ₂ , and obtain 6 new model structures of MV _i →CV ₂ through 6 monitoring times;

d) Repeat the operation of sub-steps a) and b) in the first step, but change the control variable from CV ₁ to CV ₃ , and obtain 6 new model structures of MV _i →CV ₃ through 6 monitoring times;

Step two:

a) The system returns to the original steady state, and at the same time give new set values to the 6 process operating variables (the disturbance amplitudes can be completely different from each other within the safe range, but they must be recorded), and observe the superposition and coupling output of MV-CV ₁ at the same time, Record the response output data of the entire process of CV ₁ ; taking CV ₁ as an example, the initial state can be selected as the steady state of the cold system, and a certain range of disturbance is added to each of MV ₁ to MV ₆ , and the disturbance range of MV ₁ to MV ₆ is tentatively determined are equal, and since each link of MV _i -CV ₁ includes time lag and inertia, in this centralized test, the sampling period is adjusted to 6min, and the total test time is 4h (240min). Under the disturbance of 6 manipulated variables, CV The recorded temperature values of ₁ are shown in Table 3.

Table 3 The temperature detection data recording table of No. 1 regenerator under the common step disturbance of 6 operating variables

The record table draws a graph according to time as shown in Figure 5.

b) Repeat the operation of sub-step a) in the second step, change the control variable from CV ₁ to CV ₂ , apply step disturbance, observe the superposition and coupling output of MV-CV ₂ at the same time, and record the various parameters of the whole process of CV ₂ All kinds of input and comprehensive response output data are ready for use;

c) Repeat the operation of sub-step b) in the second step, change the control variable from CV ₂ to CV ₃ , apply a step disturbance, observe the superposition and coupling output of MV-CV ₃ at the same time, and record the various parameters of the whole process of CV ₃ All kinds of input and comprehensive response output data are ready for use;

third step:

In view of the mutual coupling between input and output variables, one-time MV ₁ ~ MV ₆ to CV ₁ , CV ₂ , and CV ₃ models are established for CV ₁ , CV ₂ and CV ₃ respectively. It is worth noting that refer to the first step The model structure in is the basis for the number of parameters to be optimized.

The optimization modeling method of fluid catalytic cracking process based on the interaction force of charged ions in the membrane is as follows:

a) First perform batch model parameter estimation for the 6 processes of MV-CV ₁ , and obtain 40 sets of input and output sampling data of the actual process (as shown in Table 1) through experiments, and use them as training samples for parameter estimation, and optimize the index function as where CV _{1, k} is the output data of the model, and the calculation formula is Output sampled values for processes at the same MV. This optimization index is used as the objective function during the optimization search of the fluid catalytic cracking process optimization modeling method inspired by the mutual force of charged ions in the membrane;

b) Set the operating environment for program operation, wherein the number of layers of the nested membrane structure of the P system is m=5, the number of object sets in each membrane is n=10, the cycle number G=500, and the adaptive mutation probability is a _p =0.01, b _p =0.29, c _p =15/G, G ₀ =G/2, interaction factor K _d =0.2, size of acceptable error ε=1×10 ^-4 and termination rule, where g is the current running algebra; the termination criterion of the algorithm is as described in the previous section, and it only needs to meet one of the two conditions;

c) Run the P system optimization algorithm inspired by the interaction force of charged ions in the membrane to estimate the unknown parameters in the fluid catalytic cracking process model, and determine the best fit for the actual sampling data for the first sub-process MV _i →CV ₁ 36 parameters, complete the model of 6 operating variables to the temperature of No. 1 regenerator, as shown in Table 4.

The modeling results have been compared with the well-known genetic algorithm. The genetic algorithm directly adopts the ga toolbox in the international standard software Matlab, and sets the matching interface. The search range, initial population size, and cycle algebra settings are consistent, and the default settings are used for others. The comparison of the fitting results of different algorithms is shown in Fig. 6.

Table 4 Model parameters of the 6 operating variables to the temperature of No. 1 regenerator

the fourth step:

Similarly, by loading the sampling point of MV-CV ₂ , the mathematical model of the temperature of the No. 2 regenerator in the FCC reaction-regeneration process can be obtained;

the fifth step:

In the same way, by loading the sampling point of MV-CV ₃ , the mathematical model of the temperature of the No. 2 regenerator in the FCC reaction-regeneration process can be obtained;

It can be seen from Fig. 6 that the temperature mathematical model prediction of the No. 1 regenerator obtained based on the proposed optimal modeling method of the micro-charge mutual force P system FCC reaction regeneration process in the membrane and the sampling temperature series of the actual regenerator The sum of squared errors is the smallest, and the accuracy of prediction is higher than other methods.

Claims (6)

1. micro- mutual power P system FCC reaction regenerations process optimum modeling method of charge, feature comprise the following steps in a kind of film：

1) the crucial sampled data that refinery fluid catalytic cracking process is obtained by execute-in-place or experiment, for each group The various inputs of reaction-regenerator predict 3 kinds of reaction-regenerator main outputs in conjunction with initial imprecise computational, by its with it is anti- Answer-regenerator under equivalent input variable actual monitoring output error sum of squares as object function；

2) according to cell membrane, to passing in and out, the transhipment effect of cytotrophy substance, having been enter into charged ion in film, (like charges are mutual Repel, xenogenesis charge attracts each other) interaction, take out efficient oil fluid catalytic cracking reaction-regenerative process Optimization Modeling method；It is inspired by charged ion interaction in cell membrane, proposes to be inspired by the mutual power of charged ion in film as follows P system optimization method rule：The mutual power rule of selection rule, transhipment rule, micro- charge, dynamic variation rule, termination rules (wherein micro- mutual power rule of charge is this patent core to be protected)；

3) initial environment and parameter of P system optimization algorithm are set：Topological structure, P system operation cycle-index G=500, dynamic State mutation probability isExchange Probability p_c=0.2, Probability p is transported_sc=0.2, superspace is apart from the factor K_d=0.2, size ε=1 × 10 of acceptable error^-4And termination rules, wherein a_p=0.01, b_p=0.29, c_p=15/G, G₀ =G/2, g are current operation algebraically；

4) it runs in the mutual power P system optimization algorithm estimation fluid catalytic cracking of micro- charge in catalyst reaction-regenerative process model Unknown parameter：Wherein 6 groups of input variables exist with 3 groups of output variables intersects transitive relation, the z changing patterns of transmission function Type is represented by such as drag：

As it can be seen that each subprocess MV_i(z)→CV_j(z), as long as determining parameter a_ij(1)、a_ij(2)、a_ij(3)、b_ij(1)、b_ij (2) and d_ijThe mathematical model of (i=1,2 ..., 6), FCC reactions-regenerative process can describe the process characteristic of institute's research object. Each control variable is up to 36 parameters and needs to recognize.；

5) it is the precision of prediction for ensureing model, it would be desirable to select one group of optimal parameter, under this parameter group, a certain mesh for model Offer of tender numerical value can minimize (or maximization), by minimizing following object function：

In formula, f_jFor the SSE of j-th of control variable；Ns is to estimate each CV_jUsed sampled point number when parameter；CV_{J, k}For K-th of model prediction output being calculated according to formula (1)；For with CV_{J, k}It is true under corresponding same given operation Real process sampled value.The estimated value of unknown parameter in FCC reactions-regenerative process model is obtained, estimated value is substituted into FCC reactions- In regenerative process model, formed can three of high-precision forecast fluid catalytic cracking reaction-regenerative process crucial output quantities number Learn model.

2. micro- mutual power P system FCC reaction regenerations process optimum modeling method of charge in a film according to claim 1, It is characterized in that the individual update rule of the P system optimization algorithm inspired by charged ion in film is：Micro- charge force is similarly hereinafter Kind of charge is mutually exclusive, and xenogenesis charge is attracted each other and superspace distance draws that repulsion is non-linear to be formed by individual update to be optimized Rule.

3. micro- mutual power P system FCC reaction regenerations process optimum modeling method of charge in film according to claim 1, special Sign is that the fluidized catalytic process is in petroleum catalytic cracking reaction now, and using powdery catalyst converter, and catalytic cracking is given birth to Production device makes catalyst in reacting and regenerating two equipment in " fluidisation recurrent state ", and actual process and device refer to Yang Et al. S.H. it is entitled to be published in U.S.'s periodical《Chemical Engineering Science》Upper volume 51, phase 11, page number 2977- 2982 technical process and device¹, under technique same case, actual device can have subtle difference.

4. micro- mutual power P system FCC reaction regenerations process optimum modeling method of charge in a kind of film according to claim 1, It is characterized in that the film topological structure of the P system optimization algorithm is a nested shape membrane structure, skim-coat film and innermost layer film Outside, other films all close on inner layer film comprising it and include by its outer membrane simultaneously.

5. micro- mutual power P system FCC reaction regenerations process optimum modeling method of charge in a kind of film according to claim 1, It is characterized in that the step of P system optimization algorithm estimates the unknown parameter in fluid catalytic cracking process model For：

1) initial environment and each Key output value model group of fluid catalytic cracking model of the P system optimization method are set In 36 (maximum 36) parameters to be estimated search range, randomly generate object set；

2) fluid catalytic cracking for being derived from reaction regeneration device is reacted in feeding preheating temperature MV₁, recycle oil preset flow MV₂, it is remaining Oily preset flow MV₃, charge raw material oil preset flow MV₄, feedstock oil be supplied to riser preset flow MV₅, leg outlet Preset temperature MV₆Under No. 1 regenerator actual temperature value CV₁, No. 2 regenerators actual temperature CV₂, riser reaction heat CV₃ The model predication value of these three variables 3 output variable sampled values corresponding under 6 performance variables of practical FCCU processes Error sum of squares is as object function；

3) the P system optimization algorithm with nested membrane structure is calculated since innermost layer film, in all films outside skim-coat film The object in portion region executes the mutual power rule operation of selection rule, exchange rule, TSP question rule, micro- charge successively；

4) it is entered step 5) if all objects update of each intra-membrane area finishes, otherwise return to step 3)；

5) object executes TSP question rule, selection rule operation in the film of surface layer；If meeting termination rules, it is transferred to step 6)；Otherwise by its suboptimal solution, i.e. part candidate is sent to by exchanging rule in innermost layer film, continues step 3), Complete follow-on optimizing search；

6) the P system optimization algorithm operation that the mutual power of charged ion inspires in by film reaches the stop criterion of algorithm, and gained is optimal It is worth the estimated value as fluid catalytic cracking process unknown-model parameter, estimated value is substituted into fluid catalytic cracking process model In, form the mathematical model of fluid catalytic cracking process.

6. a kind of P system FCCU process modeling approach inspired by charged ion in film according to claim 1, feature Be it is described select the mutual power rule of rule, TSP question rule, micro- charge, exchange rule, termination rules for：

1) selection rule

It is described according to the physical characteristic of cell membrane, cell membrane is the jujube cake model of lecithin lipid bilayer and protein mosaic, ovum Phospholipid bilayer have mobility, can selectively allow substance needed for the life of part by active transport, collaboration transport and Freely transport disengaging cell membrane；Select the regular object for being defined as possessing Small object functional value will be as candidate communicatee；Institute It is the combination of feasible solution in search space, current the type with micro- charge and its target function value to have object；

2) exchange rule

Exchange rule is expressed as：If working as the non-outermost tunic of cephacoria, face when part of it candidate target is sent to it by cephacoria In close outer membrane；If when cephacoria is outermost tunic, judge whether algorithm meets termination rules, if it is satisfied, then directly will Optimum individual output in the outermost tunic；Otherwise, then the optimal individual in current outermost tunic is sent to innermost layer film Interior equivalent replaces the worst individual of equivalent in innermost layer film.Scale=ceil of exchange is (when quantity * exchanges individual in cephacoria Probability), ceil () function is to round up.

3) TSP question rule

In the P system optimization algorithm that charged ion inspires in by film, individual can generate new individual by monomer local variations, It guides the mutual power optimization algorithm of micro- charge to jump out local minimum, proposes following burst variation rule：

P=rand (m, n) ζ=rand v=1,2 ..., n l=1,2 ..., m

In formula (3), p is to randomly generate m * n matrix, ζ &#91;0,1&#93;Between random number, m is the setting number of plies of nested membrane structure, N is the setting object number in every tunic, p_mBurst mutation probability is indicated respectively；Parameter in formula (4) is respectively set as follows： a_p=0.01, b_p=0.29, c_p=15/G, G₀=G/2, G are the maximum operation set by the mutual power P system optimization algorithm of micro- charge Algebraically, g are current operation algebraically；

4) the mutual power rule of micro- charge

According to cell biology viewpoint, cell membrane can identify the type of various ionic species, at the same have be directed to these from Son carries out active transport (lasting transfer of the low concentration containing part ion to high concentration, this is different from own diffusion).Such as it is real Surveying proves：Na⁺、K⁺、Ca²⁺、Cl^-There are huge differences for concentration of the plasma inside and outside cell membrane, with potassium ion K⁺And sodium ion Na⁺For, cell membrane is carried out out of film to K+ to outside film or to Na⁺It carries out in film to after the pumping procedure outside film, gesture It must cause the micro variation of other ionic environments outside film inner membrance.There are one with existing charged particle inside and outside film for these charged ions Fixed interaction, the concept that, xenogenesis charge mutually exclusive according to like charges is attracted each other, micro- mutual power rule statement of charge For：

A. a kind of charge "+" or "-", L=1 are assigned at random to each object in the algorithm incipient stage.

B. when in cephacoria individual evolve and finish, it is assumed that by exchanging Rule excellent individual number be C (>=1 C), then needle It to this C excellent individual, sorts successively according to respective fitness value, l-th excellent individual is dissolved into its outer membrane；

C. individual retains its original K=round (n*P after fitness value sorts in its outer membrane_ch) individual is (generally P_chValue finally make K number less, for example, 3), since k=K+1 individual：Calculate it is current the K+1 individual with into Enter to the Euclidean space distance when the individual in cephacoriaIf current K+1 Individual current electrically charged polarity with enter when the charge polarity in cephacoria is identical, according to the mutually exclusive original of like charges Reason, then for current the K+1 individual be scanned by variable degree of freedom, j=1；

D. if x_{C1, j}< x_{K+1, j}, thenOtherwise

E. if the current current electrically charged polarity of the K+1 individual with enter when the charge polarity in cephacoria is different, root It attracts each other principle, then proceeds as follows according to xenogenesis charge：

F. if x_{C1, j}< x_{K+1, j}, thenOtherwise

G. if the other dimensions of current individual do not complete traversal, j=j+1, return to step d；Otherwise by the band of current K+1 individuals Electrical properties negate that (strategy prevents preferable initial candidate solution from being repelled always or being attracted always poly- by constant current optimal solution Collection), while k=k+1；If k beyond individual amount largest index in monofilm, is directly entered in next step, otherwise return to step c；Until update is completed to n-th of individual in the K+1 individual；

H.L=L+1, while deleting the worst individual of the film endoadaptation angle value and (compensating l-th individual and enter in film caused The case where body number increases ensures that individual amount is constant in monofilm).If L exceeds C excellent individual quantity largest index, directly It taps into next step, otherwise return to step b；

I. terminate.

What is referred to is pointed out that：1) when the individual equal band like charges of two comparisons, and therebetween the repulsion under make it is to be updated right Some variable of elephant executesWhen operation, it is possible to so that updated object is more than current variable Boundary is searched for, needs to detect in practice, ensures that its updated individual is every using equidistant rebound (may repeatedly spring back) is crossed the border One-dimensional variable does not cross the border；2) in the case of algorithm allows, for eachEvaluation new individual is suitable Angle value, elite retention strategy is answered to ensure only to retain effectively update.

Algorithm under micro- mutual power rule of charge by according between object band electrical properties and the mutual power principle of charge make individual collection whole A search space random distribution again；

5) termination rules

Termination rules are expressed as algorithm and reach maximum operation algebraically or meet following formula：

F in formula^*And f^bestThe globally optimal solution of the optimal solution and optimization problem that have currently found, ε=1 × 10 are indicated respectively^-4For Acceptable error.