CN101587562B - Complex chemical process modeling method having astroid topologic structural film calculation - Google Patents

Abstract

The invention discloses a complex chemical process modeling method having astroid topologic structural film calculation, comprising steps of: 1) acquiring input and output sampling data of chemical process via on-site operation or experiment; for input sampling data of same group of chemical process, using error absolute value sum of estimated output chemical process mould and actual sampling output of chemical process as a target function; 2) setting control parameters having astroid topologic structural film calculation; 3) operating the astroid topologic structural film calculation to estimate unknown parameters in the chemical process mould to acquire estimation value of unknown parameters in the chemical process mould via minimal target function; introducing the estimation value of unknown parameters to the chemical process mould to form a mathematic mould of chemical process. The invention uses astroid topologic structure, which protects object with smaller function value in optimizing process having high precision and rapid speed of optimization. Diversity of group is added due to rule of conversion.

Description

A kind of complicated chemical procedure modelling approach with the calculating of star topology film

Technical field

The present invention relates to modeling method, relate in particular to a kind of complicated chemical procedure modelling approach that the star topology film calculates that has.

Background technology

Because complicated day by day and a lot of chemical industry objects of chemical process system have the severe nonlinear characteristic, so modeling chemical engineering processes is the research difficult point and the focus of chemical field always, are the industrial essential links of modern chemistry.The modern chemical industry process control is all based on its corresponding mathematical modeling, and the mathematical model that is to say chemical process is the core of control.Setting up mathematical model accurately has great importance to the theoretical research and the practical application of chemical process.

Traditionally, the foundation of the mathematical model of chemical plant installations mainly contains two big class methods, and a class is the mechanism model method, and another kind of is system identifying method.The mechanism model method need rely on reliable rule and experimental knowledge to set up the original differential equation, and these rules and experimental knowledge must be general form by table.This method exists very big defective and limitation, more and more is not suitable for modern large-scale chemical process system.To be the researcher proposed corresponding chemical process model structure according to the reaction mechanism of chemical process to system identifying method, selects suitable experimental technique, asks for model parameter, calculates its optimum estimated value, the correctness of last verification model.Existing in the chemical process model of the structure of these known models much can't be by the parameter that directly measures, need the operation parameter method of estimation to estimate to obtain the value of these unknown parameters, and, obtain the mathematical model of chemical process with this with in the corresponding chemical process model of the estimated value substitution of these parameters.The estimated value of these unknown parameters has material impact to the accuracy of the chemical process mathematical model that obtains.Thereby, the parameter estimation problem is converted into optimization problem, a lot of traditional optimized Algorithm all are used to estimate the parameter of complicated chemical process.

The origin of optimum theory and method can be traced back to the age that infinitesimal analysis produces, then up to the thirties in 20th century, because pressing for of aspects such as military and commercial production just makes the research of optimization method obtain vigorous growth.Optimization method commonly used mainly contains: analytical method, enumerative technique and random search.Analytical method is absorbed in local optimum easily, and requires objective function and constrained domain to represent with analytic expression, be difficult to use in and find the solution that objective function is discontinuous, constrained domain is not communicated with, objective function be difficult to analytic expression express, solution space has situations such as multimodal characteristic.Enumerative technique has simple characteristics, but it need calculate the value of each point in the search volume, so inefficiency, bad adaptability.In fact, the search volume of many optimization problems is all very big, does not allow to search for bit by bit.Stochastic search methods then is to note obtained best result at any time by roaming also at random in the search volume, and its efficient is still not high, and only separates when the tight distribution of search volume, just can find optimum solution, and this condition generally is difficult to satisfy.At characteristics such as the complicacy of the optimization problem in the engineering, binding character, non-linear, many local minimum points, modeling difficulties, seek and be suitable for the large-scale parallel search and have intelligentized optimization method having become an important research direction.

In recent years, be subjected to the inspiration of bio-science technical research achievement, rapid based on the optimization method development of biological computation.The researchist uses for reference bionic thought, has proposed some and has had efficient optimizing ability and extensive adaptability intelligent optimization method.For example, artificial neural network (ANN) has been simulated the institutional framework and the function of human brain to a certain extent; Genetic algorithm (GA) has been used for reference the evolutionary mechanism of nature biotechnology " struggle for existence, eugenic bad eliminating, the survival of the fittest "; Ant group algorithm (ACO) then is subjected to ant colony to seek the inspiration of food optimal path ^[5]Tabu search has been simulated the mental process of human mind.These are finding the solution the advantage of having showed them aspect the complicated optimum problem based on the optimization method of biological computation.Wherein conventional genetic algorithm (SGA) is wide as a kind of adaptation, and the stochastic search methods of strong robustness has stronger ability of searching optimum, be fit to solve this type of problem, yet this method search efficiency is lower, the relatively poor and easy precocity of local search ability.The film that inspired by eucaryotic cell structure and function calculates, and has the feature of powerful parallel and Distributed Calculation, has been used to the modeling of biological medicine, population problem's modeling, and the finding the solution of password and economic dispatch problem.Film calculation optimization method (see figure 1) with star topology is to be subjected to the inspiration of biological cell and computer network architecture and a kind of film optimized Algorithm of proposing, can be used for solving multivariate, nonlinear optimization problem, the parameter that is controlled effectively and chemical industry model.This kind method is used for solving heavy oit pyrolysis three lumped model structural parameters estimates, obtain comparatively ideal effect.

Summary of the invention

The objective of the invention is to overcome the prior art deficiency, a kind of complicated chemical procedure modelling approach that the star topology film calculates that has is provided.

Complicated chemical procedure modelling approach with the calculating of star topology film comprises the steps:

1) passes through the data that execute-in-place or experiment obtain the input and output of chemical process sampling, for the input sampling data of same group of chemical process, the Error Absolute Value sum of the estimation output of chemical process model and the actual samples output data of chemical process is as objective function;

2) set and to have the evolutionary generation G that the star topology film calculates operation, what have that the star topology film calculates has three subsystems and a top layer film, and each subsystem comprises the n tunic, has m object in every tunic, and the rewriting probability is p _m, crossover probability is 1, the transposition probability is p _tAnd exchange the size of scale, the stop criterion of algorithm;

3) operation has the star topology film and calculates the unknown parameter in worker's process model is estimated, by minimizing objective function, obtain the estimated value of unknown parameter in the chemical process model, in the estimated value substitution chemical process model with unknown parameter in the chemical process model, form the mathematical model of chemical process.

The stop criterion of described algorithm is: the operation algebraically of algorithm reaches maximum algebraically.

Described operation has the calculating of star topology film the unknown parameter in the chemical process model is estimated, by minimizing objective function, obtain the estimated value of unknown parameter in the chemical process model, in the estimated value substitution chemical process model with unknown parameter in the chemical process model, the mathematical model step that forms chemical process is:

1) setting has the initial parameter of star topology film calculating and the hunting zone of model parameter, produces object set at random;

2) the Error Absolute Value sum of the actual samples output data of the estimation output of chemical process model and chemical process is as objective function;

3) first tunic with first subsystem that the star topology film calculates begins to calculate, and the object of all film interior zones that the skim-coat film is outer experiences rewriting rule, rule of conversion, intersection regular operation successively;

4), otherwise return step 3) if all objects evolution of three interior each film inner regions of subsystem finish then enter step 5);

5) first subsystem exchanges with second subsystem and obtains the part optimal value, and the 3rd subsystem and part optimal value compare and obtain the final optimization pass value once more, carries out the optimizing search of a generation, repeating step 2) to step 5), finish follow-on optimizing search;

6) calculate the stop criterion that operation reaches algorithm when having the star topology film, the estimation output optimal value of chemical process model is as the estimated value of chemical process model unknown parameter, in the estimated value substitution chemical process model with unknown parameter, form the mathematical model of chemical process.

Described rewriting rule, rule of conversion, interlace operation, interchange rule formula are as follows:

1) rewriting rule

S→S′ifp ₀≤p _m

p ₀＝rand

$\left\{\begin{array}{c}S=x_1{x}_1...x_l\\ S^{\′}=y_1{y}_2....y_l\end{array}\right.$

$\left\{\begin{array}{c}{y}_i=x_i\\ \mathrm{or}{y}_i=x_i+{\mathrm{\η}}_i\end{array}\right.$

i＝1，2，...l

Wherein, l is the number of optimization variable or the length of character string, η _iBe the random variation amount, S represents a strand of evolution object.According to this rule, the Probability p of Chan Shenging at random ₀The Probability p that is less than or equal to setting _mThe time, the character x among the S _iBe converted to y _i=x _i+ η _iThereby, constitute S ', otherwise keep intact;

2) rule of conversion

$\left\{\begin{array}{c}S\&RightArrow;S^{\&prime;}\mathrm{if}{\mathrm{<figure-callout\; id="0"\; label="p"\; filenames="H2009101003078E00022.png,H2009101003078E00031.png"\; state="\{\{state\}\}">p</figure-callout>}}_0\&le;{\mathrm{<figure-callout\; id="0"\; label="p\; t\; p"\; filenames="H2009101003078E00022.png,H2009101003078E00031.png"\; state="\{\{state\}\}">p</figure-callout>}}_t& \\ p_{}{}_0=\mathrm{rand}\end{array}\right.$

$\left\{\begin{array}{c}S=x_1,x_2,...x_i...x_j...x_l\\ S^{\&prime;}=x_1,{\stackrel{\&CenterDot;}{x}}_2,...x_i...x_j...x_l\end{array}\right.$

Wherein, S is the string of picked at random in the film, i, and (i≤j) is the position of character in this character string of selecting at random to j, and l is the number of optimization variable or the length of character string, and S ' is the new character strings object after rule of conversion is carried out.When the Probability p that produces at random ₀The Probability p that is less than or equal to setting _tThe time, this rule will take place in string;

3) intersection rule

This rule is divided into the numerical value intersection and single-point intersects:

$\left\{\begin{array}{c}S=x_1,x_2,...x_i,...x_j...x_l\\ S_r=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...y_i,...y_j...y_l\end{array}\right.$

For S, picked at random character string S _r, and i＜j in the formula.z _i=α x _i+ (1-α) y _i, z _j=α x _j+ (1-α) y _jα ∈ (0,1) wherein.S became S ' after numerical value intersected: S '=x ₁, x ₂... z _i... x _j... y _l.

(S，S _r)→(S′，S _r′)

$\left\{\begin{array}{c}S=x_1,x_2,...u_i,u_{i+1}...x_l\\ S_r=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...u_j,u_{j+1}...y_l\\ S^{\&prime;}=x_1,x_2,...u_i,u_{j+1}...x_l\\ S_r^{\&prime;}=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...u_j,u_{i+1}...y_l\end{array}\right.$

String S, S _rRespectively at u _i, u _I+1And u _j, u _J+1The position cutting is come.Then, fragment is respectively at u _i, u _J+1And u _j, u _I+1Be reassembled into two new character strings;

4) exchange rule

To be each film send to the outside of film with the reasonable character string of its fitness value to this rule, just enter in the upper layer film of film,

[S _min1，...，S _minq] _iS _max1，...S _maxq→[...] _iS _min1，...，S _minq

i＝1，2，...l

In the formula, q is the object number that exchanges, and promptly exchanges the size of scale, S _Max1..., S _MaxqBe q poor character object of i+1 zone endoadaptation degree.S _Min1..., S _MinqBe q the character object that fitness is better than other objects in the one's respective area.

The present invention adopts star topology, has protected the object that has less functional value in the optimizing process, has optimizing precision height, fireballing characteristics, and because the existence of rule of conversion, has increased the diversity of population.

Description of drawings

Fig. 1 is the film computation structure figure with star topology;

Fig. 2 is heavy oil heat three lumped reaction synoptic diagram;

Fig. 3 is the comparison diagram of the output and the estimation model output of ideal model;

Fig. 4 is the comparison diagram based on the output of the ideal model of test data and estimation model output;

Fig. 5 is the comparison diagram of this method and SGA;

Fig. 6 is based on this method of test data and the comparison diagram of SGA.

Embodiment

Complicated chemical procedure modelling approach with the calculating of star topology film comprises the steps:

1) by the input and output sampled data in execute-in-place or the experiment acquisition chemical process, for the input sampling data of same group of chemical process, the Error Absolute Value sum of the estimation output of chemical process model and the output of the actual samples of chemical process is as objective function;

2) set the evolutionary generation G with star topology film calculating operation, have the calculating of star topology film and have three subsystems, each subsystem comprises the n tunic, has m object in every tunic, and the rewriting probability is p _m, crossover probability is 1, the transposition probability is p _tAnd exchange the size of scale, the stop criterion of algorithm;

3) operation has the star topology film and calculates the unknown parameter in the chemical process model is estimated, by minimizing objective function, obtain the estimated value of unknown parameter in the chemical process model, in the estimated value substitution chemical process model with unknown parameter in the chemical process model, form the mathematical model of chemical process.

The stop criterion of described algorithm is: the operation algebraically of algorithm reaches maximum algebraically.

Described operation has the calculating of star topology film the unknown parameter in the chemical process model is estimated, by minimizing objective function, obtain the estimated value of unknown parameter in the chemical process model, in the estimated value substitution chemical process model with unknown parameter in the chemical process model, the mathematical model step that forms chemical process is:

1) setting has the initial parameter of star topology film calculating and the hunting zone of model parameter, produces object set at random;

2) the Error Absolute Value sum of the actual samples output data of the estimation output of chemical process model and chemical process is as objective function;

3) first tunic with first subsystem that the star topology film calculates begins to calculate, and the object of all film interior zones that the skim-coat film is outer experiences rewriting rule, rule of conversion, intersection regular operation successively;

4), otherwise return step 3) if all objects evolution of three interior each film inner regions of subsystem finish then enter step 5);

5) first subsystem exchanges with second subsystem and obtains the part optimal value, and the 3rd subsystem and part optimal value compare and obtain the final optimization pass value once more, carries out the optimizing search of a generation, repeating step 2) to step 5), finish follow-on optimizing search;

6) calculate the stop criterion that operation reaches algorithm when having the star topology film, the estimation output optimal value of chemical process model is as the estimated value of chemical process model unknown parameter, in the estimated value substitution chemical process model with unknown parameter, form the mathematical model of chemical process.

Described rewriting rule, rule of conversion, interlace operation, interchange rule formula are as follows:

1) rewriting rule

S?→S′ifp ₀≤p _m

p ₀＝rand

$\left\{\begin{array}{c}S=x_1{x}_1...x_l\\ S^{\&prime;}=y_1{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2....y_l\end{array}\right.$

$\left\{\begin{array}{c}{y}_i=x_i\\ \mathrm{or}{y}_i=x_i+{\mathrm{\&eta;}}_i\end{array}\right.$

i＝1，2，...l

Wherein, l is the number of optimization variable or the length of character string, η _iBe the random variation amount, S represents a strand of evolution object.According to this rule, the Probability p of Chan Shenging at random ₀The Probability p that is less than or equal to setting _mThe time, the character x among the S _iBe converted to y _i=x _i+ η _iThereby, constitute S ', otherwise keep intact;

2) rule of conversion

$\left\{\begin{array}{c}S\&RightArrow;S^{\&prime;}\mathrm{if}{\mathrm{<figure-callout\; id="0"\; label="p"\; filenames="H2009101003078E00022.png,H2009101003078E00031.png"\; state="\{\{state\}\}">p</figure-callout>}}_0\&le;{\mathrm{<figure-callout\; id="0"\; label="p\; t\; p"\; filenames="H2009101003078E00022.png,H2009101003078E00031.png"\; state="\{\{state\}\}">p</figure-callout>}}_t& \\ p_{}{}_0=\mathrm{rand}\end{array}\right.$

$\left\{\begin{array}{c}S=x_1,x_2,...x_i...x_j...x_l\\ S^{\&prime;}=x_1,x_2,...x_i...x_j...x_l\end{array}\right.$

Wherein, S is the string of picked at random in the film, i, and (i≤j) is the position of character in this character string of selecting at random to j, and l is the number of optimization variable or the length of character string, and S ' is the new character strings object after rule of conversion is carried out.When the Probability p that produces at random ₀The Probability p that is less than or equal to setting _tThe time, this rule will take place in string;

Intersect regular

This rule is divided into the numerical value intersection and single-point intersects:

$\left\{\begin{array}{c}S=x_1,x_2,...x_i,...x_j...x_l\\ S_r=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...y_i,...y_j...y_l\end{array}\right.$

For S, picked at random character string S _r, and i＜j in the formula.z _i=α x _i+ (1-α) y _i, z _j=α x _j+ (1-α) y _jα ∈ (0,1) wherein.S became S ' after numerical value intersected: S '=x ₁, x ₂... z _i... z _j... y _l.

(S，S _r)→(S′，S _r′)

$\left\{\begin{array}{c}S=x_1,x_2,...u_i,u_{i+1}...x_l\\ S_r=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...u_j,u_{j+1}...y_l\\ S^{\&prime;}=x_1,x_2,...u_i,u_{j+1}...x_l\\ S_r^{\&prime;}=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...u_j,u_{i+1}...y_l\end{array}\right.$

String S, S _rRespectively at u _i, u _I+1And u _j, u _J+1The position cutting is come.Then, fragment is respectively at u _i, u _J+1And u _j, u _I+1Be reassembled into two new character strings;

4) exchange rule

To be each film send to the outside of film with the reasonable character string of its fitness value to this rule, just enter in the upper layer film of film,

[S _min1，...，S _minq] _iS _max1，...S _maxq→[...] _iS _min1，...，S _minq

i＝1，2，...l

In the formula, q is the object number that exchanges, and promptly exchanges the size of scale, S _Max1..., S _MaxqBe q poor character object of i+1 zone endoadaptation degree.S _Min1..., S _MinqBe q the character object that fitness is better than other objects in the one's respective area.

Embodiment

Below the inventive method be used for heavy oit pyrolysis three lumped models determine, describe in further detail:

The lumped reaction model is adopted in cracking to heavy oil, can simplify the various reaction system of component, be about to numerous simplification compound in the reactive system, press the similar principle of dynamics, merger is several virtual components, to set up the kinetic model of the lumped reaction of simplifying.(see figure 2)

If x is that heavy oil H is as the summation of raw material generation less than the productive rate of 510 ℃ of all cuts, pyrolysis gas and toluene insolubles; x _wIt is 390 ℃-510 ℃ heavy intermediate distillate yied; x _LLight ends and condensation product productive rate sum for pyrolysis, 210-390 ℃.Suppose:

(1) pyrolytic reaction raw material H be in the residual oil greater than 510 ℃ toluene solvend, middle the heavy end W that the reaction back generates can further be converted into pyrolysis gas, light ends and condensation product L. the while again

(2) secondary reaction that is converted into lump thing L by W is a first order reaction.

Then under isothermy, overall reaction order is 1 o'clock, can derive the reaction yield equation according to reaction mechanism and be:

$x_L\text{=}\frac{C_L}{n_L}[1-{(1-x)}^{n_L}]+\frac{C_W{C}_{\mathrm{WL}}}{n_W-C_{\mathrm{WL}}}\{\frac{1}{C_{\mathrm{WL}}}(1-{(1-x)}^{C_{\mathrm{WL}}})+\frac{1}{n_W}[{(1-x)}^{n_W}-1\left]\right\}$

Consider Temperature Influence, introduce Arrhenius equation: K=K _Oe ^-E/RTTherefore, $K_P=K_{\mathrm{PO}}{e}^{-E_P/\mathrm{RT}},$ $K_L=K_{\mathrm{LO}}{e}^{-E_L/\mathrm{RT}},$ $K_W=K_{\mathrm{WO}}{e}^{-E_W/\mathrm{RT}},$ And $K_{\mathrm{WL}}=K_{\mathrm{WLO}}{e}^{-E_{\mathrm{WL}}/\mathrm{RT}};$ Thereby obtain:

$x_L=\frac{K_{\mathrm{LPO}}{e}^{-E_{\mathrm{LP}}/T}}{n_L}[1-{(1-z)}^{n_L}]$

$+\frac{L_{\mathrm{WPO}}{K}_{\mathrm{WLPO}}{e}^{-(E_{\mathrm{WP}}+E_{\mathrm{WLP}})/T}}{n_W-K_{\mathrm{WLPO}}{e}^{-E_{\mathrm{WLP}}/T}}\{\frac{1-{(1-z)}^{K_{\mathrm{WLPO}}{e}^{-E_{\mathrm{WLP}}/T}}}{K_{\mathrm{WLPO}}{e}^{-E_{\mathrm{WLP}}/T}}+\frac{1}{n_W}[{(1-z)}^{n_W}-1\left]\right\}$

Wherein z and T (temperature of reaction) are independent variable, x _LBe dependent variable.The parameter that needs in the following formula to estimate has K _LPO, K _WPO, K _WLPO, E _LP, E _WP, n _L, n _WTotally 8.

The implication of each variable representative is in the formula:

E: the energy of activation of reaction, J/mol;

E _P: the energy of activation of total thermal conversion reaction, J/mol;

E _L: transient intermediates M generates lump thing L reaction activity, J/mol;

E _W: transient intermediates M generates the energy of activation of cut W reaction, J/mol;

n _W: transient intermediates M generates the cut W order of reaction;

K _P: the rate constant of total thermal conversion reaction;

K _L: transient intermediates M generates the reaction rate constant of lump thing L;

K _W: transient intermediates M generates the reaction rate constant of cut W;

K _WL: W changes into the reaction rate constant of lumped reaction thing L;

K _PO: the frequency factor of total thermal conversion reaction;

K _LO: moment intermediate product M generates the frequency factor of lump thing L reaction;

K _WO: transient intermediates M generates the frequency factor of the reaction of cut W;

K _WLO: cut W changes into the frequency factor of lump thing L reaction;

n _L: transient intermediates M generates the order of reaction of lump thing L;

X: generate≤510 ℃ of cuts, pyrolysis gas and toluene insolubles productive rate sum;

T: temperature of reaction, K;

R: gas law constant, R=8.314J/ (mol K);

x _w: 390 ℃-510 ℃ heavy intermediate distillate yied;

x _L: pyrolysis gas, 210-390 ℃ light ends and condensation product productive rate sum;

Calculate as follows based on film to heavy oil pyrolysis three lumped reaction modeling methods with star topology:

1) obtain actual 56 groups of sampling inputoutput datas by experiment, choose the training sample of 20 groups of data as parameter estimation, the optimization target function is $f=\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}|x_L\left(i\right)-{\hat{x}}_L\left(i\right)|,$ N is a number of training in the formula,

Be the model output valve that substitution estimated parameter value obtains, x _LBe the theoretical model output valve.Objective function when this optimizes index as star topology film calculating optimizing search;

2) the algebraically G of setting program operation, film inside, top layer has three subsystems, comprises the n tunic in each system, has m object in every tunic, and the rewriting probability is p _m, crossover probability is 1, the transposition probability is p _tAnd exchange scale size,, the stop criterion of algorithm promptly runs to till the maximum algebraically;

3) operation has the star topology film and calculates the unknown parameter in heavy oil pyrolysis three lumped models is estimated, by minimizing objective function, obtains unknown parameter K _LPO, K _WPO, K _WLPO, E _LP, E _WP, n _L, n _WEstimated value, in estimated value substitution heavy oil pyrolysis three lumped models with unknown parameter, form the mathematical model of heavy oil pyrolysis.

Described algorithm stop criterion is: the operation algebraically of algorithm reaches maximum algebraically.

Described operation has the unknown parameter K of star topology film calculating to heavy oil pyrolysis three lumped models _LPO, K _WPO, K _WLPO, E _LP, E _WP, n _L, n _WCarry out estimating step:

The film that described operation has star topology calculates the step that the unknown parameter in the heavy oil pyrolysis model is estimated:

Step 1: the initial parameter of set algorithm in this modelling and the hunting zone of model parameter produce object set at random.The algorithm initial parameter has comprised that evolutionary generation is made as 50 generations, three subsystems, the number of objects of each subsystem the inside two membranes, every tunic is 15, the regular probability of variation is 0.05, the regular probability that intersects is 1, the rule of conversion probability is 0.05, participates in the scale 40% that exchanges in each film.The hunting zone of this model parameter is:

K _LPO∈(0，10)，K _WPO∈(0，10)，K _WLPO∈(0，10)，E _LP∈(800，1500)，E _WP∈(1500，4000)，E _WLP∈(1500，4500)，n _L∈(0，5)，n _W∈(0，5)

According to number of parameters to be estimated, determine that decision variable length is 8;

Step 2: the fitness value of calculating object, i.e. the Error Absolute Value sum of exporting in the estimation output of chemical process model and the chemical industry real process: $f=\underset{i=0}{\overset{n-1}{\mathrm{\&Sigma;}}}|x_L\left(i\right)-{\hat{x}}_L\left(i\right)|,$ N is a number of training in the formula,

Be the model output valve that substitution estimated parameter value obtains, x _LBe the theoretical model output valve;

Step 3: first tunic with first subsystem of star topology film calculating begins to calculate, and the object of all film interior zones that the skim-coat film is outer experiences rewriting rule, rule of conversion, intersection regular operation successively;

Step 4:, otherwise return step 3) if all objects evolution of three interior each film inner regions of subsystem finish then enter step 5);

Step 5: first subsystem exchanges with second subsystem and obtains the part optimal value, the 3rd subsystem and part optimal value compare and obtain the final optimization pass value once more, carry out the optimizing search of a generation, repeating step 2) to step 5), finish follow-on optimizing search;

Step 6: when program run reaches end condition, promptly reach the maximum algebraically of setting, the output optimal value in this model of estimated value substitution with unknown parameter, forms the mathematical model of heavy oil pyrolysis as the estimated value of heavy oil pyrolysis three lumped model unknown parameters.

A kind of have a complicated chemical procedure modelling approach that the star topology film calculates, and the rewriting rule that uses in this method, rule of conversion, interlace operation, interchange rule formula are as follows:

1) rewriting rule

S?→S′ifp ₀≤p _m

p ₀＝rand

$\left\{\begin{array}{c}S=x_1{x}_2...x_l\\ S^{\&prime;}=y_1{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2....y_l\end{array}\right.$

$\left\{\begin{array}{c}{y}_i=x_i\\ \mathrm{or}{y}_i=x_i+{\mathrm{\&eta;}}_i\end{array}\right.$

i＝1，2，...l

Wherein, l is the number of optimization variable or the length of character string, η _iBe the random variation amount, S represents a strand of evolution object.According to this rule, the Probability p of Chan Shenging at random ₀The Probability p that is less than or equal to setting _mThe time, the character x among the S _iBe converted to y _i=x _i+ η _iThereby, constitute S ', otherwise keep intact.

2) rule of conversion

$\left\{\begin{array}{c}S\&RightArrow;S^{\&prime;}\mathrm{if}{\mathrm{<figure-callout\; id="0"\; label="p"\; filenames="H2009101003078E00022.png,H2009101003078E00031.png"\; state="\{\{state\}\}">p</figure-callout>}}_0\&le;{\mathrm{<figure-callout\; id="0"\; label="p\; t\; p"\; filenames="H2009101003078E00022.png,H2009101003078E00031.png"\; state="\{\{state\}\}">p</figure-callout>}}_t& \\ p_{}{}_0=\mathrm{rand}\end{array}\right.$

$\left\{\begin{array}{c}S=x_1,x_2,...x_i...x_j...x_l\\ S^{\&prime;}=x_1,x_2,...x_i...x_j...x_l\end{array}\right.$

Wherein, S is the string of picked at random in the film, i, and (i≤j) is the position of character in this character string of selecting at random to j, and l is the number of optimization variable or the length of character string, and S ' is the new character strings object after rule of conversion is carried out.When the Probability p that produces at random ₀The Probability p that is less than or equal to setting _tThe time, this rule will take place in string.

3) intersection rule

This rule is divided into the numerical value intersection and single-point intersects:

$\left\{\begin{array}{c}S=x_1,x_2,...x_i,...x_j...x_l\\ S_r=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...y_i,...y_j...y_l\end{array}\right.$

For S, picked at random character string S _r, and i＜j in the formula.z _i＝αx _i+(1-α)y _i，z _j＝αx _j+(1-α)y _j

α ∈ (0,1) wherein.S became S ' after numerical value intersected: S '=x ₁, x ₂... z _i... z _j... y _l.

(S，S _r)→(S′，S _r′)

$\left\{\begin{array}{c}S=x_1,x_2,...u_i,u_{i+1}...x_l\\ S_r=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...u_j,u_{j+1}...y_l\\ S^{\&prime;}=x_1,x_2,...u_i,u_{j+1}...x_l\\ S_r^{\&prime;}=y_1,{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="H2009101003078E00011.png,H2009101003078E00021.png"\; state="\{\{state\}\}">y</figure-callout>}}_2,...u_j,u_{i+1}...y_l\end{array}\right.$

String S, S _rRespectively at u _i, u _I+1And u _j, u _J+1The position cutting is come.Then, fragment is respectively at u _i, u _J+1And u _j, u _I+1Be reassembled into two new character strings.

4) exchange rule

To be each film send to the outside of film with the reasonable character string of its fitness value to this rule, just enters in the upper layer film of film.

[S _min1，...，S _minq] _iS _max1，...S _maxq→[...] _iS _min1，...，S _minq

i＝1，2，...l

In the formula, q is the object number that exchanges, S _Max1..., S _MaxqBe q poor character object of i+1 zone endoadaptation degree.S _Min1..., S _MinqBe q the character object that fitness is better than other objects in the one's respective area.

According to step 3), the estimates of parameters that obtains heavy oil pyrolysis three lumped reaction models is as follows:

Method	K LP0	K WP0	K WLP0	E LP	E WP	E WLP	n L	n W	f
										This patent method	2.6705	5.6058	6.0352	862.3870	2600.2	2150.7	1.5493	2.9184	0.2872

Above-mentioned estimated parameter substitution heavy oil pyrolysis three lumped reaction models are heavy, obtain corresponding mathematical model.Under identical input data, the comparison diagram of the output of ideal model and estimation model output is seen Fig. 3. in order to verify the validity of estimation model, 56 groups of inputoutput datas are tested.Based on test data, the comparison diagram of the output of ideal model and the output of estimation model is seen Fig. 4. simultaneously the described method of patent of the present invention and SGA have been done at same model that data relatively see the following form and Fig. 5 and Fig. 6, further verified this side's validity.

Method	K LP0	K WP0	K WLP0	E LP	E WP	E WLP	n L	n W	f
										This patent method	2.6705	5.6058	6.0352	862.3870	2600.2	2150.7	1.5493	2.9184	0.2872
SGA	4.680	5.155	4.197	1257	1850	3776	1.191	1.488	1.2116

Claims (3)

1. one kind has the complicated chemical procedure modelling approach that the star topology film calculates, and its feature comprises the steps:

1) passes through the data that execute-in-place or experiment obtain the input and output of chemical process sampling, for the input sampling data of same group of chemical process, the Error Absolute Value sum of the estimation output of chemical process model and the actual samples output data of chemical process is as objective function;

2) set the evolutionary generation G with star topology film calculating operation, have the calculating of star topology film and have three subsystems and a top layer film, each subsystem comprises the n tunic, has m object in every tunic, and the rewriting probability is p _m, crossover probability is 1, the transposition probability is p _tAnd exchange the size of scale, the stop criterion of algorithm;

3) operation has the star topology film and calculates the unknown parameter in the complicated chemical industry process model is estimated, by minimizing objective function, obtain the estimated value of unknown parameter in the chemical process model, in the estimated value substitution chemical process model with unknown parameter in the chemical process model, form the mathematical model of chemical process;

Described operation has the calculating of star topology film the unknown parameter in the chemical process model is estimated, by minimizing objective function, obtain the estimated value of unknown parameter in the chemical process model, in the estimated value substitution chemical process model with unknown parameter in the chemical process model, the mathematical model step that forms chemical process is:

(1) setting has the initial parameter of star topology film calculating and the hunting zone of model parameter, produces object set at random;

(2) the Error Absolute Value sum of the actual samples output data of the estimation output of chemical process model and chemical process is as objective function;

(3) first tunic with first subsystem that the star topology film calculates begins to calculate, and the object of all film interior zones that the skim-coat film is outer experiences rewriting rule, rule of conversion, intersection regular operation successively;

(4), otherwise return step (3) if all objects evolution of three interior each film inner regions of subsystem finish then enter step (5);

(5) first subsystems exchange with second subsystem and obtain the part optimal value, the 3rd subsystem and part optimal value compare and obtain the final optimization pass value once more, carry out the optimizing search of a generation, repeating step (2) is finished follow-on optimizing search to step (5);

(6) calculate the stop criterion that operation reaches algorithm when having the star topology film, the estimation output optimal value of chemical process model is as the estimated value of chemical process model unknown parameter, in the estimated value substitution chemical process model with unknown parameter, form the mathematical model of chemical process.

2. a kind of complicated chemical procedure modelling approach that the star topology film calculates that has according to claim 1, it is characterized in that the stop criterion of described algorithm is: the operation algebraically of algorithm reaches maximum algebraically.

3. a kind of complicated chemical procedure modelling approach that the star topology film calculates that has according to claim 1 is characterized in that described rewriting rule, rule of conversion, interlace operation, interchange rule formula are as follows:

1) rewriting rule

S?→S′if?p ₀≤p _m

p ₀＝rand

$\left\{\begin{array}{c}S=x_1{x}_2...x_l\\ S^{\&prime;}=y_1{y}_2...y_l\end{array}\right.$

$\left\{\begin{array}{c}{y}_i=x_i\\ \mathrm{or}{y}_i=x_i+{\mathrm{\&eta;}}_i\end{array}\right.$

i＝1，2，...l

Wherein, l is the number of optimization variable or the length of character string, η _iBe the random variation amount, S represents a strand of evolution object, according to this rule, and the Probability p of Chan Shenging at random ₀The Probability p that is less than or equal to setting _mThe time, the character x among the S _iBe converted to y _i=x _i+ η _iThereby, constitute S ', otherwise keep intact;

2) rule of conversion

$\left\{\begin{array}{c}S\&RightArrow;S^{\&prime;}\mathrm{if}{p}_0\&le;p_t\\ p_0=\mathrm{rand}\end{array}\right.$

$\left\{\begin{array}{c}S=x_1,x_2,...x_i...x_j...x_l\\ S^{\&prime;}=x_1,x_2,...x_i...x_j...x_l\end{array}\right.$

Wherein, S is the string of picked at random in the film, i, (i≤j) is the position of character in this character string of selecting at random to j, l is the number of optimization variable or the length of character string, and S ' is the new character strings object after rule of conversion is carried out, when the Probability p that produces at random ₀The Probability p that is less than or equal to setting _tThe time, this rule will take place in string;

3) intersection rule

This rule is divided into the numerical value intersection and single-point intersects:

$\left\{\begin{array}{c}S=x_1,x_2,...x_i,...x_j...x_l\\ S_r=y_1,y_2,...y_i,...y_j...y_l\end{array}\right.$

For S, picked at random character string S _r, and i＜j in the formula, z _i=α x _i+ (1-α) y _i, z _j=α x _j+ (1-α) y _jα ∈ (0,1) wherein, numerical value intersect back S and become S ': S '=x ₁, x ₂... z _i... z _j... y _l.

(S，S _r)→(S′，S′ _r)

$\left\{\begin{array}{c}S=x_1,x_2,...u_i,u_{i+1}...x_l\\ S_r=y_1,y_2,...u_j,u_{j+1}...y_l\\ S^{\&prime;}=x_1,x_2,...u_i,u_{j+1}...x_l\\ {S_r}^{\&prime;}=y_1,y_2,...u_j,u_{i+1}...y_l\end{array}\right.$

String S, S _rRespectively at u _i, u _I+1And u _j, u _J+1The position cutting is come, and then, fragment is respectively at u _i, u _J+1And u _j, u _I+1Be reassembled into two new character strings;

4) exchange rule

To be each film send to the outside of film with the reasonable character string of its fitness value to this rule, just enter in the upper layer film of film,

[S _min1，...，S _minq] _iS _max1，...S _maxq→[...] _iS _min1，...，S _minq

i＝1，2，...l

In the formula, q is the object number that exchanges, and promptly exchanges the size of scale, S _Max1..., S _MaxqBe q poor character object of i+1 zone endoadaptation degree, S _Min1..., S _MinqBe q the character object that fitness is better than other objects in the one's respective area.

Images