CN105447589A - Control method and apparatus for reducing NOx emission load of coal-fired unit - Google Patents

Abstract

The invention discloses a control method and apparatus for reducing a NOx emission load of a coal-fired unit. The method and the apparatus are optimized and implemented based on a chaotic distribution estimation algorithm. On the basis of the chaotic algorithm, chaotic mutation and evaluation are carried out on a population; the processed population is sent into a distributed estimation algorithm for optimization and thus a new individual is generate; evaluation is carried out and a next-generation population is generated; and if the next-generation population does not meet a termination condition, the population is sent back to the chaotic algorithm to carry out screening and mutation continuously until the termination condition is met. According to the improved distributed estimation algorithm, optimized selection of adjustable parameters of the model is realized; an optimal value of the input parameter is provided for the running staff; and thus the NOx emission load of the coal-fired boiler can be minimized on the same condition.

Description

A kind of control method and device lowering coal unit NOx discharge

Technical field

The present invention relates to NOx emission control field, be specifically related to a kind of control method and the device that lower coal unit NOx discharge.

Background technology

Along with the fast development of China's economy, the demand of people to electric power also grows with each passing day, and this also means that the more energy of needs supports economic growth.At present, the thermal power generation of China is still in the leading position of power industry, and the singularity of energy structure due to China, coal will be in the leading position of energy-consuming within coming few decades.While coal burns in a large number, also bring serious problem of environmental pollution, oxides of nitrogen (NOx) is one of atmosphere pollution of producing in combustion of coal.The generation of acid rain, photo-chemical smog etc. all has direct relation with the oxides of nitrogen in dusty gas.Therefore, corresponding laws and rules has all been formulated to limit the discharge of NOx in countries in the world, and especially with Germany and Japan the most strictly, its emission standard is 200mg/Nm respectively ³and 400mg/Nm ³.

China is populous nation, is also coal production maximum in the world and country of consumption simultaneously.The NOx emission produced because of coal burning causes huge destruction to the Ecological environment of China.According to data, in the with serious pollution line urban atmosphere of China, the NOx of more than 65% is produced by coal combustion, and the NOx that wherein coal-fired power plant produces occupies larger proportion again.Along with the enforcement of the strategy of sustainable development and the pay attention to day by day of Environmental protection, country starts to put into effect the relevant law about restriction power plant NOx emission.State Bureau of Environmental Protection has promulgated " thermal power plant's air pollution emission standard " (GB13223-2011) on July 18th, 2011, clear stipulaties was from 1 day January in 2012, and thermal power generation boiler and the NOx discharge of gas turbine group of newly-built, enlarging and reconstruction must not more than 200mg/m ³, be that the emission of NOx of boiler amount of fuel must not more than 100mg/m with oil ³.These provide quantifiable index all to high-efficient low polluting combustion technology.

In recent years, intellectual technology for coal-fired boiler NOx discharge characteristic adopts the Estimation of Distribution Algorithm improved to be optimized Controlling model, but there is following problem: although Estimation of Distribution Algorithm can solve low-dimensional single mode optimization problem, and there is good ability of searching optimum, but it is easily early ripe when solving the multi-modal problem of higher-dimension, convergence locally optimal solution, causes optimization ability to decline.

Summary of the invention

For the problems referred to above, the invention provides a kind of control method and the device that lower coal unit NOx discharge, Estimation of Distribution Algorithm is improved, is applied to choosing of Optimized model parameter.

Object of the present invention realizes by the following technical solutions:

Lower a control method for coal unit NOx discharge, it is characterized in that, comprise the following steps: (1) initialization affects the input parameter population of NOx discharge; (2) initial population is evaluated, prediction NOx value; (3) fitness is distributed; (4) chaotic mutation is performed; (5) chaotic mutation population is assessed; (6) population of future generation is produced with Estimation of Distribution Algorithm; (7) judge whether to meet end condition, if satisfy condition, stop iteration, and export for reference input value affecting NOx discharge, otherwise re-execute step (3).

Preferably, step (1) specifically comprises the following steps:

A () sets up the 1 dimensional Logistic Map model of chaotic mutation: α _t+1=λ α _t(1-α _t);

Wherein, λ represents controling parameters, λ ∈ [0,4]; α is a random number between [0,1]; T represents iteration time;

B () sets up the individual volume coordinate form of boiler operating parameter vector to be optimized: X _j,G=(x _{j, 1, G}..., x _{j, i, G}..., x _{j, n, G});

Wherein, i=1,2 ..., n, n are the number wanting Optimal Parameters; J=1,2 ..., NP; G=1,2 ..., Gmax, Gmax are maximum evolutionary generation;

C () produces one group of number α at random ₀as initial value, α ₀=rand (1, n), and α ₀∈ [0,1];

D () is by vectorial α ₀in the mapping model that substitution step (a) is set up, given λ value, then through NP iterative computation, obtain chaos time sequence [α ₁, α ₂..., α _nP]; Make A=[α ₁, α ₂..., α _nP] ^t, then A is the matrix of NP × n;

E () is by the chaos time sequence [α in step (d) ₁, α ₂..., α _nP] scope of problem to be optimized is expanded to from [0,1], obtain i-th gene location: x of an initial population jth parameter individuality to be optimized _{j, i, 1}=low+ (high-low) A (j, i);

Wherein, low is the minimum value of solution space; High is the maximal value of solution space.

Preferably, step (2) specifically performs: obtain each individual corresponding desired value in population, target setting function f (x), and output valve function minf (x) obtaining the minimum model of f (x), the NOx value namely predicted; X is boiler operating parameter vector to be optimized, and x=[x1, x2, ..., x17, x18], wherein x1 is total coal-air ratio, and x2 is primary air ratio, and x3 is SOFA wind rate, x4 is OFA baffle opening, x5-x8 is auxiliary air baffle plate aperture, and x9-x12 is fuel air baffle opening, and x13-x16 is the primary air flow of coal pulverizer and the ratio of coal-supplying amount, x17 is hot secondary air temperature, and x18 is burner hearth bellows pressure reduction.

Preferably, step (3) specifically performs: linearly distribute the value of objective function f (x) and carry out descending sort, and functional value maximum for desired value is placed on first position, and the minimum functional value of desired value is placed on NP position.

Preferably, step (4) specifically performs: set up chaotic mutation formula: x ' _{j, i, G}=x _{j, i, G}+ r (j, i) × 2 (1-2A (j, i)), wherein, A (j, i) is the element of the matrix A according to the 1 dimensional Logistic Map model generation in step (a); R (j, i) is variation radius.

Preferably, step (5) specifically comprises the following steps:

F () selects more excellent individuality, set up single argument Gauss model;

G () is sampled and is produced new individuality from model, if the valuation functions value of new population individuality is less, then replace old individuality with newly individual.

Lower a control device for coal unit NOx discharge, it is characterized in that, comprising:

Initialization module, affects the input parameter population of NOx discharge for initialization;

Evaluation module, for evaluating initial population, prediction NOx value;

Distribution module, for distributing fitness;

Chaotic mutation module, for performing chaotic mutation;

Evaluation module, for assessment of chaotic mutation population;

Distribution estimation module, for producing population of future generation with Estimation of Distribution Algorithm;

Iteration module, meets end condition for judging whether, if satisfy condition, stops iteration, and exports for reference input value affecting NOx discharge, otherwise re-executes distribution module.

Preferably, initialization module specifically comprises:

Set up the 1 dimensional Logistic Map model of chaotic mutation: α _t+1=λ α _t(1-α _t), wherein, λ represents controling parameters, λ ∈ [0,4]; α is a random number between [0,1]; T represents iteration time;

Set up the volume coordinate form that boiler operating parameter vector to be optimized is individual: X _j,G=(x _{j, 1, G}..., x _{j, i, G}..., x _{j, n, G})

Wherein, i=1,2 ..., n, n are the number wanting Optimal Parameters, j=1,2 ..., NP, G=1,2 ..., Gmax, Gmax are maximum evolutionary generation;

Random generation one group of number α ₀as initial value, α ₀=rand (1, n), and α ₀∈ [0,1];

By vectorial α ₀substitute in 1 dimensional Logistic Map model, given λ value, then through NP iterative computation, obtain chaos time sequence [α ₁, α ₂..., α _nP], make A=[α ₁, α ₂..., α _nP] ^t, then A is the matrix of NP × n;

By chaos time sequence [α ₁, α ₂..., α _nP] scope of problem to be optimized is expanded to from [0,1], obtain i-th gene location: x of an initial population jth parameter individuality to be optimized _{j, i, 1}=low+ (high-low) A (j, i), wherein, low is the minimum value of solution space; High is the maximal value of solution space.

Preferably, evaluation module specifically comprises: obtain each individual corresponding desired value in population, target setting function f (x), and output valve function minf (x) obtaining the minimum model of f (x), the NOx value namely predicted; X is boiler operating parameter vector to be optimized, and x=[x1, x2, ..., x17, x18], wherein x1 is total coal-air ratio, and x2 is primary air ratio, and x3 is SOFA wind rate, x4 is OFA baffle opening, x5-x8 is auxiliary air baffle plate aperture, and x9-x12 is fuel air baffle opening, and x13-x16 is the primary air flow of coal pulverizer and the ratio of coal-supplying amount, x17 is hot secondary air temperature, and x18 is burner hearth bellows pressure reduction.

Preferably, chaotic mutation module specifically comprises: set up chaotic mutation formula: x ' _{j, i, G}=x _{j, i, G}+ r (j, i) × 2 (1-2A (j, i)),

Wherein, A (j, i) is the element of the matrix A according to the generation of 1 dimensional Logistic Map model; R (j, i) is variation radius.Beneficial effect of the present invention is:

1. adopt chaos Estimation of Distribution Algorithm to optimize NOx emission model, utilize the peculiar property of chaos phenomenon to be optimized search in space, make it jump out local extremum district, find globally optimal solution fast, thus solve Estimation of Distribution Algorithm premature problem.2. the Estimation of Distribution Algorithm by improving realizes the optimum option to model adjustable parameter, and for operations staff provides the optimum value of input parameter, under making identical operating mode, coal-fired boiler NOx discharge amount is minimum.

Accompanying drawing explanation

The invention will be further described to utilize accompanying drawing, but the embodiment in accompanying drawing does not form any limitation of the invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, can also obtain other accompanying drawing according to the following drawings.

Fig. 1 is a kind of FB(flow block) lowering the control method of coal unit NOx discharge;

Fig. 2 a kind ofly lowers the control method of coal unit NOx discharge and the test result comparison diagram of device one embodiment.

Embodiment

The invention will be further described with the following Examples.

A kind of control method lowering coal unit NOx discharge as shown in Figure 1, comprises the following steps: (1) initialization affects the input parameter population of NOx discharge; (2) initial population is evaluated, prediction NOx value; (3) fitness is distributed; (4) chaotic mutation is performed; (5) chaotic mutation population is assessed; (6) population of future generation is produced with Estimation of Distribution Algorithm; (7) judge whether to meet end condition, if satisfy condition, stop iteration, and export for reference input value affecting NOx discharge, otherwise re-execute step (3).

Step (1) specifically comprises the following steps:

A () sets up the 1 dimensional Logistic Map model of chaotic mutation: α _t+1=λ α _t(1-α _t);

Wherein, λ represents controling parameters, and λ ∈ [0,4], α are a random number between [0,1], and t represents iteration time;

When λ obtains different value, the system obtained has different qualities:

1) when λ ∈ [0,1], the system obtained is very simple, only has fixed point zero, does not have other periodic point.

2) when λ ∈ [1,3], the system of gained is still very simple, only has two fixed points, and 1 and 1-1/ λ.

3) when λ ∈ [3,4], there is aperiodicity in the system very complex obtained, the value of λ is larger, and it is larger that system travels through scope.

B () sets up the individual volume coordinate form of boiler operating parameter vector to be optimized: X _j,G=(x _{j, 1, G}..., x _{j, i, G}..., x _{j, n, G})

Wherein, i=1,2 ..., n, n are will the number of Optimal Parameters; J=1,2 ..., NP; G=1,2 ..., Gmax, Gmax are maximum evolutionary generation;

Make the initialized population not repeatedly whole space of random ergodic, just need to adopt dimensional Logistic model, specific practice is as follows:

C () produces one group of number α at random ₀as initial value, α ₀=rand (1, n), and α ₀∈ [0,1];

D () is by vectorial α ₀in the mapping model that substitution step (a) is set up, given λ value, then through NP iterative computation, just chaos time sequence [α can be obtained ₁, α ₂..., α _nP]; Make A=[α ₁, α ₂..., α _nP] ^t, then A is the matrix of NP × n;

E () is by the chaos time sequence [α in step (d) ₁, α ₂..., α _nP] scope of problem to be optimized is expanded to from [0,1], obtain initial population jth i-th individual gene location: x _{j, i, 1}=low+ (high-low) A (j, i);

Wherein, low is the minimum value of solution space; High is the maximal value of solution space.

Step (2) specifically performs: obtain each individual corresponding desired value in population, target setting function f (x), and output valve function minf (x) obtaining the minimum model of f (x), the NOx value namely predicted, x=[x1, x2...x17, x18], x is boiler operating parameter vector to be optimized, for concrete operating condition, due to boiler load, use coal certain, operations staff is mainly First air for reducing the operation adjustable parameter of NOx emission, hot secondary air temperature, the parameters such as blast, therefore x1 is total coal-air ratio, x2 is primary air ratio, x3 is SOFA wind rate, x4 is OFA baffle opening, x5-x8 is auxiliary air baffle plate aperture, x9-x12 is fuel air baffle opening, x13-x16 is the primary air flow of coal pulverizer and the ratio of coal-supplying amount, x17 is hot secondary air temperature, x18 is burner hearth bellows pressure reduction.

Step (3) specifically performs: linearly distribute the value of objective function f (x) and carry out descending sort, functional value maximum for desired value is placed on first position, the minimum functional value of desired value is placed on NP position, and so effective fitness function can accelerate algorithm the convergence speed effectively.

Step (4) specifically performs: set up chaotic mutation formula: x ' _{j, i, G}=x _{j, i, G}+ r (j, i) × 2 (1-2A (j, i)),

Wherein, A (j, i) is the element of the matrix A according to the 1 dimensional Logistic Map model generation in step (a); R (j, i) is variation radius.

Step (5) specifically comprises the following steps:

F () selects more excellent individuality, set up single argument Gauss model.

G () is sampled and is produced new individuality from model, if the valuation functions value of new population individuality is less, then replace old individuality with newly individual.

Lower a control device for coal unit NOx discharge, comprising:

Initialization module, affects the input parameter population of NOx discharge for initialization;

Evaluation module, for evaluating initial population, prediction NOx value;

Distribution module, for distributing fitness;

Chaotic mutation module, for performing chaotic mutation;

Evaluation module, for assessment of chaotic mutation population;

Distribution estimation module, for producing population of future generation with Estimation of Distribution Algorithm;

Iteration module, meets end condition for judging whether, if satisfy condition, stops iteration, and exports for reference input value affecting NOx discharge, otherwise re-executes distribution module.

Initialization module specifically comprises:

Set up the 1 dimensional Logistic Map model of chaotic mutation: α _t+1=λ α _t(1-α _t), wherein, λ represents controling parameters, λ ∈ [0,4]; α is a random number between [0,1]; T represents iteration time;

Set up the volume coordinate form that boiler operating parameter vector to be optimized is individual: X _j,G=(x _{j, 1, G}..., x _{j, i, G}..., x _{j, n, G})

Wherein, i=1,2 ..., n, n are the number wanting Optimal Parameters, j=1,2 ..., NP, G=1,2 ..., Gmax, Gmax are maximum evolutionary generation;

Random generation one group of number α ₀as initial value, α ₀=rand (1, n), and α ₀∈ [0,1];

By vectorial α ₀substitute in 1 dimensional Logistic Map model, given λ value, then through NP iterative computation, obtain chaos time sequence [α ₁, α ₂..., α _nP], make A=[α ₁, α ₂..., α _nP] ^t, then A is the matrix of NP × n;

By chaos time sequence [α ₁, α ₂..., α _nP] scope of problem to be optimized is expanded to from [0,1], obtain i-th gene location: x of an initial population jth parameter individuality to be optimized _{j, i, 1}=low+ (high-low) A (j, i), wherein, low is the minimum value of solution space; High is the maximal value of solution space.

Evaluation module specifically comprises: obtain each individual corresponding desired value in population, target setting function f (x), and output valve function minf (x) obtaining the minimum model of f (x), the NOx value namely predicted; X is boiler operating parameter vector to be optimized, and x=[x1, x2, ..., x17, x18], wherein x1 is total coal-air ratio, and x2 is primary air ratio, and x3 is SOFA wind rate, x4 is OFA baffle opening, x5-x8 is auxiliary air baffle plate aperture, and x9-x12 is fuel air baffle opening, and x13-x16 is the primary air flow of coal pulverizer and the ratio of coal-supplying amount, x17 is hot secondary air temperature, and x18 is burner hearth bellows pressure reduction.

Distribution module specifically comprises: linearly distribute the value of objective function f (x) and carry out descending sort, and functional value maximum for desired value is placed on first position, and the minimum functional value of desired value is placed on NP position.

Chaotic mutation module specifically comprises: set up chaotic mutation formula: x ' _{j, i, G}=x _{j, i, G}+ r (j, i) × 2 (1-2A (j, i)),

Wherein, A (j, i) is the element of the matrix A according to the generation of 1 dimensional Logistic Map model; R (j, i) is variation radius.

Evaluation module specifically comprises:

F () selects more excellent individuality, set up single argument Gauss model;

G () is sampled and is produced new individuality from model, if the valuation functions value of new population individuality is less, then replace old individuality with newly individual.

Adopt chaos Estimation of Distribution Algorithm to optimize NOx emission model, utilize the peculiar property of chaos phenomenon to be optimized search in space, make it jump out local extremum district, find globally optimal solution fast, thus solve Estimation of Distribution Algorithm premature problem.Realize the optimum option to model adjustable parameter by the Estimation of Distribution Algorithm improved, for operations staff provides the optimum value of input parameter, under making identical operating mode, coal-fired boiler NOx discharge amount is minimum.For certain power plant's operating mode, under certain load, utilize chaos distribution Estimation Optimization algorithm to reduce 8.42% to low NOx drainage amount NOx within the scope of variable bound that regulated variable is optimized acquisition, demonstrate the feasibility of the method.

As shown in Figure 2, under 30 different operating modes, estimate to using chaos distribution respectively the NOx discharge optimized under computing method, genetic algorithm and initial condition to contrast, it is better that test proves to adopt the emission effect of chaos Estimation of Distribution Algorithm optimization more hereditary calculation optimization NOx, effectively can reduce the discharge capacity of NOx.

Finally should be noted that; above embodiment is only in order to illustrate technical scheme of the present invention; but not limiting the scope of the invention; although done to explain to the present invention with reference to preferred embodiment; those of ordinary skill in the art is to be understood that; can modify to technical scheme of the present invention or equivalent replacement, and not depart from essence and the scope of technical solution of the present invention.

Claims (10)

1. lower a control method for coal unit NOx discharge, it is characterized in that, comprise the following steps: (1) initialization affects the input parameter population of NOx discharge; (2) initial population is evaluated, prediction NOx value; (3) fitness is distributed; (4) chaotic mutation is performed; (5) chaotic mutation population is assessed; (6) population of future generation is produced with Estimation of Distribution Algorithm; (7) judge whether to meet end condition, if satisfy condition, stop iteration, and export for reference input value affecting NOx discharge, otherwise re-execute step (3).

2. a kind of control method lowering coal unit NOx discharge according to claim 1, is characterized in that, step (1) specifically comprises the following steps:

A () sets up the 1 dimensional Logistic Map model of chaotic mutation: α _t+1=λ α _t(1-α _t);

Wherein, λ represents controling parameters, λ ∈ [0,4]; α is a random number between [0,1]; T represents iteration time;

B () sets up the individual volume coordinate form of boiler operating parameter vector to be optimized: X _j,G=(x _{j, 1, G}..., x _{j, i, G}..., x _{j, n, G});

Wherein, i=1,2 ..., n, n are the number wanting Optimal Parameters; J=1,2 ..., NP; G=1,2 ..., Gmax, Gmax are maximum evolutionary generation;

C () produces one group of number α at random ₀as initial value, α ₀=rand (1, n), and α ₀∈ [0,1];

D () is by vectorial α ₀in the mapping model that substitution step (a) is set up, given λ value, then through NP iterative computation, obtain chaos time sequence [α ₁, α ₂..., α _nP]; Make A=[α ₁, α ₂..., α _nP] ^t, then A is the matrix of NP × n;

E () is by the chaos time sequence [α in step (d) ₁, α ₂..., α _nP] scope of problem to be optimized is expanded to from [0,1], obtain i-th gene location: x of an initial population jth parameter individuality to be optimized _{j, i, 1}=low+ (high-low) A (j, i);

Wherein, low is the minimum value of solution space; High is the maximal value of solution space.

3. a kind of control method lowering coal unit NOx discharge according to claim 2, it is characterized in that, step (2) specifically performs: obtain each individual corresponding desired value in population, target setting function f (x), and obtain output valve function minf (x) of the minimum model of f (x), the NOx value namely predicted; X is boiler operating parameter vector to be optimized, and x=[x1, x2,, x17, x18], wherein x1 is total coal-air ratio, and x2 is primary air ratio, and x3 is SOFA wind rate, x4 is OFA baffle opening, x5-x8 is auxiliary air baffle plate aperture, and x9-x12 is fuel air baffle opening, and x13-x16 is the primary air flow of coal pulverizer and the ratio of coal-supplying amount, x17 is hot secondary air temperature, and x18 is burner hearth bellows pressure reduction.

4. a kind of control method lowering coal unit NOx discharge according to claim 3, it is characterized in that, step (3) specifically performs: linearly distribute the value of objective function f (x) and carry out descending sort, functional value maximum for desired value is placed on first position, and the minimum functional value of desired value is placed on NP position.

5. a kind of control method lowering coal unit NOx discharge according to claim 4, is characterized in that, step (4) specifically performs: set up chaotic mutation formula: x ' _{j, i, G}=x _{j, i, G}+ r (j, i) × 2 (1-2A (j, i)),

Wherein, A (j, i) is the element of the matrix A according to the 1 dimensional Logistic Map model generation in step (a); R (j, i) is variation radius.

6. a kind of control method lowering coal unit NOx discharge according to claim 5, is characterized in that, step (5) specifically comprises the following steps:

F () selects more excellent individuality, set up single argument Gauss model;

G () is sampled and is produced new individuality from model, if the valuation functions value of new population individuality is less, then replace old individuality with newly individual.

7. lower a control device for coal unit NOx discharge, it is characterized in that, comprising:

Initialization module, affects the input parameter population of NOx discharge for initialization;

Evaluation module, for evaluating initial population, prediction NOx value;

Distribution module, for distributing fitness;

Chaotic mutation module, for performing chaotic mutation;

Evaluation module, for assessment of chaotic mutation population;

Distribution estimation module, for producing population of future generation with Estimation of Distribution Algorithm;

Iteration module, meets end condition for judging whether, if satisfy condition, stops iteration, and exports impact for reference

The input value of NOx discharge, otherwise re-execute distribution module.

8. a kind of control device lowering coal unit NOx discharge according to claim 7, it is characterized in that, initialization module specifically comprises:

Set up the 1 dimensional Logistic Map model of chaotic mutation: α _t+1=λ α _t(1-α _t), wherein, λ represents controling parameters, λ ∈ [0,4]; α is a random number between [0,1]; T represents iteration time;

Set up the volume coordinate form that boiler operating parameter vector to be optimized is individual: X _j,G=(x _{j, 1, G}..., x _{j, i, G}..., x _{j, n, G})

Wherein, i=1,2 ..., n, n are the number wanting Optimal Parameters, j=1,2 ..., NP, G=1,2 ..., Gmax, Gmax are maximum evolutionary generation;

Random generation one group of number α ₀as initial value, α ₀=rand (1, n), and α ₀∈ [0,1];

By vectorial α ₀substitute in 1 dimensional Logistic Map model, given λ value, then through NP iterative computation, obtain chaos time sequence [α ₁, α ₂..., α _nP], make A=[α ₁, α ₂..., α _nP] ^t, then A is the matrix of NP × n;

By chaos time sequence [α ₁, α ₂..., α _nP] scope of problem to be optimized is expanded to from [0,1], obtain i-th gene location: x of an initial population jth parameter individuality to be optimized _{j, i, 1}=low+ (high-low) A (j, i), wherein, low is the minimum value of solution space; High is the maximal value of solution space.

9. a kind of control device lowering coal unit NOx discharge according to claim 8, it is characterized in that, evaluation module specifically comprises: obtain each individual corresponding desired value in population, target setting function f (x), and obtain output valve function minf (x) of the minimum model of f (x), the NOx value namely predicted; X is boiler operating parameter vector to be optimized, and x=[x1, x2, ..., x17, x18], wherein x1 is total coal-air ratio, and x2 is primary air ratio, and x3 is SOFA wind rate, x4 is OFA baffle opening, x5-x8 is auxiliary air baffle plate aperture, and x9-x12 is fuel air baffle opening, and x13-x16 is the primary air flow of coal pulverizer and the ratio of coal-supplying amount, x17 is hot secondary air temperature, and x18 is burner hearth bellows pressure reduction.

10. a kind of control device lowering coal unit NOx discharge according to claim 9, it is characterized in that, chaotic mutation module specifically comprises: set up chaotic mutation formula: x ' _{j, i, G}=x _{j, i, G}+ r (j, i) × 2 (1-2A (j, i)),

Wherein, A (j, i) is the element of the matrix A according to the generation of 1 dimensional Logistic Map model; R (j, i) is variation radius.