CN102622530A - Improved genetic algorithm-based method for distributing and optimizing thermal and electrical load of steam extraction and heating unit - Google Patents

Abstract

The invention provides an improved genetic algorithm-based method for distributing and optimizing thermal and electrical load of a steam extraction and heating unit, and belongs to the technical field of energy saving and monitoring of a power station. The invention aims to distribute and optimize the thermal and electrical load of a plurality of heating units of the power station, optimally distribute the electrical load and the thermal load of the power station when the electrical load and the thermal load meet requirements of users, and reduce total energy consumption and save energy. An actual heat consumption and consumption difference curve of the unit is set, and a design heat consumption curve of the unit is obtained; the electrical load and steam extraction amount of each unit are acquired; and the electrical load and the steam extraction amount value of each unit are obtained based on the improved genetic algorithm when the total thermal consumption values of the units are minimum, and the electrical load, the steam extraction amount value optimal solution and the corresponding minimum total thermal consumption of each unit are output when the total thermal consumption values of the units are minimum by improving genetic coding and fitness function, and selecting, crossing and mutating the genetic algorithm under the condition that the optimization process meets constraint conditions. By the method, the speed of optimization process and accuracy of an optimization result are improved.

Description

Based on the thermoelectric load distribution optimization method of the extraction for heat supply unit that improves genetic algorithm

Technical field

The present invention relates to the thermoelectric load distribution optimization method of many heat supply units of a kind of power plant, belong to generating plant energy-saving monitoring technical field.

Background technology

Along with the raising of expanding economy and people's living standard, urban district heating system is developed rapidly, and wherein the cogeneration of heat and power energy conversion efficiency has clear superiority, and therefore, the heat supply unit that draws gas has obtained development energetically.

The extraction for heat supply unit provides electric power and heating with heat to the user, and what of the heating power that power plant provides and electric power are controlled by hot user and electric user's demand, and therefore, power plant must load according to the draw gas thermoelectricity of unit of hot user and electric user's demand adjustment heat supply.

For the thermoelectricity load of confirming; How power plant between each unit carries out thermoelectric burden apportionment according to the type of unit and the difference of unit efficiency, makes the heat consumption rate of entire power plant minimum; Making the economic benefit of entire power plant best, is the problem that faces in power plant's production run.This just need carry out allocation optimized to electric load and thermal load that power plant's heat supply is drawn gas between unit, confirms the electric load and the thermal load of every unit.

The optimized distribution of load is meant, at the total scheduling load of full factory down, confirms that according to the thermodynamic property of each unit each unit agrees the load of band, thereby makes a kind of Optimization Dispatching of the coal consumption amount minimum of full factory.

Load optimized distribution to power plant; Early carrying out also comparatively ripe is the electric load allocation optimized research of pure condensate unit; Obtained using widely etc. little gaining rate method; Because extraction for heat supply unit thermal load also need participate in optimized distribution, therefore, no matter be from heat consumption curve obtain, or all the electric load allocation optimized than the pure condensate unit is complicated to optimize the complicacy angle.At present, to heat, the electric load allocation optimized of extraction for heat supply unit, many researchs have been carried out.

Document [1] (Wei Hao; Song Baofeng; Zhao Weidong; Wang Yi; " Jilin electric power "; 2002 the 5th phases; " exploitation and the application of the thermoelectric load of heat supply steam turbine group Optimal Distributing System ") the middle thermoelectric load distribution optimization system of introducing: the mathematical model that adopts " point by point method " to distribute; Utilize equivalent heat drop theory that the major parameter that influences the Steam Turbine economy is carried out variance analysis, can realize that economic target is calculated and the function of energy loss analysis, parameter demonstration, inquiry and warning and steam turbine analog quantity system figure demonstration.The mathematical model poor stability that " point by point method " in this document distributed, computing velocity is slower, can't optimize continuously.It is comparatively complicated that simultaneously Equivalent Entropy Drop Method is applied to the heat supply unit that draws gas.

Document [2] (Ran Peng, Zhang Shufang, " steam turbine technology "; 2006 the 48th the 1st phases of volume: " based on cogeneration plant's load optimized calculation method of genetic algorithm ") use the method that genetic algorithm is set up cogeneration plant's load Optimization Model in; Solved when problem scale expansion, when the variable constraints is a lot, can be easy to be absorbed in local optimum; And numerical stability is reduced, finally cause restraining the problem of difficulty.Document [2] is though part has solved the problem in the document [1]; It is excessive that but initial population is worked as in the genetic algorithm existence; The problem that computing velocity is slower; This algorithm is not realized heat supply unit online in real time optimizational function simultaneously, can not widespread use in actual production, do not consider of the influence of the actual motion condition of unit to hear rate yet.

There are some problems in the solution that can find out the thermoelectric load distribution on-line optimization of present heat supply unit; Therefore need further study the thermoelectric load distribution on-line optimization of heat supply unit problem to these problems; Make many heat supply units of power plant can realize the optimum allocation of thermoelectric load, reach energy saving purposes.

Summary of the invention

The present invention is in order to realize the thermoelectric load distribution optimization of many heat supply units of power plant, and the electric load, thermal load that makes power plant can optimum allocation when meeting consumers' demand, and reduces total energy consumption and reach purpose of energy saving; And then provide a kind of based on the thermoelectric load distribution optimization method of the extraction for heat supply unit that improves genetic algorithm.

The present invention solves the problems of the technologies described above the technical scheme of taking to be:

Detailed process based on the thermoelectric load distribution optimization method of the extraction for heat supply unit that improves genetic algorithm of the present invention is:

Step 1, the actual heat consumption curve of unit is set: the actual heat consumption curve that obtains every unit according to test; Said actual heat consumption curve is meant that be that Q is independent variable (horizontal ordinate) with power P with the amount of drawing gas, and hear rate value R is gang's curve of dependent variable (ordinate), promptly

The 1st unit: R ₁=f (P ₁, Q ₁);

The 2nd unit: R ₂=f (P ₂, Q ₂);

……

N platform unit: R _n=f (P _n, Q _n);

Step 2, unit consumption difference fair curve is set, confirms that unit consumption difference revises overall coefficient θ _iI is the machine group #, i=1,2; N, n represent the unit number: all can exert an influence to hear rate during based on condenser back pressure, main steam pressure, main steam temperature, reheat pressure, reheat steam temperature, these six factor off-design values of feed temperature, provide according to producer then or the consumption difference fair curve of power plant checks in the hear rate correction factor Δ of each influence factor of every unit _1iΔ _2iΔ _3iΔ _6i, Δ _1iΔ _2iΔ _3iΔ _6iBe respectively condenser back pressure, main steam pressure, main steam temperature, reheat pressure, the reheat steam temperature of every unit, the hear rate correction factor of feed temperature; Make θ _i=Δ _1iΔ _2iΔ _3iΔ _6i

Step 3, obtain unit design heat consumption curve: revise overall coefficient θ according to every unit consumption difference _iThe actual hear rate correction of unit is obtained unit design heat consumption curve (being the actual heat consumption curve of revised unit), and the design heat consumption curve of each unit is:

The 1st unit: R ₁=θ ₁F (P ₁, Q ₁);

The 2nd unit: R ₂=θ ₂F (P ₂, Q ₂);

……

N platform unit: R _n=θ _nF (P _n, Q _n);

Step 4, (from system, obtaining data) are obtained the amount of the drawing gas Q of each unit _i(characterizing thermal load) and electric load P with it _i: the amount of the drawing gas Q that records each unit earlier _iWith electric load P _i, obtain corresponding hear rate R through the described unit design of step 3 heat consumption curve then _i, i ∈ [1, n],

The electric load that obtains the n of power plant platform unit is respectively P ₁, P ₂..., P _n, the amount of drawing gas is respectively Q ₁, Q ₂..., Q _n(characterizing thermal load with it), hear rate value are R ₁, R ₂..., R _n, n is the unit number;

The purpose of optimizing is the overall heat consumption value that obtains making all units

P hour ₁, P ₂..., P _n, Q ₁, Q ₂..., Q _nAllocative decision, wherein objective function is:

$\stackrel{\‾}{R}=\frac{P_1\×R_1+P_2\×R_2+\·\·\·P_n\×R_n}{P_1+P_2+\·\·\·+P_n}---\left(1\right)$

Set constraint condition:

First constraint condition is: Q _z=Q ₁+ Q ₂+ ... + Q _n=const, P _z=P ₁+ P ₂+ ... + P _n=const (2)

The i.e. Q of the amount of drawing gas always of all units _zWith total electric load P _zBe respectively constant;

Second constraint condition is: Q _i∈ (Q _Imin, Q _Imax), P _i∈ (P _Imin, P _Imin) (3)

The minimax electric load that is every unit is respectively: P _1min, P _1maxP _2min, P _2max P _Nmin, P _NmaxThe minimax amount of drawing gas is respectively Q _1min, Q _1maxQ _2min, Q _2max Q _Nmin, Q _Nnax);

Step 5, based on improving electric load that genetic algorithm obtains overall heat consumption value each unit hour that satisfies all units and drawing gas value: detailed process is following

1, initial population is set

Matrix with 2n * m then can be represented initial population:

$\left(\begin{array}{c}{P}_{11}{P}_{12}......P_{1n}{Q}_{11}{Q}_{12}......Q_{1n}\\ P_{21}{P}_{22}......P_{2n}{Q}_{21}{Q}_{22}......Q_{2n}\\ ............................\\ P_{m1}{P}_{m2}......P_{\mathrm{mn}}{Q}_{m1}{Q}_{m2}......Q_{\mathrm{mn}}\end{array}\right)$

The individual number of m for setting, the amount of drawing gas Q _iWith electric load P _iBe the random number that satisfies second constraint condition; Above-mentioned initial population adopts the formal construction of the constraint coding that satisfies second constraint condition;

Electric load to preceding (n-1) individual unit in the above-mentioned initial population is encoded with the constraint that the amount of drawing gas satisfies second constraint condition, and last unit passes through computes:

$\left\{\begin{array}{c}{P}_n=P_z-P_1-P_2-\·\·\·-P_{n-1}\\ Q_n=Q_z-Q_1-Q_2-\·\·\·-Q_{n-1}\end{array}\right.---\left(4\right)$

Can obtain the initial population that all units satisfy first constraint condition and satisfied second constraint condition of preceding (n-1) individual unit like this:

$\left(\begin{array}{c}{P}_{11}{P}_{12}......P_{1(n-1)}{Q}_{11}{Q}_{12}......Q_{1(n-1)}\\ P_{21}{P}_{22}......P_{2(n-1)}{Q}_{21}{Q}_{22}......Q_{2(n-1)}\\ ............................\\ P_{m1}{P}_{m2}......P_{m(n-1)}{Q}_{m1}{Q}_{m2}......Q_{m(n-1)}\end{array}\right)$

Following formula n platform unit is to satisfy difference maximum and the maximum heating load of its electric load maximal value and minimum value and the unit that the minimum thermal load does not wait, that is:

P _nmax-P _nmin＞P _imax-P _imin

Q _nmax-Q _nmin≠0

P _Nmax, P _NminThe maximum electrical load and minimum electric load of representing the selected n platform unit that comes out; P _Imax, P _IminThe maximum electrical load and the minimum electric load of expression residue unit; Q _Nmax, Q _NminMaximum heating load and the minimum thermal load of representing the selected n platform unit that comes out;

2, make up fitness function: calculate through fitness, realize individual optimized choice, make also satisfied second constraint condition of n platform unit in the Optimization result simultaneously;

Ineligible individuality is:

P _n＜P _min OR?P _n＞P _max

Q _n＜Q _min OR?Q _n＞Q _max

Thermoelectric load distribution optimization is the minimum value of asking objective function, and the optimization aim of genetic algorithm is to find the individuality with maximum adaptation degree, so definition fitness function ObjV definition is as follows:

1), for qualified individuality:

P _min＜P _n＜P _max&Q _min＜Q _n＜Q _max

$\mathrm{ObjV}=\frac{1}{\stackrel{\‾}{R}}---\left(5\right)$

2), for ineligible individuality: P _n＜P _MinOR P _n＞P _Max, Q _n＜Q _MinOR Q _n＞Q _Max. adopt index measure transform (1) objective function

$\mathrm{ObjV}=\frac{1}{\stackrel{\‾}{R_{\max }}\×\exp \left(\mathrm{\α\β}\right)}---\left(6\right)$

Wherein: work as P _n＜P _MinThe time, $\mathrm{\β}={\mathrm{\β}}_p=\frac{P_{\mathrm{Min}}-P_n}{P_{\mathrm{Min}}}$

Work as P _n＞P _MaxThe time, $\mathrm{\β}={\mathrm{\β}}_p=\frac{P_n-P_{\mathrm{Max}}}{P_{\mathrm{Max}}}$

For Q _nIn like manner can get:

Work as Q _n＜Q _MinThe time, $\mathrm{\β}={\mathrm{\β}}_Q=\frac{Q_{\mathrm{Min}}-Q_n}{Q_{\mathrm{Min}}}$

Work as Q _n＞Q _MaxThe time, $\mathrm{\β}={\mathrm{\β}}_Q=\frac{Q_n-Q_{\mathrm{Max}}}{Q_{\mathrm{Max}}}$

For β _P, β _QWhen existing simultaneously, β=max (β _P, β _Q)

is the maximum hear rate of unit in operational process, obtains through power plant's production and test figure;

α is a constant coefficient, and target is to make to work as the P that calculates _n, Q _nWhen surpassing setting threshold 100%, its fitness value is greater than 100 times of the fitness value that satisfies condition down, i.e. β=1, exp (α)＞100; α gets 5 in experiment;

Like this in selection course; The individuality that fitness is little will have very big probability to be eliminated; The individuality that does not satisfy second constraint condition simultaneously also will have very big probability to be eliminated; Thereby realize individual optimized choice, obtaining the maximum adaptation degree under the ideal state is the minimum individuality of overall heat consumption value ;

3, in the completion after the step, carry out selection, intersection, mutation process again based on the traditional genetic algorithm; When genetic algebra reach end condition N for the time; Genetic process stops, the electric load of the overall heat consumption value

that all units are satisfied in output each unit hour and the hear rate value of value optimum solution, each unit and the minimum overall heat consumption of all units accordingly of drawing gas.

The invention has the beneficial effects as follows:

1, in whole optimizing process, utilized power consumption analysis.Taken into full account of the influence of factors such as the variation of condenser back pressure, main steam pressure, main steam temperature, reheat pressure, reheat steam temperature, feed temperature, fully improved the accuracy of Optimization result hear rate.

2, the improvement of optimized Algorithm.In optimizing process, solved problem with equality constraint, simultaneously, fitness function is revised, improved the optimal speed and the accuracy of system greatly.

When utilizing this method on hardware, specifically to realize, following beneficial effect is arranged:

1, can realize providing unit heat consumption curve and consumption difference curve interface are set, can access the actual heat consumption curve of revising through the consumption difference of unit.The user is provided with the actual heat consumption curve of unit through power plant's real data.System is according to the deviation of present parameter and design parameter, and inquiry consumption difference curve obtains the influence value of each parameter to the unit hear rate, and actual heat consumption curve is revised, and obtains the actual heat consumption curve of unit through correction.

2, managerial personnel can limit the thermal load of every unit and the scope of electric load.

3, two kinds of patterns of online in real time optimization and offline optimization have been realized.Native system is through the whole optimizing process of mutual completion of computing machine and mis system; Computing machine reads the current status data that needs to optimize unit from mis system; Again Optimization result is write mis system after optimizing, realized the online in real time optimization of thermoelectric load distribution.The user also can carry out offline optimization to unit through manual input, so that historical data is analyzed comparison.

Concrete quantification effect: taken all factors into consideration under the situation of each parameter influence such as the variation of condenser back pressure, main steam pressure, main steam temperature, reheat pressure, reheat steam temperature, feed temperature; With two 200MW in certain generating plant and two 600MW unit heat supply phases and non-heat supply phase operating mode is example; The heat supply phase; The unit hear rate is by the 7632.4441KJ/KWh that optimizes after preceding 7654.2609KJ/KWh reduces to optimization, and hear rate reduces 21.817KJ/KWh; The non-heat supply phase, the unit hear rate is by the 8373.638KJ/KWh that optimizes after preceding 8429.5307KJ/KWh reduces to optimization, and hear rate reduces 55.893KJ/KWh, has huge energy-saving potential.

Description of drawings

Fig. 1 is based on the logic diagram (frame of broken lines representes to improve genetic algorithm) of the thermoelectric load distribution method for on-line optimization of power plant of improving genetic algorithm;

Fig. 2 utilizes the power plant's thermoelectric load distribution optimization system structural representation based on the improvement genetic algorithm of the present invention;

Fig. 3 is that certain Power Plant heat consumption curve is provided with sectional drawing;

Fig. 4 is a thermoelectric load distribution Optimization Platform interface sectional drawing.

Embodiment

Embodiment one: as shown in Figure 1, the described detailed process based on the thermoelectric load distribution optimization method of the extraction for heat supply unit that improves genetic algorithm of this embodiment is:

Step 1, the actual heat consumption curve of unit is set: the actual heat consumption curve that obtains every unit according to test; Said actual heat consumption curve is meant that be that Q is independent variable (horizontal ordinate) with power P with the amount of drawing gas, and hear rate value R is gang's curve of dependent variable (ordinate), promptly

The 1st unit: R ₁=f (P ₁, Q ₁);

The 2nd unit: R ₂=f (P ₂, Q ₂);

……

N platform unit: R _n=f (P _n, Q _n);

Step 2, unit consumption difference fair curve is set, confirms that unit consumption difference revises overall coefficient θ _iI is the machine group #, i=1,2; N, n represent the unit number: all can exert an influence to hear rate during based on condenser back pressure, main steam pressure, main steam temperature, reheat pressure, reheat steam temperature, these six factor off-design values of feed temperature, provide according to producer then or the consumption difference fair curve of power plant checks in the hear rate correction factor Δ of each influence factor of every unit _1iΔ _2iΔ _3iΔ _6i, Δ _1iΔ _2iΔ _3iΔ _6iBe respectively condenser back pressure, main steam pressure, main steam temperature, reheat pressure, the reheat steam temperature of every unit, the hear rate correction factor of feed temperature; Make θ _i=Δ _1iΔ _2iΔ _3iΔ _6i

Step 3, obtain unit design heat consumption curve: revise overall coefficient θ according to every unit consumption difference _iThe actual hear rate correction of unit is obtained unit design heat consumption curve (being the actual heat consumption curve of revised unit), and the design heat consumption curve of each unit is:

The 1st unit: R ₁=θ ₁F (P ₁, Q ₁);

The 2nd unit: R ₂=θ ₂F (P ₂, Q ₂);

……

N platform unit: R _n=θ _nF (P _n, Q _n);

Step 4, (from system, obtaining data) are obtained the amount of the drawing gas Q of each unit _i(characterizing thermal load) and electric load P with it _i: the amount of the drawing gas Q that records each unit earlier _iWith electric load P _i, obtain corresponding hear rate R through the described unit design of step 3 heat consumption curve then _i, i ∈ [1, n],

The electric load that obtains the n of power plant platform unit is respectively P ₁, P ₂..., P _n, the amount of drawing gas is respectively Q ₁, Q ₂..., Q _n(characterizing thermal load with it), hear rate value are R ₁, R ₂..., R _n, n is the unit number;

The purpose of optimizing is the overall heat consumption value that obtains making all units

P hour ₁, P ₂..., P _n, Q ₁, Q ₂..., Q _nAllocative decision, wherein objective function is:

$\stackrel{\‾}{R}=\frac{P_1\×R_1+P_2\×R_2+\·\·\·P_n\×R_n}{P_1+P_2+\·\·\·+P_n}---\left(1\right)$

Set constraint condition:

First constraint condition is: Q _z=Q ₁+ Q ₂+ ... + Q _n=const, P _z=P ₁+ P ₂+ ... + P _n=const (2)

The i.e. Q of the amount of drawing gas always of all units _zWith total electric load P _zBe respectively constant.

Second constraint condition is: Q _i∈ (Q _Imin, Q _Imax), P _i∈ (P _Imin, P _Imin) (3)

The minimax electric load that is every unit is respectively: P _1min, P _1maxP _2min, P _2max P _Nmin, P _NmaxThe minimax amount of drawing gas is respectively Q _1min, Q _1maxQ _2min, Q _2max Q _Nmin, Q _Nmax

Step 5, based on improving electric load that genetic algorithm obtains overall heat consumption value

each unit hour that satisfies all units and drawing gas value: detailed process is following

1, initial population is set

Matrix with 2n * m then can be represented initial population:

$\left(\begin{array}{c}{P}_{11}{P}_{12}......P_{1n}{Q}_{11}{Q}_{12}......Q_{1n}\\ P_{21}{P}_{22}......P_{2n}{Q}_{21}{Q}_{22}......Q_{2n}\\ ............................\\ P_{m1}{P}_{m2}......P_{\mathrm{mn}}{Q}_{m1}{Q}_{m2}......Q_{\mathrm{mn}}\end{array}\right)$

The individual number of m for setting, the amount of drawing gas Q _iWith electric load P _iBe the random number that satisfies second constraint condition; Above-mentioned initial population adopts the formal construction of the constraint coding that satisfies second constraint condition;

The electric load of (n-1) individual unit before in the above-mentioned initial population and the amount of drawing gas are satisfied the coding of second constraint condition, and last unit passes through computes:

$\left\{\begin{array}{c}{P}_n=P_z-P_1-P_2-\·\·\·-P_{n-1}\\ Q_n=Q_z-Q_1-Q_2-\·\·\·-Q_{n-1}\end{array}\right.---\left(4\right)$

Can obtain the initial population that all units satisfy first constraint condition and satisfied second constraint condition of preceding (n-1) individual unit like this:

$\left(\begin{array}{c}{P}_{11}{P}_{12}......P_{1(n-1)}{Q}_{11}{Q}_{12}......Q_{1(n-1)}\\ P_{21}{P}_{22}......P_{2(n-1)}{Q}_{21}{Q}_{22}......Q_{2(n-1)}\\ ............................\\ P_{m1}{P}_{m2}......P_{m(n-1)}{Q}_{m1}{Q}_{m2}......Q_{m(n-1)}\end{array}\right)$

Following formula n platform unit is to satisfy difference maximum and the maximum heating load of its electric load maximal value and minimum value and the unit that the minimum thermal load does not wait, that is:

P _nmax-P _nmin＞P _imax-P _imin

Q _nmax-Q _nmin≠0

P _Nmax, P _NminThe maximum electrical load and minimum electric load of representing the selected n platform unit that comes out; P _Imax, P _IminThe maximum electrical load and the minimum electric load of expression residue unit; Q _Nmax, Q _NminMaximum heating load and the minimum thermal load of representing the selected n platform unit that comes out;

2, make up fitness function: calculate through fitness, realize individual optimized choice, make also satisfied second constraint condition of n platform unit in the Optimization result simultaneously;

Ineligible individuality is:

P _n＜P _min OR?P _n＞P _max

Q _n＜Q _min OR?Q _n＞Q _max

Thermoelectric load distribution optimization is the minimum value of asking objective function, and the optimization aim of genetic algorithm is to find the individuality with maximum adaptation degree, so definition fitness function ObjV definition is as follows:

1), for qualified individuality; P _Min＜P _n＜P _Max&Q _Min＜Q _n＜Q _Max

$\mathrm{ObjV}=\frac{1}{\stackrel{\‾}{R}}---\left(5\right)$

2), for ineligible individuality: P _n＜P _MinOR P _n＞P _Max, Q _n＜Q _MinOR Q _n＞Q _Max, adopt index measure transform (1) objective function

$\mathrm{ObjV}=\frac{1}{\stackrel{\‾}{R_{\max }}\×\exp \left(\mathrm{\α\β}\right)}---\left(6\right)$

Wherein: work as P _n＜P _MinThe time, $\mathrm{\β}={\mathrm{\β}}_p=\frac{P_{\mathrm{Min}}-P_n}{P_{\mathrm{Min}}}$

Work as P _n＞P _MaxThe time, $\mathrm{\β}={\mathrm{\β}}_p=\frac{P_n-P_{\mathrm{Max}}}{P_{\mathrm{Max}}}$

For Q _nIn like manner can get:

Work as Q _n＜Q _MinThe time, $\mathrm{\β}={\mathrm{\β}}_Q=\frac{Q_{\mathrm{Min}}-Q_n}{Q_{\mathrm{Min}}}$

Work as Q _n＞Q _MaxThe time, $\mathrm{\β}={\mathrm{\β}}_Q=\frac{Q_n-Q_{\mathrm{Max}}}{Q_{\mathrm{Max}}}$

For β _P, β _QWhen existing simultaneously, β=max (β _P, β _Q)

is the maximum hear rate of unit in operational process, obtains through power plant's production and test figure.

α is a constant coefficient, and target is to make to work as the P that calculates _n, Q _nWhen surpassing setting threshold 100%, its fitness value is greater than 100 times of the fitness value that satisfies condition down, i.e. β=1, exp α)＞100; α gets 5 in experiment;

Like this in selection course; The individuality that fitness is little will have very big probability to be eliminated; The individuality that does not satisfy second constraint condition simultaneously also will have very big probability to be eliminated; Thereby realize individual optimized choice, obtaining the maximum adaptation degree under the ideal state is the minimum individuality of overall heat consumption value

;

3, in the completion after the step, carry out selection, intersection, mutation process again based on the traditional genetic algorithm; When genetic algebra reach end condition N for the time; Genetic process stops, the electric load of the overall heat consumption value

that all units are satisfied in output each unit hour and the hear rate value of value optimum solution, each unit and the minimum overall heat consumption of all units accordingly of drawing gas.

Be described further (step 5) again to the inventive method:

Genetic algorithm is to be the basis with natural selection and theory of heredity, the efficient global optimizing searching algorithm that survival of the fittest rule in the biological evolution process is combined with the colony chromosomal random information exchanging mechanism in inside.Genetic algorithm has been abandoned traditional way of search, and the evolutionary process of simulation organic sphere adopts artificial evolution's mode that object space is carried out the random optimization search.It regards the feasible solution in the problem as in the colony body one by one, and each coding is weaved into the form of symbol string, simulates Darwinian heredity and selects the evolutionary process with natural selection, and colony is carried out the operation (heredity, intersect, make a variation) based on heredity repeatedly.Target fitness function according to intended target is estimated each individuality; Evolutionary rule according to the survival of the fittest, the survival of the fittest; Constantly obtain optimum colony, search the optimum individual of optimizing in the colony with overall parallel search mode simultaneously, in the hope of the optimum solution that satisfies condition.

The general process of genetic algorithm is: initial population is set, calculates fitness, select; Intersect, variation produces new population; Recomputate fitness, loop iteration reaches initial set value up to iterations successively; Heredity finishes, and last population that obtains is optimum population in generation, and the individuality in the population is an optimum individual.In this example; The purpose that we optimize is overall heat consumption value

minimum that makes all units, wherein

$\stackrel{\‾}{R}=\frac{P_1\×R_1+P_2\×R_2+......P_n\×R_n}{P_1+P_2+......+P_n}---\left(1\right)$

The amount of the drawing gas Q of each unit is perhaps obtained in user's input in real time from system in the interface _i(characterizing thermal load) and electric load P with it _i, can obtain each electric load and the corresponding hear rate R of the amount of drawing gas through checking and verify the border heat consumption curve _i(i ∈ [1, n]) is through R _iJust can calculate the overall heat consumption of all units with following formula.Optimization method is following:

1, initial population is set

The parameter in space because genetic algorithm can not directly be handled problems therefore must be perhaps individual the chromosome that the feasible solution that requires problem is expressed as hereditary space through coding.Coding method commonly used has binary coding, gray encoding, multistage parameter coding, orderly string encoding etc.Because this optimization problem is the continuous function optimization problem of multidimensional, high-precision requirement; Codings such as use scale-of-two represent that individuality has some disadvantages; The lot of valuable experience that people are summed up in the research of some classic optimisation algorithms also just can't be used, and also is not easy to handle the constraint condition of non-trivial.In order to overcome the shortcoming of binary coding method, this paper adopts floating-point encoding.For the requirement that population is provided with, each individuality must be the feasible solution of this optimization problem, optimizes so just to be of practical significance.Illustrate: if formula (1) is the equation of a pure mathematics, P _i, Q _iAny value all are feasible solutions of formula (1), then cataloged procedure then can be represented initial population as long as generate the matrix of a 2n*m:

$\left(\begin{array}{c}{P}_{11}{P}_{12}......P_{1n}{Q}_{11}{Q}_{12}......Q_{1n}\\ P_{21}{P}_{22}......P_{2n}{Q}_{21}{Q}_{22}......Q_{2n}\\ ............................\\ P_{m1}{P}_{m2}......P_{\mathrm{mn}}{Q}_{m1}{Q}_{m2}......Q_{\mathrm{mn}}\end{array}\right)$

The individual number of m for setting, P _i, Q _iBe random number.Such coding is called for there not being the constraint coding.

But the electric load of each unit has bound with the amount of drawing gas in real process, i.e. Q _Max, Q _MinAnd P _Max, P _Min, the overall electric load of all units is a definite value with the amount of drawing gas, promptly

Q _z＝Q ₁+Q ₂+…+Q _n＝const，P _z＝P ₁+P ₂+…+P _n＝const

So electric load is an equality constraint with the amount of drawing gas; And genetic algorithm is because its randomness is difficult to solve equality constraint, even perhaps adopt constrained coding; Then initial population then produces under a qualifications; So just run counter to the principle of genetic algorithm simulation biological evolution, individual randomness and diversity are restricted, and the optimization Algorithm effect can be had a greatly reduced quality.So our innovation is to have solved through a series of conversion the problem of equality constraint, and can guarantee that the final individuality of optimizing out is optimum solution.

Known constraints condition 1 is: Q _z=Q ₁+ Q ₂+ ... + Q _n=const, P _z=P ₁+ P ₂+ ... + P _n=const (2)

The i.e. Q of the amount of drawing gas always of all units _zWith total electric load P _zBe respectively constant.

Constraint condition 2 is: Q _i∈ (Q _Min, Q _Max), P ∈ (P _Min, P _Max) (3)

The minimax electric load that is every unit is respectively: P _1min, P _1maxP _2min, P _2max P _Nmin, P _NmaxThe minimax amount of drawing gas is respectively Q _1min, Q _1maxQ _2min, Q _2max Q _Nmin, Q _Nmax

In order to make coding satisfy the actual conditions of unit operation; Promptly satisfy constraint condition 1; We are revised as the coding that the electric load of (n-1) individual unit before in the above-mentioned initial population and the amount of drawing gas is satisfied second constraint condition with coding, and last unit passes through computes:

$\left\{\begin{array}{c}{P}_n=P_z-P_1-P_2-\·\·\·-P_{n-1}\\ Q_n=Q_z-Q_1-Q_2-\·\·\·-Q_{n-1}\end{array}\right.---\left(4\right)$

The initial population that obtains is:

$\left(\begin{array}{c}{P}_{11}{P}_{12}......P_{1(n-1)}{Q}_{11}{Q}_{12}......Q_{1(n-1)}\\ P_{21}{P}_{22}......P_{2(n-1)}{Q}_{21}{Q}_{22}......Q_{2(n-1)}\\ ............................\\ P_{m1}{P}_{m2}......P_{m(n-1)}{Q}_{m1}{Q}_{m2}......Q_{m(n-1)}\end{array}\right)$

Following formula n platform unit is to satisfy difference maximum and the maximum heating load of its electric load maximal value and minimum value and the unit that the minimum thermal load does not wait, that is:

P _nmax-P _nmin＞P _imax-P _imin

Q _nmax-Q _nmin≠0

P _Nmax, P _NminThe maximum electrical load and minimum electric load of representing the selected n platform unit that comes out; P _Imax, P _IminThe maximum electrical load and the minimum electric load of expression residue unit; Q _Nmax, Q _NminMaximum heating load and the minimum thermal load of representing the selected n platform unit that comes out.

Completely random during being provided with of initial population so not only, first constraint condition (formula (2)) has also well satisfied.But for formula (3), can not well satisfy when initial population is encoded, that is to say, can produce some insignificant individualities in the cataloged procedure and make:

P _n＜P _min OR?P _n＞P _max

Q _n＜Q _min OR?Q _n＞Q _max

We need cast out these individualities in the process of optimization at the back as far as possible, and left is exactly qualified individuality.

2, fitness function

Fitness function is the standard that is used for distinguishing the individual in population quality, is unique foundation of carrying out natural selection.Thermoelectric load distribution optimization is to ask the minimum of a function value, and the optimization aim of genetic algorithm is to find the individuality with maximum adaptation degree, so definition fitness function ObjV definition is as follows:

1), for qualified individuality: P _Min＜P _n＜P _Max&Q _Min＜Q _n＜Q _Max

$\mathrm{ObjV}=\frac{1}{\stackrel{\‾}{R}}---\left(5\right)$

2), for ineligible individuality: P _n＜P _MinOR P _n＞P _Max, Q _n＜Q _MinOR Q _n＞Q _Max. adopt index measure transform (1) objective function

$\mathrm{ObjV}=\frac{1}{\stackrel{\‾}{R_{\max }}\×\exp \left(\mathrm{\α\β}\right)}---\left(6\right)$

Wherein: work as P _n＜P _MinThe time, $\mathrm{\β}={\mathrm{\β}}_p=\frac{P_{\mathrm{Min}}-P_n}{P_{\mathrm{Min}}}$

Work as P _n＞P _MaxThe time, $\mathrm{\β}={\mathrm{\β}}_p=\frac{P_n-P_{\mathrm{Max}}}{P_{\mathrm{Max}}}$

For Q _nIn like manner can get:

Work as Q _n＜Q _MinThe time, $\mathrm{\β}={\mathrm{\β}}_Q=\frac{Q_{\mathrm{Min}}-Q_n}{Q_{\mathrm{Min}}}$

Work as Q _n＞Q _MaxThe time, $\mathrm{\β}={\mathrm{\β}}_Q=\frac{Q_n-Q_{\mathrm{Max}}}{Q_{\mathrm{Max}}}$

For β _P, β _QWhen existing simultaneously, β=max (β _P, β _Q)

is the maximum hear rate of unit in operational process, obtains through power plant's production and test figure.

α is a constant coefficient, and target is to make to work as the P that calculates _n, Q _nWhen surpassing setting threshold 100%, its fitness value is greater than 100 times of the fitness value that satisfies condition down, i.e. β=1, exp (α)＞100; α gets 5 in experiment.

We just can draw such conclusion like this; When the electric load and the amount of drawing gas Pn of last unit that calculates through population, when Qn did not satisfy condition, its fitness value can amplify according to its degree that exceeds setting threshold; Exceed manyly more, amplify more severe (exponential increase).(can know by formula (5) and formula (6); The degree that exceeds setting threshold is amplified more; The fitness value is just more little) like this in selection course; The individuality that fitness is little will have very big probability to be eliminated; Thereby realize individual optimized choice, obtaining the maximum adaptation degree under the ideal state is the minimum individuality of overall heat consumption value

.

3, select

Selection operation selects defect individual to form new population with certain probability from old colony, obtains of future generation individual with breeding.Individual selected probability is relevant with fitness value, and the ideal adaptation degree is high more, and selected probability is big more.This paper adopts the roulette method, and promptly based on the selection strategy of fitness ratio, individual selected probability is:

$p_i=\frac{F_i}{{\∑}_j^N{F}_j}$

Fi is this individual fitness value, be all ideal adaptation degree value sums.

4, interlace operation

Because this paper adopts floating-point encoding, so choosing arithmetic, corresponding Crossover Strategy intersects, be to produce two new individualities by the linear combination of two individuals.Suppose at two individuals X _A, X _BBetween carry out arithmetic and intersect, then be by two new individualities that produce after the arithmetical operation:

X′ _A＝aX _B+(1-a)X _A

X′ _B＝aX _A+(1-a)X _B

Wherein a is a parameter, and a can be a constant, also can be the variable that is determined by evolutionary generation.This paper adopts a is set is a constant 0.8.

5, variation

Variation can improve the local search ability of genetic algorithm and can keep the diversity of population.Variation strategy commonly used has the variation of basic position, evenly variation, border variation etc.Thermoelectric load distribution problem is complicated nonlinear problem; Good effect can be arranged near optimum solution; Must search optimum solution but be difficult to determinacy, in order to address this problem, this paper has adopted the Gaussian approximation variation; Can improve the local search ability of genetic algorithm, and have certain probability to make algorithm jump out local minimum point the focussing search zone.Use during concrete operations that to meet average be original parameter value, variance is that a random number of the normal distribution of original parameter value square is replaced original genic value, can be known that by the characteristic of normal distribution Gaussian mutation also is near the regional area the original individuality of focussing search.Concrete formula is following:

$f\left(x\right)=\frac{1}{\sqrt{2\mathrm{\π}}{q}^2}{e}^{-\frac{{(x-q)}^2}{2q^2}}$

Wherein q is original genic value.

6, end condition

When genetic algebra reached for 50 generations, genetic process stopped automatically, output optimum solution and corresponding minimum overall heat consumption.

Utilize the embodiment (with reference to figure 1～4) of the inventive method:

With 4 units of certain power plant is example:

1 pair every unit heat consumption curve is provided with.Carry out site test, draw every heat consumption curve under the unit difference amount of drawing gas.According to the size of factor off-design values such as the variation of every unit condenser back pressure, main steam pressure, main steam temperature, reheat pressure, reheat steam temperature, feed temperature, inquiry consumption difference curve is revised heat consumption curve.Obtain the final heat consumption curve of revising that advanced, as shown in Figure 3.

2 pairs of optimal conditions are provided with.Import the maximum/miniwatt of every unit, maximum/amount of drawing gas for a short time, as shown in Figure 4.

3 select " Automatic Optimal ", are optimized.Optimization result is as shown in Figure 4, can find out that power plant's hear rate has reduced 45.1kJ/kWh through after this system optimization.Result after the optimization can write mis system, and the operations staff controls steam turbine according to the result.

4 supervisory computers read current set state data (variation of condenser back pressure, main steam pressure, main steam temperature, reheat pressure, reheat steam temperature, feed temperature) from mis system; Get electric load Q, the thermal load P of user's request up till now, obtain current each unit hear rate value R and power plant's overall heat consumption value

with consumption difference fair curve through heat consumption curve C under each the unit design conditions in the searching and managing computing machine

5 managerial personnel can limit through the optimal conditions (like the scope of electric load, thermal load) of computing machine to every unit.

6 utilize the improvement genetic algorithm to be optimized.Output Optimization result, the i.e. optimum thermoelectric load distribution scheme of every unit and power plant's hear rate value, reduction hear rate value.

7 Optimization result input mis system, the operations staff makes amendment to unit parameter according to Optimization result, realizes the control to unit.

As shown in Figure 2, utilize the software of the method for the invention to be written in the supervisory computer among Fig. 2, the electric power supply plant managerial personnel carry out energy-saving monitoring and adjustment.Supervisory computer reads the parameters of current unit from mis system: electric load, thermal load, condenser back pressure, main steam pressure, main steam temperature, reheat pressure, reheat steam temperature, feed temperature.Through inquiry unit design heat consumption curve and consumption difference fair curve, obtain current each unit hear rate value R ₁, R ₂..., R _nWith power plant's overall heat consumption value

Managerial personnel can be through optimal conditions (the minimax electric load P of supervisory computer to every unit _1min, P _1max, P _2min, P _2max..., P _Nmin, P _NmaxWith the minimax amount of drawing gas Q _1min, Q _1max, Q _2min, Q _2max..., Q _Nmin, Q _Nmax) limit.

Claims (3)

1. one kind based on the thermoelectric load distribution optimization method of the extraction for heat supply unit that improves genetic algorithm, and it is characterized in that: the detailed process of said method is:

Step 1, the actual heat consumption curve of unit is set: the actual heat consumption curve that obtains every unit according to test; Said actual heat consumption curve is meant that be that Q is an independent variable with power P with the amount of drawing gas, and hear rate value R is gang's curve of dependent variable, promptly

The 1st unit: R ₁=f (P ₁, Q ₁);

The 2nd unit: R ₂=f (P ₂, Q ₂);

……

N platform unit: R _n=f (P _n, Q _n);

Step 2, unit consumption difference fair curve is set, confirms that unit consumption difference revises overall coefficient θ _iI is the machine group #, i=1,2; N, n represent the unit number: all can exert an influence to hear rate during based on condenser back pressure, main steam pressure, main steam temperature, reheat pressure, reheat steam temperature, these six factor off-design values of feed temperature, provide according to producer then or the consumption difference fair curve of power plant checks in the hear rate correction factor Δ of each influence factor of every unit _1iΔ _2iΔ _3iΔ _6i, Δ _1iΔ _2iΔ _3iΔ _6iBe respectively condenser back pressure, main steam pressure, main steam temperature, reheat pressure, the reheat steam temperature of every unit, the hear rate correction factor of feed temperature; Make θ _i=Δ _1iΔ _2iΔ _3iΔ _6i

Step 3, obtain unit design heat consumption curve: revise overall coefficient θ according to every unit consumption difference _iThe actual hear rate correction of unit is obtained unit design heat consumption curve, and the design heat consumption curve of each unit is:

The 1st unit: R ₁=θ ₁F (P ₁, Q ₁);

The 2nd unit: R ₂=θ ₂F (P ₂, Q ₂);

……

N platform unit: R _n=θ _nF (P _n, Q _n);

Step 4, obtain the amount of the drawing gas Q of each unit _i(characterizing thermal load) and electric load P with it _i: the amount of the drawing gas Q that records each unit earlier _iWith electric load P _i, obtain corresponding hear rate R through the described unit design of step 3 heat consumption curve then _i, i ∈ [1, n],

The electric load that obtains the n of power plant platform unit is respectively P ₁, P ₂..., P _n, the amount of drawing gas is respectively Q ₁, Q ₂..., Q _n, the hear rate value is R ₁, R ₂..., R _n, n is the unit number;

The purpose of optimizing is the overall heat consumption value that obtains making all units

P hour ₁, P ₂..., P _n, Q ₁, Q ₂..., Q _nAllocative decision, wherein objective function is:

Set constraint condition:

First constraint condition is: Q _z=Q ₁+ Q ₂+ ... + Q _n=const, P _z=P ₁+ P ₂+ ... + P _n=const (2)

The i.e. Q of the amount of drawing gas always of all units _zWith total electric load P _zBe respectively constant;

Second constraint condition is: Q _i∈ (Q _Imin, Q _Imax), P _i∈ (P _Imin, P _Imin) (3)

The minimax electric load that is every unit is respectively: P _1min, P _1maxP _2min, P _2max P _Nmin, P _NmaxThe minimax amount of drawing gas is respectively Q _1min, Q _1maxQ _2min, Q _2max Q _Nmin, Q _Nmax);

Step 5, based on improving electric load that genetic algorithm obtains overall heat consumption value each unit hour that satisfies all units and drawing gas value: detailed process is following

1, initial population is set

Matrix with 2n * m then can be represented initial population:

The individual number of m for setting, the amount of drawing gas Q _iWith electric load P _iBe the random number that satisfies second constraint condition; Above-mentioned initial population adopts the formal construction of the constraint coding that satisfies second constraint condition;

Electric load to preceding (n-1) individual unit in the above-mentioned initial population is encoded with the constraint that the amount of drawing gas satisfies second constraint condition, and last unit passes through computes:

Can obtain the initial population that all units satisfy first constraint condition and satisfied second constraint condition of preceding (n-1) individual unit like this:

Following formula n platform unit is to satisfy difference maximum and the maximum heating load of its electric load maximal value and minimum value and the unit that the minimum thermal load does not wait, that is:

P _nmax-P _nmin＞P _imax-P _imin

Q _nmax-Q _nmin≠0

P _Nmax, P _NminThe maximum electrical load and minimum electric load of representing the selected n platform unit that comes out; P _Imax, P _IminThe maximum electrical load and the minimum electric load of expression residue unit; Q _Nmax, Q _NminMaximum heating load and the minimum thermal load of representing the selected n platform unit that comes out.

2. structure fitness function: calculate through fitness, realize individual optimized choice, make simultaneously that n platform unit also satisfies second constraint condition in the Optimization result;

Ineligible individuality is:

P _n＜P _min OR?P _n＞P _max

Q _n＜Q _min OR?Q _n＞Q _max

Thermoelectric load distribution optimization is the minimum value of asking objective function, and the optimization aim of genetic algorithm is to find the individuality with maximum adaptation degree, so definition fitness function ObjV definition is as follows:

1), for qualified individuality: P _Min＜P _n＜P _Max&Q _Min＜Q _n＜Q _Max

2), for ineligible individuality: P _n＜P _MinOR P _n＞P _Max, Q _n＜Q _MinOR Q _n＞Q _Max, adopt the objective function shown in the index measure transform (1),

Wherein: work as P _n＜P _MinThe time,

Work as P _n＞P _MaxThe time,

For Q _nIn like manner can get:

Work as Q _n＜Q _MinThe time,

Work as Q _n＞Q _MaxThe time,

For β _P, β _QWhen existing simultaneously, β=max (β _P, β _Q)

is the maximum hear rate of unit in operational process, obtains through power plant's production and test figure;

α is a constant coefficient, and target is to make to work as the P that calculates _n, Q _nWhen surpassing setting threshold 100%, its fitness value is greater than 100 times of the fitness value that satisfies condition down, i.e. β=1, exp (α)＞100; α gets 5 in experiment;

Like this in selection course; The individuality that fitness is little will have very big probability to be eliminated; The individuality that does not satisfy second constraint condition simultaneously also will have very big probability to be eliminated; Thereby realize individual optimized choice, obtaining the maximum adaptation degree under the ideal state is the minimum individuality of overall heat consumption value

.

3. after step is gone up in completion, carry out selection, intersection, mutation process again based on the traditional genetic algorithm; When genetic algebra reach end condition N for the time; Genetic process stops, the electric load of the overall heat consumption value

that all units are satisfied in output each unit hour and the hear rate value of value optimum solution, each unit and the minimum overall heat consumption of all units accordingly of drawing gas.

Images