CN106227927A - The optimization method of tower type solar thermoelectricity systems, spot collection thermal sub-system - Google Patents

Abstract

The invention discloses the optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system, implement step as follows: (1) is by heliostat field piecemeal, with receptor surface focus position as decision variable, try one's best a height of target with receptor outlet temperature of molten salt, it is less than threshold value for constraint, constitution optimization problem with receptor surface energy distribution standard deviation；(2) improvement complex method solving-optimizing problem is utilized；(3) utilization is based on MBDFO algorithm improves and optimizates method solving-optimizing problem.In the present invention; the optimization method of tower type solar thermoelectricity systems, spot collection thermal sub-system; under conditions of guarantee receptor outlet temperature of molten salt tries one's best height; make the distribution of receptor surface energy more uniformly to avoid hot-spot; being conducive to protecting receptor, improving heat exchange efficiency, the operation for tower type solar thermo-power station provides reference.

Description

The optimization method of tower type solar thermoelectricity systems, spot collection thermal sub-system

Technical field

The present invention relates to tower-type solar thermal power generating system field, particularly relate to a kind of tower type solar thermoelectricity system and gather The optimization method of light collection thermal sub-system.

Background technology

Tower-type solar thermal power generating system is the heliostat utilizing the independently tracked sun in a lot of face, focuses light rays at one It is fixed on the receptor of top of tower, and is used with the form of heat energy, drive steam turbine, electromotor to generate electricity.Whole tower Formula solar energy system is divided into optically focused, thermal-arrest, accumulation of heat, supplementary energy and 5 subsystems of generating, has the sun during energy transmission The conversion of energy-heat energy-mechanical energy-electric energy.Wherein light and heat collection subsystem is the critical system that solar energy is converted into heat energy, research The Optimization Work of light and heat collection subsystem is conducive to improving generating efficiency, and the security operations of receptor is had important guidance Effect.

About the optimization of tower type solar light and heat collection system, existing research is design optimization mostly；Performance optimizes big Part is the local optimum for certain single subsystem, lacks the optimizing research on the basis of total system.

Summary of the invention

The invention provides the optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system, the technology of employing Scheme is as follows:

The optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system comprises the steps:

1) by heliostat field piecemeal, with receptor surface focus position as decision variable, fused salt temperature is exported with receptor Degree as far as possible a height of target in the case of nonvolatile, with receptor surface energy distribution standard deviation less than threshold value for constraint, structure Optimization problem；

2) utilize and improve complex method or based on MBDFO algorithm improve and optimizate method solving-optimizing problem；According to solving The position of result adjustment focus point realizes the optimization of tower type solar thermoelectricity systems, spot collection thermal sub-system.

Further, described step 1) particularly as follows:

Jing Chang being divided into two regions, uses double focus strategy, the most a part of heliostat is irradiated to focus point 1, another Part heliostat is irradiated to focus point 2, and two focus point coordinates are respectively (0,0, z₁), (0,0, z₂), export fused salt with receptor Temperature as far as possible a height of target in the case of nonvolatile, with receptor surface energy distribution standard deviation less than threshold value for constraint, structure Make shown in optimization problem such as formula (1):

Wherein, T_vRepresent the volatilization temperature value of fused salt, for constant, T_fssTemperature of molten salt steady-state value, its table is exported for receptor It is shown as z₁, z₂Function fitness (z₁,z₂), z₁And z₂Representing two focus point z-axis coordinates, its value must not exceed receptor Vertically height h_t, F is the standard deviation of receptor surface energy values, is expressed as independent variable z₁, z₂Function g (z₁,z₂), F₀For setting Maximum standard deviation.

Further, when optimization method is for improving complex method, the process of described solving-optimizing problem particularly as follows:

Optimized model shown in formula (1) is expressed as minf (x), x ∈ Rⁿ, s.t.g (x) >=0, wherein, x=[z₁；z₂], Representing the z coordinate of two focus points, object function f (x) represents receptor outlet temperature steady-state value T_fss, constraint function g (x) wraps Include T_v-T_fss、F₀-F andThree constraints, the algorithm steps of complex method is as follows:

3.1) intial compound form { x is chosen⁰,x¹,···,xⁿ, reflection coefficient α ＞ 1, spreading coefficient γ ＞ 1, shrink system Number

β ∈ (0,1) and precision ε ＞ 0；

3.2) judge that intial compound form, whether in feasible zone, if not in feasible zone, is adjusted according to given scaling criterion Whole initial value position, until intial compound form is in feasible zone；If in feasible zone, by n+1 the summit according to target letter of complex The size of numerical value renumbers, and makes the numbering on summit meet f (x⁰)≤f(x¹)≤…≤f(x^n-1)≤f(xⁿ)；

3.3) orderIfStop iteration, export x⁰, otherwise turn Enter step 3.4)；

3.4) x is calculatedⁿ⁺²=xⁿ⁺¹+α(xⁿ⁺¹-xⁿ), check xⁿ⁺²Whether in feasible zone, the most whether meet g (xⁿ⁺²) >=0, If not in feasible zone, reflection coefficient α is decreased up to xⁿ⁺²In feasible zone, calculate f (xⁿ⁺²), if f is (xⁿ⁺²) ＜ f (x⁰), turn Enter step 3.5), otherwise as f (xⁿ⁺²) ＜ f (x^n-1) time proceed to step 3.6), as f (xⁿ⁺²)≥f(xⁿ-¹) proceed to step 3.7)；Institute The minishing method stating reflection coefficient α is the extraction of root or is multiplied by a coefficient more than 0 less than 1；

3.5) x is calculatedⁿ⁺³=xⁿ⁺¹+γ(xⁿ⁺²-xⁿ⁺¹), check xⁿ⁺³Whether in feasible zone, if not in feasible zone, will Spreading coefficient γ decreases up to xⁿ⁺³In feasible zone, if f is (xⁿ⁺³) ＜ f (x0), make xⁿ=xⁿ⁺³, proceed to step 3.2), otherwise turn Enter step 3.6)；Described spreading coefficient γ minishing method is the extraction of root or is multiplied by a coefficient more than 0 less than 1；

3.6) x is madeⁿ=xⁿ⁺², proceed to step 3.2)；

3.7) x is madeⁿ={ xⁱ|f(xⁱ)=min (f (xⁿ),f(xⁿ⁺²)), calculate xⁿ⁺⁴=xⁿ⁺¹+β(xⁿ-xⁿ⁺¹), check xⁿ⁺⁴ Whether in feasible zone, if not in feasible zone, constriction coefficient β is decreased up to xⁿ⁺⁴In feasible zone, if f is (xⁿ⁺⁴) ＜ f (xⁿ), make xⁿ=xⁿ⁺⁴, proceed to step 3.2), otherwise proceed to step 3.8)；The minishing method of described constriction coefficient β be the extraction of root or Person is multiplied by a coefficient more than 0 less than 1；

3.8) x is made^j=x⁰+θ(x^j-x⁰), j=0,1, n, proceed to step 3.2).

Further, it is based on MBDFO algorithm when improving and optimizating method, described solving-optimizing method when optimization method Principle be:

Wherein, owing to the gradient information of model is unavailable,WithIt is false.So, by inserting Value conditional definition scalar c, vector g ∈ Rⁿ, symmetrical matrix G ∈ R^n×n, interpolation condition definition method is discussed below.

m_k(y^l)=f (y^l), l=1,2 ..., q. (3)

Model shown in formula (2) comprisesIndividual coefficient (the i.e. coefficient sum of c, g and G, it is contemplated that G's Symmetry), the interpolation condition shown in formula (3) defines model m_kUniqueness, as long as q meets formula (4).

So, formula (3) can be converted into a quadratic linear equation group.If selecting interpolation point y¹,y²,…,y^q, then This linear system is nonsingular, model m_kIt is unique.

Model m_kOnce set up, then carry out material calculation p value by approximate solution Trust-region subproblem.

If x_k+ p has fully reappeared object function, then next iteration is defined as x_k+1=x_k+ p, and update trusted zones Radius, new iteration starts；Otherwise, reject this step-length, improve interpolation territory Y or reduce trusted zones.

In order to reduce algorithm complex, every iteration once updates a model m_kRather than start anew to recalculate.Choosing Select a convenient and simple method and simulate the curve of quadratic polynomial, be generally selected Lagrange or newton multinomial.These It is the most suitable that the feature of basic method may serve to weigh interpolation collection Y exactly, and can improve slotting needs when Value collection.

In Trust Region Algorithm, it is real that the most suitable and Trust Region Radius the more New Policy of step-length is all based on object function The ratio of border difference and the difference of forecast model, i.e.

Wherein, x_k ⁺Represent test point.

If ρ >=η meets, then it represents that obtaining the abundant reduction of object function, this is simplest situation.In this situation Under, test point x_k ⁺An element in Y can be deleted as new iteration point, makes x_k ⁺It is included in Y.

When ρ >=η is unsatisfactory for, then may be caused by two reasons, one is that interpolated sample collection Y sample is not abundant, separately One is that trusted zones is the biggest.First reason typically occurs in iteration and is limited in a lower dimensional space not including feasible zoneIn.This algorithm can converge to this minimal subset.Above-mentioned this situation can by monitoring formula (3) define linear The interpolation condition of system detects.If conditional number is the highest, then change Y set, it is common that replace one with a new element Geju City element is so that interplotation system (3) is the most nonsingular.If the interpolation condition of Y meets condition, then can subtract simply Little Trust Region Radius.

Polynomial interopolation theoretical and update interpolation collection method be described below:

(1) polynomial interopolation

First it is discussed in detail and how to use interpolation method to set up target function model.Consider a linear model, such as formula (7) institute Show:

m_k(x_k+ p)=f (x_k)+g^Tp (7)

In order to determine vector matrixUtilize interpolation condition m_k(y^l)=f (y^l), l=1,2 ..., n, can be write as Shown in formula (8):

(s^l)^TG=f (y^l)-f(x_k), l=1,2 ..., n (8)

Wherein:

s^l=y^l-x_k, l=1,2 ..., n (9)

Conditional (7) represents a system of linear equations, and the row of coefficient matrix is by vector (s^l)^TBe given.As can be seen here, model (7) uniquely determined by formula (8), and if only if interpolation point { y¹,y²,…,y^qIt is included in set { s^l: l=1,2 ..., in n} and It it is linear independence.If this condition is set up, then by an x_k,y¹,y²,...,yⁿThe simplex of composition is nondegenerate (nondegenerate).

Consider now how to build the secondary model of one such as formula (2) form, as f=f (x_k) time.Rewrite model such as formula (10) shown in:

Wherein, the element in g Yu G is the unknown vector of q-1 dimension, and as shown in formula (11), p is the unknown vector of q-1 dimension, as Shown in formula (12):

Model (10) is identical with the form of model (7), unknown vectorCoefficient can be able to calculate under linear case.

Multiple quadratic function can represent by different modes.The advantage of monomial base (10) can be simply by setting Element in G is 0 to obtain the structure of Hessian matrix.But, other bases can avoid the singular point of model (3) more easily.

Pass throughOne group of base in the n dimensional linear space represented.Function (3) therefore can be expressed as formula (13) institute Show:

For some factor alpha_i, interpolation collection Y={y¹,y²,…,y^qThe definition that can pass through formula (14) uniquely determines α_i, and Determinant defined in formula (14) is non-zero:

(2) interpolation collection Y is updated

When MBDFO algorithm is iterated, determinant δ (Y) may be close to zero, and numerical solution difficulty can be caused even to lose Lose.Therefore some algorithms comprise the mechanism keeping interpolation point correctly to configure.One of which mechanism is described below.Row is waited until with it Column δ (Y) becomes less than a threshold value, might as well call a geometry when test point can not fully reduce object function f Regulation mechanism, the target of this geometry Regulation mechanism is to replace an interpolation point to make the value of determinant (14) increase.Wherein, use The following character of δ (Y), this states according to Lagrangian.

For any y ∈ Y, definition LagrangianL (, multinomial y), degree up to 2, namely L (y, y)= 1,AndY ∈ Y, shown in two dimension Lagrangian such as formula (15).

Assume that the renewal gathering Y is by deleting a some y_-Increase a new some y₊, thus obtain new set Y₊.That May certify that formula (16) (under a suitable standardization and given specified conditions).

|δ(Y⁺)|≤|L(y₊,y_-)||δ(Y)| (16)

MBDFO algorithm can make full use of this inequality and update interpolation collection.

First the first situation (ρ >=η), test point x are considered⁺Can fully reduce object function f.Use x⁺Update the point in Y y_-。

For (16), the some y of deletion_-Definition is as shown in formula (17):

It follows that consider the second situation (ρ >=η), i.e. it is the most abundant that object function f reduces.First determine whether Y needs Improve, use following rule to determine.At current iteration point x_kFor any yⁱ∈ Y meets | | x_k-yⁱ| | during≤Δ it is considered that Y is suitable, in this case, reduces Trust Region Radius and starts next iteration.

If Y is inappropriate, call geometry Regulation mechanism.Select a some y_-∈ Y replaces, substitution point y⁺Can Increase the value of determinant (14).For any yⁱ∈ Y, it is possible to replace its potential valueBe given for formula (18):

The point y that will delete_-For

Preferred, when optimization method is based on MBDFO algorithm when improving and optimizating method, and described solves as the present invention The process of optimization problem particularly as follows:

Optimized model shown in formula (1) is expressed as minf (x), x ∈ Rⁿ, s.t.g (x) >=0, wherein, x=[z₁；z₂], Representing the z coordinate of two focus points, object function f (x) represents receptor outlet temperature steady-state value T_fss, constraint function g (x) wraps Include T_v-T_fss、F₀-F andThree constraints, the MBDFO algorithm steps of improvement is as follows:

5.1) an initial interpolated sample collection Y={y is selected¹,y²,···,y^q, wherein yⁱ∈Rⁿ, i=1,2 ..., q, Select a some x₀, it is allowed to for any one yⁱ∈ Y meets f (x₀)≤f(yⁱ), a selected initial Trust Region Radius Δ₀, one Individual constant η ∈ (0,1), wherein, for the Optimized model shown in formula (1), yⁱ=[z₁ ⁱ；z₂ ⁱ], q calculates according to formula (4) can ?

5.2) when meeting the condition of convergence, when i.e. the temperature value N continuous time change of iteration point position is less than convergence precision, i.e. Flag=N, N are setup parameter, stop iteration；Otherwise, interpolation is utilized to integrate Y by object function interpolation as quadratic function m_k(x), tool Bodily form formula is m_k([z₁；z₂])=A+Bz₁+Cz₂+Dz₁ ²+Ez₁ ²+Fz₁z₂, wherein A, B, C, D, E, F are the ginseng that interpolation method is tried to achieve Number；Constraint function is fitted to quadratic function g_k(x)；Solve in Trust Region Radius as shown in formula (19) by trust region method Optimization problem obtains step-length p, obtains model optimal value x_p=x_k+p；

5.3) x is judged_pWhether in feasible zone, i.e. g (x_p) whether more than 0, if g is (x_p) >=0, then calculate ratio according to formula (6) Rate ρ；If g is (x_p) ＜ 0, then adjust x according to given scaling criterion_pUntil g (x_p) >=0, according to formula (6) calculating ratio ρ；

5.4) if meeting ρ >=η, then by data y in interpolation collection Y_-Replace with x_p, data y that are replaced_-MeetWherein L (x_p, y) represent Lagrange interpolation polynomial, shown in its computational methods such as formula (15), more New Trust Region Radius is Δ_k+1, meet Δ_k+1≥Δ_k, update next iteration point x_k+1For x_p, proceed to next iteration, i.e. step 5.2)；If being unsatisfactory for ρ >=η, then judge interpolation collection Y the need of renewal, if | | x_k-yⁱ||≤Δ_k,yⁱ∈ Y, illustrates interpolation collection Y Need not update, proceed to step 5.5)；If being unsatisfactory for | | x_k-yⁱ||≤Δ_k,yⁱ∈ Y, illustrates that interpolation collection Y needs to update, proceeds to Step 5.6)；

5.5) Trust Region Radius Δ is updated_k+1, meet Δ_k+1＜ Δ_k, next iteration point x_k+1=x_k, proceed to change next time Generation, i.e. step 5.2)；

5.6) update interpolation collection Y: obtain with Trust Region Algorithm meet formula (18) be inserted into valueThen obtain to be deleted PointTrusted zones Δ_k+1=Δ_k；The element making object function minimum is selected in YAs the next one Iteration pointProceed to step 5.2)；

The optimization method of above-mentioned tower type solar thermoelectricity systems, spot collection thermal sub-system can take into account energy and safety two Requirement, the operation to actual tower solar energy power plant has guiding significance.

Accompanying drawing explanation

Fig. 1 is the optimization method flow chart of tower type solar thermoelectricity systems, spot collection thermal sub-system；

Fig. 2 is divided into two-part mirror field schematic diagram；

Fig. 3 is to improve complex method flow chart；

Fig. 4 is to improve MBDFO algorithm flow chart；

Fig. 5 is the focus point coordinate iterative process improving complex method；

Fig. 6 is the receptor outlet temperature of molten salt iterative process improving complex method；

Fig. 7 is to improve complex method to solve the imaging X-Y scheme of local best points；

Fig. 8 is to improve complex method to solve the imaging three-dimensional figure of local best points；

Fig. 9 is the focus point coordinate iterative process improving MBDFO algorithm；

Figure 10 is the receptor outlet temperature of molten salt iterative process improving MBDFO algorithm.

Detailed description of the invention

Below in conjunction with embodiment and Figure of description, the application is described further.

As it is shown in figure 1, the optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system, implement step such as Under:

(1) by heliostat field piecemeal, with receptor surface focus position as decision variable, fused salt temperature is exported with receptor Degree a height of target as far as possible, is less than threshold value for constraint, constitution optimization problem with receptor surface energy distribution standard deviation.

At the present embodiment, T_vThe volatilization temperature value of expression fused salt is 600 DEG C, and receptor vertically height is 3.1m.

Jing Chang is divided into two regions, as in figure 2 it is shown, the heliostat using double focus strategy, i.e. filled circles to represent shines Being mapped to focus point 1, the heliostat that open circles represents is irradiated to focus point 2.Two focus point coordinates are respectively (0,0, z₁), (0, 0,z₂), try one's best a height of target with receptor outlet temperature of molten salt, be about less than threshold value with receptor surface energy distribution standard deviation Bundle, constitution optimization problem is shown below:

min|600+273.15-T_fss|

Wherein, T_fssRepresenting receptor outlet temperature of molten salt steady-state value, it can be expressed as z₁, z₂Function fitness (z₁,z₂), z₁And z₂Representing two focus point z-axis coordinates, its value must not exceed receptor vertically height 3.1m.F is receptor The standard deviation of surface energy values, can be expressed as independent variable z₁, z₂Function g (z₁,z₂), F₀For the maximum standard deviation set.On Object function and the constraints of stating Optimized model can take into account energy and two requirements of safety.

(2) improvement complex method solving-optimizing problem is utilized.

Optimized model shown in formula (1) is expressed as minf (x), x ∈ Rⁿ, s.t.g (x) >=0, wherein, x=[z₁；z₂], Representing the z coordinate of two focus points, object function f (x) represents receptor outlet temperature steady-state value T_fss, constraint function g (x) wraps Include 600-T_fss、F₀-F and-3.1≤x≤3.1 3 retrain.The algorithm flow chart of complex method is as it is shown on figure 3, algorithm steps As follows:

2.1) intial compound form { x is chosen⁰,x¹,···,xⁿ, reflection coefficient α ＞ 1, spreading coefficient γ ＞ 1, shrink system Number β ∈ (0,1) and precision ε ＞ 0；

2.2) judge that intial compound form, whether in feasible zone, if not in feasible zone, is adjusted according to given scaling criterion Whole initial value position, renumbers the size of n+1 the summit according to target functional value of complex, makes the numbering on summit meet f (x⁰)≤f(x¹)≤···≤f(x^n-1)≤f(xⁿ)；

Wherein said scaling criterion is as follows:

1) for current initial value x₀=[z₁；z₂], if F₀-F ＞ 1000, then z₁=z₁+ 0.1, z₂=z₂-0.1；

2) if 100 ＜ F₀-F ＜=1000, then z₁=z₁+ 0.01, z₂=z₂-0.01；

3) if 10 ＜ F₀-F ＜=100, then z₁=z₁+ 0.001, z₂=z₂-0.001；

4) if 0 ＜ F₀-F ＜=10, then z₁=z₁+ 0.0001, z₂=z₂-0.0001；

5) if F₀-F ＜=0, does not scales.

2.3) orderIfStop iteration output x⁰, otherwise proceed to 2.4)；

2.4) x is calculatedⁿ⁺²=xⁿ⁺¹+α(xⁿ⁺¹-xⁿ), check xⁿ⁺²Whether in feasible zone, the most whether meet g (xⁿ⁺²) >=0, If not in feasible zone, reflection coefficient α is decreased up to xⁿ⁺²In feasible zone.Minishing method is the extraction of root or is multiplied by one It is less than the coefficient of 1 more than 0, calculates f (xⁿ⁺²), if f is (xⁿ⁺²) ＜ f (x⁰), proceed to 2.5), otherwise as f (xⁿ⁺²) ＜ f (x^n-1) time turn Enter 2.6), as f (xⁿ⁺²)≥f(x^n-1) proceed to (7)；

2.5) x is calculatedⁿ⁺³=xⁿ⁺¹+γ(xⁿ⁺²-xⁿ⁺¹), check xⁿ⁺³Whether in feasible zone, if not in feasible zone, will Spreading coefficient γ decreases up to xⁿ⁺³In feasible zone, minishing method is the extraction of root or is multiplied by a coefficient more than 0 less than 1. If f is (xⁿ⁺³) ＜ f (x⁰), make xⁿ=xⁿ⁺³, proceed to 2.2), otherwise proceed to 2.6)；

2.6) x is madeⁿ=xⁿ⁺², proceed to 2.2)；

2.7) x is madeⁿ={ xⁱ|f(xⁱ)=min (f (xⁿ),f(xⁿ⁺²)), calculate xⁿ⁺⁴=xⁿ⁺¹+β(xⁿ-xⁿ⁺¹), check xⁿ⁺⁴ Whether in feasible zone, if not in feasible zone, constriction coefficient β is decreased up to xⁿ⁺⁴In feasible zone, minishing method is evolution Method or be multiplied by one more than 0 less than 1 coefficient.If f is (xⁿ⁺⁴) ＜ f (xⁿ), make xⁿ=xⁿ⁺⁴, proceed to 2.2), otherwise proceed to 2.8)；

2.8) x is made^j=x⁰+θ(x^j-x⁰), j=0,1, n, proceed to 2.2).

(3) utilization is based on MBDFO algorithm improves and optimizates method solving-optimizing problem.

Optimized model shown in formula (1) is expressed as minf (x), x ∈ Rⁿ, s.t.g (x) >=0, wherein, x=[z₁；z₂], Representing the z coordinate of two focus points, object function f (x) represents receptor outlet temperature steady-state value T_fss, constraint function g (x) wraps Include T_v-T_fss、F₀-F andThree constraints, the MBDFO algorithm steps of improvement is as follows:

3.1) an initial interpolated sample collection Y={y is selected¹,y²,···,y^q, wherein yⁱ∈Rⁿ, i=1,2 ..., q, Select a some x₀, it is allowed to for any one yⁱ∈ Y meets f (x₀)≤f(yⁱ), a selected initial Trust Region Radius Δ₀, one Individual constant η ∈ (0,1), wherein, for the Optimized model shown in formula (1), yⁱ=[z₁ ⁱ；z₂ ⁱ], q calculates according to formula (4) can ?

3.2) when meeting the condition of convergence, when i.e. the temperature value N continuous time change of iteration point position is less than convergence precision, i.e. Flag=N, N are setup parameter, stop iteration；Otherwise, interpolation is utilized to integrate Y by object function interpolation as quadratic function m_k(x), tool Bodily form formula is m_k([z₁；z₂])=A+Bz₁+Cz₂+Dz₁ ²+Ez₁ ²+Fz₁z₂, wherein A, B, C, D, E, F are the ginseng that interpolation method is tried to achieve Number；Constraint function is fitted to quadratic function g_k(x)；Solve in Trust Region Radius as shown in formula (19) by trust region method Optimization problem obtain step-length p, obtain model optimal value x_p=x_k+p；

3.3) x is judged_pWhether in feasible zone, i.e. g (x_p) whether more than 0, if g is (x_p) >=0, then calculate ratio according to formula (6) Rate ρ；If g is (x_p) ＜ 0, then adjust x according to given scaling criterion_pUntil g (x_p) >=0, according to formula (6) calculating ratio ρ；

Wherein step 3.3) described in scaling criterion as follows:

1) for current initial value x₀=[z₁；z₂], if F₀-F ＞ 1000, then z₁=z₁+ 0.1, z₂=z₂-0.1；

2) if 100 ＜ F₀-F ＜=1000, then z₁=z₁+ 0.01, z₂=z₂-0.01；

3) if 10 ＜ F₀-F ＜=100, then z₁=z₁+ 0.001, z₂=z₂-0.001；

4) if 0 ＜ F₀-F ＜=10, then z₁=z₁+ 0.0001, z₂=z₂-0.0001；

5) if F₀-F ＜=0, does not scales.

3.4) if meeting ρ >=η, then by data y in interpolation collection Y_-Replace with x_p, data y that are replaced_-MeetWherein L (x_p, y) represent Lagrange interpolation polynomial, shown in its computational methods such as formula (15), more New Trust Region Radius is Δ_k+1, meet Δ_k+1≥Δ_k, update next iteration point x_k+1For x_p, proceed to next iteration, i.e. step 3.2)；If being unsatisfactory for ρ >=η, then judge interpolation collection Y the need of renewal, if | | x_k-yⁱ||≤Δ_k,yⁱ∈ Y, illustrates interpolation collection Y Need not update, proceed to step 3.5)；If being unsatisfactory for | | x_k-yⁱ||≤Δ_k,yⁱ∈ Y, illustrates that interpolation collection Y needs to update, proceeds to Step 3.6)；

3.5) Trust Region Radius Δ is updated_k+1, meet Δ_k+1＜ Δ_k, next iteration point x_k+1=x_k, proceed to change next time Generation, i.e. step 3.2)；

3.6) update interpolation collection Y: obtain with Trust Region Algorithm meet formula (18) be inserted into valueThen obtain to be deleted PointTrusted zones Δ_k+1=Δ_k；The element making object function minimum is selected in YAs the next one Iteration pointProceed to step 3.2)；

Present example is applied to a radial pattern Jing Chang (coordinate is known) comprising 1909 heliostats；Emulation place warp Latitude is (34 ,-116), and the time is 10:00 in the morning on October 1st, 1997；Receptor center overhead height is 76.4m, receives The a diameter of 5.1m of device, height is 6.2m；The long 7.5m of heliostat, wide 5m, overhead height 6m；Spread at random a little on every heliostat 10000,100 light of trace in the light cone of each random point reflection, when using improvement complex method solving-optimizing problem, just Beginning complex takes [0.5,0.6,1.2；-0.5 ,-2.8 ,-3], reflection coefficient α=1.2, spreading coefficient γ=8, constriction coefficient β= 0.3 and precision ε=1e-3, energy variance threshold values setting value F₀Take 2.0000e+04, optimum results focus point iterative process such as Fig. 5 Shown in, as shown in Figure 6, wherein, local best points is [z1 to receptor outlet temperature of molten salt iterative process；Z2]=[1.1347；- 0.2932], the local optimum of object function is 5.1687 DEG C, i.e. local optimum temperature value T_fssIt it is 594.8313 DEG C；Energy mark Quasi-difference is 1.9999e+04；The optimization time is 5.2h, and the local best points emulation utilizing optimization to obtain obtains receptor surface Heat is distributed as Figure 7-8.Utilize improve MBDFO Algorithm for Solving optimization problem time, take initial interpolation collection for [0.5,1.5, 2.2,1.4,1.5,2；-0.5 ,-2.5 ,-2 ,-1.3 ,-0.5 ,-1.5], object function reduces degree threshold value η=0.01, initially believes Rely territory radius Δ₀=0.1 precision ε=1e-3, takes coefficient 1.01 during expansion Trust Region Radius, takes coefficient when reducing Trust Region Radius 0.9, energy variance threshold values setting value F₀Taking 2.0000e+04, optimum results focus point iterative process is as it is shown in figure 9, receptor goes out As shown in Figure 10, wherein, local best points is [z1 to mouth temperature of molten salt iterative process；Z2]=[1.1461；-0.2831], target The local optimum of function is 5.1508, i.e. local optimum temperature value T_fssIt it is 594.8492 DEG C；Energy scale difference is 1.9999e+ 04；The optimization time is 44 minutes, and the local best points emulation utilizing optimization to obtain obtains heat distribution and the figure on receptor surface 7-8 is similar to.It can be seen that the local best points that two kinds of optimization methods obtain is basically identical, but improve MBDFO Algorithm for Solving time Between be greatly reduced.Use computer operating system be Windows7, emulation platform is MATLAB, have invoked gPROMS with CUDA interface.

Claims (6)

1. the optimization method of a tower type solar thermoelectricity systems, spot collection thermal sub-system, it is characterised in that comprise the steps:

1) by heliostat field piecemeal, with receptor surface focus position as decision variable, exist with receptor outlet temperature of molten salt As far as possible a height of target in the case of nonvolatile, with receptor surface energy distribution standard deviation less than threshold value for constraint, constitution optimization Problem；

2) utilize and improve complex method or based on MBDFO algorithm improve and optimizate method solving-optimizing problem；According to solving result The position adjusting focus point realizes the optimization of tower type solar thermoelectricity systems, spot collection thermal sub-system.

The optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system the most as claimed in claim 1, its feature It is described step 1) particularly as follows:

Jing Chang being divided into two regions, uses double focus strategy, the most a part of heliostat is irradiated to focus point 1, another part Heliostat is irradiated to focus point 2, and two focus point coordinates are respectively (0,0, z₁), (0,0, z₂), export temperature of molten salt with receptor As far as possible a height of target in the case of nonvolatile, with receptor surface energy distribution standard deviation less than threshold value for constraint, construct excellent Shown in change problem such as formula (1):

$\begin{array}{c}\min |T_v-T_{fss}|\\ s.t.\left\{\begin{array}{c}{T}_{fss}=fitness(z_1,z_2)\\ -\frac{h_t}{2}<z_1<\frac{h_t}{2}\\ -\frac{h_t}{2}<z_2<\frac{h_t}{2}\\ F=g(z_1,z_2)\\ F\&le;F_0\end{array}\right.\end{array}---\left(1\right)$

Wherein, T_vRepresent the volatilization temperature value of fused salt, for constant, T_fssExporting temperature of molten salt steady-state value for receptor, it is expressed as z₁, z₂Function fitness (z₁,z₂), z₁And z₂Representing two focus point z-axis coordinates, it is vertical that its value must not exceed receptor Highly h_t, F is the standard deviation of receptor surface energy values, is expressed as independent variable z₁, z₂Function g (z₁,z₂), F₀For setting Big standard deviation.

The optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system the most as claimed in claim 1, its feature Be when optimization method is for improving complex method, the process of described solving-optimizing problem particularly as follows:

Optimized model shown in formula (1) is expressed as minf (x), x ∈ Rⁿ, s.t.g (x) >=0, wherein, x=[z₁；z₂], represent The z coordinate of two focus points, object function f (x) represents receptor outlet temperature steady-state value T_fss, constraint function g (x) includes T_v- T_fss、F₀-F andThree constraints, the algorithm steps of complex method is as follows:

3.1) intial compound form { x is chosen⁰,x¹,…,xⁿ, reflection coefficient α ＞ 1, spreading coefficient γ ＞ 1, constriction coefficient β ∈ (0, 1) and precision ε ＞ 0；

3.2) judge that intial compound form, whether in feasible zone, if not in feasible zone, adjusts initial value position according to scaling criterion, Until intial compound form is in feasible zone；If in feasible zone, by the size weight of n+1 the summit according to target functional value of complex New numbering, makes the numbering on summit meet f (x⁰)≤f(x¹)≤…≤f(xⁿ-¹)≤f(xⁿ)；

3.3) orderIfStop iteration, export x⁰, otherwise proceed to step Rapid 3.4)；

3.4) x is calculatedⁿ⁺²=xⁿ⁺¹+α(xⁿ⁺¹-xⁿ), check xⁿ⁺²Whether in feasible zone, the most whether meet g (xⁿ⁺²) >=0, if not In feasible zone, reflection coefficient α is decreased up to xⁿ⁺²In feasible zone, calculate f (xⁿ⁺²), if f is (xⁿ⁺²) ＜ f (x⁰), proceed to step Rapid 3.5), otherwise as f (xⁿ⁺²) ＜ f (xⁿ-¹) time proceed to step 3.6), as f (xⁿ⁺²)≥f(xⁿ-¹) proceed to step 3.7)；Described The minishing method of reflection coefficient α is the extraction of root or is multiplied by a coefficient more than 0 less than 1；

3.5) x is calculatedⁿ⁺³=xⁿ⁺¹+γ(xⁿ⁺²-xⁿ⁺¹), check xⁿ⁺³Whether in feasible zone, if not in feasible zone, will extension Coefficient gamma decreases up to xⁿ⁺³In feasible zone, if f is (xⁿ⁺³) ＜ f (x⁰), make xⁿ=xⁿ⁺³, proceed to step 3.2), otherwise proceed to step Rapid 3.6)；Described spreading coefficient γ minishing method is the extraction of root or is multiplied by a coefficient more than 0 less than 1；

3.6) x is madeⁿ=xⁿ⁺², proceed to step 3.2)；

3.7) x is madeⁿ={ xⁱ|f(xⁱ)=min (f (xⁿ),f(xⁿ⁺²)), calculate xⁿ⁺⁴=xⁿ⁺¹+β(xⁿ-xⁿ⁺¹), check xⁿ⁺⁴Whether In feasible zone, if not in feasible zone, constriction coefficient β is decreased up to xⁿ⁺⁴In feasible zone, if f is (xⁿ⁺⁴) ＜ f (xⁿ), order xⁿ=xⁿ⁺⁴, proceed to step 3.2), otherwise proceed to step 3.8)；The minishing method of described constriction coefficient β is the extraction of root or is multiplied by One coefficient more than 0 less than 1；

3.8) x is made^j=x⁰+θ(x^j-x⁰), j=0,1 ..., n, proceed to step 3.2).

The optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system the most as claimed in claim 3, its feature Be described step 3.2) scaling criterion be:

1) for current initial value x₀=[z₁；z₂], if F₀-F ＞ 1000, then z₁=z₁+ 0.1, z₂=z₂-0.1；

2) if 100 ＜ F₀-F ＜=1000, then z₁=z₁+ 0.01, z₂=z₂-0.01；

3) if 10 ＜ F₀-F ＜=100, then z₁=z₁+ 0.001, z₂=z₂-0.001；

4) if 0 ＜ F₀-F ＜=10, then z₁=z₁+ 0.0001, z₂=z₂-0.0001；

5) if F₀-F ＜=0, does not scales.

The optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system the most as claimed in claim 1, its feature Be when the process that optimization method is based on MBDFO algorithm when improving and optimizating method, described solving-optimizing problem particularly as follows:

Optimized model shown in formula (1) is expressed as minf (x), x ∈ Rⁿ, s.t.g (x) >=0, wherein, x=[z₁；z₂], represent The z coordinate of two focus points, object function f (x) represents receptor outlet temperature steady-state value T_fss, constraint function g (x) includes T_v- T_fss、F₀-F andThree constraints, the MBDFO algorithm steps of improvement is as follows:

5.1) an initial interpolated sample collection Y={y is selected¹,y²,…,y^q, wherein yⁱ∈Rⁿ, i=1,2 ..., q, select one Point x₀, it is allowed to for any one yⁱ∈ Y meets f (x₀)≤f(yⁱ), a selected initial Trust Region Radius Δ₀, a constant η ∈ (0,1), wherein, for the Optimized model shown in formula (1), yⁱ=[z₁ ⁱ；z₂ ⁱ], q can be calculated according to formula (4)

$q=\frac{1}{2}(n+1)(n+2)---\left(4\right)$

5.2) when meeting the condition of convergence, when i.e. the temperature value N continuous time change of iteration point position is less than convergence precision, i.e. flag =N, N are setup parameter, stop iteration；Otherwise, interpolation is utilized to integrate Y by object function interpolation as quadratic function m_k(x), concrete shape Formula is m_k([z₁；z₂])=A+Bz₁+Cz₂+Dz₁ ²+Ez₁ ²+Fz₁z₂, wherein A, B, C, D, E, F are the parameter that interpolation method is tried to achieve；Will Constraint function fits to quadratic function g_k(x)；In Trust Region Radius, the optimization as shown in formula (19) is solved by trust region method Problem obtains step-length p, obtains model optimal value x_p=x_k+p；

$\begin{array}{c}\underset{p}{\min }{m}_k\left(x_k+p\right)\\ s.t.\left\{\begin{array}{c}\left|\right|p|{|}_2\&le;\mathrm{\&Delta;}\\ g_k(x_k+p)>0\end{array}\right.\end{array}---\left(19\right)$

5.3) x is judged_pWhether in feasible zone, i.e. g (x_p) whether more than 0, if g is (x_p) >=0, then according to formula (6) calculating ratio ρ； If g is (x_p) ＜ 0, then adjust x according to scaling criterion_pUntil g (x_p) >=0, according to formula (6) calculating ratio ρ；

$\mathrm{\&rho;}=\frac{f\left(x_k\right)-f\left({x_k}^{+}\right)}{m\left(x_k\right)-m\left({x_k}^{+}\right)}---\left(6\right)$

5.4) if meeting ρ >=η, then by data y in interpolation collection Y_-Replace with x_p, data y that are replaced_-MeetWherein L (x_p, y) represent Lagrange interpolation polynomial, shown in its computational methods such as formula (15), more New Trust Region Radius is Δ_k+1, meet Δ_k+1≥Δ_k, update next iteration point x_k+1For x_p, proceed to next iteration, i.e. step 5.2)；If being unsatisfactory for ρ >=η, then judge interpolation collection Y the need of renewal, if | | x_k-yⁱ||≤Δ_k,yⁱ∈ Y, illustrates interpolation collection Y Need not update, proceed to step 5.5)；If being unsatisfactory for | | x_k-yⁱ||≤Δ_k,yⁱ∈ Y, illustrates that interpolation collection Y needs to update, proceeds to Step 5.6)；

$L(y,y^i)=\underset{j\&NotEqual;i}{\overset{m}{\underset{j=1}{\mathrm{\&Pi;}}}}\frac{y\left(1\right)-y^j\left(1\right)}{y^i\left(1\right)-y^j\left(1\right)}\underset{j\&NotEqual;i}{\overset{m}{\underset{j=1}{\mathrm{\&Pi;}}}}\frac{y\left(2\right)-y^j\left(2\right)}{y^i\left(2\right)-y^j\left(2\right)}---\left(15\right)$

5.5) Trust Region Radius Δ is updated_k+1, meet Δ_k+1＜ Δ_k, next iteration point x_k+1=x_k, proceed to next iteration, i.e. Step 5.2)；

5.6) update interpolation collection Y: obtain with Trust Region Algorithm meet formula (18) be inserted into valueThen point to be deleted is obtainedTrusted zones Δ_k+1=Δ_k；The element making object function minimum is selected in YChange as the next one Dai DianProceed to step 5.2)；

$y_r^i=\arg \underset{\left|\right|y-x_k\left|\right|\&le;\mathrm{\&Delta;}}{max}\left|L(y,y^i)\right|---\left(18\right).$

The optimization method of a kind of tower type solar thermoelectricity systems, spot collection thermal sub-system the most as claimed in claim 5, its feature Be described step 5.3) described in scaling criterion as follows:

1) for current initial value x₀=[z₁；z₂], if F₀-F ＞ 1000, then z₁=z₁+ 0.1, z₂=z₂-0.1；

2) if 100 ＜ F₀-F ＜=1000, then z₁=z₁+ 0.01, z₂=z₂-0.01；

3) if 10 ＜ F₀-F ＜=100, then z₁=z₁+ 0.001, z₂=z₂-0.001；

4) if 0 ＜ F₀-F ＜=10, then z₁=z₁+ 0.0001, z₂=z₂-0.0001；

5) if F₀-F ＜=0, does not scales.