CN105913161A - Method of acquiring maximum power point of photovoltaic system based on multi-objective optimization - Google Patents

Abstract

The invention discloses a method of acquiring the maximum power point of a photovoltaic system based on multi-objective optimization. The method is characterized by comprising the following steps: 1, the photovoltaic system maximum power point tracking problem is converted into a multi-objective optimization problem shown as in a formula (1); and 2, the multi-objective optimization method is used for optimizing the multi-objective optimization problem to obtain voltage and current corresponding to the maximum power point. The maximum power point of the photovoltaic system in the environment can be acquired by only detecting the external environment, and the purpose of improving the precision of the maximum power point can be achieved.

Description

A kind of acquisition methods of photovoltaic system maximum power point based on multiple-objection optimization

Technical field

The present invention relates to photovoltaic system maximum power point technical field, a kind of photovoltaic system based on multiple-objection optimization The acquisition methods of system maximum power point.

Background technology

Photovoltaic generating system under certain environmental conditions, only exists a peak power output point.Once its operating point is deviateed High-power, the output of solaode will strongly reduce.Therefore, in order to utilize photovoltaic generating system to greatest extent Energy, improve system efficiency it is necessary to constantly adjust the running voltage of photovoltaic system according to the change of external environment so that Photovoltaic generating system always works near maximum power point, and this technology is referred to as MPPT maximum power point tracking.

Along with developing rapidly of photovoltaic generation, the effect of MPPT maximum power point tracking algorithm seems and is even more important.At present, peak power Point track algorithm is broadly divided into two categories below:

(1) approximation method of adjustment based on engineer applied.The Typical Representative of this kind of method is constant voltage tracing: owing to not sharing the same light Some constant magnitude of voltage U always it is approximately close to according to the maximum power point under intensity_max, so according to the structure of photovoltaic cell U is calculated with parameter_max, and make photovoltaic system become a manostat, just can stabilize it and be operated near maximum power point. Constant voltage tracing control method is easily achieved, and control system is simple.But its steady-state error is bigger, the merit of photovoltaic system loss Rate also compares many.

(2) Mathematical Method based on output characteristics.This kind of method mainly includes disturbance observational method, conductance increment method etc., is Use a widest class MPPT maximum power point tracking algorithm at present.This type of method by the output voltage disturbance in addition to photovoltaic array, Observe the change of its output characteristics to judge current duty, thus determine the direction of disturbance next time, finally make system stability Operate near maximum power point.But the disturbance step-length that this kind of MPPT maximum power point tracking algorithm is selected is for tracking accuracy and sound Answering speed to take into account, disturbance step-length is the biggest, and speed is the fastest, but precision is the poorest；Vice versa.

Summary of the invention

The present invention is for avoiding the weak point existing for above-mentioned prior art, it is provided that a kind of photovoltaic system based on multiple-objection optimization The acquisition methods of maximum power point, to only needing to detect the external environment of photovoltaic system, can trace into photovoltaic system effectively Maximum power point, and reach to improve the purpose of maximum power point precision.

In order to achieve the above object, the technical solution adopted in the present invention is:

The feature of the acquisition methods of a kind of photovoltaic system maximum power point based on multiple-objection optimization of the present invention is to enter as follows OK:

Step 1, the tracking problem of described photovoltaic system maximum power point is converted to the multi-objective optimization question as shown in formula (1):

$\begin{array}{c}\min \left(f_1\right(x),f_2(x\left)\right)\\ \begin{array}{cc}s.t.& x=(V_r,I_r),V_r\∈\[0,V_{\max }\],I_r\∈\[0,I_{\max }\]\end{array}\end{array}---\left(1\right)$

In formula (1), V_rAnd I_rRepresent the accessed voltage and current corresponding to maximum power point respectively；V_maxAnd I_maxRespectively Represent given illumination and at a temperature of maximum voltage and maximum current；f₁X () expression minimizes power function, and have:

f₁(x)=V_max×I_max-V_r×I_r (2)

In formula (1), f₂X () represents current error function, and have:

$f_2\left(x\right)=|I_r-I_{PV}|+I_0\{\exp \[\frac{q}{AkT}(V_r+I_r{R}_s)\]-1\}+\frac{V_r+I_r{R}_s}{R_{sh}}---\left(3\right)$

In formula (3), I_PVFor the photovoltaic electric current generated by illumination；I₀For diode saturation current；Q is the quantity of electric charge of electronics；k For Boltzmann constant；T is the absolute temperature of solaode；A is desirability figure, A ∈ [1,2]；R_SFor series equivalent electricity Resistance；R_shFor parallel equivalent resistance；

Step 2, utilize Multipurpose Optimal Method that described multi-objective optimization question is optimized, obtain corresponding to maximum power point Voltage V_rWith electric current I_r。

The feature of acquisition methods of the present invention lies also in,

Described Multipurpose Optimal Method be preference multi-objective optimization algorithm based on weight vector be to carry out as follows:

Step 2.1, definition select probability are δ；Definition cyclic variable is L；Definition maximum cycle is L_max；Delimiting period For T；Defined variable m, initializes m=1；

In m-th cycle T_mUnder, produce initial population P₀M () and corresponding initial weight vector set w, be designated asW={w₁,w₂,…,w_i,…,w_n}；Represent the I initial individuals, andWithRepresent i-th initial individuals respectivelyInitial voltage and initial current；w_iRepresent i-th individuality p_iThe initial weight vector of (m)；And w_i=(w_i1,w_i2)； w_i1And w_i2Represent i-th individuality p respectively_iFirst weights of (m) and the second weights；

Step 2.2, initialization L=1；Binary search sampling site method is utilized to reduce initial population P₀The region of search [0, V of (m)_max] and [0,I_max], thus obtain the interval [V of Optimizing Search_k-1,V_k+1] and [I_k-1,I_k+1]；

Step 2.3, to arrange preference point be p (m)=(p₁(m),p₂(m))；p₁M () is m-th cycle T_mUnder minimize power Function f₁The expected value of (x), p₂M () is m-th cycle T_mLower current error function f₂The expected value of (x)；Preference neighborhood is set B=(b₁,b₂)；b₁And b₂Represent length and the width of preference neighborhood respectively；

Step 2.4, utilize formula (4) by m-th cycle T_mFollowing i-th initial individualsInitial weight vector w_iReflect It is mapped to the i-th initial individuals in the preference neighborhood b of preference point p (m), after being mappedWeight vector w_i', and w_i'=(w_i′₁,w_i′₂)；Thus the weight vector set w '={ w after being mapped₁′,w′₂,…,w_i′,…,w_n' }:

w_ia'=2 × b_a×w_ia+p_a(m)-b_a, a=1,2 (4)

Step 2.5, utilize formula (5) by i-th initial individualsWeight vector w_i' be converted to i-th initial individualsThe weight vector λ with preference information_i, thus obtain the weight vector set with preference information λ={ λ₁,λ₂,…,λ_i,…,λ_n}:

${\mathrm{\λ}}_i=\frac{1}{w_i^{\′}}---\left(5\right)$

Step 2.5, calculating i-th initial individualsInitial weight vector w_iAt the beginning of individual with remaining n-1 respectively Euclidean distance between beginning weight vector, obtains i-th initial individualsEuclidean distance set, from described Euclidean distance Set is chosen minimum front M apart from corresponding individuality as i-th initial individualsM nearest-neighbors Individual；

Step 2.6, individual from described i-th with described select probability δM nearest-neighbors individuality in choose three Individual or probability with 1-δ is from described initial population P_L-1M () is chosen three individualities, and utilize differential evolution operator to selected Three individualities taken carry out cross and variation, thus produce m-th cycle T_mThe i-th of lower the L time circulation is the most individual

Step 2.7, repetition step 2.6 are to population P_L-1M in (), each individuality carries out cross and variation, thus produce the L time circulation N the new individual new population constituted

Step 2.8, the new population Q that the L time is circulated_L(m) and the population P of the L-1 time circulation_L-1M () merges, obtain M-th cycle T_mThe merging population R of lower the L time circulation_L(m)；

Step 2.9, utilize Chebyshev's formula from described the L time circulation merging population R_LM () selects and there is preference information Mate most n of weight vector set λ individual constitute the population P circulated the L time_L(m)；

Step 2.10, L+1 is assigned to L, and judges L ＞ L_maxWhether set up, if setting up, then it represents that obtain the m-th cycle T_mLower L_maxThe population P of secondary circulation_Lmax(m), and from described population P_LmaxM () is chosen maximum power point, with described maximum work Voltage and current corresponding to rate point is m-th cycle T to be obtained_mUnder voltage V_r(m) and electric current I_r(m)；If no Set up, then return step 2.6；

Step 2.11, by voltage V_r(m) and electric current I_rM () is as V_rAnd I_rSubstitution minimizes power function f₁In (x), obtain f₁' (x), it is judged that f₁' (x) ＜ p₁M whether () set up, if setting up, then by f₁' (x) is as the m+1 cycle T_m+1Under power Littleization expected value；Again by m-th cycle T_mUnder L_maxThe population P of secondary circulation_LmaxM () is as the m+1 cycle T_m+1Under Initial population P₀(m+1)；

Step 2.12, make m+1 be assigned to m, and return step 2.2.

Described binary search sampling site method is to carry out as follows:

Step 2.2.1, to [0, V_max] carry out uniform sampling, and after calculating the output of each sampled point, choose maximum work output Employing point corresponding to rate is designated as V_k；

Step 2.2.2, choose [V_k-1,V_k+1] and [I of correspondence_k-1,I_k+1] interval as Optimizing Search.

Compared with the prior art, beneficial effects of the present invention is embodied in:

1, the present invention only needs to detect the environmental change that photovoltaic system is extraneous, can get the maximum power point of photovoltaic system, it is easy to Operation realizes, and can realize the peak power output point of off-line tracking photovoltaic system, it is to avoid because oscillating around at maximum power point And cause energy loss；Meanwhile, external environment generation cataclysm or photovoltaic array be positioned at local shelter from heat or light when, the present invention Also the maximum power point of photovoltaic system can be acquired.

2, the existing method under external environment generation transition or photovoltaic array are positioned at situation of locally sheltering from heat or light that the present invention is directed to obtains maximum work The deficiency of rate point, it is proposed that the method for multiple-objection optimization optimizes maximum power point problem, thus improves the precision of peak power With its accuracy.

3, the present invention by propose binary search sampling site method search the hunting zone gone to the lavatory so that algorithm obtain one less and more smart True scope, improves the time efficiency of algorithm.

4, the present invention incorporates decisionmaker's preference information in search procedure, and the preference that final Search Results is limited in policymaker is adjacent In territory so that it is the constraint of photovoltaic system model can be met more accurately.

5, the present invention is by the most dynamically updating the preference information of policymaker, improves the precision of algorithm.

6, the present invention is by the most dynamically Population Regeneration, accelerates the convergence rate of population in algorithm so that the speed of service of algorithm Faster, the operation time of algorithm is decreased.

Accompanying drawing explanation

Fig. 1 is the equivalent model of photovoltaic cell in prior art；

Fig. 2 is the structural representation of the present invention；

Fig. 3 is weight vector mapping graph of the present invention；

Fig. 4 is the flow chart of present invention preference based on weight vector Multipurpose Optimal Method.

Detailed description of the invention

In the present embodiment, the acquisition methods of a kind of photovoltaic system maximum power point based on multiple-objection optimization is to carry out as follows:

Step 1, the tracking problem of described photovoltaic system maximum power point is converted to the multi-objective optimization question as shown in formula (1):

$\begin{array}{c}\min \left(f_1\right(x),f_2(x\left)\right)\\ \begin{array}{cc}s.t.& x=(V_r,I_r),V_r\∈\[0,V_{\max }\],I_r\∈\[0,I_{\max }\]\end{array}\end{array}---\left(1\right)$

In formula (1), V_rAnd I_rRepresent the accessed voltage and current corresponding to maximum power point respectively；V_maxAnd I_maxRespectively Represent given illumination and at a temperature of maximum voltage and maximum current；f₁X () expression minimizes power function, and have:

f₁(x)=V_max×I_max-V_r×I_r (2)

In formula (1), f₂X () represents current error function, and have:

$f_2\left(x\right)=|I_r-I_{PV}|+I_0\{\exp \[\frac{q}{AkT}(V_r+I_r{R}_s)\]-1\}+\frac{V_r+I_r{R}_s}{R_{sh}}---\left(3\right)$

In formula (3), I_PVFor the photovoltaic electric current generated by illumination；I₀For diode saturation current；Q is the quantity of electric charge of electronics；K is Boltzmann constant；T is the absolute temperature of solaode；A is desirability figure, A ∈ [1,2]；R_SFor series equivalent resistance； R_shFor parallel equivalent resistance；

According to the equivalent model of the photovoltaic cell shown in Fig. 1, its output characteristic equation can be drawn:

$I=I_{PV}-I_0\{\exp \[\frac{q}{AkT}(V+{\mathrm{IR}}_s)\]-1\}-\frac{V+{\mathrm{IR}}_s}{R_{sh}}---\left(4\right)$

The purpose of the maximum power point of tracking photovoltaic system seeks to find (V, the I) that make V*I value maximum right, and to make simultaneously The real current I' obtained by the output characteristic equation of model by running voltage V meets photovoltaic cell with the electric current I traced into The error constraints of model.Therefore, the peak power of tracking photovoltaic system should consider power, considers the current error of model again, This meets the character of multi-objective optimization question, therefore can be converted into the tracking problem of the maximum power point of photovoltaic system such as formula (1) multi-objective optimization question shown in:

Step 2, utilize Multipurpose Optimal Method that the multi-objective optimization question in embodiment is optimized, obtain maximum power point Corresponding voltage V_rWith electric current I_r。

In being embodied as, Multipurpose Optimal Method be preference multi-objective optimization algorithm based on weight vector be to carry out as follows: As in figure 2 it is shown, after maximum power point problem is converted to the multi-objective optimization question shown in formula (1), the present embodiment use based on The preference multi-objective optimization algorithm of weight vector optimizes this multi-objective optimization question, can get the power points of correspondence.

Step 2.1, definition select probability are δ；Definition cyclic variable is L；Definition maximum cycle is L_max；Delimiting period For T；Defined variable m, initializes m=1；

In m-th cycle T_mUnder, produce initial population P₀M () and corresponding initial weight vector set w, be designated asW={w₁,w₂,…,w_i,…,w_n}；Represent the I initial individuals, andRepresent i-th initial individuals respectivelyInitial voltage and initial current；w_iRepresent i-th individuality p_iThe initial weight vector of (m)；And w_i=(w_i1,w_i2)； w_i1And w_i2Represent i-th individuality p respectively_iFirst weights of (m) and the second weights；

Step 2.2, initialization L=1；Binary search sampling site method is utilized to reduce initial population P₀The region of search [0, V of (m)_max] and [0,I_max], thus obtain the interval [V of Optimizing Search_k-1,V_k+1] and [I_k-1,I_k+1]；

Step 2.2.1, to [0, V_max] carry out uniform sampling, and after calculating the output of each sampled point, choose maximum work output Employing point corresponding to rate is designated as V_k；

Step 2.2.2, choose [V_k-1,V_k+1] and [I of correspondence_k-1,I_k+1] interval as Optimizing Search.

Output characteristic equation according to model, error function F (V, I) builds formula (5), and the purpose of this function is to find applicable photovoltaic (V, the I) of system model is right；

$F(V,I)=I_0\left(\exp \right(\frac{q}{AkT}\left(V+{\mathrm{IR}}_s\right))-1)+\frac{V+{\mathrm{IR}}_s}{R_{sh}}+I-I_{PV}---\left(5\right)$

OrderThis external voltage V is constant, it therefore follows that the function about Z as shown in formula (6)；

$F\left(Z\right)=I_0\left(\exp \right(Z)-1)+\frac{AkT}{{\mathrm{qR}}_{sh}}Z+\frac{AkTZ-qV}{R_s}-I_{PV}---\left(6\right)$

Z in formula (6) is carried out derivation such as formula (7), it is possible to find F ' (Z) >=0, shows that function F (Z) is a monotonically increasing function, As Z=I=V=0, can obtainTherefore because finding Z ＞ 0 to meet F (Z) ＞ 0.

$F^{\′}\left(Z\right)=I_0\exp \left(Z\right)+\frac{AkT}{{\mathrm{qR}}_{sh}}+\frac{AkT}{R_s}---\left(7\right)$

Analytical formula (6) understands, as Z ＞ 0, and I₀(exp (Z)-1) ＞ 0；Therefore, if metTime, just can meet F (Z) ＞ 0；That is be appreciated that when Z meetsTime, F (Z) ＞ 0；Meanwhile, F (Z) is a monotonically increasing function, interval [0, Z_x] Have and an only zero root, therefore binary search is sent out, and to can be used for searching for (V, I) right；

Therefore, under given external environment, as I=0 and F (Z)=0, can get photovoltaic system by binary search Big running voltage V_max.Work as V_maxAfter determining, the hunting zone of the initial solution in Multipurpose Optimal Method is fixed on [0, V_max], for Obtain approximate optimal solution, can be right by choosing optimal candidate (V, I) from some row solutions, to this, the present embodiment is in interval [0,V_max] to choose 50 (V, I) uniformly right, is defined as (V_i,I_i), i=1,2 ..., 50；Corresponding output P_i=V_i*I_i, Assume that peak power output is P_{near_opt}=P_k, corresponding (V, I) is to for (V_k,I_k), therefore optimum running voltage is in interval [V_k-1,V_k+1] (the P-V curve of photovoltaic system is unimodal curve), corresponding electric current interval is [I_k-1,I_k+1], wherein electric current can be with The increase of voltage and reduce；Therefore, by the decision space [0, V of initial multi-objective optimization question_max]×[0,I_max] reduce [V_k-1,V_k+1]×[I_k-1,I_k+1]；So can reduce the search volume of solution, obtain a less and more accurate scope, improve The time efficiency of algorithm.

Step 2.3, to arrange preference point be p (m)=(p₁(m),p₂(m))；p₁M () is m-th cycle T_mUnder minimize power Function f₁The expected value of (x), p₂M () is m-th cycle T_mLower current error function f₂The expected value of (x)；Preference neighborhood is set B=(b₁,b₂)；b₁And b₂Represent length and the width of preference neighborhood respectively；

Step 2.4, utilize formula (8) by m-th cycle T_mFollowing i-th initial individualsInitial weight vector w_iReflect It is mapped to the i-th initial individuals in the preference neighborhood b of preference point p (m), after being mappedWeight vector w_i', and w_i'=(w_i′₁,w_i′₂)；Thus the weight vector set w '={ w after being mapped₁′,w′₂,…,w_i′,…,w_n' }:

w_ia'=2 × b_a×w_ia+p_a(m)-b_a, a=1,2 (8)

The weight vector that Fig. 3 gives in two objective optimisation problems is mapped in the neighborhood of preference point p=(0.5,0.2) by formula 8, Weight vector after being mapped, wherein Size of Neighborhood is b=(0.25,0.25).

Step 2.5, utilize formula (9) by i-th initial individualsWeight vector w_i' be converted to i-th initial individualsThe weight vector λ with preference information_i, thus obtain the weight vector set with preference information λ={ λ₁,λ₂,…,λ_i,…,λ_n}:

${\mathrm{\λ}}_i=\frac{1}{w_i^{\′}}---\left(9\right)$

Step 2.5, calculating i-th initial individualsInitial weight vector w_iAt the beginning of individual with remaining n-1 respectively Euclidean distance between beginning weight vector, obtains i-th initial individualsEuclidean distance set, from described Euclidean distance Set is chosen minimum front M apart from corresponding individuality as i-th initial individualsM nearest-neighbors Individual；

Step 2.6, individual from described i-th with described select probability δM nearest-neighbors individuality in choose three Individual or probability with 1-δ is from described initial population P_L-1M () is chosen three individualities, and utilize differential evolution operator to selected Three individualities taken carry out cross and variation, thus produce m-th cycle T_mThe i-th of lower the L time circulation is the most individual

Step 2.7, repetition step 2.6 are to population P_L-1M in (), each individuality carries out cross and variation, thus produce the L time circulation N the new individual new population constituted

Step 2.8, the new population Q that the L time is circulated_L(m) and the population P of the L-1 time circulation_L-1M () merges, obtain M-th cycle T_mThe merging population R of lower the L time circulation_L(m)；

Step 2.9, utilize Chebyshev's formula from described the L time circulation merging population R_LM () selects and there is preference information Mate most n of weight vector set λ individual constitute the population P circulated the L time_L(m)；

Step 2.10, L+1 is assigned to L, and judges L ＞ L_maxWhether set up, if setting up, then it represents that obtain the m-th cycle T_mLower L_maxThe population P of secondary circulation_Lmax(m), and from described population P_LmaxM () is chosen maximum power point, with described maximum work Voltage and current corresponding to rate point is m-th cycle T to be obtained_mUnder voltage V_r(m) and electric current I_r(m)；If no Set up, then return step 2.6；

Step 2.11, by voltage V_r(m) and electric current I_rM () is as V_rAnd I_rSubstitution minimizes power function f₁In (x), obtain f₁' (x), it is judged that f₁' (x) ＜ p₁M whether () set up, if setting up, then by f₁' (x) is as the m+1 cycle T_m+1Under power Littleization expected value；So can constantly update the preference point of initial setting up in algorithm, dynamically updating of preference information improves this reality Execute the precision of example.Again by m-th cycle T_mUnder L_maxThe population P of secondary circulation_LmaxM () is as the m+1 cycle T_m+1Under Initial population P₀(m+1)；Owing to the number of times of environmental transients extraneous in a day is little, therefore the kind under a upper moment maximum power point Group has directive significance to subsequent time, therefore in the present embodiment using the final population in a upper moment as the initial population of subsequent time. The convergence rate of population during dynamically more novel species accelerates algorithm so that the speed of service of algorithm faster, when decreasing the operation of algorithm Between；

Step 2.12, make m+1 be assigned to m, and return step 2.2.

Fig. 4 gives the preference multi-objective optimization algorithm based on weight vector algorithm flow chart under a cycle.

Claims (3)

1. an acquisition methods for photovoltaic system maximum power point based on multiple-objection optimization, is characterized in that carrying out as follows:

Step 1, the tracking problem of described photovoltaic system maximum power point is converted to the multi-objective optimization question as shown in formula (1):

$\begin{array}{c}min\left(f_1\right(x),f_2(x\left)\right)\\ \begin{array}{cc}s.t.& x=(V_r,I_r),V_r\∈\[0,V_{max}\],I_r\∈\[0,I_{max}\]\end{array}\end{array}---\left(1\right)$

In formula (1), V_rAnd I_rRepresent the accessed voltage and current corresponding to maximum power point respectively；V_maxAnd I_maxRespectively Represent given illumination and at a temperature of maximum voltage and maximum current；f₁X () expression minimizes power function, and have:

f₁(x)=V_max×I_max-V_r×I_r (2)

In formula (1), f₂X () represents current error function, and have:

$f_2\left(x\right)=|I_r-I_{PV}|+I_0\{\exp \[\frac{q}{AkT}(V_r+I_r{R}_s)\]-1\}+\frac{V_r+I_r{R}_s}{R_{sh}}---\left(3\right)$

In formula (3), I_PVFor the photovoltaic electric current generated by illumination；I₀For diode saturation current；Q is the quantity of electric charge of electronics；k For Boltzmann constant；T is the absolute temperature of solaode；A is desirability figure, A ∈ [1,2]；R_SFor series equivalent electricity Resistance；R_shFor parallel equivalent resistance；

Step 2, utilize Multipurpose Optimal Method that described multi-objective optimization question is optimized, obtain corresponding to maximum power point Voltage V_rWith electric current I_r。

2. according to the acquisition methods described in claim, it is characterized in that, described Multipurpose Optimal Method is based on weight vector inclined Good multi-objective optimization algorithm, and carry out as follows:

Step 2.1, definition select probability are δ；Definition cyclic variable is L；Definition maximum cycle is L_max；Delimiting period For T；Defined variable m, initializes m=1；

In m-th cycle T_mUnder, produce initial population P₀M () and corresponding initial weight vector set w, be designated asW={w₁,w₂,…,w_i,…,w_n}；Represent the I initial individuals, and WithRepresent i-th initial individuals respectivelyInitial voltage and initial current；w_iRepresent i-th individuality p_iThe initial weight vector of (m)；And w_i=(w_i1,w_i2)； w_i1And w_i2Represent i-th individuality p respectively_iFirst weights of (m) and the second weights；

Step 2.2, initialization L=1；Binary search sampling site method is utilized to reduce initial population P₀The region of search [0, V of (m)_max] and [0,I_max], thus obtain the interval [V of Optimizing Search_k-1,V_k+1] and [I_k-1,I_k+1]；

Step 2.3, to arrange preference point be p (m)=(p₁(m),p₂(m))；p₁M () is m-th cycle T_mUnder minimize power Function f₁The expected value of (x), p₂M () is m-th cycle T_mLower current error function f₂The expected value of (x)；Preference neighborhood is set B=(b₁,b₂)；b₁And b₂Represent length and the width of preference neighborhood respectively；

Step 2.4, utilize formula (4) by m-th cycle T_mFollowing i-th initial individualsInitial weight vector w_iReflect It is mapped to the i-th initial individuals in the preference neighborhood b of preference point p (m), after being mappedWeight vector w '_i, and w′_i=(w '_i1,w′_i2)；Thus the weight vector set w '={ w ' after being mapped₁,w′₂,…,w′_i,…,w′_n}:

w_ia'=2 × b_a×w_ia+p_a(m)-b_a, a=1,2 (4)

Step 2.5, utilize formula (5) by i-th initial individualsWeight vector w '_iBe converted to i-th initial individualsThe weight vector λ with preference information_i, thus obtain the weight vector set with preference information λ={ λ₁,λ₂,…,λ_i,…,λ_n}:

${\mathrm{\λ}}_i=\frac{1}{w_i^{\′}}---\left(5\right)$

Step 2.5, calculating i-th initial individualsInitial weight vector w_iAt the beginning of individual with remaining n-1 respectively Euclidean distance between beginning weight vector, obtains i-th initial individualsEuclidean distance set, from described Euclidean distance Set is chosen minimum front M apart from corresponding individuality as i-th initial individualsM nearest-neighbors Individual；

Step 2.6, individual from described i-th with described select probability δM nearest-neighbors individuality in choose three Individual or probability with 1-δ is from described initial population P_L-1M () is chosen three individualities, and utilize differential evolution operator to selected Three individualities taken carry out cross and variation, thus produce m-th cycle T_mThe i-th of lower the L time circulation is the most individual

Step 2.7, repetition step 2.6 are to population P_L-1M in (), each individuality carries out cross and variation, thus produce the L time circulation N the new individual new population constituted

Step 2.8, the new population Q that the L time is circulated_L(m) and the population P of the L-1 time circulation_L-1M () merges, obtain M-th cycle T_mThe merging population R of lower the L time circulation_L(m)；

Step 2.9, utilize Chebyshev's formula from described the L time circulation merging population R_LM () selects and there is preference information Mate most n of weight vector set λ individual constitute the population P circulated the L time_L(m)；

Step 2.10, L+1 is assigned to L, and judges L ＞ L_maxWhether set up, if setting up, then it represents that obtain the m-th cycle T_mLower L_maxThe population P of secondary circulation_Lmax(m), and from described population P_LmaxM () is chosen maximum power point, with described maximum work Voltage and current corresponding to rate point is m-th cycle T to be obtained_mUnder voltage V_r(m) and electric current I_r(m)；If no Set up, then return step 2.6；

Step 2.11, by voltage V_r(m) and electric current I_rM () is as V_rAnd I_rSubstitution minimizes power function f₁In (x), obtain f₁' (x), it is judged that f₁' (x) ＜ p₁M whether () set up, if setting up, then by f₁' (x) is as the m+1 cycle T_m+1Under power Littleization expected value；Again by m-th cycle T_mUnder L_maxThe population P of secondary circulation_LmaxM () is as the m+1 cycle T_m+1Under Initial population P₀(m+1)；

Step 2.12, make m+1 be assigned to m, and return step 2.2.

3. according to the acquisition methods described in claim, it is characterized in that, described binary search sampling site method is to carry out as follows:

Step 2.2.1, to [0, V_max] carry out uniform sampling, and after calculating the output of each sampled point, choose maximum work output Employing point corresponding to rate is designated as V_k；

Step 2.2.2, choose [V_k-1,V_k+1] and [I of correspondence_k-1,I_k+1] interval as Optimizing Search.