CN103996081B - Mixed multiattribute group decision making method for network frame reconstruction scheme evaluation - Google Patents

Abstract

The invention discloses a mixed multiattribute group decision making method for network frame reconstruction scheme evaluation. The mixed multiattribute group decision making method comprises the steps that the quantitative attribute and the qualitative attribute of each subsystem in power system subareas are determined, a decision matrix is built, and the object weight of the attributes is determined according to the decision matrix; different decision makers in the group decision making process judge the relative importance of the attributes by using a pairwise comparison matrix or a fuzzy complementary judgment matrix respectively, the subjective weight is obtained, the subjective weight and the objective weight are combined to form a combination weight, a network frame reconstruction scheme is evaluated based on the combination weight and the decision matrix, and an optimal scheme and an alternative scheme are obtained. The method has the advantages that certain variables incapable of being given precisely and variables incapable of being precisely quantized in the restoring process are fully considered, the reliability and flexibility of the decision are improved, and restoring delay or even restoring failures caused by immature quantification methods and attribute precision are avoided.

Description

A kind of mixed multi-attribute groups Decision Method for the assessment of rack reconfiguration scheme

Technical field

The invention belongs to power system recovery field, specifically relate to a kind of mixed multi-attribute groups Decision Method for the assessment of rack reconfiguration scheme in a kind of power system recovery stage.

Background technology

Rack reconstruct is the important stage of power system recovery, and main target utilizes black starting-up power supply to build stable trunk rack, ready for recovering load comprehensively.This stage needs to strengthen the contact between generating plant, improves the power supply reliability of station service and carries out the cyclization arranged side by side and intrasystem of subsystem.Power system recovery is the combinatorial optimization problem of a multistage, multivariate, multiple goal, multiple constraint, and being pressed for time, in pressure is large, task is heavy recovering process, traffic control personnel are in the urgent need to the reliable aid decision-making method of one.Current power system recovery is generally that layering and zoning is carried out, subsystem is dynamically determined to recover target according to recovering process, may be there are several and recover target in each recovering step, thus produce multiple recovery scheme, and several factors such as switching overvoltage, self-excitation etc. all can have an impact to the security and stability of resume speed and system in rejuvenation, therefore the assessment of rack reconfiguration scheme belongs to typical Multiple Attribute Decision Problems.

The part attribute of multiple attribute decision making (MADM) matrix can use perfect number to quantize, and the frequency and voltage started as switch motion number of times, transient overvoltage multiple, large-scale subsidiary engine falls.These attributes can be obtained by site test or emulation.But to the reliability of quantitative attributes as equipment, the importance etc. of node, if adopt perfect number quantification can bring the loss of information, do not fully demonstrate the complicacy of human cognitive and the uncertainty of objective things, adopt linguistic variable to describe more suitable.Rejuvenation can be subject to the impact of a lot of uncertain factor, and part attribute is accurately estimated as release time etc. is very difficult, and this type of parameter adopts fuzzy number, interval number, Vague collection etc. can reflect that probabilistic variable description is more suitable.Therefore the assessment of rack reconfiguration scheme is a hybrid multi-attribute decision making problem, needs the relation considered between qualitative attribute and quantitative attributes, and needs the combination taking into account determinacy variable and uncertain attribute.

The Multiple Attribute Decision Problems research of current power system recovery concentrates on the black starting-up stage mostly, and because stage system structure is simple for this reason, recovery scheme is less, and research complexity is low, and the scheme evaluation research of rack reconstruction stage is less.Be directed to that quantitative attributes and qualitative attribute combine, situation that determinacy variable and uncertain attribute combine, most research is all be that same type carries out multiple attribute decision making (MADM) by dissimilar variables transformations at present.This disposal route can bring much meaningless workload, extends the decision-making time, and can lose partial information in attribute unitizes process, is unfavorable for the carrying out of decision-making, and decision-making has lacked general applicability; At present for unit and load importance, recover the factors such as risk not very ripe method quantize, rashly use perfect number to quantize to cause recovering to incur loss through delay even falling flat because of algorithm errors.

Summary of the invention

The object of the invention is to overcome the above-mentioned deficiency of existing methods, providing a kind of mixed multi-attribute groups Decision Method for the assessment of rack reconfiguration scheme.The method go for electric system have a power failure on a large scale after partition recovery or after brown-outs System recover scheme evaluation sequence.Decrease the information loss in attribute conversion process, eliminate adverse effect quantitative attributes and uncertain attribute precision brought.To verify and be combined with evaluation part, improve the efficiency of decision-making, save the decision-making time.Can carry out comprehensively, embodying the attribute bias of colony, and considering the significance level of different decision maker to the suggestion of decision maker different in group decision.In decision-making, the objective information of comprehensive colony's attribute bias and decision matrix, makes decision process more reasonable.Be conducive to auxiliary traffic control personnel and formulate recovery plan, improve reliability and dirigibility that rejuvenation judges decision-making.

The present invention adopts following technical scheme to achieve these goals:

A kind of mixed multi-attribute groups Decision Method for the assessment of rack reconfiguration scheme, comprise: determine the qualitative attribute of subsystems in electric system subregion, quantitative attributes and uncertain genus, set up decision matrix, determine the objective weight of described attribute according to decision matrix; In group decision, different decision maker uses pairwise comparison matrix or Fuzzy Complementary Judgment Matrices to provide the judgement of described relative significance of attribute respectively, obtain subjective weight, described subjective weight and objective weight are combined and generates combining weights, based on combining weights and decision matrix, rack reconfiguration scheme is assessed, obtain optimal case and alternative scheme.

Described concrete steps are:

S1: network and the equipment state of determining subsystems in electric system subregion, dynamically determines current recovery target, uses dijkstra's algorithm to obtain the Weighted Shortest Path Problem footpath of having recovered network and having recovered between target;

S2: emulate according to the network parameter of subsystem and state, determine the numerical value of qualitative attribute, if every attribute value is all verified by power system security, then the program meets the demands; If a certain attribute does not meet safe stability of power system requirement, then network parameter and state are adjusted, if the scheme after adjustment still can not meet the demands, then give up the program;

S3: group decision-making negotiation is carried out to all schemes remained, provides the method for expressing of quantitative attributes describing mode and uncertain attribute;

S4: the decision maker participating in group decision adopts pairwise comparison matrix or Fuzzy Complementary Judgment Matrices to be described the relative importance between each attribute respectively, the description then adopting the method for mathematical programming to assemble each decision maker generates the subjective weight of attribute;

S5: decision maker consults to set up unique decision matrix, carries out standardization to decision matrix;

S6: according to the decision matrix after standardization, based on the objective weight of maximum deviation model determination attribute; Based on the combining weights of minimum relative information entropy model computation attribute;

S7: according to combinations of attributes weight and decision matrix, carries out assessment sequence to optional program, optimum scheme comparison and alternative scheme.

In described step S2, the out-of-limit adjustment means of attribute comprise:

For superpotential and reactive balance problem, adopting the switching of reactive power compensator, adjustment generator terminal voltage, the mode of Load adjustment or adjustment load tap changer position adjusts, reducing overvoltage level by reducing system reference voltage level;

The transient voltage dip that subsidiary engine starts is adjusted by terminal voltage with raising by the boot sequence of adjustment subsidiary engine;

Line transmission capacity is out-of-limit to be adjusted by adjustment generator output and load;

The frequency of load restoration is fallen and is adjusted by Load adjustment amount of recovery;

If need the amount of projects of adjustment more, the mode adopting the circuit weight changed in route searching to arrange adjusts.

In described step S4, the mathematical model of subjective weight is as follows:

$\min F={[\underset{l=1}{\overset{k_1}{\mathrm{\Σ}}}\left({\mathrm{\α}}_{\left(l\right)}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left|{{\mathrm{\ϵ}}_i}^{\left(l\right)}\right|}^p\right)+\underset{m=1}{\overset{k_2}{\mathrm{\Σ}}}\left({\mathrm{\β}}_{\left(m\right)}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{\left|{r_j}^{\left(m\right)}\right|}^p\right)]}^{\frac{1}{p}}$

$\text{s.t.}\left\{\begin{array}{cc}(B^{\left(l\right)}-\mathrm{nI})W-E^{\left(l\right)}=0,l=\mathrm{1,2},...,k_1& \\ (H^{\left(m\right)}-\mathrm{nI})W-R^{\left(m\right)}=0,m=\mathrm{1,2},...,k_2& \\ e^{T}W=1& \\ W\≥0& \end{array}\right.$

Wherein, k ₁individual decision maker uses pairwise comparison matrix to carry out the determination of relative importance, k ₂individual decision maker uses Fuzzy Complementary Judgment Matrices to carry out the determination of relative importance; N represents attribute number; I represents n rank unit matrix; W is the subjective weight after assembling; B ^(l)the pairwise comparison matrix adopting AHP 1-9 scale to determine, E ^(l)=(ε ₁ ^(l), ε ₂ ^(l)..., ε _n ^(l)) represent the error vector of pairwise comparison matrix; H ^(m)fuzzy Complementary Judgment Matrices P ^(m)through processing the matrix obtained, wherein h _ij=p _ij/ p _ji, Fuzzy Complementary Judgment Matrices P ^(m)determine according to AHP 0.1-0.9 scale; R ^(m)=(r ₁ ^(m), r ₂ ^(m)..., r _n ^(m)) represent the error vector of Fuzzy Complementary Judgment Matrices; E=(1 ..., 1) ^t, expression be subjective weight sum be the constraint of 1; α _(l)>=0, l=1,2 ..., k ₁, β _(m)>=0, m=1,2 ..., k ₂, α _(l), β _(m)represent the weight of different decision maker in group decision respectively, parameter p is positive integer, and usual value is p=1,2 ..., ∞.

Qualitative attribute, quantitative attributes and uncertain attribute is comprised in the decision matrix of described step S5; The data type of decision matrix comprises one or more in perfect number, interval number, fuzzy number, linguistic variable, Vague collection and cloud model, and the difference between attribute uses Euclidean distance to measure.

In described step S5 to the method that decision matrix carries out standardization be:

To perfect number:

$f_{\mathrm{ij}}=\left\{\begin{array}{c}\frac{\underset{i}{\min }{a}_{\mathrm{ij}}}{a_{\mathrm{ij}}},a_{\mathrm{ij}}\∈J_1\\ \frac{a_{\mathrm{ij}}}{\underset{i}{\max }{a}_{\mathrm{ij}}},a_{\mathrm{ij}}\∈J_2\end{array}\right.$

To interval number:

$f_{\mathrm{ij}}=\left\{\begin{array}{c}[\frac{\underset{i}{\min }{a_{\mathrm{ij}}}^L}{{a_{\mathrm{ij}}}^U},\frac{\underset{i}{\min }{a_{\mathrm{ij}}}^L}{{a_{\mathrm{ij}}}^L}],a_{\mathrm{ij}}\∈J_1\\ [\frac{{a_{\mathrm{ij}}}^L}{\underset{i}{\max {a_{\mathrm{ij}}}^U}},\frac{{a_{\mathrm{ij}}}^U}{\underset{i}{\max {a_{\mathrm{ij}}}^U}}],a_{\mathrm{ij}}\∈J_2\end{array}\right.$

To fuzzy number:

$f_{\mathrm{ij}}=\left\{\begin{array}{c}[\frac{\underset{i}{\min {a_{\mathrm{ij}}}^L}}{{a_{\mathrm{ij}}}^U},\frac{\underset{i}{\min {a_{\mathrm{ij}}}^L}}{{a_{\mathrm{ij}}}^M},\frac{\underset{i}{\min {a_{\mathrm{ij}}}^L}}{{a_{\mathrm{ij}}}^L}],a_{\mathrm{ij}}\∈J_1\\ [\frac{{a_{\mathrm{ij}}}^L}{\underset{i}{\max {a_{\mathrm{ij}}}^U}},\frac{{a_{\mathrm{ij}}}^M}{\underset{i}{\max {a_{\mathrm{ij}}}^U}},\frac{{a_{\mathrm{ij}}}^U}{\underset{i}{\max }{a_{\mathrm{ij}}}^U}]{,a}_{\mathrm{ij}}\∈J_2\end{array}\right.$

Wherein perfect number is expressed as a _ij; Interval number is expressed as [a _ij ^l, a _ij ^u]; Fuzzy number is expressed as [a _ij ^l, a _ij ^m, a _ij ^u]; J ₁, J ₂cost type property set and profit evaluation model property set respectively;

Variable Vague collection, cloud model and linguistic variable being converted into other form processes.

In described step S6, based on maximum deviation model, the method calculating objective weight is:

$\max J=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{w_j}^{\′\′}D(f_{\mathrm{ij}},f_{\mathrm{kj}})$

$\left\{\begin{array}{cc}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{w_j}^{\′\′2}=1,j=\mathrm{1,2},...,n& \\ {w_j}^{\′\′}\≥0& \end{array}\right.$

Wherein, optional program number is m, and attribute number is n; w _j" represent the objective weight of attribute; Euclidean distance between D () function representation two assessed values; F=(f _ij) _{m × n}represent standardization decision matrix; J is the Weighted distance between different schemes attribute.

Based on minimum relative information entropy model, the method for calculation combination weight is:

$\min J=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j\left[\ln \frac{w_j}{{w_j}^{\′}}\right]+\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j[ln\frac{w_j}{{w_j}^{\′\′}}]$

$s.t.\left\{\begin{array}{c}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1\\ w_j\≥0,j=\mathrm{1,2},...,n\end{array}\right.$

Wherein w _j', w _j", j=1,2 ..., n represents subjective weight and the objective weight of evaluation index respectively; w _j, j=1,2 ..., n is the combining weights finally obtained.

In described step S7, assessment algorithm is VIKOR algorithm, and its step is as follows:

If standardization decision matrix F=is (f _ij) _{m × n}, total m decision scheme, A={a _i, 1≤i≤m; Decision scheme has n attribute, G={G _j, 1≤j≤n; Scheme a _ito attribute G _jstandardised assessment value be f _ij;

If G _j{ j ∈ N ₁=(1,2 ..., h ₁) be Real-valued index set, G _j{ j ∈ N ₂=(h ₁+ 1, h ₁+ 2 ..., h ₂) be interval type index set, G _j{ j ∈ N ₃=(h ₂+ 1, h ₂+ 2 ..., h ₃) be fuzzy number index set, G _j{ j ∈ N ₄=(h ₃+ 1, h ₃+ 2 ..., h ₄) be Vague collection index set, G _j{ j ∈ N ₅=(h ₄+ 1, h ₄+ 2 ..., h ₅) be linguistic variable index set, G _j{ j ∈ N ₆=(h ₅+ 1, h ₅+ 2 ..., h ₆) be cloud model index set;

1) preparation of data: the data type comprising attribute in standardization decision matrix, the combining weights of attribute and decision matrix;

2) the positive ideal solution F of all optional programs is determined ⁺with minus ideal result F ^-, so-called positive ideal solution is made up of the optimum evaluation value of each index, and minus ideal result is the most bad assessed value composition;

3) the comprehensive assessment optimum solution S of numerical procedure _iinferior solution R most with comprehensive assessment _i;

4) the advantage ratio Q of numerical procedure generation _i;

5) sequence is determined;

According to S _i, R _i, Q _inumerical value sort respectively according to order from small to large, obtain three collating sequences, it is generally acknowledged that in ascending order arrangement, the scheme that numerical value is less has and preferably sorts;

6) compromise proposal or compromise proposal disaggregation is determined;

If meet following two conditions, the scheme a that Q value is minimum ⁽¹⁾be exactly that final compromise sequence is separated:

Condition 1: wherein a ⁽²⁾for arranging deputy optional program by Q value ascending order;

Condition 2:a ⁽¹⁾come the scheme of foremost in the sequence according to S value or the arrangement of R value ascending order;

If one of them condition can not meet, then the compromise solution obtained is not unique, but obtains compromise proposal disaggregation; If wherein condition 1 does not meet, then scheme a ⁽¹⁾, a ⁽²⁾..., a ^(r)be its compromise solution, wherein a ^(r)meet if condition 2 does not meet, scheme a ⁽¹⁾and a ⁽²⁾for compromise solution;

If obtain multiple compromise solution, traffic control personnel are according to any recovery scheme in actual conditions selection compromise solution as optimal case, and other schemes are as subsequent use; Or the comprehensive selection of recovery scheme and alternative scheme is carried out according to S, R, Q tri-collating sequences.

Beneficial effect of the present invention:

1) each factor affecting rack reconstruct security and stability and resume speed can be considered, qualitative attribute and quantitative attributes are combined, and can assemble determinacy variable and uncertain attribute, decrease the loss of information, avoid the recovery that uncertainty attribute deterministic and quantitative attributes quantify defects cause and postpone even failed.

2) use combining weights to weigh the relative importance of attribute in decision-making, both can consider the subjective preferences of decision maker to attribute, and can consider again and be conducive to the reliability of decision-making by the objective information that decision matrix provides.

3) go for, in group decision environment, the significance level of different decision maker can being considered.Can according to decision maker hobby and familiarity chooses pairwise comparison matrix or Fuzzy Complementary Judgment Matrices carries out relative importance description, mandatory requirement decision maker does not provide the relative importance judgment matrix of particular form, be conducive to the accuracy of decision-making, utilize colony's preference determination weight to be conducive to reliability and the dirigibility of decision-making.

4) assessment algorithm can determine one or more compromise solution scheme, and can using other compromise solutions as alternative scheme, or comprehensively determine recovery scheme and alternative scheme according to three kinds of different sequences, if recovery scheme cannot be implemented can enable alternative scheme rapidly smoothly, reduce the decision-making time, improve the efficiency of decision-making.

5) safety check and attribute are assessed combine, simplify decision process.

6) the mixed multi-attribute groups Decision Method proposed both can consider the maximization of group effectiveness, can minimize again individual sorry value, was conducive to considering worst factor in recovery.

7) this method is applicable to the formulation of recovery scheme, and can recover to carry out aid decision making to the rack reconstruct after having a power failure on a large scale or the rack after brown-outs.

Accompanying drawing explanation

Fig. 1 is process flow diagram of the present invention;

Fig. 2 is the above rank stem network of the western electrical network subsystem 220kV in Shandong.

Embodiment

Below in conjunction with drawings and Examples, the present invention will be further described.

As shown in Figure 1, the key step of the mixed multi-attribute groups Decision Method of rack reconfiguration scheme assessment is as follows:

S1: network and the equipment state of determining subsystems in electric system subregion, dynamically determine current recovery target, circuit weight is set to line length, electrical distance, one of line charging electric capacity or line loop operation time, uses dijkstra's algorithm to obtain the Weighted Shortest Path Problem footpath of having recovered network and having recovered between target;

S2: emulate according to the network parameter of subsystem and state, determine the numerical value of qualitative attribute, if every attribute value is all by safety check, then the program meets the demands; If a certain attribute does not meet safe stability of power system requirement, then network parameter and state are adjusted, if the scheme after adjustment still can not meet the demands, then give up the program;

S3: group decision-making negotiation is carried out to all schemes remained, the method for expressing of uncertain attribute providing quantitative attributes describing mode and can not accurately estimate;

S4: the decision maker participating in group decision adopts pairwise comparison matrix or Fuzzy Complementary Judgment Matrices to be described the relative importance between attribute respectively, the description then adopting the method for mathematical programming to assemble each decision maker generates the subjective weight of attribute;

S5: decision maker consults to set up unique decision matrix, and carries out standardization to decision matrix;

S6: according to the decision matrix after standardization, based on the objective weight of maximum deviation model determination attribute; Based on minimum relative information entropy model calculation combination weight;

S7: according to combinations of attributes weight and standardization decision matrix, carries out assessment sequence to optional program, optimum scheme comparison and alternative scheme.

In described step S1, can by mixed multi-attribute groups Decision Method and the existing EMS/DTS system integration, Real-time Obtaining network structure and equipment state in rejuvenation, judge the recovery extent of network accordingly.The network of Power resumption is thought a node, then determines to recover target according to the state of recovery plan, target trunk rack and concrete recovery policy.Determine to recover the method that target generally can adopt graph theory, search spread subsystem, first check in 220kV and above network whether there is the main force to be restored unit that capacity is 100MW ~ 600MW, be set to recover target, be preferentially set to recover target if there is Hydropower Unit.After unit has recovered or recovers failure, 220kV and other important transformer station of higher level in search rack, be set to recover target.If unit exists meritorious equilibrium problem, near unit, search for load, load restoration is set to recover target.If subsystem has possessed certain stability, subsystem and trunk rack, external power are supported, other subsystems arranged side by side as recovering target.

In described step S2 qualitative attribute by negotiation determine, general selectable variable comprise switching overvoltage, continue power-frequency overvoltage, voltage transitions number of times, subsidiary engine start transient voltage and frequency is fallen, the frequency of load restoration is fallen, the available transfer capability, line length, unit capacity etc. of circuit.Adjustment means for superpotential and reactive balance problem comprise the adjustment of the switching of various reactive power compensator, the adjustment of generator terminal voltage, load adjustment and load tap changer position, and overvoltage level can be reduced by reducing system reference voltage level, wherein 220kV transformer station does not generally adjust load tap changer position.If need the amount of projects of adjustment more, the set-up mode of the circuit weight changed in route searching can be considered.The transient voltage dip that subsidiary engine starts can be improved by terminal voltage by the boot sequence and suitably raising of adjustment subsidiary engine.Line transmission capacity is out-of-limit can be solved by adjustment generator output and load.The frequency of load restoration is fallen and can be improved by Load adjustment amount of recovery.

Need to determine choosing and corresponding appraisal procedure of quantitative attributes, uncertainty attribute in described step S3, these attributes can adopt the one in linguistic variable, interval number, fuzzy number, Vague collection, cloud model or severally to represent.The variable determined is needed mainly to be divided into two classes in this step, one class is uncertainty due to objective things and complicacy, be difficult to attribute variable-value being carried out to accurately estimation, such as release time etc., direct use perfect number quantification can have an impact to the reliability of decision-making, and uncertain attribute can be used to be described the scope of its possibility value and numeric distribution; Another kind of is describe the language comment of attribute degree, the reliability of the importance of such as transformer station, the importance of unit, equipment, the possibility etc. of unit generation self-excitation, use exact magnitude workload large, also consideration is difficult to thorough, recovery is easily caused to be incured loss through delay even failed, adopt the linguistic variable such as important, very important to be described the workload that can reduce decision maker, avoid the incorrect decision quantizing to cause.General employing be 9 ± 2 linguistic scale compare the custom meeting human cognitive, the corresponding relation of dissimilar variable is as shown in table 1.

Table 1 different variable comment corresponding relation

Linguistic variable	Interval number	Fuzzy number	Vague collection	Cloud model
					Extreme difference (inessential, unreliable)	[0,0.05]	[0,0,0.1]	[0,0]	[0,0.0167,0.0001]
Very poor (inessential, unreliable)	[0.05,0.15]	[0,0.1,0.2]	[0.1,0.15]	[0.1,0.0167,0.0001]
					Difference (inessential, unreliable)	[0.15,0.25]	[0.1,0.2,0.3]	[0.2,0.3]	[0.2,0.0167,0.0001]
Poor (inessential, unreliable)	[0.25,0.35]	[0.2,0.3,0.4]	[0.3,0.45]	[0.3,0.0167,0.0001]
					Slightly poor (inessential, unreliable)	[0.35,0.45]	[0.3,0.4,0.5]	[0.4,0.6]	[0.4,0.0167,0.0001]
Generally	[0.45,0.55]	[0.4,0.5,0.6]	[0.5,0.5]	[0.5,0.0167,0.0001]
					Slightly good (important, reliable)	[0.55,0.65]	[0.5,0.6,0.7]	[0.6,0.8]	[0.6,0.0167,0.0001]
Better (important, reliable)	[0.65,0.75]	[0.6,0.7,0.8]	[0.7,0.85]	[0.7,0.0167,0.0001]
					Good (important, reliable)	[0.75,0.85]	[0.7,0.8,0.9]	[0.8,0.9]	[0.8,0.0167,0.0001]
Very well (important, reliable)	[0.85,0.95]	[0.8,0.9,1]	[0.9,0.95]	[0.9,0.0167,0.0001]
					Fabulous (important, reliable)	[0.95,1]	[0.9,1,1]	[1,1]	[1,0.0167,0.0001]

In described step S4, it is more difficult that decision maker directly provides attribute weight numeric ratio, and the relative importance of attribute ratio is easier to provide.The status of different decision maker, education background, place tissue, position, incomplete same to the familiarity of problem, expertise, can not be identical to the understanding of relative significance of attribute, also may there is difference in the expression-form of the relative importance provided.Relative importance expression-form common at present has pairwise comparison matrix and Fuzzy Complementary Judgment Matrices two kinds, employing AHP1-9 scale and AHP 0.1-0.9 scale describe the relative importance between attribute respectively, the requirement of these two kinds of matrixes and expression-form difference are very large, need to adopt the method for mathematical programming to assemble.

Judgment matrix refers to the importance matrix between two between element of the same attribute layer of structure.The judgment matrix form of pairwise comparison matrix is:

Wherein for 1≤i, j, k≤n, b _ij=1/b _ji, b _ij=b _ik/ b _jk, b _ij> 0, b _ijrepresent that attribute i is to the significance level of attribute j, adopt AHP1-9 scale to represent.

Table 2.AHP 1-9 scale

Fuzzy Complementary Judgment Matrices employing AHP 0.1-0.9 scale represents the importance between attribute, and its expression-form is:

Wherein p _ij+ p _ji=1, p _ii=0.5, p _ij>=0, pi _jrepresent that attribute i is to the significance level of attribute j.

Table 3.AHP 0.1-0.9 scale

K is had in assumed group decision-making ₁individual decision maker uses pairwise comparison matrix to carry out the determination of relative importance, k ₂individual decision maker uses Fuzzy Complementary Judgment Matrices.The mathematical model then calculating subjective weight is:

$\min F={[\underset{l=1}{\overset{k_1}{\mathrm{\Σ}}}\left({\mathrm{\α}}_{\left(l\right)}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left|{{\mathrm{\ϵ}}_i}^{\left(l\right)}\right|}^p\right)+\underset{m=1}{\overset{k_2}{\mathrm{\Σ}}}\left({\mathrm{\β}}_{\left(m\right)}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{\left|{r_j}^{\left(m\right)}\right|}^p\right)]}^{\frac{1}{p}}$

$\text{s.t.}\left\{\begin{array}{cc}(B^{\left(l\right)}-\mathrm{nI})W-E^{\left(l\right)}=0,l=\mathrm{1,2},...,k_1& \\ (H^{\left(m\right)}-\mathrm{nI})W-R^{\left(m\right)}=0,m=\mathrm{1,2},...,k_2& \\ e^{T}W=1& \\ W\≥0& \end{array}\right.$

Wherein, n represents attribute number; I represents n rank unit matrix; W is the subjective weight after assembling; B ^(l)the pairwise comparison matrix adopting AHP 1-9 scale to determine, E ^(l)=(ε ₁ ^(l), ε ₂ ^(l)..., ε _n ^(l)) represent the error vector that pairwise comparison matrix brings because of disturbance and inconsistency; H ^(m)fuzzy Complementary Judgment Matrices P ^(m)through processing the matrix obtained, wherein h _ij=p _ij/ p _ji, Fuzzy Complementary Judgment Matrices is determined according to AHP 0.1-0.9 scale; R ^(m)=(r ₁ ^(m), r ₂ ^(m)..., r _n ^(m)) represent the error vector that Fuzzy Complementary Judgment Matrices brings due to disturbance and inconsistency in judgement; E=(1 ..., 1) ^twhat represent be subjective weight sum is the constraint of 1; α _(l)>=0, l=1,2 ..., k ₁, β _(m)>=0, m=1,2 ..., k ₂, represent the weight of different decision maker in group decision, the weight of decision maker can consult to determine or use AHP method to determine, revises and determine after also can carrying out mutual negotiation according to the consistance judged between different decision maker.

Mathematical programming model is above solved and namely can obtain subjective weight.The value of parameter p can get all positive integers in theory, but is p=1 to calculate easy usual value, and 2 ..., ∞.

As p=1, Optimized model can be converted into linear programming model:

$\min F=\underset{l=1}{\overset{k_1}{\mathrm{\Σ}}}{\mathrm{\α}}_{\left(l\right)}{e}^T(E^{\left(l\right)+}+E^{\left(l\right)-})+\underset{m=1}{\overset{k_2}{\mathrm{\Σ}}}{\mathrm{\β}}_{\left(m\right)}{e}^T(R^{\left(m\right)+}+R^{\left(m\right)-})$

$\text{s.t.}\left\{\begin{array}{cc}(B^{\left(l\right)}-\mathrm{nI})W-E^{\left(l\right)+}+E^{\left(l\right)-}=0,l=\mathrm{1,2},...,k_1& \\ (H^{\left(m\right)}-\mathrm{nI})W-R^{\left(m\right)+}+R^{\left(m\right)-}=0,m=\mathrm{1,2},...,k_2& \\ e^{T}W=1& \\ W,E^{\left(l\right)+},E^{\left(l\right)-},R^{\left(m\right)+},R^{\left(m\right)-}\≥0& \end{array}\right.$

Wherein E ^(l)+=(ε ₁ ^(l)+, ε ₂ ^(l)+..., ε _n ^(l)+), E ^-=(ε ₁ ^(l)-, ε ₂ ^(l)-..., ε _n ^(l)-), R ^(m)+=(r ₁ ^(m)+, r ₂ ^(m)+..., r _n ^(m)+), R ^(m)-=(r ₁ ^(m)-, r ₂ ^(m)-..., r _n ^(m)-), be by error vector E ^(l), R ^(m)generate through following change:

${{\mathrm{\ϵ}}_i}^{\left(l\right)+}=\frac{{{\mathrm{\ϵ}}_i}^{\left(l\right)}+\left|{{\mathrm{\ϵ}}_i}^{\left(l\right)}\right|}{2},{{\mathrm{\ϵ}}_i}^{-}=\frac{-{{\mathrm{\ϵ}}_i}^{\left(l\right)}+\left|{{\mathrm{\ϵ}}_i}^{\left(l\right)}\right|}{2},i=\mathrm{1,2},...,n,l=\mathrm{1,2},...,k_1$

${r_i}^{\left(m\right)+}=\frac{{r_i}^{\left(i\right)}+\left|{r_i}^{\left(m\right)}\right|}{2},{{\mathrm{\ϵ}}_i}^{-}=\frac{-{r_i}^{\left(m\right)}+\left|{r_i}^{\left(m\right)}\right|}{2},i=\mathrm{1,2},...,n,m=\mathrm{1,2},...,k_2$

As p=2, Optimized model can be converted into quadratic programming model:

min F＝W ^TGW

$s.t.\left\{\begin{array}{c}G=\underset{l=1}{\overset{k_1}{\mathrm{\Σ}}}{\mathrm{\α}}_{\left(l\right)}{(B^{\left(l\right)}-\mathrm{nI})}^T(B^{\left(l\right)}-\mathrm{nI})+\underset{m=1}{\overset{k_2}{\mathrm{\Σ}}}{(H^{\left(m\right)}-\mathrm{nI})}^T(H^{\left(m\right)}-\mathrm{nI})\\ e^{T}W=1\\ W\≥0\end{array}\right.$

As p=∞, Optimized model is also linear programming model:

$\min F=\underset{l=1}{\overset{k_1}{\mathrm{\Σ}}}{\mathrm{\α}}_{\left(l\right)}{\mathrm{\ϵ}}^{\left(l\right)}+\underset{m=1}{\overset{k_2}{\mathrm{\Σ}}}{\mathrm{\β}}_{\left(m\right)}{r}^{\left(m\right)}$

$s.t.\left\{\begin{array}{c}-{\mathrm{\ϵ}}^{\left(l\right)}\·e\≤(B^{\left(l\right)}-\mathrm{nI})W\≤{\mathrm{\ϵ}}^{\left(l\right)}\·e\\ -r^{\left(m\right)}\·e\≤(H^{\left(m\right)}-\mathrm{nI})W\≤r^{\left(m\right)}\·e\\ e^{T}W=1\\ W,{\mathrm{\ϵ}}^{\left(l\right)},r^{\left(m\right)}\≥0\\ {\mathrm{\ϵ}}^{\left(l\right)}=\max \left|{{\mathrm{\ϵ}}_i}^{\left(l\right)}\right|\\ r^{\left(m\right)}=\max \left|{r_i}^{\left(m\right)}\right|\end{array}\right.$

Above model all can use Lingo, Matlab or other software to solve, and calculated amount is little, and solving speed is fast.

In described step S5, if decision matrix D=is (d _ij) _{m × n}, total m decision scheme, A={a _i, 1≤i≤m; Decision scheme has n attribute, G={G _j, 1≤j≤n; Scheme a _ito attribute G _jassessed value be d _ij; If G _j{ j ∈ N ₁=(1,2 ..., h ₁) be Real-valued index set, G _j{ j ∈ N ₂=(h ₁+ 1, h ₁+ 2 ..., h ₂) be interval type index set, G _j{ j ∈ N ₃=(h ₂+ 1, h ₂+ 2 ..., h ₃) be fuzzy number index set, G _j{ j ∈ N ₄=(h ₃+ 1, h ₃+ 2 ..., h ₄) be Vague collection index set, G _j{ j ∈ N ₅=(h ₄+ 1, h ₄+ 2 ..., h ₅) be linguistic variable index set, G _j{ j ∈ N ₆=(h ₅+ 1, h ₅+ 2 ..., h ₆) be cloud model index set;

Need impact decision matrix standardization being eliminated dimension, make, between each attribute, there is comparability.Attribute type gets cost type and profit evaluation model two class, and the attribute of other types need carry out transforming carrying out standardization again.The variable that Vague collection, cloud model and linguistic variable can be converted into other form processes again, and standardized method is as shown in table 4, and wherein perfect number is expressed as a _ij, interval number is expressed as [a _ij ^l, a _ij ^u], fuzzy number is expressed as [a _ij ^l, a _ij ^m, a _ij ^u].

Table 4 standardized method

Objective weight in described step S6 the information depending on decision matrix and provide is provided, it is generally acknowledged that the information that the attribute that change is larger provides in decision-making is more, larger weight should be given, if certain attribute is all identical in all schemes, so its for decision-making be do not have helpful.Difference between decision-making uses Euclidean distance to measure.

Euclidean distance between two ATTRIBUTE INDEX is:

$d(A,B)=\left\{\begin{array}{c}|a-b|,J\∈N_1\\ \sqrt{\frac{1}{2}[{(a^L-b^L)}^2+{(a^U-b^U)}^2]},J\∈N_2\\ \sqrt{\frac{1}{3}[{(a^L-b^L)}^2+{(a^M-b^M)}^2+{(a^U-b^U)}^2]},J\∈N_3\\ \sqrt{\frac{1}{2}[{(t_A-t_B)}^2+{(f_A-f_B)}^2+{({\mathrm{\π}}_A-{\mathrm{\π}}_B)}^2]},J\∈N_4\\ \sqrt{\frac{1}{2}[{(E{x}_A-E{x}_B)}^2+{(E{n}_A-E{n}_B)}^2+{(H{e}_A-H{e}_B)}^2]},J\∈N_6\end{array}\right.$

Wherein perfect number is a, b, and interval number is [a ^l, a ^u], [b ^l, b ^u], fuzzy number is [a ^l, a ^m, a ^u], [b ^l, b ^m, b ^u], cloud model is (Ex _a, En _a, He _a), (Ex _b, En _b, He _b), Vague collection is [t _a,-1f _a], t _b[-, f1 _b, π] _a=1-f _a-t _a, π _b=1-f _b-t _b.Distance calculating method due to linguistic variable is not very ripe, so the variable being generally translated into other types calculates Euclidean distance.In order to improve the confidence level of decision-making, be generally converted into identical attribute type with linguistic variable in a decision-making, but the attribute type transformed in different decision-making does not require identical.

The mathematical model asking for objective weight based on maximum deviation model is:

$\max J=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{w_j}^{\′\′}D(f_{\mathrm{ij}},f_{\mathrm{kj}})$

$s.t.\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{w^{\′\′2}}_j=1,j=\mathrm{1,2},...,n$

w _j″≥0

Wherein w _j" represent the objective weight of attribute; Euclidean distance between D () function representation two assessed values; F=(f _ij) _{m × n}represent standardization decision matrix; Unitization constraint should be met without weight under preference profile decision maker.

The subjective weight obtained in objective weight and above-mentioned steps S5 obtains combining weights as final weight according to minimum Relative Entropy models coupling:

$\min J=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j\left[\ln \frac{w_j}{{w_j}^{\′}}\right]+\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j[ln\frac{w_j}{{w_j}^{\′\′}}]$

$s.t.\left\{\begin{array}{c}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1\\ w_j\≥0,j=\mathrm{1,2},...,n\end{array}\right.$

Wherein w _j', w _j", j=1,2 ..., n represents subjective weight and the objective weight of evaluation index respectively; w _j, j=1,2 ..., n represents combining weights.

The step of the assessment algorithm in described step S7 is:

1) preparation of data.Data needed for evaluation process mainly comprise the data type of combinations of attributes weight in normalized matrix F, the S6 in above-mentioned steps S5 and each attribute

2) the positive ideal solution F of all optional programs is determined ⁺with minus ideal result F ^-, so-called positive ideal solution is made up of the optimum evaluation value of each index, and minus ideal result is the most bad assessed value composition;

3) the comprehensive assessment optimum solution S of numerical procedure _iinferior solution R most with comprehensive assessment _i;

$S_i=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j\frac{D({F_i}^{+},f_{\mathrm{ij}})}{D({F_i}^{+},{F_i}^{-})}$

$R_i=\max \left(w_j\frac{D({F_i}^{+},f_{\mathrm{ij}})}{D({F_i}^{+},{F_i}^{-})}\right)$

4) the advantage ratio Q of numerical procedure generation _i;

$Q_i=v\frac{S_i-{\min S}_i}{\max S_i-{\min S}_i}+(1-v)\frac{R_i-\min R_i}{\max R_i-\min R_i}$

Wherein the value of parameter v represents that decision maker more notes most of suggestion or more notes individual sorry value, generally gets v=0.5.Just represent the suggestion more noting most people as v > 0.5, otherwise more note individual sorry value.

5) sequence is determined;

According to S _i, R _i, Q _inumerical value sort respectively according to order from small to large, obtain three collating sequences, it is generally acknowledged that in ascending order arrangement, the scheme that numerical value is less has and preferably sorts;

6) compromise proposal or compromise proposal disaggregation is determined

If meet following two conditions, the scheme a that Q value is minimum ⁽¹⁾be exactly that final compromise sequence is separated:

Condition 1: wherein a ⁽²⁾for arranging deputy optional program by Q value ascending order;

Condition 2:a ⁽¹⁾come the scheme of foremost in the sequence according to S value or the arrangement of R value ascending order;

If one of them condition can not meet, then the compromise solution obtained is not unique, but obtains compromise proposal disaggregation.If wherein condition 1 does not meet, then scheme a ⁽¹⁾, a ⁽²⁾..., a ^(r)be its compromise solution, wherein a ^(r)meet if condition 2 does not meet, scheme a ⁽¹⁾and a ⁽²⁾for compromise solution;

If obtain multiple compromise solution, represent that the optimal case of Q sequence significantly can not be better than other schemes in compromise solution, this probably exists in decision-making.Now traffic control personnel can according to self on being in the push of different regions grid structure and equipment, recovery plan, select any recovery scheme in compromise solution as optimal case on the impact of follow-up recovery, and other schemes are as subsequent use.Or according to S _i, R _i, Q _ithree collating sequence comprehensive selection optimal cases and alternative scheme.Alternative scheme can be enabled rapidly when uncertain factor affects when optimal case direct motion is carried out, reduce the decision-making time, improve the efficiency of decision-making.

After system is had a power failure on a large scale, need first recognition network and equipment state, then the division of subsystem is carried out as the case may be, in each subsystem, dynamically determine to recover target according to the recovery plan of formulation in advance, the structure of target trunk rack and concrete recovery policy, according to critical path method (CPM) determination recovery scheme, its margin of safety value may be determined to the factor of influential system security and stability and resume speed as evaluation attribute, just adjust if can not meet the demands.Adjustment means for superpotential and reactive balance problem comprise the adjustment of the switching of various reactive power compensator, the adjustment of generator terminal voltage, load adjustment and load tap changer position, and reduction system reference voltage level can be adopted to reduce overvoltage level, wherein 220kV transformer station does not generally adjust load tap changer position.If need the amount of projects of adjustment more, the set-up mode of the circuit weight changed in route searching can be considered.The transient voltage dip that subsidiary engine starts can be improved by terminal voltage by the boot sequence and suitably raising of adjustment subsidiary engine.Line transmission capacity is out-of-limit can be solved by adjustment generator output and load.The frequency of load restoration falls out-of-limit can improvement by Load adjustment amount of recovery.The scheme do not met the demands after adjustment is given up, and then group decision is consulted to determine that assessment factor forms decision matrix, and decision maker determines the assessment of relative significance of attribute separately.Corresponding attribute weight can be drawn according to the objective information combination of decision maker to colony's preference of attribute and decision matrix, just can carry out comprehensive assessment sequence to optional program in conjunction with attribute weight and decision matrix.

The present invention with the western electrical network subsystem in Shandong for example illustrates embody rule method.Instance system grid structure is as shown in Fig. 2 in accompanying drawing.According to the plan of Shandong Power black starting-up, rack reconstruct start after, the state of this system is that Taishan Pumped Storage Power Station has started the #5 unit of Shi Heng second power plant as black starting-up unit, and this unit success grid-connected can stablizing exert oneself.

Comprehensive consideration is carried out according to the structure of recovery plan, target bulk transmission grid and concrete recovery policy, determine that the main target recovered in this step is main force's unit of subsystem, mainly comprise Heze Plant #5 unit, Pump of Zhou County Power Plant ' #3 unit, Huang Tai power plant #5 unit, the Holy City cogeneration plant #1 unit, canal power plant #5 unit.Can show that corresponding recovery scheme is as shown in table 5 according to critical path method (CPM):

The set of table 5 candidate scheme

Decision maker determines that assessment factor comprises transient overvoltage margin of safety f through consultation ₁, the transient voltage dip margin of safety f that large-scale subsidiary engine starts ₂, power-frequency overvoltage margin of safety f ₃, release time f ₄, unit importance f ₅risk f is recovered with unit ₆.Wherein transient overvoltage margin of safety uses the difference of the maximum superpotential of reality in maximum permission superpotential and system and the superpotential ratio of maximum permission to represent, wherein according to the rules, the voltage multiplication that exceeds most of 220kV circuit is that the superpotential multiple of 3p.u., 500kV circuit is no more than 2p.u.; The ratio that the transient voltage dip margin of safety that large-scale subsidiary engine starts utilizes subsidiary engine to start the transient voltage maximum difference and low-voltage variation setting valve of falling the low-voltage variation setting valve of value and system represents, wherein the setting valve of domestic subsidiary engine low-voltage variation is (0.65p.u., 0.5s), the setting valve of import subsidiary engine low-voltage variation is (0.75p.u., 9s); The margin of safety of power-frequency overvoltage represents with the difference of the maximum power-frequency overvoltage of reality in permission power-frequency overvoltage numerical value maximum in circuit and system and the ratio of maximum permission power-frequency overvoltage, and wherein in China's supertension line, General Requirements power-frequency overvoltage multiple is no more than 1.1p.u.; Release time need be estimated the running time required for unit starting and the running time required for restoration path, need to consider set state, marginal time restriction, machine set type, unit capacity, subsidiary engine start-up time, voltage transitions number of times etc., therefore be difficult to provide accurate release time, the interval estimated can only be provided; The importance of unit is that decision maker discusses the comment determined and represents, needs to consider unit capacity, impact, unit geographic position, unit institute on-load priority etc. on follow-up recovery; Recover that risk is electric pressure according to restoration path, the reliability, voltage transitions number of times, line length etc. of equipment consider and determine.Simulation result and the assessed value of each attribute are as shown in table 6 below.

The attribute assessed value of table 6 optional program

Because the computing method of language comment are not also very perfect at present, therefore language comment is converted into other types variable, corresponding relation is as shown in table 1, then carries out data processing according to standardized method and obtains the decision matrix after standardization as shown in table 7.

Decision matrix after table 7 standardization

Suppose there are two decision maker's participative decision makings, use pairwise comparison matrix and Fuzzy Complementary Judgment Matrices provide the relative importance between attribute respectively, and judgment matrix is respectively C ₁, C ₂.

$C_1=\left[\begin{array}{cccccc}1& 2& \frac{1}{2}& \frac{1}{7}& \frac{1}{8}& \frac{1}{7}\\ \frac{1}{2}& 1& \frac{1}{3}& \frac{1}{8}& \frac{1}{9}& \frac{1}{8}\\ 2& 3& 1& \frac{1}{5}& \frac{1}{7}& \frac{1}{6}\\ 7& 8& 5& 1& \frac{1}{2}& \frac{1}{2}\\ 8& 9& 7& 2& 1& 2\\ 7& 8& 6& 2& \frac{1}{2}& 1\end{array}\right],C_2=\left[\begin{array}{cccccc}0.5& 0.6& 0.4& 0.2& 0.1& 0.2\\ 0.4& 0.5& 0.4& 0.1& 0.1& 0.1\\ 0.6& 0.6& 0.5& 0.3& 0.2& 0.2\\ 0.8& 0.9& 0.7& 0.5& 0.4& 0.2\\ 0.9& 0.9& 0.8& 0.6& 0.5& 0.4\\ 0.8& 0.9& 0.8& 0.6& 0.4& 0.5\end{array}\right]$

The weight of getting two decision maker is respectively 0.3,07, then the method in described step S4 can be used to calculate subjective weight vectors for w'=(0.0440,0.0352,0.0593,0.2200,0.3643,0.2772).Can calculate objective weight according to the decision matrix shown in table 7 is w, and "=(0.2662; 0.2598,0.1872,0.1028; 0.0818; 0.1021) can the comprehensive subjective and objective weight final weight that obtains attribute be then w=(0.1040,0.0812; 0.0985; 0.2007,0.2645,0.2512).

Corresponding S, R, Q numerical value is calculated according to above-mentioned assessment algorithm, as shown in table 8:

Table 8 scheme evaluation value (v=0.5)

According to Q sequence can obtain scheme optimal sequencing for ( ), the compromise solution obtained is scheme 4, meets the requirement of assessment algorithm, and namely preferentially adopt scheme 4, other scheme is in order as alternative scheme.Therefore optimum main force's unit recovery order is the Holy City cogeneration plant #1 unit, Pump of Zhou County Power Plant ' #3 unit, Huang Tai power plant #5 unit, canal power plant #5 unit, Heze Plant #5 unit.The main cause of scheme 4 optimum is that the Holy City cogeneration plant #1 unit capacity is relatively little, and circuit is relatively short, and margin of safety is higher, recovers risk little, and when the initial stage, network was weaker in rack reconstruct, priority restores is conducive to stable grid structure.The reason that scheme 2 priority ratio is higher is that Pump of Zhou County Power Plant ' is higher at the important ratio of Shandong Power, and does black startup test, and traffic control personnel are familiar, and margin of safety also meets the demands.The priority of scheme 3 is higher mainly because decision maker, to the preference of unit importance, powers because Huang Tai power plant primary responsibility is Jinan, and the political significance and the economic results in society that recover Huang Tai power plant are higher.Scheme 1 preferentially do not adopt be because station, Mount Taishan need through comparatively long transmission line be that Heze Plant is powered, transformer station's quantity of process is more, resume speed is slower, and the subsidiary engine of scheme 1 starts transient voltage dip close to limit value, be easy to because of some uncertain disturbances out-of-limit in real process, be therefore not suitable for recovering in the weak situation of recovery initial stage rack.

By reference to the accompanying drawings the specific embodiment of the present invention is described although above-mentioned; but not limiting the scope of the invention; one of ordinary skill in the art should be understood that; on the basis of technical scheme of the present invention, those skilled in the art do not need to pay various amendment or distortion that creative work can make still within protection scope of the present invention.

Claims (7)

1., for a mixed multi-attribute groups Decision Method for rack reconfiguration scheme assessment, it is characterized in that, comprise the steps:

S1: network and the equipment state of determining subsystems in electric system subregion, dynamically determines current recovery target, uses dijkstra's algorithm to obtain the Weighted Shortest Path Problem footpath of having recovered network and having recovered between target;

S2: emulate according to the network parameter of subsystem and state, determine the numerical value of qualitative attribute, if every attribute value is all verified by power system security, then the program meets the demands; If a certain attribute does not meet safe stability of power system requirement, then network parameter and state are adjusted, if the scheme after adjustment still can not meet the demands, then give up the program;

S3: group decision-making negotiation is carried out to all schemes remained, provides the method for expressing of quantitative attributes describing mode and uncertain attribute;

S4: the decision maker participating in group decision adopts pairwise comparison matrix or Fuzzy Complementary Judgment Matrices to be described the relative importance between each attribute respectively, the description then adopting the method for mathematical programming to assemble each decision maker generates the subjective weight of attribute;

S5: decision maker consults to set up unique decision matrix, carries out standardization to decision matrix;

S6: according to the decision matrix after standardization, based on the objective weight of maximum deviation model determination attribute; Based on the combining weights of minimum relative information entropy model computation attribute;

S7: according to combinations of attributes weight and decision matrix, carries out assessment sequence to optional program, optimum scheme comparison and alternative scheme.

2. a kind of mixed multi-attribute groups Decision Method for the assessment of rack reconfiguration scheme as claimed in claim 1, it is characterized in that, in described step S2, the out-of-limit adjustment means of attribute comprise:

For superpotential and reactive balance problem, adopting the switching of reactive power compensator, adjustment generator terminal voltage, the mode of Load adjustment or adjustment load tap changer position adjusts, reducing overvoltage level by reducing system reference voltage level;

The transient voltage dip that subsidiary engine starts is adjusted by terminal voltage with raising by the boot sequence of adjustment subsidiary engine;

Line transmission capacity is out-of-limit to be adjusted by adjustment generator output and load;

The frequency of load restoration is fallen and is adjusted by Load adjustment amount of recovery;

If need the amount of projects of adjustment more, the mode adopting the circuit weight changed in route searching to arrange adjusts.

3. as claimed in claim 1 a kind of for rack reconfiguration scheme assessment mixed multi-attribute groups Decision Method, it is characterized in that, in described step S4, the mathematical model of subjective weight is as follows:

$\min F={[\underset{l=1}{\overset{k_1}{\mathrm{\Σ}}}\left({\mathrm{\α}}_{\left(l\right)}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\left|{{\mathrm{\ϵ}}_i}^{\left(l\right)}\right|}^p\right)+\underset{m=1}{\overset{k_2}{\mathrm{\Σ}}}\left({\mathrm{\β}}_{\left(m\right)}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{\left|{r_j}^{\left(m\right)}\right|}^p\right)]}^{\frac{1}{p}}$

$s.t.\begin{array}{}\end{array}\left\{\begin{array}{cc}(B^{\left(l\right)}-\mathrm{nI})W-E^{\left(l\right)}=0,& l=\mathrm{1,2},...,k_1\\ (H^{\left(m\right)}-\mathrm{nI})W-R^{\left(m\right)}=0,& m=\mathrm{1,2},...,k_2\\ e^{T}W=1& \\ W\≥0& \end{array}\right.$

Wherein, k ₁individual decision maker uses pairwise comparison matrix to carry out the determination of relative importance, k ₂individual decision maker uses Fuzzy Complementary Judgment Matrices to carry out the determination of relative importance; N represents attribute number; I represents n rank unit matrix; W is the subjective weight after assembling; B ^(l)the pairwise comparison matrix adopting AHP 1-9 scale to determine, E ^(l)=(ε ₁ ^(l), ε ₂ ^(l)..., ε _n ^(l)) represent the error vector of pairwise comparison matrix; H ^(m)fuzzy Complementary Judgment Matrices P ^(m)through processing the matrix obtained, wherein h _ij=p _ij/ p _ji, Fuzzy Complementary Judgment Matrices P ^(m)determine according to AHP 0.1-0.9 scale; R ^(m)=(r ₁ ^(m), r ₂ ^(m)..., r _n ^(m)) represent the error vector of Fuzzy Complementary Judgment Matrices; E=(1 ..., 1) ^t, expression be subjective weight sum be the constraint of 1; α _(l)>=0, l=1,2 ..., k ₁, β _(m)>=0, m=1,2 ..., k ₂, α _(l), β _(m)represent the weight of different decision maker in group decision respectively, parameter p is positive integer, and usual value is p=1,2 ..., ∞.

4. as claimed in claim 1 a kind of for rack reconfiguration scheme assessment mixed multi-attribute groups Decision Method, it is characterized in that, in the decision matrix of described step S5, comprise qualitative attribute, quantitative attributes and uncertain attribute; The data type of decision matrix comprises one or more in perfect number, interval number, fuzzy number, linguistic variable, Vague collection and cloud model, and the difference between attribute uses Euclidean distance to measure.

5. a kind of mixed multi-attribute groups Decision Method for the assessment of rack reconfiguration scheme as described in claim 1 or 4, is characterized in that, in described step S5 to the method that decision matrix carries out standardization be:

To perfect number:

$f_{\mathrm{ij}}=\left\{\begin{array}{c}\frac{\underset{i}{\min }{a}_{\mathrm{ij}}}{a_{\mathrm{ij}}},a_{\mathrm{ij}}\∈J_1\\ \frac{a_{\mathrm{ij}}}{\underset{i}{\max }{a}_{\mathrm{ij}}},a_{\mathrm{ij}}\∈J_2\end{array}\right.$

To interval number:

$f_{\mathrm{ij}}=\left\{\begin{array}{c}[\frac{\underset{i}{\min }{a_{\mathrm{ij}}}^L}{{a_{\mathrm{ij}}}^U},\frac{\underset{i}{\min }{a_{\mathrm{ij}}}^L}{{a_{\mathrm{ij}}}^L}],a_{\mathrm{ij}}\∈J_1\\ [\frac{{a_{\mathrm{ij}}}^L}{\underset{i}{\max {a_{\mathrm{ij}}}^U}},\frac{{a_{\mathrm{ij}}}^U}{\underset{i}{\max {a_{\mathrm{ij}}}^U}}],a_{\mathrm{ij}}\∈J_2\end{array}\right.$

To fuzzy number:

$f_{\mathrm{ij}}=\left\{\begin{array}{c}[\frac{\underset{i}{\min }{a_{\mathrm{ij}}}^L}{{a_{\mathrm{ij}}}^U},\frac{\underset{i}{\min }{a_{\mathrm{ij}}}^L}{{a_{\mathrm{ij}}}^M},\frac{\underset{i}{\min }{a_{\mathrm{ij}}}^L}{{a_{\mathrm{ij}}}^L}],a_{\mathrm{ij}}\∈J_1\\ [\frac{{a_{\mathrm{ij}}}^L}{\underset{i}{\max }{a_{\mathrm{ij}}}^U},\frac{{a_{\mathrm{ij}}}^M}{\underset{i}{\max }{a_{\mathrm{ij}}}^U},\frac{{a_{\mathrm{ij}}}^U}{\underset{i}{\max }{a_{\mathrm{ij}}}^U}],a_{\mathrm{ij}}\∈J_2\end{array}\right.$

Wherein perfect number is expressed as a _ij; Interval number is expressed as [a _ij ^l, a _ij ^u]; Fuzzy number is expressed as [a _ij ^l, a _ij ^m, a _ij ^u]; J ₁, J ₂cost type property set and profit evaluation model property set respectively;

Variable Vague collection, cloud model, linguistic variable being converted into other form processes.

6. a kind of mixed multi-attribute groups Decision Method for the assessment of rack reconfiguration scheme as claimed in claim 1, it is characterized in that, in described step S6, based on maximum deviation model, calculating the method for objective weight is:

$\max J=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}\underset{k=1}{\overset{m}{\mathrm{\Σ}}}{w_j}^{\′\′}D(f_{\mathrm{ij}},f_{\mathrm{kj}})$

$s.t.\left\{\begin{array}{cc}\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{w_j}^{\′\′2}=1& j=\mathrm{1,2},...,n\\ {w_j}^{\′\′}\≥0& \end{array}\right.$

Wherein, optional program number is m, and attribute number is n; w _j" represent the objective weight of attribute; Euclidean distance between D () function representation two assessed values; F=(f _ij) _{m × n}represent standardization decision matrix; J is the Weighted distance between different schemes attribute;

Based on minimum relative information entropy model, the method for calculation combination weight is:

$\min J=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j\left[\ln \frac{w_j}{{w_j}^{\′}}\right]+\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j\left[\ln \frac{w_j}{{w_j}^{\′\′}}\right]$

$s.t.\left\{\begin{array}{cc}\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{w}_j=1& \\ w_j\≥0& j=\mathrm{1,2},...,n\end{array}\right.$

Wherein w _j', w _j", j=1,2 ..., n represents subjective weight and the objective weight of evaluation index respectively; w _j, j=1,2 ..., n is the combining weights finally obtained.

7. as claimed in claim 1 a kind of for rack reconfiguration scheme assessment mixed multi-attribute groups Decision Method, it is characterized in that, in described step S7, assessment algorithm is VIKOR algorithm, and its step is as follows:

If standardization decision matrix F=is (f _ij) _{m × n}, total m decision scheme, A={a _i, 1≤i≤m; Decision scheme has n attribute, G={G _j, 1≤j≤n; Scheme a _ito attribute G _jstandardised assessment value be f _ij;

If G _j{ j ∈ N ₁=(1,2 ..., h ₁) be Real-valued index set, G _j{ j ∈ N ₂=(h ₁+ 1, h ₁+ 2 ..., h ₂) be interval type index set, G _j{ j ∈ N ₃=(h ₂+ 1, h ₂+ 2 ..., h ₃) be fuzzy number index set, G _j{ j ∈ N ₄=(h ₃+ 1, h ₃+ 2 ..., h ₄) be Vague collection index set, G _j{ j ∈ N ₅=(h ₄+ 1, h ₄+ 2 ..., h ₅) be linguistic variable index set, G _j{ j ∈ N ₆=(h ₅+ 1, h ₅+ 2 ..., h ₆) be cloud model index set;

1) preparation of data: the data type comprising attribute in standardization decision matrix, the combining weights of attribute and decision matrix;

2) the positive ideal solution F of all optional programs is determined ⁺with minus ideal result F ^-, so-called positive ideal solution is made up of the optimum evaluation value of each index, and minus ideal result is the most bad assessed value composition;

3) the comprehensive assessment optimum solution S of numerical procedure _iinferior solution R most with comprehensive assessment _i;

4) the advantage ratio Q of numerical procedure generation _i;

5) sequence is determined;

According to S _i, R _i, Q _inumerical value sort respectively according to order from small to large, obtain three collating sequences, it is generally acknowledged that in ascending order arrangement, the scheme that numerical value is less has and preferably sorts;

6) compromise proposal or compromise proposal disaggregation is determined

If meet following two conditions, the scheme a that Q value is minimum ⁽¹⁾be exactly that final compromise sequence is separated:

Condition 1: wherein a ⁽²⁾for arranging deputy optional program by Q value ascending order;

Condition 2:a ⁽¹⁾come the scheme of foremost in the sequence according to S value or the arrangement of R value ascending order;

If one of them condition can not meet, then the compromise solution obtained is not unique, but obtains compromise proposal disaggregation; If wherein condition 1 does not meet, then scheme a ⁽¹⁾, a ⁽²⁾..., a ^(r)be its compromise solution, wherein a ^(r)meet if condition 2 does not meet, scheme a ⁽¹⁾and a ⁽²⁾for compromise solution;

If obtain multiple compromise solution, traffic control personnel are according to any recovery scheme in actual conditions selection compromise solution as optimal case, and other schemes are as subsequent use; Or the comprehensive selection of recovery scheme and alternative scheme is carried out according to S, R, Q tri-collating sequences.