CN103440398A - Pattern recognition-based power grid branch importance estimation method - Google Patents

Abstract

The invention discloses a pattern recognition-based power grid branch importance estimation method. The method comprises the following steps of: based on three risk indicators for constructing branch return, performing self-organizing clustering on a three-dimensional risk by using an ISODATA (Iterative Self-organizing Data Analysis Techniques Algorithm); determining the safety level of each branch; analyzing a three-dimensional risk vector of each branch by using a PCA (Principal Component Analysis) method; determining a principal component of all branch risks and taking the first principal component as reference for comprehensive risk evaluation indicator design so as to realize design of a comprehensive evaluation indicator after data dimension reduction, importance sorting of the branches and the importance grading of the branches. According to the method, the importance of each branch is evaluated by using the branch return risk; compared with the evaluation only from the point of topological structure, the method has the advantage of being more persuasive; the ISODATA clustering algorithm-based branch importance grading method and the PCA-based branch importance sorting are provided, so that a theoretical basis is provided for search and construction of a core backbone network frame.

Description

A kind of grid branch importance appraisal procedure of Schema-based identification

Technical field

The invention belongs to the differentiation planning technology field of electric system, particularly a kind of grid branch importance appraisal procedure of Schema-based identification.

Background technology

In recent years, extreme frequent natural calamity occurs both at home and abroad, may cause electric system to occur from part to large-area power outage, has had a strong impact on the safe operation of electrical network.Differentiation is planned the ability that guarantees the important load continued power when constructed core backbone frame possesses reply disaster and catastrophic failure.The key that builds the core backbone frame is the importance assessment of branch road.Traditional method mainly concentrates on the analysis to topological structure of electric.Because the analytical approach of topological structure is not considered the probability of topologies change, and seldom relate to Operation of Electric Systems characteristic and himself constraint.Yet methods of risk assessment has considered the probability of accident generation and the consequence of generation, and branch road is carried out to importance assessment overall scientific more.

In the Study of Risk Evaluation Analysis for Power System field, the integrated approach of multi-risk System index generally adopts the appraisal procedure of weighted type, and the key of various methods of weighting is determining of each index weights." the power system security risk assessment and fragility " that Sun Fei delivers at PhD dissertation 2011 utilizes 3 scaling laws of analytical hierarchy process and 9 scaling laws to be combined expert survey again and realizes the distribution of index weights; Liu Xindong etc. are at Electric Power Automation Equipment 2009,29(2): " the Transient Security for Power Systems risk assessment based on Risk Theory and fuzzy reasoning " that 15-20 delivers adopts the integrated approach of norm weighting to carry out the comprehensive of security risk index.It is larger that the weight coefficient of these risk integrated approachs is affected by subjective factor.At present deep not enough for the research of branch road safety classification problem, Wang Ning etc. are at east china electric power 2008,36(3): " low voltage security forewarning for power systems based on risk " that 66-69 delivers is divided into 5 grades by voltage security, and principle of grading is directly according to the Risk Calculation desired value, to come the artificial classification of risks interval of evenly setting every one-level.Obviously, this stage division lacks necessary Theoretical Proof, and form is simple, and subjectivity is strong.

Summary of the invention

The objective of the invention is to explore a kind of grid branch importance appraisal procedure of Schema-based identification.The method is moved back on the basis of fortune risk assessment index in the structure grid branch, the three-dimensional risk of using the ISODATA algorithm to move back fortune to branch road is carried out self-organizing clustering, determine the importance rate of each branch road, then utilize the PCA principal component analytical method to determine the major component of all branch road risks, realize importance ranking and the importance degree classification thereof of branch road after Data Dimensionality Reduction, for the core backbone framework, build theoretical foundation is provided.

For solving the problems of the technologies described above, the present invention adopts following technical scheme:

A kind of grid branch importance appraisal procedure of Schema-based identification, the method includes the steps of:

Step 1, structure grid branch importance evaluation index;

The importance of branch road judges the consequence that in whole electric system, other all branch roads or node cause after can moving back fortune according to single branch road, considers probability and corresponding consequence thereof that accident occurs;

Branch road moves back the fortune risk:

It is the probability and the product that moves back the consequence produced after fortune that branch road moves back fortune that the branch road of definition electric system moves back the fortune risk, that is:

$\mathrm{Risk}\left(Y\right|E_i)=P\left(E_i\right)\×\underset{j\≠i}{\mathrm{\Σ}}\∫f\left(Y\right|E_{i,}{L}_j)\×\mathrm{Eev}\left(Y\right)\mathrm{dY}---\left(1\right)$

In formula, E _ithe i bar branch road that refers to power grid accident-electrical network moves back fortune, P (E _i) refer to accident E _ithe probability occurred, generally obey Poisson distribution; L _jit is j bar branch road; F (Y|E _i, L _j) be after branch road i moves back fortune, the probability distribution of branch road j in specific run state Y in system; Sev (Y) has described the severity of accident when specific run state Y.Specific run state Y comprises Branch Power Flow, node voltage and node load.∫ f (Y|E _i, L _j) * Sev (Y) dY refers to accident E _ithe corresponding consequence after generation, branch road j produced.

therefore self-explanatory characters are E _iafter generation to the consequence summation of other all branch roads, Risk (Y|E _i) refer to that the branch road of electric system moves back the fortune risk.

Branch road moves back the fortune risk indicator:

Move back after fortune the difference of other all branch roads or consequence that node causes in whole electric system according to branch road, branch road is moved back to the fortune risk and be divided into overload risk, low-voltage risk and lose three kinds of load risks.

(1) overload risk

After the overload risk is power system accident, cause other not the active power of fault branch surpass the possibility of its ratings and the combination of the order of severity, that is:

$\mathrm{Risk}\left(P_L\right|E_i)=P\left(E_i\right)\underset{j\≠i}{\mathrm{\Σ}}\∫f\left(P_L\right|E_i,L_j)\×\mathrm{Sve}\left(P_L\right){\mathrm{dP}}_L---\left(2\right)$

Wherein, i=1,2 ..., n, j=1,2 ..., n, n is the way in system, f (P _l| E _i, L _j) be after branch road i moves back fortune, the branch power relative value P of branch road j in system _lprobability distribution; Sev (P _l) to have described the branch power relative value be P _lthe time accident severity.∫ f (P _l| E _i, L _j) * Sev (P _l) dP _lrefer to accident E _ithe overladen consequence after generation, branch road j produced.

therefore self-explanatory characters are E _iafter generation to the overload consequence summation of other all branch roads, Risk (P _l| E _i) refer to accident E _ioverload risk after generation.Overload severity Sev (P _l) depending on that the trend of other all branch roads after accident distributes, the severity function is suc as formula shown in (3), and curve is as shown in Figure 2.

$\mathrm{Sev}\left(P_L\right)\left\{\begin{array}{cc}1& P_L>1\\ 10P_L-9& 0.9<P_L\&le;1\\ 0& P_L\&le;0.9\end{array}\right.---\left(3\right)$

In formula, P is branch power, P _efor the rated power of this branch road, P _l=P/P _efor this branch power relative value.

(2) low-voltage risk

The low-voltage risk is after power system accident, to cause in system node voltage lower than the combination of possibility and the seriousness of ratings, that is:

$\mathrm{Risk}\left(V_B\right|E_i)=P\left(E_i\right)\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\&Integral;f\left(V_B\right|E_i,L_j)\&times;\mathrm{Eve}\left(V_B\right){\mathrm{dV}}_B---\left(4\right)$

Wherein, m is node number in system, f (V _b| E _i, L _j) be after branch road i moves back fortune, the node voltage relative value V of node j in system _bprobability distribution; Sev (V _b) to have described the node voltage relative value be V _bthe time accident severity.∫ f (V _b| E _i, L _j) * Sev (V _b) dV _brefer to accident E _ithe consequence of the low-voltage after generation, node j produced.

therefore self-explanatory characters are E _iafter generation to the low-voltage consequence summation of all nodes, Risk (V _b| E _i) refer to accident E _ilow-voltage risk after generation.Low-voltage severity Sev (V _b) depending on the voltage of accident posterior nodal point, the severity function is suc as formula shown in (5), and curve is as shown in Figure 3.

$\mathrm{Sev}\left(V_B\right)=\left\{\begin{array}{cc}1& V_B\&GreaterEqual;0.9\\ 10-10V_B& 0.9<V_B\&le;1\\ 0& V_B>1\end{array}\right.---\left(5\right)$

In formula, V is node voltage, V _efor the rated voltage of this node, V _b=V/V _efor this node voltage relative value.

(3) lose the load risk

Losing the load risk is the combination that after accident, the system loading node loses possibility and the order of severity of load, that is:

$\mathrm{Risk}\left(P_q\right|E_i)=P\left(E_i\right)\underset{i=1}{\overset{m_d}{\mathrm{\&Sigma;}}}{P}_{\mathrm{qi}\&times;\mathrm{Sev}\left(P_q\right)---\left(6\right)}$

In formula, m _dfor load bus number, P _qifor accident E _irear i the load that load bus loses, P _qfor the load that system loses, Sev (P _q) for losing load P in system _qseverity, Risk (P _q| E _i) be accident E _imistake load risk after generation.

Wherein, lose the load severity and depend on the ratio that loses load, its severity function is suc as formula shown in (7).

$\mathrm{Sev}\left(P_q\right)=\left\{\begin{array}{cc}1& 0.3<P_q/P_{\mathrm{fh}}\&le;1\\ \frac{10}{3}{P}_q/{\mathrm{<figure-callout\; id="0"\; label="P\; fh"\; filenames="130710170101.png,130710170103.png"\; state="\{\{state\}\}">P</figure-callout>}}_{\mathrm{fh}}& 0<P_q/P_{\mathrm{fh}}\&le;0.3\end{array}\right.---\left(7\right)$

In formula, P _fhfor the original loads of load bus in system, this numerical value is known quantity, and each load bus is that original loads is different, P _qmistake load for load bus in system.

Step 2, build three-dimensional risk vector

By the overload risk, low-voltage risk and the three-dimensional risk vector x of mistake load the risk forms _i, be expressed as:

x _i=(Risk(P _L|E _i),Risk(V _B|E _i),Risk(P _q|E _i)) (8)

Step 3, based on the ISODATA clustering algorithm, carry out the classification of branch road importance

The ISODATA cluster, i.e. iteration self-organization data analysis algorithm, be a kind of mode identification method of non-supervisory Dynamic Clustering Algorithm.ISODATA adopts Euclidean distance to analyze the similarity of data itself, and its core concept is that Euclidean distance is less, and similarity is larger, and the data that similarity is high flock together automatically.

The i.e. setting accident of the thought collection of branch road importance classification is that single branch road moves back fortune, calculate respectively this overload risk of moving back the transport line road, low-voltage risk and lose the load risk, three value-at-risks of every branch road are expressed as to a three-dimensional risk vector, adopt the ISODATA clustering algorithm by the three-dimensional risk vector of all branch roads according to data similarity cluster, obtain cluster centre and the branch number of every one-level, realize every preliminary automatic classification of branch road importance rate; Size judgement branch road importance rate with cluster centre and initial point Euclidean distance, cluster centre is far away apart from initial point, the three-dimensional risk point that this rank comprises is far away apart from initial point, so the risk of these risk points is just larger, and under it, importance information is just more important; The initial cluster center set due to this patent is random, and branch road importance classification results is also incomplete same, but occurs that the inconsistent place of classification results is all at boundary.

The basic step of ISODATA clustering algorithm is as follows:

1) parameters: sample x to be sorted _i; The cluster centre number K of expection; Initial cluster centre number N _c; Minimum number of samples θ in each Clustering Domain _n; The standard deviation θ of sample range distribution in Clustering Domain _s; Minor increment θ between two cluster centres _c; The number of times I of the interative computation that the judgement circulation stops _p; Distance B between two cluster centres _ij.

2) choose at random N _cindividual sample is as the center of initial clustering;

3) by sample x _ibe assigned to nearest cluster S _j.Rule is: if D _j=min (|| i-C _j||), i=1,2 ..., p, j=1,2 ..., c, by x _ibe grouped into cluster S _j.Wherein, C _jbe j cluster centre, D _jfor sample x _ito the distance of j cluster centre, this distance is the shortest.

4) calculate the center of each cluster:

$Z_i=\frac{1}{N_i}\underset{x\&Element;S_j}{\mathrm{\&Sigma;}}{x}_j---\left(9\right)$

By new central value Z _ithe center C that is decided to be cluster _i=Z _i, x _jbe j three-dimensional risk vector, j=1 wherein, 2 ..., S _j, N _ifor cluster S _jthe class number.

5) division.If the number of current cluster is less than the clusters number K of expection, start to carry out the cluster division.

6) merge.When the centre distance of two clusters is less than the minor increment θ at the two center _cthe time, two Cluster mergings are a new cluster.If the distance between whole cluster centres

just start to merge.New cluster centre is:

$C_n=\frac{1}{N_i+N_j}[N_i{C}_i+N_j{C}_j]---\left(10\right)$

Wherein, C _iand C _jbe respectively the cluster centre of i class and j class, N _iand N _jbe respectively cluster S _iand S _jthe class number.

7) if iterations reaches maximum iteration time I _p, or the process convergence, iterative process finishes, otherwise I _p=I _p+ 1, get back to step 3.

Step 4, based on the PCA method, carry out the branch road importance ranking

Principal component analysis (PCA) (PCA) method is a kind of statistical analysis technique that area of pattern recognition is found the multidimensional data key property that is widely used in, its fundamental purpose is that high dimensional data is projected to lower dimensional space, parse principal ingredient from multidimensional data, disclose the multidimensional data effective information, simplify the Analysis of Complex problem.

The main thought of PCA is that former N dimensional vector is used to linear transformation, the new N column vector that obtains sorting from high to low according to importance.Choose the M(M<N of importance maximum in the new column vector obtained) the dimension subvector, as the major component of former column vector.

Note x ₁..., x _pfor p component of original column vector, establish the component ξ listd after conversion _i, i=1,2 ..., p, be the linear combination of former column vector subcomponent, the mould of setting linear combination coefficient is 1,

${\mathrm{\&alpha;}}_i^T{\mathrm{\&alpha;}}_i=1---\left(11\right)$

α _ifor the linear combination coefficient vector, it is column vector.This p α _iconstitutive characteristic transformation matrix A.Each component α of optimum orthogonal transformation A _imake corresponding ξ _ivariance reach extreme value, data will more discrete dividing, similarity is lower, has also just represented more information.Require to form each column vector pairwise orthogonal of A, two pairwise uncorrelateds between the new component that guarantees to obtain simultaneously.In addition, if the variance of certain dimension component is larger, this component is just more important, has more information.

Using three-dimensional risk vector as source data, utilize the PCA method to carry out dimensionality reduction to three-dimensional risk vector, retain the main information in source data, obtain the maximum principal direction of all three-dimensional risk data, then by three-dimensional risk vector projection to just can obtain on the principal direction axis can clear resolution different classes of (being ISODATA automatic classification result) one-dimensional data, and then carry out the calculating of integrated risk index with reference to PCA dimensionality reduction result.First principal component ξ ₁be integrated risk, computing formula is suc as formula shown in (12).

${\mathrm{\&xi;}}_1={\mathrm{\&alpha;}}_1^{T}x---\left(12\right)$

Wherein,

for the weight vectors of three-dimensional risk indicator, mean the normalization importance proportion of three-dimensional risk, the three-dimensional risk vector that x is branch road.ξ ₁for integrated risk, i.e. the distance that project to initial point of space risk point on the principal direction axle.

Obtain the integrated risk of all branch roads according to formula (12), according to the integrated risk value, carry out the branch road importance ranking, integrated risk is larger, and the impact on whole electric system after this branch trouble is larger, so this branch road is more important.

The judgement of step 5, branch road importance ranking and classification adjustment

According to branch road importance ranking result, the up-and-down boundary branch road of mark spatial scalability projection, if there is not intersection in the projection of adjacent two important rank branch roads, can set the risk average that the classification point of this two-stage is the border branch road of this two-stage; If there is the intersection in allowed band, with important level, high border branch road risk is as the criterion, drop on the border branch road and all be automatically made to lay particular stress on away from the subpoint of initial point direction and want grade, the classification point according to the risk average of new two-stage border branch road as this two-stage.Set classification and intersect number percent for intersecting a way N _jcaccount for all way N _znumber percent, shown in (13).

$\mathrm{\&psi;}=\frac{N_{\mathrm{jc}}}{N_z}\&times;100\%---\left(13\right)$

When classification intersects number percent while meeting Ψ<10%, can think, adopt first principal component can comprehensively embody the three dimensions rating information of p index, i.e. classification intersects in allowed band.If do not meet Ψ<10%, jump to step 3, Reparametrization.

The present invention has the following advantages:

1, the present invention adopts branch road to move back the importance that the fortune risk is assessed branch road, more can react the importance of branch road from the angle of electrical network characteristic and self constraint, and the list of comparing is assessed and had more cogency from the angle of topological structure.

2, the present invention proposes the branch road importance stage division based on the ISODATA clustering algorithm, built the risk indicator of vector form, the three-dimensional risk vector gathering that similarity is higher is a class, and this method is compared existing stage division and had more theoretical foundation.

3, the present invention proposes the branch road importance ranking based on PCA, fortune Rate of aggregative risk function is moved back in design, using the risky vectorial first principal component of institute as the integrated risk value, one dimension is fallen into in the three-dimensional risk vector of electrical network, can reduce comparatively accurately three-dimensional classification results simultaneously.

4, the present invention has carried out sequence and classification to branch road importance, for the search of key rack provides precondition.

The accompanying drawing explanation

The schematic flow sheet of the grid branch importance assessment that Fig. 1 is Schema-based identification;

The grid branch importance classification figure that Fig. 2 is Schema-based identification;

Fig. 3 is the branch road importance ranking figure based on principal component analysis (PCA).

Embodiment

The grid branch importance appraisal procedure of a kind of Schema-based identification, carry out prominence score level and sequence to 46 branch roads of IEEE39 node system.The method comprises the following step:

A, structure grid branch importance evaluation index; In the present embodiment, the fluctuation of assumed load allocation factor is 5%, and usings this variation as system operational parameters.If contingency set is each branch road in system, disconnect successively, the year of these branch roads is cut-off rate λ _yall 0.3.

A1, calculating branch road move back the probability of fortune;

A2, according to three kinds of risk severity functions, ask for and consider that single branch road that operational factor changes moves back the order of severity of three kinds of risks after fortune;

A3, utilization move back based on branch road three kinds of value-at-risks that the fortune methods of risk assessment calculates each branch road, and its result is as shown in table 1, and accompanying drawing 2 has provided three kinds of Risk Calculation results of all branch roads.

A4, three risk indicators that branch road is moved back to fortune form three-dimensional risk vector, obtain the three-dimensional risk vector of 46 branch roads of 39 node systems.

Table 1

B, adopt the ISODATA algorithm, the three-dimensional risk of 46 branch roads carried out to the importance classification, be divided into one-level, secondary and three grades, respectively with * ,+and mean, classification results in Table 1 and Fig. 2 shown in.

C, employing PCA methods analyst branch road move back the three-dimensional risk vector of fortune, obtain the first principal component of all three-dimensional risk vectors, all three dimensions risk points are done to projection to the principal direction axle corresponding to first principal component, as shown in Figure 2.The normalized weight coefficient vector of three-dimensional risk is:

,

D, basis

with the three-dimensional risk vector of all branch roads, can calculate the integrated risk value after all branch road risk dimensionality reductions.Branch road integrated risk value is listed in table 1.

E, the branch road integrated risk is sorted, obtained branch road importance ranking result, as shown in Fig. 3 and table 2.

F, preliminary classification and importance ranking result are analyzed, the branch road that obtains intersecting is L30, and the intersection branch road number percent Ψ of existence=2.2<10%, in allowed band, verified validity and the feasibility of this risk integrated approach.Table 3 has provided the value-at-risk of border branch road and the branch road importance classification point of 39 node systems.

Table 2

Table 3

Claims (2)

1. the grid branch importance appraisal procedure of Schema-based identification is characterized in that: comprises the following steps,

Step 1, structure grid branch importance evaluation index;

The importance of branch road judges the consequence that in whole electric system, other all branch roads or node cause after moving back fortune according to single branch road, considers probability and corresponding consequence thereof that accident occurs; The branch road of definition electric system moves back fortune risk Risk (Y|E _i) for branch road moves back the probability of fortune and the product of the consequence of moving back the rear generation of fortune, that is:

$\mathrm{Risk}\left(Y\right|E_i)=P\left(E_i\right)\&times;\underset{j\&NotEqual;i}{\mathrm{\&Sigma;}}\&Integral;f\left(Y\right|E_{i,}{L}_j)\&times;\mathrm{Eev}\left(Y\right)\mathrm{dY}$

Wherein, E _ithe i bar branch road that refers to electrical network moves back fortune, P (E _i) refer to accident E _ithe probability occurred, obey Poisson distribution; L _jit is j bar branch road; F (Y|E _i, L _j) be after branch road i moves back fortune, the probability distribution of branch road j in specific run state Y in system, the severity of accident when Sev (Y) is illustrated in specific run state Y, specific run state Y comprises Branch Power Flow, node voltage and node load, ∫ f (Y|E _i, L _j) * Sev (Y) dY refers to accident E _ithe corresponding consequence after generation, branch road j produced,

therefore self-explanatory characters are E _iafter generation to the consequence summation of other all branch roads;

According to branch road, move back after fortune the difference of other all branch roads or consequence that node causes in whole electric system, branch road moves back the fortune risk and comprises overload risk, low-voltage risk, loses the load risk, and circular is as follows:

Overload risk Risk (P _l| E _i) calculating:

$\mathrm{Risk}\left(P_L\right|E_i)=P\left(E_i\right)\underset{j\&NotEqual;i}{\mathrm{\&Sigma;}}\&Integral;f\left(P_L\right|E_i,L_j)\&times;\mathrm{Sve}\left(P_L\right){\mathrm{dP}}_L$

Wherein, i=1,2 ..., n, j=1,2 ..., n, n is the way in system, f (P _l| E _i, L _j) be after branch road i moves back fortune, the branch power relative value P of branch road j in system _lprobability distribution; Sev (P _l) to have described the branch power relative value be P _lthe time accident severity; ∫ f (P _l| E _i, L _j) * Sev (P _l) dP _lrefer to accident E _ithe overladen consequence after generation, branch road j produced;

therefore self-explanatory characters are E _iafter generation to the overload consequence summation of other all branch roads; Overload severity Sev (P _l) depend on that the trend of other all branch roads after accident distributes, and specifically is expressed as:

$\mathrm{Sev}\left(P_L\right)\left\{\begin{array}{cc}1& P_L>1\\ 10P_L-9& 0.9<P_L\&le;1\\ 0& P_L\&le;0.9\end{array}\right.$

In formula, P _l=P/P _efor this branch power relative value, P is branch power, P _erated power for this branch road;

Low-voltage risk Risk (V _b| E _i) calculating:

$\mathrm{Risk}\left(V_B\right|E_i)=P\left(E_i\right)\underset{j=1}{\overset{m}{\mathrm{\&Sigma;}}}\&Integral;f\left(V_B\right|E_i,L_j)\&times;\mathrm{Eve}\left(V_B\right){\mathrm{dV}}_B$

Wherein, m is node number in system, f (V _b| E _i, L _j) be after branch road i moves back fortune, the node voltage relative value V of node j in system _bprobability distribution; Sev (V _b) to have described the node voltage relative value be V _bthe time accident severity, ∫ f (V _b| E _i, L _j) * Sev (V _b) dV _brefer to accident E _ithe consequence of the low-voltage after generation, node j produced,

therefore self-explanatory characters are E _iafter generation to the low-voltage consequence summation of all nodes, low-voltage severity Sev (V _b) depend on and specifically be expressed as the voltage of accident posterior nodal point:

$\mathrm{Sev}\left(V_B\right)=\left\{\begin{array}{cc}1& V_B\&GreaterEqual;0.9\\ 10-10V_B& 0.9<V_B\&le;1\\ 0& V_B>1\end{array}\right.$

In formula, V _b=V/V _efor this node voltage relative value, V is node voltage, V _erated voltage for this node;

Lose load risk Risk (P _q| E _i) calculating:

$\mathrm{Risk}\left(P_q\right|E_i)=P\left(E_i\right)\underset{i=1}{\overset{m_d}{\mathrm{\&Sigma;}}}{P}_{\mathrm{qi}\&times;\mathrm{Sev}\left(P_q\right)}$

In formula, m _dfor load bus number, P _qifor accident E _irear i the load that load bus loses, P _qfor the load that system loses, Sev (P _q) for losing load P in system _qseverity, specifically be expressed as:

$\mathrm{Sev}\left(P_q\right)=\left\{\begin{array}{cc}1& 0.3<P_q/P_{\mathrm{fh}}\&le;1\\ \frac{10}{3}{P}_q/P_{\mathrm{fh}}& 0<P_q/P_{\mathrm{fh}}\&le;0.3\end{array}\right.$

In formula, P _fhfor the original loads of load bus in system, P _qmistake load for load bus in system.

Step 2, by overload risk, low-voltage risk with lose the three-dimensional risk vector x of load the risk forms _i, specifically be expressed as:

x _i=(Risk(P _L|E _i),Risk(V _B|E _i),Risk(P _q|E _i))

X _iit is the three-dimensional risk vector of i bar branch road;

Step 3, based on the ISODATA clustering algorithm, carry out the classification of branch road importance;

The process of carrying out the classification of branch road importance based on the ISODATA clustering algorithm is specially: adopt the ISODATA clustering algorithm by the three-dimensional risk vector of all branch roads according to data similarity cluster, obtain cluster centre and the branch number of every one-level, realize every preliminary automatic classification of branch road importance rate; Size judgement branch road importance rate with cluster centre and initial point Euclidean distance, cluster centre is far away apart from initial point, the three-dimensional risk point that this rank comprises is far away apart from initial point, so the risk of these risk points is just larger, and under it, importance information is just more important;

Step 4, based on the PCA method, carry out the branch road importance ranking; Detailed process is as follows:

Using three-dimensional risk vector as source data, utilize the PCA method to carry out dimensionality reduction to three-dimensional risk vector, retain the main information in source data, obtain the maximum principal direction of all three-dimensional risk data, then by three-dimensional risk vector projection to can the different classes of one-dimensional data of clear resolution to obtain on the principal direction axis, and then carry out the calculating of integrated risk index with reference to PCA dimensionality reduction result; First principal component ξ ₁be integrated risk, computing formula is as follows;

${\mathrm{\&xi;}}_1={\mathrm{\&alpha;}}_1^{T}x$

Wherein, for the weight vectors of three-dimensional risk indicator, mean the normalization importance proportion of three-dimensional risk, the three-dimensional risk vector that x is branch road, ξ ₁for integrated risk, i.e. the distance that project to initial point of space risk point on the principal direction axle;

Obtain the integrated risk of all branch roads according to above formula, according to the integrated risk value, carry out the branch road importance ranking, integrated risk is larger, and the impact on whole electric system after this branch trouble is larger, so this branch road is more important;

Step 5, judgement branch road importance ranking and classification adjustment;

According to branch road importance ranking result, the up-and-down boundary branch road of mark spatial scalability projection, if there is not intersection in the projection of adjacent two important rank branch roads, can set the risk average that the classification point of this two-stage is the border branch road of this two-stage; If there is the intersection in allowed band, with important level, high border branch road risk is as the criterion, drop on the border branch road and all be automatically made to lay particular stress on away from the subpoint of initial point direction and want grade, the classification point according to the risk average of new two-stage border branch road as this two-stage.Set classification and intersect number percent for intersecting a way N _jcaccount for all way N _znumber percent, as follows,

$\mathrm{\&psi;}=\frac{N_{\mathrm{jc}}}{N_z}\&times;100\%$

When classification intersects number percent while meeting Ψ<10%, adopt first principal component can comprehensively embody the three dimensions rating information of p index, i.e. classification intersects in allowed band; If do not meet Ψ<10%, jump to step 3, Reparametrization.

2. the grid branch importance appraisal procedure that a kind of Schema-based according to claim 1 is identified, it is characterized in that: the ISODATA clustering algorithm in described step 3 comprises the following steps;

Step 3.1, parameters: sample x to be sorted _i; The cluster centre number K of expection; Initial cluster centre number N _c; Minimum number of samples θ in each Clustering Domain _n; The standard deviation θ of sample range distribution in Clustering Domain _s; Minor increment θ between two cluster centres _c; The number of times I of the interative computation that the judgement circulation stops _p; Distance B between two cluster centres _ij;

Step 3.2, choose N at random _cindividual sample is as the center of initial clustering;

Step 3.3, by sample x _ibe assigned to nearest cluster S _j; Rule is: if D _j=min (|| x _i-C _j||), i=1,2 ..., p, j=1,2 ..., c, by x _ibe grouped into cluster S _j; Wherein, C _jbe j cluster centre, D _jfor sample x _ito the distance of j cluster centre, this distance is the shortest;

Step 3.4, the center of calculating each cluster:

$Z_i=\frac{1}{N_i}\underset{x\&Element;S_j}{\mathrm{\&Sigma;}}{x}_j$

By new central value Z _ithe center C that is decided to be cluster _i=Z _i, x _jbe j three-dimensional risk vector, j=1 wherein, 2 ..., S _j, N _ifor cluster S _jthe class number;

Step 3.5, division; If the number of current cluster is less than the clusters number K of expection, start to carry out the cluster division;

Step 3.6, merging; When the centre distance of two clusters is less than the minor increment θ at the two center _cthe time, two Cluster mergings are a new cluster; If the distance between whole cluster centres

just start to merge, new cluster centre is:

$C_n=\frac{1}{N_i+N_j}[N_i{C}_i+N_j{C}_j]$

Wherein, C _iand C _jbe respectively the cluster centre of i class and j class, N _iand N _jbe respectively cluster S _iand S _jthe class number;

If step 3.7 iterations reaches maximum iteration time I _p, or the process convergence, iterative process finishes, otherwise I _p=I _p+ 1, get back to step 3.

Images