CN104123678A - Electricity relay protection status overhaul method based on status grade evaluation model - Google Patents

Abstract

Disclosed is an electricity relay protection status overhaul method based on a status grade evaluation model. The electricity relay protection status overhaul method based on the status grade evaluation model is characterized by including the steps: proposing a standard for constructing a power system relay protection evaluation grade based on a K-means clustering algorithm of hierarchical system clustering; drawing up a relay protection status evaluation standard according to DL/T623-1997 power system relay protection and automatic safety device running evaluation rules; constructing the status grade evaluation model on the basis of establishing a status grade standard, and proposing a status grade evaluation model construction algorithm based on a K- nearest neighbor classification; drawing up a corresponding classification status overhaul strategy according to a classification status grade of a relay protection device; dynamically drawing up an overhaul plan of the relay protection device, reasonably arranging the status overhaul plan and status overhaul content, guaranteeing safe running of the relay protection device, and substantially promoting status overhaul work according to the classification status overhaul strategy which is optimized and a device status evaluation result. The electricity relay protection status overhaul method based on the status grade evaluation model can effectively reduce power failure rate for device overhauls and improve power supply reliability.

Description

A kind of electric relay protection repair based on condition of component method based on state grade assessment models

Technical field

The present invention relates to relay protection of power system state estimation technical field, a kind of electric relay protection repair based on condition of component method based on state grade assessment models.

Background technology

Along with the development of electric system, relay protection repair based on condition of component becomes more and more important in system, the assessment of relay protection state grade is the prerequisite of implementing relay protection repair based on condition of component, can this link be directly connected to repair based on condition of component and normally carry out, the rigid equivalent divided rank methods of marking of the many employings of most relay protection state grade assessment, the traditional k-means clustering algorithm of the general employing of division of classification standard, k-means clustering algorithm is as classical clustering algorithm, there is the plurality of advantages such as simple, quick and effective processing large-scale data;

But, tradition k-means clustering algorithm is a kind of rigid partitioning algorithm, it is strict each factor to be analyzed being divided in a certain class, and select at random initial cluster center, choose and may cause different cluster results for random initial cluster center, even exist without the situation of separating, ignored the different cluster effects of the each attribute of data object to cluster result performance simultaneously, these cause algorithm to be difficult to obtain stable and accurate cluster result;

Protective relaying device and secondary circuit Strategies of Maintenance and plan ingredient important in production work especially, maintenance system also develops into planned maintenance by emergency maintenance, and to develop into preventative maintenance be repair based on condition of component again.Repair based on condition of component is to solve to overhaul that strength deficiency, mistake in bad repair are repaiied under current prophylactic repair pattern and the effective means of the outstanding problem such as maintenance blindly, but current state maintenance is taking on-line monitoring as basis, only to find out faulty equipment to keep in repair, do not carry out self-adaptation adjustment according to state estimation result, that has reduced greatly Strategies of Maintenance can operating force.

Summary of the invention

The object of the invention is to overcome the deficiency that existing traditional k-means clustering algorithm exists, a kind of electric relay protection repair based on condition of component method based on state grade assessment models is provided.Build relay protection of power system evaluation grade standard by the k-means clustering algorithm based on hierarchical system cluster.Use the structure state grade assessment models based on the classification of k-arest neighbors, set up the Continuous Mappings relation of condition grading to state grade, realize sizing of equipment sexual state grade precise evaluation.Formulate relay protection classification status Strategies of Maintenance; and according to equipment state evaluation grade result, the classification Strategies of Maintenance of formulating is optimized to the turnaround plan of dynamic system locking equipment; can effectively reduce the interruption maintenance number of times of equipment, improve the reliability of power supply.

The object of the invention is to be realized by following technical scheme: a kind of electric relay protection repair based on condition of component method based on state grade assessment models, it is characterized in that, it comprises the following steps:

(1) propose a kind of k-means clustering algorithm based on hierarchical system cluster and build relay protection of power system state estimation classification standard; the initial cluster center of described k-means algorithm adopts hierarchical system clustering algorithm to determine; cluster centre adopts the high-density region center representation of class to represent; with the weights of the each factor of Information Entropy computational data sample; the employing regulatory factor that goes forward one by one is dynamically controlled merging; set up relay protection state evaluation classification standard

Wherein the between class distance of initial cluster center merging threshold epsilon computing formula is:

ε＝D _min+γ(D _max-D _min)

D _maxfor current maximum kind spacing, D _minfor current infima species spacing, γ is regulatory factor,

Wherein between class distance D (V _i, V _j) adopt the Euclidean distance between cluster centre to calculate:

$D(V_i,V_j)=\sqrt{\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{(V_{i_k}-V_{j_k})}^2\·w_i}$

V _ifor cluster C _icenter,

W _ibe the weights of i factor, adopt Information Entropy to calculate:

$w_i=\frac{1-h_i}{N-{\mathrm{\Σ}}_{i=1}^N{h}_i}(0\≤w_i\≤1,);$

(2) according to DL/T623-1997 relay protection of power system and automatic safety device postitallation evaluation procedure making relay protection state evaluation standard, for protective relaying device current state, each factor is marked, then be weighted the scoring that COMPREHENSIVE CALCULATING obtains factor class, obtain the overall score of protective device through final weighting joint account;

(3) in state grade standard base, build state estimation model setting up, propose to build algorithm based on the state grade assessment models of k-arest neighbors classification, utilize multifactor optimum seeking method to determine the distributed degrees of state factor between state area, set up the Continuous Mappings relation of condition grading to state grade, set up state grade assessment models, realize sizing of equipment sexual state grade precise evaluation;

(4) according to the graded properties state grade of relay protection device, formulate the classification status Strategies of Maintenance quantizing, obtain prolongation cycle, normal cycle, shortening cycle, time limit maintenance, overhaul five kinds of classification status Strategies of Maintenances immediately; And according to state estimation result computing mode membership function, formulate the optimization graded properties repair based on condition of component strategy based on state estimation,

Wherein the degree of membership computing formula of current state distribution of grades is:

(5) according to the classification Strategies of Maintenance of optimizing, the result of evaluating according to equipment state, the turnaround plan of dynamic system locking equipment, according to state evaluation grade severely subnormal, extremely, note, normally, good priority principle from low to high, equipment in same state grade situation is according to equipment shape grade state degree of membership assessment result principle from small to large, in same state grade and degree of membership situation, according to the maintenance weight of device history accumulation record of examination computing equipment, according to the prolongation cycle, normal cycle, the shortening cycle, time limit maintenance, power W is composed in maintenance immediately _reparrange maintenance,

A kind of electric relay protection repair based on condition of component method based on state grade assessment models of the present invention beneficial effect is compared with prior art: built more accurate classification standard, the state grade assessment models that has proposed the classification of k-arest neighbors builds algorithm, utilize multifactor optimum seeking method to determine the distributed degrees of state factor between state area, avoid traditional rigid division, set up the Continuous Mappings relation of data factor to state grade, from original preventative prophylactic repair to formulating scientifically and rationally graded properties Strategies of Maintenance, dynamic system locking equipment turnaround plan, can effectively reduce the maintenance frequency of power cut of equipment, improve the reliability of power supply, overhaul targetedly, improve the reliability of relay protection device.

Brief description of the drawings

Fig. 1 is a kind of FB(flow block) of the electric relay protection repair based on condition of component method based on state grade assessment models;

Fig. 2 is the graph of a relation between distributions and degree of membership.

Embodiment

Below in conjunction with accompanying drawing, a kind of electric relay protection repair based on condition of component method based on state grade assessment models of the present invention is described in detail.

With reference to Fig. 1, a kind of electric relay protection repair based on condition of component method based on state grade assessment models, comprises the steps:

(1) step of the described k-means clustering algorithm structure relay protection of power system evaluation grade standard based on hierarchical system cluster is:

Propose a kind of k-means clustering algorithm based on hierarchical system cluster and build relay protection of power system evaluation grade standard, utilize the weights of the each factor of Information Entropy computational data sample, adopt and compose the foundation of power Euclidean distance as similarity measurement, choose more reasonably initial cluster center and carry out cluster, obtain more accurate cluster result, set up relay protection state evaluation classification standard;

Choose k ₁individual, make k < k ₁< N, the number that N is contained sample point makes selected initial cluster center belong to respectively different classes as far as possible, uses hierarchical system clustering procedure, k ₁individual sample point constitutes a class by itself, two particles that distance is less than threshold epsilon are merged into a class, until when merging into k class, the k choosing a like this initial center is guaranteed in each different class, between class distance merges threshold epsilon and adopts the regulatory factor that goes forward one by one dynamically to control merging

ε＝D _min+γ(D _max-D _min)

(1)

D _maxfor current maximum kind spacing;

D _minfor current infima species spacing;

γ is regulatory factor, and γ ∈ [0,1] gets the initial value of quartile as γ;

Calculate the similarity of all sample points to k cluster centre, measuring similarity adopts composes power Euclidean distance d _w(x _a, x _b), incorporate respectively these sample points into cluster that similarity is the highest;

$d_w(x_a,x_b)=\sqrt{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{(x_{\mathrm{ai}}-x_{\mathrm{bi}})}^2\&CenterDot;w_i}---\left(2\right)$

Wherein between class distance D (V _i, V _j) adopt the Euclidean distance between cluster centre to calculate:

$D(V_i,V_j)=\sqrt{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{(V_{i_k}-V_{j_k})}^2\&CenterDot;w_i}---\left(3\right)$

V _ifor cluster C _icenter;

Adopt Information Entropy to calculate the weight w of each factor attribute _i, w _ibe the weights of i factor;

Information Entropy calculates each factor weights step:

1) raw data normalization, raw data matrix A=(a _ij) _{m × n}, to obtaining R=(r after its normalization _ij) _{m × n}, normalization formula:

$r_{\mathrm{ij}}=\left|\frac{a_{\mathrm{ij}}-\min \left\{a_{\mathrm{ij}}\right\}}{\max \left\{a_{\mathrm{ij}}\right\}-\min \left\{a_{\mathrm{ij}}\right\}}\right|---\left(4\right)$

2) entropy of i factor is:

$h_i=-k\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{f}_{\mathrm{ij}}\ln f_{\mathrm{ij}}---\left(5\right)$

In formula $f_{\mathrm{ij}}=r_{\mathrm{ij}}/{\mathrm{\&Sigma;}}_{j=1}^N{r}_{\mathrm{ij}},k=1/\ln N;$

3), after having defined the entropy of i factor, obtain the entropy power w of i factor _i;

$w_i=\frac{1-h_i}{N-{\mathrm{\&Sigma;}}_{i=1}^N{h}_i}(0\&le;w_i\&le;1,)---\left(6\right)$

According to cluster result, recalculate k cluster center separately, in order to calculate more accurately two distances between cluster, cluster centre adopts the high-density region center of gravity of class to represent, and high density center of gravity more can represent the general character of data in most of classes.The center of gravity of compute classes, then centered by center of gravity, outwards spread, form multidimensional clustering density region, the distance of each diffusion is maximum kind inside radius, the speed of control hierarchy cluster, calculates the density that each density area is corresponding, sorts by density, the zone definitions of density maximum is high-density region, using the center of gravity of high-density region as class center;

According to new cluster centre, recalculate the similarity of remaining sample point to k cluster centre, obtain new cluster;

Repeat above-mentioned two steps, until cluster centre no longer changes;

Calculate standard deviation sigma of all categories _idetect the objectivity of cluster, if standard deviation sigma _i> ε, cluster again;

${\mathrm{\&sigma;}}_i=\sqrt{\frac{{\mathrm{\&Sigma;}}_{x_i\&Element;N}{d}_w(x_i,c\left(N\right))}{N(N-1)}}i=\mathrm{1,2},...,n---\left(7\right)$

The standard deviation denominator of sample distance in class field is write as form, ensure that molecule denominator is all secondary, truly eliminate the impact of data number, reflected sample is apart from the degree of scatter of cluster centre more accurately;

According to above-mentioned algorithm, status data clustering is gone out to five classes, form state grade distribution and divide C ₁, C ₂, C ₃, C ₄, C ₅, respectively good with the state grade of relay protection device, normal, note, abnormal, severely subnormal is corresponding

(2) describedly according to relay protection state evaluation standard, each factor of evaluation is carried out to condition grading step and is:

According to DL/T623-1997 relay protection of power system and automatic safety device postitallation evaluation procedure making relay protection state evaluation standards of grading, according to project evaluation standards of grading, carry out condition grading for the each factor of evaluation of current state;

With the unit of being spaced apart, carry out quantitatively evaluating protective relaying device and secondary circuit state, corresponding factor of evaluation and factor class are respectively as table 1 and table 2;

Protective relaying device factor of evaluation and factor class are in table 1

Table 1

Secondary circuit factor of evaluation and factor class are in table 2

Table 2

First, for protective relaying device current state, each factor A in his-and-hers watches 1 _umark, use Information Entropy in step (1) to try to achieve u factor weight w _u, i factor class AC corresponding to u factor _iinner factor weight α _ikformula is:

${\mathrm{\&alpha;}}_{\mathrm{ik}}=w_u/\underset{A_j\&Element;{\mathrm{AC}}_i}{\overset{N_i}{\mathrm{\&Sigma;}}}{w}_j---\left(8\right)$

N _iit is the factor number that i factor class comprises;

α _ikfor factor A _ufactor class class in weight;

By weight α in the factor class of factor _ik, can try to achieve the factor class weight beta of i factor class _i, and ${\mathrm{\&Sigma;}}_{i=1}^{N_C}{\mathrm{\&beta;}}_i=1;$

N _cfor the number of factor of evaluation class;

Be weighted the scoring that COMPREHENSIVE CALCULATING obtains factor class, obtain the overall score of protective device through final weighting joint account, secondary circuit scoring process is identical with protective relaying device;

Reference table 1, taking running environment scoring as example, concrete steps comprise:

Device running environment evaluation criterion: environment temperature-5～40 DEG C, ambient humidity < 95%;

According to temperature acquisition data, temperature final score K ₁₁calculate according to its length of time of living in evaluation cycle:

$K_{11}=\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{g}_{1i}*B_i---\left(9\right)$

G _1ifor the score of corresponding moment temperature;

$g_{1i}=\left\{\begin{array}{cc}0.25T+3.75& -5<T\&le;5\\ 5& 5<T\&le;30\\ -0.25T+12.5& 30<T\&le;40\\ 0& T<-5,T>40\end{array}\right.---\left(10\right)$

B _ifor the residing time accounts for the number percent of whole evaluation time;

P is for evaluating number of times;

Temperature can not be suddenlyd change, and therefore, as long as there is 0 point of situation, devices illustrated breaks down, and reduces whole evaluation state score value, shows as temperature individual event scoring low;

According to humidity collection data, humidity final score K ₁₂calculate according to its length of time of living in evaluation cycle:

$K_{12}=\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{g}_{2i}*B_i---\left(11\right)$

G _2ifor the score of corresponding moment humidity;

$g_{2i}=\left\{\begin{array}{cc}5& 0\&le;T\&le;75\\ -0.125T+14.375& 75<T\&le;95\\ 0& T>95\end{array}\right.---\left(12\right)$

B _ifor the residing time accounts for the number percent of whole evaluation time;

P is for evaluating number of times;

Humidity can not be suddenlyd change, and therefore, as long as there is 0 point of situation, devices illustrated breaks down, and reduces whole evaluation state score value, shows as humidity individual event scoring low;

Device running environment is evaluated integrate score K ₁=(K ₁₁× α ₁₁+ K ₁₂× α ₁₂);

α ₁₁for the shared weight of temperature;

α ₁₂for the shared weight of humidity;

In table 1, each other factors and factor class are marked roughly the same, final calculation element overall score:

$K=\underset{i=1}{\overset{N_i}{\mathrm{\&Sigma;}}}{K}_i\&CenterDot;{\mathrm{\&beta;}}_i---\left(13\right)$

K _iit is i factor class integrate score;

In protective relaying device and secondary circuit on-line monitoring process, if the defect of finding can not directly cause relay protection function to lose efficacy, belong to potential faults, in state evaluation, provide corresponding score value; If the defect of finding can cause relay protection function to lose efficacy, now no longer Consideration class is marked and comprehensive grading, and should reduce the score value of whole state evaluation.

(3) described sizing of equipment sexual state grade precise evaluation step is:

Setting up in state grade standard base, build state grade assessment models, propose to build algorithm based on the state grade assessment models of k-arest neighbors classification, utilize multifactor optimum seeking method to determine the distributed degrees of state factor between state area, avoid rigid equivalent division, set up state vector to state grade be subordinate to continuously mapping relations, realize sizing of equipment sexual state grade precise evaluation;

For ensureing the continuity of mapping range, design the classifying and dividing algorithm between the state factor proximity based on the classification of k-arest neighbors, factor of evaluation A _ibe divided into between the proximity in interval carry out consecutive sort processing, adopt algorithm pair be classified to lower grade and form adjacent states data vector Z ₁, adopt algorithm pair be classified to lower grade and form lower adjacent states data vector Z ₂, carry out state Z ₁and Z ₂classification;

Wherein, use relative Weighted distance to measure as classifying distance

W _ifor state evaluation index system factor weight coefficient, X _j∈ C _i, j ∈ (1, | C _i|), i ∈ (1, | C|);

If the k finding out a nearest neighbor distance state vector is subordinate to same class C _j, by state vector Z ₁be divided into standard cluster C _jin, realize interval Equicontinuous is drawn

If the k finding out a nearest neighbor distance state vector is subordinate to same class C _j-1, by state vector Z ₁be divided into standard cluster C _j-1in, realize equicontinuous is drawn

If the k finding out a minimum distance state vector is under the jurisdiction of inhomogeneity, according to relative weighting quantity computation method, calculate the state vector number that is subordinate to same class in k-arest neighbors status data collection, A in k-arest neighbors status data collection _kif with state vector A _ibe subordinate to same class, relative distance is composed power method and is adopted calculate, calculate relative weighting state vector number by Z ₁state be divided into the most contiguous most state grade C _jin;

Determine state set of factors grading system interval according to classification standard, obtain state factor A _iattribute is to mapping the demarcation interval { [A of state grade _i1, A _i2), [A _i2, A _i3), [A _i3, A _i4), [A _i4, A _i5), [A _i5, A _i6), utilize multifactor optimum seeking method to determine the distributed degrees of state factor between state area;

Factor A _istate codomain be [A _ij-1, A _ij), establish { [e ₀, e ₁], [e ₁, e ₂] ..., [e _r-1, e _r] be to factor A _ia division of state codomain, e ₀=A _ij-1, e ₅=A _ij;

If Q _jbe expressed as follows one group of observed result: { q _sr| s=1,2 ..., 5, r=1,2 ..., 5}, q _srrepresent that it is c that sample point is concentrated classification _s, factor A _ivalue drop on [e _r-1, e _r] in all sample point numbers, to arbitrary interval division { [e ₀, e ₁], [e ₁, e ₂] ..., [e _r-1, e _r], can obtain a bivariate table 3,

Table 3

If μ is class C _swith factor A _ithe degree of correlation, 0≤μ≤1, μ depends on q _sr, q _s+and q _{+ r}, and q _sr, q _s+and q _{+ r}value depend on again the distribution situation of sample point, different interval division, the distribution difference of sample point, q _sr, q _s+and q _{+ r}value different, the value of μ also can change thereupon, first give an interval division, an initial value that obtains μ, then uses multifactor optimum seeking method constantly to adjust interval border, and μ is progressively increased, until obtain maximal value, to the adjustment of interval border, be actually and change e ₀, e ₁..., e _rthereby value change the process of μ, therefore μ is regarded as about e ₀, e ₁..., e _rfunction, establish μ=f (e ₀, e ₁..., e _r), obtain one and ask μ maximum value nonlinear programming problem.

The value of attribute is carried out to initial interval division { [e ₀, e ₁], [e ₁, e ₂] ..., [e _r-1, e _r], and obtain border point set { e ₀, e ₁..., e _rafter, obtain the initial value of μ, be designated as μ ₀, then fix e ₀, e ₂..., e _rat [e ₀, e ₂] interior to e ₁search for and make μ reach optimum, the optimal value of note μ is μ ₁, make it reach optimum e ₁be designated as , interval division now becomes again with μ ₁for starting point, fixing ? inside search and make μ reach optimum e ₂value, be designated as , the optimal value of note μ is μ ₂, interval division now becomes , iteration is gone down successively, until inside search , obtain the optimal value μ of μ _r, find optimal value μ _i, determine state factor A _iat interval [A _ij-1, A _ij) in distribution situation;

To between the adjacent country between two closed intervals, classify by k-nearest neighbor classification, form set of factors grading system the interval { [A of continuous semi-closure _i1, A _i2), [A _i2, A _i3), [A _i3, A _i4), [A _i4, A _i5), [A _i5, A _i6), recycle multifactor optimum seeking method and determine state factor [A between state area _ij-1, A _ij) distributed degrees, realize sizing of equipment sexual state grade precise evaluation.

Utilize the mark mapping algorithm in state grade interval of the state factor class based on arest neighbors sorting technique proposing, set up the mark mapping range of state grade of state factor class.The state vector of different brackets is divided into different scoring intervals, interval division is processed continuously, form the set of state grade factor class scoring interval division.Obtain state evaluation factor class scoring G _idemarcation interval { [0, G _i2), [G _i2, G _i3), [G _i3, G _i4), [G _i4, G _i5), [G _i5, w _i* 100] }.

State comprehensive grading based on arest neighbors classification is divided, take into full account the propinquity between interval interior status data, the state vector that integrate score is contiguous is divided into same grade as far as possible, and state vector far away integrate score relative distance is divided into different grade intervals.Interval division is processed continuously, formed relay protection state evaluation comprehensive grading demarcation interval { [0, G _i2), [G _i2, G _i3), [G _i3, G _i4), [G _i4, G _i5), [G _i5, 100] }, score value is distinguished corresponding severely subnormal, abnormal, attention, normal, good five state grades from low to high.

Based on relay protection condition grading score value and the state grade mapping method of arest neighbors classification, obtain the state grade assessment models of protective relaying device and secondary circuit, realize the assessment of relay protection state grade.

Protective relaying device score value and state grade relation are in table 4

Table 4

Secondary circuit score value and state relation are in table 5

Table 5

(4) describedly formulate classification status Strategies of Maintenance step according to state grade degree of membership and be:

According to the graded properties state grade of relay protection device, formulate corresponding classification status Strategies of Maintenance, make every effort to Strategies of Maintenance refinement, concrete, determine, can carry out; According to state estimation result computing mode membership function, formulate the optimization graded properties repair based on condition of component strategy based on state estimation, the relay protection repair based on condition of component strategy based on state estimation is specific as follows:

Formulate the relay protection state classification Strategies of Maintenance quantizing, the new adjustment of five kinds of obtaining recommends strategy to be respectively prolongation cycle, normal cycle, shortening cycle, time limit maintenance, maintenance immediately, as shown in table 6:

Table 6

Concrete, the prolongation cycle refers to that equipment state is good, does not have security of operation hidden danger, may waste Maintenance Resource by normal cycle maintenance, can the proper extension time between overhauls(TBO), and in the case, corresponding unit is percentage %, if now corresponding requirements is θ ₁=30, mean the cycle T of top point regulation ₁extend 30%;

Normal cycle refers to that equipment state is normal, and indivedual data variation are steady, do not have security of operation hidden danger, normal time between overhauls(TBO) T that completely can be in accordance with regulations ₂test, can not extend and also not need to shorten,

The shortening cycle refers to that equipment state is slightly poor, and existence amount approaches prescribed limits, does not still affect the safe operation of equipment, may have risk by normal cycle test, and in the case, corresponding unit is percentage %, if now corresponding requirements is θ ₃=30, mean the cycle T of top point regulation ₃shorten 30%;

Time limit maintenance refers to that equipment state is more critical, and existence amount meets or exceeds prescribed limits, affects the performance index of equipment, and equipment still can continue operation, if the risk that continuation long-play has an accident is larger, in the case, should be at the time limit T of regulation ₄inside overhaul, corresponding unit is sky;

Maintenance immediately refers to that equipment state is extremely critical, and existence amount seriously exceeds prescribed limits or severely subnormal, has at any time the risk having an accident, in the case, and the time limit T that should specify ₅inside overhaul, corresponding unit is hour;

The classification status Strategies of Maintenance of formulating, under ensureing under the safe operation condition of equipment, unnecessary maintenance is avoided in the maintenance of prolongation cycle and normal cycle maintenance, and the not urgent maintenance of shortening cycle maintenance minimizing, has avoided overhauling superfluous problem; Time limit maintenance and immediately maintenance increase important maintenance, have avoided visual plant and problem to overhaul not enough problem; Reduce the cost of overhaul, improved overhaul efficiency, the work of substantive promotion repair based on condition of component, guarantee equipment is in the safe operation of electrical network;

According to state grade degree of membership, formulate the classification status Strategies of Maintenance of optimizing, specific as follows:

Use state grade assessment models, by calculating current state vector A _ithe grade clustering cluster C that distance is subordinate to _jcenter V _jsymbolic distance calculate the degree of membership of current state vector and distributions wherein obtain current state vector and be under the jurisdiction of class C _jthe really degree that state grade is divided, state grade degree of membership

for class C under representing _jinside there is the just state vector number of distance;

for class C _jinside just distance is greater than V ₀the state vector number of just distance;

span (100%, 100%), expression state is the true possibility of particular state, wherein 0 point represents that equipment state is standard state level status, smallest limit-100 point are expressed as the left side edge of standard state level status, and limes superiors+100 point are expressed as the right side edge of standard state level status;

By calculating degree of membership , calculate the fine setting factor according to standard state classification Strategies of Maintenance, to corresponding classification Strategies of Maintenance cycle T _jquantize fine setting formulate the classification status Strategies of Maintenance of optimizing, ensure Strategies of Maintenance self-adaptation equipment state situation, improve rationality and the efficiency of maintenance, the relay protection device state optimization classification Strategies of Maintenance based on state, as shown in table 7:

Table 7

(5) describedly repair plan step and be according to the classification Strategies of Maintenance dynamic system regular inspection of optimizing:

According to the classification Strategies of Maintenance of optimizing, the result of evaluating according to equipment state, the turnaround plan of dynamic system locking equipment, the plan of reasonable arrangement repair based on condition of component and content, guarantee the safe operation of equipment, the work of substantive promotion repair based on condition of component.The dynamic turnaround plan of relay protection device is formulated and is followed following principle:

(1) equipment according to state evaluation grade severely subnormal, abnormal, note, normal, good priority principle from low to high;

(2) equipment Risk grade the give priority in arranging for principle of maintenance of Gao Zeyue more from high to low;

(3) equipment in same state grade situation is according to equipment shape grade state degree of membership assessment result principle from small to large;

(4) in same state grade and degree of membership situation, according to the maintenance weight of device history accumulation record of examination computing equipment, compose power according to prolongation cycle, normal cycle, shortening cycle, time limit maintenance, maintenance immediately, the lower weight of state grade is larger;

(15)

Wherein good maintenance is negative weight, and maintenance is positive weight immediately, and normal cycle is 0.

In sum; a kind of electric relay protection repair based on condition of component method based on state grade assessment models of the present invention; formulate corresponding maintenance and maintenance policy according to the graded properties state of equipment; what promoted Strategies of Maintenance standard can operating force; carry out self-adaptation adjustment according to state estimation result; formulate the turnaround plan of dynamic quantization, under the safe operation condition of the equipment of guarantee, provide the funds such as maintenance sequence number, maintenance rank, repair time, repair apparatus, maintenance content.Realize standardization, standardization, the institutionalization of the work of relay protection repair based on condition of component.

Claims (1)

1. the electric relay protection repair based on condition of component method based on state grade assessment models, is characterized in that, it comprises the following steps:

(1) propose a kind of k-means clustering algorithm based on hierarchical system cluster and build relay protection of power system state estimation classification standard; the initial cluster center of described k-means algorithm adopts hierarchical system clustering algorithm to determine; cluster centre adopts the high-density region center representation of class to represent; with the weights of the each factor of Information Entropy computational data sample; the employing regulatory factor that goes forward one by one is dynamically controlled merging; set up relay protection state evaluation classification standard

Wherein the between class distance of initial cluster center merging threshold epsilon computing formula is:

ε＝D _min+γ(D _max-D _min)

D _maxfor current maximum kind spacing, D _minfor current infima species spacing, γ is regulatory factor,

Wherein between class distance D (V _i, V _j) adopt the Euclidean distance between cluster centre to calculate:

$D(V_i,V_j)=\sqrt{\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{(V_{i_k}-V_{j_k})}^2\&CenterDot;w_i}$

V _ifor cluster C _icenter,

W _ibe the weights of i factor, adopt Information Entropy to calculate:

$w_i=\frac{1-h_i}{N-{\mathrm{\&Sigma;}}_{i=1}^N{h}_i}(0\&le;w_i\&le;1,);$

(2) according to DL/T623-1997 relay protection of power system and automatic safety device postitallation evaluation procedure making relay protection state evaluation standard, for protective relaying device current state, each factor is marked, then be weighted the scoring that COMPREHENSIVE CALCULATING obtains factor class, obtain the overall score of protective device through final weighting joint account;

(3) in state grade standard base, build state estimation model setting up, propose to build algorithm based on the state grade assessment models of k-arest neighbors classification, utilize multifactor optimum seeking method to determine the distributed degrees of state factor between state area, set up the Continuous Mappings relation of condition grading to state grade, set up state grade assessment models, realize sizing of equipment sexual state grade precise evaluation;

(4) according to the graded properties state grade of relay protection device, formulate the classification status Strategies of Maintenance quantizing, obtain prolongation cycle, normal cycle, shortening cycle, time limit maintenance, overhaul five kinds of classification status Strategies of Maintenances immediately; And according to state estimation result computing mode membership function, formulate the optimization graded properties repair based on condition of component strategy based on state estimation,

Wherein the degree of membership computing formula of current state distribution of grades is:

(5) according to the classification Strategies of Maintenance of optimizing, the result of evaluating according to equipment state, the turnaround plan of dynamic system locking equipment, according to state evaluation grade severely subnormal, extremely, note, normally, good priority principle from low to high, equipment in same state grade situation is according to equipment shape grade state degree of membership assessment result principle from small to large, in same state grade and degree of membership situation, according to the maintenance weight of device history accumulation record of examination computing equipment, according to the prolongation cycle, normal cycle, the shortening cycle, time limit maintenance, power W is composed in maintenance immediately _reparrange maintenance,