CN103995215B - A kind of smart power grid fault diagnostic method based on multi-level feedback adjustment - Google Patents

Abstract

The present invention provides a kind of smart power grid fault diagnostic method based on multi-level feedback adjustment, including: device each in electrical network history switching information when electric network fault, electric parameters attribute information are stored to historical failure information bank；Electrical network each device electric quantity information is obtained during electric network fault；Electrical network each device electric quantity information according to obtaining carries out rough identification layer diagnosis: adopts the minimum district's demarcation method of cut-offfing based on equivalent network to determine suspected fault element, constitutes suspected fault element Candidate Set；Carry out fuzzy decision layer diagnosis；Carry out being accurately positioned layer diagnosis；According to three layers fault diagnosis result, suspected fault element collection E1, E2 are taken common factor, finally determine fault component diagnosis result.According to the complexity that the separate sources of multi-source fault message after electric network fault and various information obtain and process, electric network fault is carried out stratification analysis, take full advantage of all kinds of fault message, improve fault diagnosis accuracy by message complementary sense.

Description

A kind of smart power grid fault diagnostic method based on multi-level feedback adjustment

Technical field

The invention belongs to power transmission and distribution technical field, be specifically related to a kind of smart power grid fault diagnostic method based on multi-level feedback adjustment.

Background technology

In recent years, along with the construction energetically of intelligent grid, the scale of electrical network constantly expands, structure is increasingly sophisticated, and the safety and stability degree of reliability of power system and the successional requirement of power supply are also increased therewith.In the branch of intelligent grid, fault diagnosis is the function that intelligent grid is most basic, if the fault occurred can not process in time, can bring very serious loss, a kind of rapidly and effectively electric network failure diagnosis method is the strong backing that intelligent grid develops.After fault occurs; the protective relaying device meeting action of power system configuration is thus sending alarm to chopper; excise fault zone as early as possible; fault is avoided to spread further; but, when the automatic safety devices such as protection and chopper go wrong, it is impossible to excision fault promptly and accurately; may result in the expansion of fault zone, even develop into cascading failure.Therefore, it is necessary to the variation characteristic of electric parameters carries out fault diagnosis after occurring in conjunction with fault, it is determined that the order of fault zone and faulty equipment and fault progression, assist the analysis to fault and judgement.

In order to ensure power supply reliability, keep continuing uninterrupted power supply, fault element can be diagnosed to be after fault occurs fast and accurately and to its isolation simultaneously, restore electricity as early as possible, it is necessary to take method for diagnosing faults effectively.At present, electric network failure diagnosis method is divided into according to general sorting technique: based on the method for analytical model, based on the method for signal, Knowledge based engineering method.But these methods all exist certain defect, some methods are only limitted to single information source, and some method complexities are higher, it is most important that do not have certain versatility in the electrical network that scale is different.Therefore; the action etc. of the change of electric parameters after occurring based on fault, relay protection and chopper all follows logic cause effect relation; layering processing method can be adopted; certain inference method is adopted for each layer of information feature obtained; so both can ensure that making full use of of resource, certain versatility can be reached again.

Summary of the invention

For the deficiencies in the prior art, the present invention provides a kind of smart power grid fault diagnostic method based on multi-level feedback adjustment.

The technical scheme is that

A kind of smart power grid fault diagnostic method based on multi-level feedback adjustment, comprises the following steps:

Step 1: device each in electrical network history switching information when electric network fault, electric parameters attribute information are stored to historical failure information bank；

Step 2: obtain electrical network each device electric quantity information during electric network fault；

Step 3: the electrical network each device electric quantity information according to obtaining carries out rough identification layer diagnosis: adopt the minimum district's demarcation method of cut-offfing based on equivalent network to determine suspected fault element, constitute suspected fault element Candidate Set；

Step 3.1: the network topology of current electric grid is set up equivalent network, is equivalent to the node of network by device each in electrical network, and switch is equivalent to the limit of network；

Step 3.2: switch change information during according to electric network fault, it is determined that the switch of disconnection, the limit in equivalent network corresponding to the switch disconnected carries out labelling, and limit is cut-off in this limit namely；

Step 3.3: scan for along direction of tide, choosing and cut-offs maximum path, limit as critical path, having same tag number if existed in mulitpath, selects any bar；

Step 3.4: for the critical path selected, divides and determines and minimum cut-off district S_v(v=1,2 ..., t)；

The described minimum division cut-offfing district, specifically comprises the following steps that

Step 3.4.1: determine according to numbers with markd limit all in critical path and minimum cut-off district's number；

For the critical path selected, it is determined that minimum cut-off district number k, it is expressed as follows:

$k=\left\{\begin{array}{cc}p+1,& q>0.5\\ p,& q\≤0.5\end{array}\right.---\left(1\right)$

In formula, p takesInteger part, characterize the minimum integer part cut-offfing district's number；Q takesFractional part, characterize the minimum control part cut-offfing district's number of decision-making；Wherein x is all numbers with markd limit in critical path, and y (2≤y≤x) cut-offs the number with markd limit that district comprises, initial value y=2 for minimum；

Step 3.4.2: along cut-offfing limit in critical path, divide with y limit for a region, obtain initial minimum cut-offfing district: when q≤0.5, according to formula (1) carry out dividing minimum cut-off district after, incorporate remaining limit of cut-offfing into last region S_pIn；As q > 0.5, remaining limit of cut-offfing independently is divided into one minimum is cut-off district S_p+1；

Step 3.4.3: to each section initial minimum cut-off in the bifurcated branch road occurred in district cut-off limit, draw to the minimum of its critical path place and cut-off in district；

Step 3.5: each minimum district of cut-offfing is set up incidence matrix and limit state vector, the minimum nodes cut-off in district of the behavior of this matrix, is classified as the minimum limit number cut-off in district, if the node in matrix and limit exist topological relation, the matrix element that then this node is corresponding puts 1, otherwise sets to 0；In the state vector of limit, the element corresponding with markd limit puts 1, otherwise sets to 0；

Step 3.6: set up incidence matrix and limit state vector according to each minimum district of cut-offfing each minimum district of cut-offfing is set up node state vector CN=[cn₁,cn₂,...,cn_m]^T, CN=Y_m×n× B, wherein, Y_m×nFor incidence matrix, B=[b₁,b₂,...,b_n]^TFor limit state vector, if cn_iIt is zero, it is judged that device corresponding to node i is not defective device；If k (k=1,2 ...) non-zero, then corresponding for decision node i device is defective device, and this defective device is put into corresponding failure device Candidate Set；

Step 3.7: each minimum fault element Candidate Set determined in district that cut-offs is taken union, obtains fault element Candidate Set E1, comprise all suspected fault devices in this fault element Candidate Set；

Step 3.8: if only comprising a suspected fault device in fault element Candidate Set E1, then this device is defective device, i.e. electric network failure diagnosis result, then terminate；Otherwise perform step 4；

Step 4: carry out fuzzy decision layer diagnosis: adopt the probability inference fuzzy Decision Making Method based on causal network to determine suspected fault element, namely determine fault element according to the size of equipment fault degree after fuzzy decision, the equipment corresponding with equipment fault degree maximum is put into fault element Candidate Set E2；

Step 4.1: determine the relevant protection information of suspected fault element; and the chopper according to protection information and each protection correspondence; determine that each element effectively activates branch accordingly, and perform chopper node, the starting relay node of action and the root node of the action effectively activated in branch is numbered；

Step 4.2: calculate and effectively activate the directed edge weight performing between chopper node and starting relay node in branch, i.e. protection act probability；

Step 4.3: calculate the directed edge weight between starting relay node and execution chopper node, i.e. breaker actuation probability in effective activation branch；

Step 4.4: for all of suspected fault element, set up matrix D, the element D in D_kgExpression chopper k and suspected fault element g (g=1,2 ..., the action probability between s)；Set up the element E in matrix E, E_ghSign suspected fault element g (g=1,2 ..., the action probability between main protection s) and in protection h；Set up matrix F, the element F in F_ghCharacterize the action probability between the first back-up protection in suspected fault element g and protection h；Set up the element G in matrix G, G_ghCharacterize the action probability between the second back-up protection in suspected fault element g and protection h；

Step 4.5: set up the element H1 in matrix H 1=E × D, H1_ggCharacterize the main protection-chopper compound prbability of suspected fault element g；Element H2 in matrix H 2=F × D, H2_ggCharacterize the first back-up protection-chopper compound prbability of suspected fault element g；Element H3 in matrix H 3=G × D, H3_ggCharacterize the second back-up protection-chopper compound prbability of suspected fault element g；

Step 4.6: according to H1_gg, H2_gg, H3_ggCalculate main protection-element respectively relative to support, the first back-up protection-element relative to support support relative to the second back-up protection-element；

$\begin{array}{ccc}{p}_i^R=\frac{H{1}_{gg}}{\underset{i=1}{\overset{s}{\mathrm{\Σ}}}H{1}_{gg}}& q_i^R=\frac{H{2}_{gg}}{\underset{i=1}{\overset{s}{\mathrm{\Σ}}}H{2}_{gg}}& t_i^R=\frac{H{3}_{gg}}{\underset{i=1}{\overset{s}{\mathrm{\Σ}}}H{3}_{gg}}\end{array}---\left(4\right)$

For the relative support of main protection-element；It it is the relative support of the first back-up protection-element；It it is the relative support of the second back-up protection-element；

Step 4.7 the: respectively main protection-element of s suspected fault element is carried out fuzzy composition relative to support, the first back-up protection-element relative to support, the relative support of the second back-up protection-element, obtain equipment fault degree Δ_g；

${\mathrm{\Δ}}_g=1-e^{-H{1}_{gg}}\·e^{-H{2}_{gg}}\·e^{-H{3}_{gg}}---\left(5\right)$

In formula, Δ_g(g=1,2 ..., s), element maximum for equipment fault degree is put in fault element Candidate Set E2；

Step 4.8: if only comprising a suspected fault device in fault element Candidate Set E2, then this device is defective device, i.e. electric network failure diagnosis result, then terminate；Otherwise perform step 5；

Step 5: carry out being accurately positioned layer diagnosis: adopt the Rough Set Reduction method based on heredity TABU search to carry out the electric parameters attribute information in yojan historical failure information bank, by whether the property value of backward reasoning method failure judgement element Candidate Set E1, fault element Candidate Set E2 suspected fault element mates with the rule in the reduction rules storehouse after attribute reduction, if mating completely, then going to step 6, diagnosis terminates；Otherwise return step 3 to repartition and minimum cut-off district, it is determined that suspected fault element；

Step 6: according to three layers fault diagnosis result, suspected fault element collection E1, E2 are taken common factor, finally determine fault component diagnosis result.

Beneficial effect:

A kind of method adopting dynamic hierarchy diagnosis, according to the complexity that the separate sources of multi-source fault message after electric network fault and various information obtain and process, electric network fault is carried out stratification analysis, takes full advantage of all kinds of fault message, improve fault diagnosis accuracy by message complementary sense.Electric network failure diagnosis strategy is divided into switching layer, feeder line layer and substation level by the method, and according to the longitudinally adjusted diagnosis scheme of matching condition, shortens Diagnostic Time, improves diagnosis efficiency.The judged result of front two-layer determines whether in substation level, effectively raises the accuracy of fault diagnosis.

Accompanying drawing explanation

Fig. 1 is the smart power grid fault diagnostic system of the specific embodiment of the invention；

Fig. 2 is the smart power grid fault diagnostic method flow chart based on multi-level feedback adjustment of the specific embodiment of the invention；

Fig. 3 is the simple partial electric grid model schematic of the specific embodiment of the invention；

Fig. 4 is the rough identification layer diagnostic flow chart of the specific embodiment of the invention；

Fig. 5 is the fuzzy decision layer diagnosis flow chart of the specific embodiment of the invention；

Fig. 6 be the specific embodiment of the invention be accurately positioned layer diagnosis flow chart；

Fig. 7 is that minimum in the examples of implementation of the specific embodiment of the invention cut-offs district's schematic diagram；

Fig. 8 is the causal network figure that the specific embodiment of the invention carries out L8 during fuzzy decision layer diagnosis；

Fig. 9 is embodiment of the present invention causal network figure of L4 when carrying out fuzzy decision layer diagnosis.

Detailed description of the invention

Below in conjunction with accompanying drawing, the specific embodiment of the present invention is elaborated.

The smart power grid fault diagnostic system that the smart power grid fault diagnostic method based on multi-level feedback adjustment of present embodiment adopts, including: data acquisition processes core cell with pretreatment unit, Database Unit, multitask system Coordination Treatment unit and data.

Data acquisition and pretreatment unit gather data, process alert data and monitoring electric network state when being used for electric network fault.

Database Unit is used for storing System History record and Failure Diagnostic Code storehouse, completing data backup and management, including online monitoring data storehouse, historical data base and fault diagnosis task list；

Online monitoring data storehouse is real-time database, carries out classification storage for the fault message in a period of time after the generation of motion capture fault and by switching information, protection information, electric quantity information, simultaneously by timely for diagnosis request Write fault diagnostic task list；Described historical data base is used for storing network topology data, cause effect relation data, Realtime Alerts information, historical failure diagnostic rule, and the fault diagnosis task list of Database Unit is for storing the fault diagnosis request generated in a period of time.

Data processing core unit includes rough identification layer diagnostic module, fuzzy decision layer diagnosis module, accurate identification layer diagnostic module, Data Reduction module and Decision Control module.

Multitask system Coordination Treatment unit realizes timing scan fault diagnosis task list and data with alert by computer, reads diagnosis request the earliest in task list, startup separator diagnosis after obtaining fault information data.

Data processing core unit is to realize the process to fault diagnosis data by a host computer, after smart power grid fault occurs, electrical network multi-source fault message is uploaded to after data acquisition and pretreatment unit carry out data prediction, then transfers corresponding failure information according to each layer diagnosis information needed；

Rough identification layer diagnostic module is for according to each device electric quantity information of electrical network obtained, adopting the minimum district's demarcation method of cut-offfing based on equivalent network to determine suspected fault element, constitute suspected fault element Candidate Set E1；Fuzzy decision layer diagnosis module is for adopting the probability inference fuzzy Decision Making Method based on causal network to determine suspected fault element, namely determine fault element according to the size of equipment fault degree after fuzzy decision, the equipment corresponding with equipment fault degree maximum is put into fault element Candidate Set E2；It is accurately positioned layer diagnosis module for adopting the Rough Set Reduction method based on heredity TABU search to carry out the electric parameters attribute information in yojan historical failure information bank, by whether the property value of backward reasoning method failure judgement element Candidate Set E1, fault element Candidate Set E2 suspected fault element mates with the rule in the reduction rules storehouse after attribute reduction, if mating completely, then start Decision Control module and determine fault component diagnosis result；Otherwise repartition at rough identification layer diagnostic module and minimum cut-off district, it is determined that suspected fault element；Each diagnostic module is controlled by Decision Control module, can independent operating again can co-ordination, and be capable of between each diagnostic module diagnostic message transmission with share.

The smart power grid fault diagnostic method based on multi-level feedback adjustment of present embodiment, as in figure 2 it is shown, comprise the following steps:

Step 1: device each in electrical network history switching information when electric network fault, electric parameters attribute information are stored to historical failure information bank；

Step 2: obtain electrical network each device electric quantity information during electric network fault；

For a simple partial electric grid model, its model as it is shown on figure 3, figure has 11 buses, 10 circuits, 20 killer switches.Assume that L8 breaks down, the main protection action of circuit L8 both sides after fault, tripping chopper CB16, CB17, circuit L4B3 side the second back-up protection misoperation simultaneously, tripping chopper CB6；By 22 electrical data of Fault Recorder Information system acquisition, including 10 magnitudes of voltage, 6 current values and 6 phase-angle datas.

Step 3: the electrical network each device electric quantity information according to obtaining carries out rough identification layer diagnosis, as shown in Figure 4: adopt the minimum district's demarcation method of cut-offfing based on equivalent network to determine suspected fault element, constitute suspected fault element Candidate Set；

Step 3.1: the network topology of current electric grid is set up equivalent network, is equivalent to the node of network by device each in electrical network, and switch is equivalent to the limit of network；

Step 3.2: switch change information during according to electric network fault, it is determined that the switch of disconnection, the limit in equivalent network corresponding to the switch disconnected carries out labelling, and limit is cut-off in this limit namely；

Step 3.3: scan for along direction of tide, choosing and cut-offs maximum path, limit as critical path, having same tag number if existed in mulitpath, selects any bar；

Critical path is defined as: in a network, along the direction of trend, and the path that open switch is maximum.In present embodiment, critical path is bus B3-B8 circuit.

Step 3.4: for the critical path selected, divides and determines and minimum cut-off district S_v(v=1,2 ..., t)；

Specifically comprising the following steps that of the minimum division cut-offfing district

Step 3.4.1: determine according to numbers with markd limit all in critical path and minimum cut-off district's number；

For the critical path selected, it is determined that minimum cut-off district number k, it is expressed as follows:

$k=\left\{\begin{array}{cc}p+1,& q>0.5\\ p,& q\≤0.5\end{array}\right.---\left(1\right)$

In formula, p takesInteger part, characterize the minimum integer part cut-offfing district's number；Q takesFractional part, characterize the minimum control part cut-offfing district's number of decision-making；Wherein x is all numbers with markd limit in critical path, and y (2≤y≤x) cut-offs the number with markd limit that district comprises, initial value y=2 for minimum；

Step 3.4.2: along cut-offfing limit in critical path, divide with y limit for a region, obtain initial minimum cut-offfing district: when q≤0.5, according to formula (1) carry out dividing minimum cut-off district after, incorporate remaining limit of cut-offfing into last region S_pIn；As q > 0.5, remaining limit of cut-offfing independently is divided into one minimum is cut-off district S_p+1；

The size of q decides minimum cut-offs district's number, when q≤0.5, represent divide minimum cut-off district after, the remaining limit number that cut-offs is not above the half of y, is not enough to from becoming a region, therefore to cut-off that limit is unified incorporates last region S into by finally remaining_pIn, this ensure that the effectiveness of division；As q > 0.5, represent remaining and cut-off the limit number half more than y, it is possible to independent become a region, therefore finally remaining limit of cut-offfing independently is divided into one minimum is cut-off district S_p+1, thus reduce data dimension for subsequent calculations.

Step 3.4.3: to each section initial minimum cut-off in the bifurcated branch road occurred in district cut-off limit, draw to the minimum of its critical path place and cut-off in district；

Owing to critical path only having 3 open switch, therefore according to formula (1), initialize y=2, x=3, substitutes into formula (1) and obtains t=2, q=0, it is 1 that residue cut-offs limit number, it is not above y and is divided into a minimum district of cut-offfing, therefore y can not take 2, takes y=3, x=3, substitution formula (1) obtains t=1, q=0, and it is 0 that residue cut-offs limit number, meet and minimum cut-off district's number Rule of judgment, this is minimum cut-offs district as it is shown in fig. 7, for fault zone in dashed region, primarily determine that suspected fault element is L4, B7, L8.

Step 3.5: minimum cut-off district S to each_vSet up incidence matrix Y_m×nWith limit state vector B=[b₁,b₂,...,b_n]^T, the row m of this matrix is the minimum nodes cut-off in district, and row n is the minimum limit number cut-off in district, if the node i (i=1 in matrix, 2 ..., m) with limit j (j=1,2, ..., n) there is topological relation, then the matrix element that this node is corresponding puts 1, y_ij=1, otherwise set to 0, yij=0；Limit state vector B=[b₁,b₂,...,b_n]^TIn, the element corresponding with markd limit puts 1, otherwise sets to 0, even limit j (j=1,2 ..., there is labelling on n), then b_j=1, otherwise b_j=0；

The incidence matrix of present embodiment $Y_{3\×4}=\left[\begin{array}{cccc}1& 1& 0& 0\\ 0& 1& 1& 0\\ 0& 0& 1& 1\end{array}\right]$

Step 3.6: set up incidence matrix and limit state vector according to each minimum district of cut-offfing each minimum district of cut-offfing is set up node state vector A=[a₁,a₂,...,a_m]^T, A=Y_m×n× B, wherein, Y_m×nFor incidence matrix, B=[b₁,b₂,...,b_n]^TFor limit state vector, if a_iIt is zero, it is judged that device corresponding to node i is not defective device；If k (k=1,2 ...) non-zero, then corresponding for decision node i device is defective device, and this defective device is put into corresponding failure device Candidate Set；

In present embodiment, B=[1011]^T, node state vector $A=Y_{m\×n}\×B=\left[\begin{array}{cccc}1& 1& 0& 0\\ 0& 1& 1& 0\\ 0& 0& 1& 1\end{array}\right]\left[\begin{array}{c}1\\ 0\\ 1\\ 1\end{array}\right]=\left[\begin{array}{c}1\\ 1\\ 2\end{array}\right]$

Due to node state vector A=[a₁,a₂,…,a_m]^TIn, a₁,a₂,a₃Equal non-zero, therefore, element L4, B7, L8 that these three elements are corresponding are can fault element.

Step 3.7: minimum cut-off district S by each_vThe fault element Candidate Set determined takes union, obtains fault element Candidate Set E1, comprises all suspected fault devices in this fault element Candidate Set；

Step 3.8: if only comprising a suspected fault device in fault element Candidate Set E1, then this device is defective device, i.e. electric network failure diagnosis result, then terminate；Otherwise perform step 4；

Step 4: carry out fuzzy decision layer diagnosis, as shown in Figure 5: adopt the probability inference fuzzy Decision Making Method based on causal network to determine suspected fault element, namely determine fault element according to the size of equipment fault degree after fuzzy decision, the equipment corresponding with equipment fault degree maximum is put into fault element Candidate Set E2；

Step 4.1: determine the relevant protection information of suspected fault element; and the chopper according to protection information and each protection correspondence; determine that each element effectively activates branch accordingly, and perform chopper node, the starting relay node of action and the root node of the action effectively activated in branch is numbered；

Second layer fault diagnosis scheme is adopted to determine fault element according to the protection act information that failure information system is uploaded; it is first depending on the protection information uploaded; the i.e. main protection of circuit L8 both sides, the second back-up protection action of circuit L4B3 side; transferring the related elements in causal network data base is L8; L4, causal network data base is as shown in Figure 8, Figure 9.

Step 4.2: calculate and effectively activate the directed edge weight Ψ performing between chopper node and starting relay node in branch, i.e. protection act probability；

$\mathrm{\Ψ}=\frac{K_{rc}\·Dis}{L}\·e^{{\mathrm{\ρ}}^r}---\left(2\right)$

In formula, K_rcRepresent protection act adaptability to changes coefficient, expert reasoning method determine according to protection level；R represents protection level；Dis represents protection length, and L represents protection place line length, and ρ represents protection act weight, and e is constant；

Step 4.3: calculate the directed edge weight Γ between starting relay node and execution chopper node, i.e. breaker actuation probability in effective activation branch；

$\mathrm{\Γ}=(\sqrt{\frac{n_1}{(2^r+1)\·n_2}}+\mathrm{\Ψ})\·\frac{1}{K_c}---\left(3\right)$

In formula, K_cRepresent the protection safety coefficient corresponding with action breaker, expert system method determine according to protection level；R represents protection level；n₁Represent the element number between chopper and fault element；n₂Represent the chopper number between chopper and fault element；

Step 4.4: for all of suspected fault element, set up matrix D, the element D in D_kgExpression chopper k and suspected fault element g (g=1,2 ..., the action probability between s)；Set up the element E in matrix E, E_ghSign suspected fault element g (g=1,2 ..., the action probability between main protection s) and in protection h；Set up matrix F, the element F in F_ghCharacterize the action probability between the first back-up protection in suspected fault element g and protection h；Set up the element G in matrix G, G_ghCharacterize the action probability between the second back-up protection in suspected fault element g and protection h；

Step 4.5: set up the element H1 in matrix H 1=E × D, H1_ggCharacterize the main protection-chopper compound prbability of suspected fault element g；Element H2 in matrix H 2=F × D, H2_ggCharacterize the first back-up protection-chopper compound prbability of suspected fault element g；Element H3 in matrix H 3=G × D, H3_ggCharacterize the second back-up protection-chopper compound prbability of suspected fault element g.

Corresponding to setting up association probability matrix D, the row of this matrix represents CB16, CB17, CB6 in order respectively, and this matrix column represents L8, L4 in order respectively；Setting up matrix E, F, G, the row of these three matrixes represents L8, L4 all in order, after these three matrix columns represent L8B7 side master, L8B8 side master, L4B3 side 2 all in order respectively, and calculates matrix H 1, H2, H3, is expressed as respectively:

$\begin{array}{cc}D=\left[\begin{array}{cc}0.968& 0\\ 0.968& 0\\ 0& 0.798\end{array}\right]& E=\left[\begin{array}{ccc}0.939& 0.939& 0\\ 0& 0& 0\end{array}\right]\end{array}$

$\begin{array}{cc}F=\left[\begin{array}{ccc}0& 0& 0\\ 0& 0& 0\end{array}\right]& G=\left[\begin{array}{ccc}0& 0& 0\\ 0& 0& 0.591\end{array}\right]\end{array}$

$\begin{array}{ccc}H1=\left[\begin{array}{cc}1.818& 0\\ 0& 0\end{array}\right]& H2=\left[\begin{array}{cc}0& 0\\ 0& 0\end{array}\right]& H3=\left[\begin{array}{cc}0& 0\\ 0& 0.591\end{array}\right]\end{array};$

Step 4.6: according to H1_gg, H2_gg, H3_ggCalculate main protection-element respectively relative to support, the first back-up protection-element relative to support support relative to the second back-up protection-element；

$\begin{array}{ccc}{p}_i^R=\frac{H{1}_{gg}}{\underset{i=1}{\overset{s}{\mathrm{\Σ}}}H{1}_{gg}}& q_i^R=\frac{H{2}_{gg}}{\underset{i=1}{\overset{s}{\mathrm{\Σ}}}H{2}_{gg}}& t_i^R=\frac{H{3}_{gg}}{\underset{i=1}{\overset{s}{\mathrm{\Σ}}}H{3}_{gg}}\end{array}---\left(4\right)$

For the relative support of main protection-element；It it is the relative support of the first back-up protection-element；It it is the relative support of the second back-up protection-element；

Step 4.7 the: respectively main protection-element of s suspected fault element is carried out fuzzy composition relative to support, the first back-up protection-element relative to support, the relative support of the second back-up protection-element, obtain equipment fault degree Δ_g；

${\mathrm{\Δ}}_g=1-e^{-H{1}_{gg}}\·e^{-H{2}_{gg}}\·e^{-H{3}_{gg}}---\left(5\right)$

In formula, Δ_g(g=1,2 ..., s), the fault degree of suspected fault element L8, L4 is calculated as follows:

The fault degree of suspected fault element L8: ${\mathrm{\Δ}}_{g}1=1-e^{-H{1}_{gg}}\·e^{-H{2}_{gg}}\·e^{-H{3}_{gg}}=0.838$

The fault degree of suspected fault element L4: ${\mathrm{\Δ}}_{g}2=1-e^{-H{1}_{gg}}\·e^{-H{2}_{gg}}\·e^{-H{3}_{gg}}=0.376$

Element L8 maximum for equipment fault degree is put in fault element Candidate Set E2；

Step 4.8: if only comprising a suspected fault device in fault element Candidate Set E2, then this device is defective device, i.e. electric network failure diagnosis result, then terminate；Otherwise perform step 5；

Step 5: carry out being accurately positioned layer diagnosis: adopt the Rough Set Reduction method based on heredity TABU search to carry out the electric parameters attribute information in yojan historical failure information bank, by whether the property value of backward reasoning method failure judgement element Candidate Set E1, fault element Candidate Set E2 suspected fault element mates with the rule in the reduction rules storehouse after attribute reduction, if mating completely, then going to step 6, diagnosis terminates；Otherwise return step 3 to repartition and minimum cut-off district, it is determined that suspected fault element；As shown in Figure 6；

Step 5.1: extract electric parameters attribute information relevant to suspected fault element in historical failure information bank, construct original decision-making table IS, by attribute discretization, and adopts rough set theory to carry out attribute reduction and value yojan, and its process is used that recognizable vector A expresses；

Step 5.2: would not exist in the element in recognizable vector A and reject, and the element wherein only comprising single attribute is taken out reservation, all the other elements participate in the coding of genetic algorithm, it is determined that individual UVR exposure length M, the sample number N of population, maximum iteration time C1, crossover probability p_c, mutation probability p_m, structure ideal adaptation value function l (a)；

Step 5.3: make initial population iterations c₁=0, randomly generate first generation initial population X₁；

Step 5.4: calculating adaptive value function l (a) of each individuality in population, and individuality bigger for adaptive value function is selected the entrance next generation, all the other individualities carry out intersecting and suddenling change, and are collectively forming a new generation population X₂；

Step 5.5: adopt tabu search algorithm, calculates and obtains taboo list length N1, searching times C2, puts taboo list for sky；

Step 5.6: make initial ranging number of times c₂=0, it is determined that adaptive value function l (a), as object function, arranges population X₂In each individuality respectively initial solution；

Step 5.7: according to initial solution, produce certain neighborhood disaggregation, often once move, c₂Value add 1, until c₂> C2, then search terminates, and records optimal solution, and these optimal solutions are collectively forming the population of a new generation.

Step 5.8: if c₁> C1, then search terminates, and goes to step 4.9；Otherwise return step 4.4, c₁Value add 1.

Step 5.9: calculate adaptive value function l (a) of each individuality in population, order is worth maximum individual UVR exposure as final attribute reduction form, and by Discretization for Continuous Attribute, forms reduction rules storehouse T_i(i=1,2 ..., n)；

Step 5.10: the electric quantity information discretization of suspected fault element will transferred in real time respectively, whether the property value after judging discretization by backward reasoning method mates completely with the rule in reduction rules storehouse, if mating completely, then diagnosis terminates, otherwise enter the diagnosis of rough identification layer, the value of y adds 1, readjusts the minimum division cut-offfing district, until determining fault element；

Step 6: according to three layers fault diagnosis result, suspected fault element collection E1, E2 are taken common factor, finally determine fault component diagnosis result；In present embodiment, E1, E2 take common factor and finally determine that fault element result is circuit L8, same with the fault phase initially set up, thus demonstrating correctness and the effectiveness of this patent.

Claims (1)

1. the smart power grid fault diagnostic method based on multi-level feedback adjustment, it is characterised in that: comprise the following steps:

Step 1: device each in electrical network history switching information when electric network fault, electric parameters attribute information are stored to historical failure information bank；

Step 2: obtain electrical network each device electric quantity information during electric network fault；

Step 3: the electrical network each device electric quantity information according to obtaining carries out rough identification layer diagnosis: adopt the minimum district's demarcation method of cut-offfing based on equivalent network to determine suspected fault element, constitute suspected fault element Candidate Set；

Step 3.1: the network topology of current electric grid is set up equivalent network, is equivalent to the node of network by device each in electrical network, and switch is equivalent to the limit of network；

Step 3.2: switch change information during according to electric network fault, it is determined that the switch of disconnection, the limit in equivalent network corresponding to the switch disconnected carries out labelling, and limit is cut-off in this limit namely；

Step 3.3: scan for along direction of tide, choosing and cut-offs maximum path, limit as critical path, having same tag number if existed in mulitpath, selects any bar；

Step 3.4: for the critical path selected, divides and determines and minimum cut-off district S_v, v=1,2 ..., t, v represents the minimum number cut-offfing district, and t represents the maximum of the minimum number cut-offfing district；

The described minimum division cut-offfing district, specifically comprises the following steps that

Step 3.4.1: determine according to numbers with markd limit all in critical path and minimum cut-off district's number；

For the critical path selected, it is determined that minimum cut-off district number k, it is expressed as follows:

$k=\left\{\begin{array}{cc}p+1,& q>0.5\\ p,& q\≤0.5\end{array}\right.---\left(1\right)$

In formula, p takesInteger part, characterize the minimum integer part cut-offfing district's number；Q takesFractional part, characterize the minimum control part cut-offfing district's number of decision-making；Wherein x is all numbers with markd limit in critical path, and y minimum cut-offs the number with markd limit that district comprises, 2≤y≤x, initial value y=2；

Step 3.4.2: along cut-offfing limit in critical path, divide with y limit for a region, obtain initial minimum cut-offfing district: when q≤0.5, according to formula (1) carry out dividing minimum cut-off district after, incorporate remaining limit of cut-offfing into last region S_pIn；As q > 0.5, remaining limit of cut-offfing independently is divided into one minimum is cut-off district S_P+1；

Step 3.4.3: to each section initial minimum cut-off in the bifurcated branch road occurred in district cut-off limit, draw to the minimum of its critical path place and cut-off in district；

Step 3.5: each minimum district of cut-offfing is set up incidence matrix and limit state vector, the minimum nodes cut-off in district of the behavior of this matrix, is classified as the minimum limit number cut-off in district, if the node in matrix and limit exist topological relation, the matrix element that then this node is corresponding puts 1, otherwise sets to 0；In the state vector of limit, the element corresponding with markd limit puts 1, otherwise sets to 0；

Step 3.6: set up incidence matrix and limit state vector according to each minimum district of cut-offfing each minimum district of cut-offfing is set up node state vector CN=[cn₁, cn₂..., cn_m]^T, CN=Y_m×n× B, wherein, Y_m×nFor incidence matrix, B=[b₁, b₂..., b_n]^TFor limit state vector, if cn_iIt is zero, it is judged that device corresponding to node i is not defective device；If k non-zero, k=1,2 ..., then corresponding for decision node i device is defective device, and this defective device is put into corresponding failure device Candidate Set；

Step 3.7: each minimum fault element Candidate Set determined in district that cut-offs is taken union, obtains fault element Candidate Set E1, comprise all suspected fault devices in this fault element Candidate Set；

Step 3.8: if only comprising a suspected fault device in fault element Candidate Set E1, then this device is defective device, i.e. electric network failure diagnosis result, then terminate；Otherwise perform step 4；

Step 4: carry out fuzzy decision layer diagnosis: adopt the probability inference fuzzy Decision Making Method based on causal network to determine suspected fault element, namely determine fault element according to the size of equipment fault degree after fuzzy decision, the equipment corresponding with equipment fault degree maximum is put into fault element Candidate Set E2；

Step 4.1: determine the relevant protection information of suspected fault element; and the chopper according to protection information and each protection correspondence; determine that each element effectively activates branch accordingly, and perform chopper node, the starting relay node of action and the root node of the action effectively activated in branch is numbered；

Step 4.2: calculate and effectively activate the directed edge weight performing between chopper node and starting relay node in branch, i.e. protection act probability；

Step 4.3: calculate the directed edge weight between starting relay node and execution chopper node, i.e. breaker actuation probability in effective activation branch；

Step 4.4: for all of suspected fault element, set up matrix D, the element D in D_kgRepresent chopper k and suspected fault element g, g=1,2 ..., the action probability between s；Set up the element E in matrix E, E_ghCharacterize suspected fault element g, g=1,2 ..., the action probability between s and the main protection in protection h；Set up matrix F, the element F in F_ghCharacterize the action probability between the first back-up protection in suspected fault element g and protection h；Set up the element G in matrix G, G_ghCharacterize the action probability between the second back-up protection in suspected fault element g and protection h；

Step 4.5: set up the element H1 in matrix H 1=E × D, H1_ggCharacterize the main protection-chopper compound prbability of suspected fault element g；Element H2 in matrix H 2=F × D, H2_ggCharacterize the first back-up protection-chopper compound prbability of suspected fault element g；Element H3 in matrix H 3=G × D, H3_ggCharacterize the second back-up protection-chopper compound prbability of suspected fault element g；

Step 4.6: according to H1_gg, H2_gg, H3_ggCalculate main protection-element respectively relative to support, the first back-up protection-element relative to support support relative to the second back-up protection-element；

$\begin{array}{ccc}{p}_i^R=\frac{H{1}_{gg}}{\underset{i=1}{\overset{s}{\Σ}}H{1}_{gg}}& q_i^R=\frac{H{2}_{gg}}{\underset{i=1}{\overset{s}{\Σ}}H{2}_{gg}}& t_i^R=\frac{H{3}_{gg}}{\underset{i=1}{\overset{s}{\Σ}}H{3}_{gg}}\end{array}---\left(4\right)$

For the relative support of main protection-element；It it is the relative support of the first back-up protection-element；It it is the relative support of the second back-up protection-element；

Step 4.7 the: respectively main protection-element of s suspected fault element is carried out fuzzy composition relative to support, the first back-up protection-element relative to support, the relative support of the second back-up protection-element, obtain equipment fault degree Δ_g；

${\mathrm{\Δ}}_g=1-e^{-H{1}_{gg}}\·e^{-H{2}_{gg}}\·e^{-H{3}_{gg}}---\left(5\right)$

In formula, Δ_g, g=1,2 ..., s, puts into element maximum for equipment fault degree in fault element Candidate Set E2；

Step 4.8: if only comprising a suspected fault device in fault element Candidate Set E2, then this device is defective device, i.e. electric network failure diagnosis result, then terminate；Otherwise perform rapid 5；

Step 5: carry out being accurately positioned layer diagnosis: adopt the Rough Set Reduction method based on heredity TABU search to carry out the electric parameters attribute information in yojan historical failure information bank, by whether the property value of backward reasoning method failure judgement element Candidate Set E1, fault element Candidate Set E2 suspected fault element mates with the rule in the reduction rules storehouse after attribute reduction, if mating completely, then going to step 6, diagnosis terminates；Otherwise return step 3 to repartition and minimum cut-off district, it is determined that suspected fault element；

Step 6: according to three layers fault diagnosis result, suspected fault element collection E1, E2 are taken common factor, finally determine fault component diagnosis result.