CN109061391A

CN109061391A - A kind of electric network failure diagnosis method and system based on computer vision tidal current chart

Info

Publication number: CN109061391A
Application number: CN201811046248.6A
Authority: CN
Inventors: 范士雄; 王松岩; 刘幸蔚; 郝博文; 卫泽晨; 王伟; 韩巍; 朱承治
Original assignee: State Grid Corp of China SGCC; State Grid Zhejiang Electric Power Co Ltd; China Electric Power Research Institute Co Ltd CEPRI
Current assignee: State Grid Corp of China SGCC; State Grid Zhejiang Electric Power Co Ltd; China Electric Power Research Institute Co Ltd CEPRI
Priority date: 2018-09-07
Filing date: 2018-09-07
Publication date: 2018-12-21
Anticipated expiration: 2038-09-07
Also published as: CN109061391B

Abstract

The present invention relates to a kind of electric network failure diagnosis method and system based on computer vision tidal current chart, the described method includes: generating the current computer vision tidal current chart of power grid according to the current flow data of power grid, then the computer vision tidal current chart current according to the power grid, fault diagnosis is carried out to power grid using the electric network failure diagnosis model being pre-designed, wherein, what the electric network failure diagnosis model being pre-designed was obtained using power grid historical trend data and its corresponding fault type as training data, by adopting the above technical scheme, it can be quick, accurately it is diagnosed to be type and the place of electric network fault, overcome the problem of portable difference in conventional diagnostic technology, and reduce the pressure of electric network data scheduling.

Description

A kind of electric network failure diagnosis method and system based on computer vision tidal current chart

Technical field

The present invention relates to electricity system fault diagnosis fields, and in particular to a kind of power grid based on computer vision tidal current chart Fault diagnosis method and system.

Background technique

Electric system contains many equipment such as generator, power transmission line and bus, and operations staff can pass through relay The operating status of protector, breaker and Communication Equipment Monitoring System；When electric system receives disturbance, operations staff can lead to Data acquisition is crossed to provide with supervisor control (Supervisory Control and Data Acquisition, SCADA) Data carry out Analysis on Fault Diagnosis；

When an error occurs, SCADA in a short period of time sends a large amount of warning information to the operation of control centre Personnel, this has aggravated the scheduling burden of power grid, so that fault diagnosis is difficult to carry out in time；In addition, actual condition is complicated, people is run Member can not determine to have broken, and cannot find fault point, or even if determining route failure, can still show on measurement equipment Show the wrong measuring value for seeming " true ", rather than meet the ideal zero of electrical rule, this leads to fault point obtained Data be it is wrong, both of these case all can largely mislead operations staff, so that efficiency of fault diagnosis is low and essence Exactness is bad.

Summary of the invention

The present invention provides a kind of electric network failure diagnosis method and system based on computer vision tidal current chart, and the purpose is to roots According to the current computer vision tidal current chart of power grid, using by training the corresponding computer visualization tidal current chart of historical trend data Obtained electric network failure diagnosis model carries out fault diagnosis to power grid, avoids for a large amount of diagnostic datas being sent to from SCADA system Control centre reduces the pressure of dispatching of power netwoks, improves the Efficiency and accuracy of electric network failure diagnosis.

The purpose of the present invention is adopt the following technical solutions realization:

A kind of electric network failure diagnosis method based on computer vision tidal current chart, it is improved in that the method packet It includes:

The current computer vision tidal current chart of power grid is generated according to the current flow data of power grid；

According to the current computer vision tidal current chart of the power grid, using the electric network failure diagnosis model being pre-designed to electricity Net carries out fault diagnosis；

Wherein, the electric network failure diagnosis model being pre-designed utilizes power grid historical trend data and its corresponding failure Type is obtained as training data.

Preferably, the flow data current according to power grid generates the current computer vision tidal current chart of power grid, comprising:

Obtain current corresponding 3 machine, 9 node system figure of flow data of power grid；

By the corresponding Matrix Power value of component in corresponding 3 machine, 9 node system figure of the current flow data of the power grid with The element replacement of the corresponding predetermined location of component, obtains current flow data matrix in 28 × 28 null matrix；

It by the current flow data matrix conversion is its corresponding computer vision by the images function of MATLAB Tidal current chart.

Further, if branch where component is not in current corresponding 3 machine, 9 node system figure of flow data of the power grid Failure, then using the active power of the component as its corresponding Matrix Power value；

If branch trouble where component in corresponding 3 machine, 9 node system figure of the current flow data of the power grid, should The random noise power value P of component_lAs its corresponding Matrix Power value, wherein determine the random noise of the component as the following formula Performance number P_l:

P_l=rand [- P_dis,P_dis]

In formula, rand [] is random number functions, P_disThe maximum value of measurement equipment noise power is corresponded to for grid branch.

Further, component is corresponding in corresponding 3 machine, 9 node system figure of the flow data that the power grid is current The element of Matrix Power value predetermined location corresponding with component in 28 × 28 null matrix is replaced, and current flow data is obtained Matrix, comprising:

By the Matrix Power value of the first generator in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The member arranged respectively with the 21st row the 14th column, the 21st row the 15th column, the 22nd row the 14th column and the 22nd row the 15th in the null matrix Element value replacement；

By the Matrix Power value of the second generator in corresponding 3 machine, 9 node system figure of the current flow data of the power grid With the element value replacement of the column of eighth row the 4th, the column of eighth row the 5th, the 9th row the 4th column and the 9th row the 5th column in the null matrix；

By the Matrix Power value of third generator in corresponding 3 machine, 9 node system figure of the current flow data of the power grid With the element value replacement of the column of eighth row the 24th, the column of eighth row the 25th, the 9th row the 24th column and the 9th row the 25th column in the null matrix；

By the Matrix Power value of the first load in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The element value of 13rd row the 7th column, the 13rd row the 8th column, the 14th row the 7th column and the 14th row the 8th column in opposite number and the null matrix Replacement；

By the Matrix Power value of the second load in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The member of 13rd row the 21st column, the 13rd row the 22nd column, the 14th row the 21st column and the 14th row the 22nd column in opposite number and the null matrix Element value replacement；

By the Matrix Power value of third load in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The element of 9th row the 14th column, the 9th row the 15th column, the 10th row the 14th column and the 10th row the 15th column in opposite number and the null matrix Value replacement；

By the Matrix Power value of first line in corresponding 3 machine, 9 node system figure of the current flow data of the power grid with The element value replacement of 13rd row the 10th column, the 14th row the 10th column and the 15th row the 10th column in the null matrix；

By the Matrix Power value of the second route in corresponding 3 machine, 9 node system figure of the current flow data of the power grid with The element value replacement of 13rd row the 19th column, the 14th row the 19th column and the 15th row the 19th column in the null matrix；

By the Matrix Power value of tertiary circuit in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The element value of the 9th row the 10th column, the 10th row the 10th column and the 11st row the 10th column is replaced in opposite number and the null matrix；

By the Matrix Power value of the 4th route in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The element value of the 9th row the 19th column, the 10th row the 19th column and the 11st row the 19th column is replaced in opposite number and the null matrix；

By the Matrix Power value of the 5th route in corresponding 3 machine, 9 node system figure of the current flow data of the power grid with The element value replacement that the column of eighth row the 11st, the column of eighth row the 12nd and eighth row the 13rd arrange in the null matrix；

By the Matrix Power value of the 6th route in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The element value replacement that the column of eighth row the 16th, the column of eighth row the 17th and eighth row the 18th arrange in null matrix described in opposite number；

By the Matrix Power value of the first transformer in corresponding 3 machine, 9 node system figure of the current flow data of the power grid It is replaced with the element value of the 17th row the 14th column, the 18th row the 14th column and the 19th row the 14th column in the null matrix；

By the Matrix Power value of the second transformer in corresponding 3 machine, 9 node system figure of the current flow data of the power grid With the element value replacement that the column of eighth row the 7th, the column of eighth row the 8th and eighth row the 9th arrange in the null matrix；

By the Matrix Power value of third transformer in corresponding 3 machine, 9 node system figure of the current flow data of the power grid With the element value replacement that the column of eighth row the 20th, the column of eighth row the 21st and eighth row the 22nd arrange in the null matrix.

Preferably, it is obtained using power grid historical trend data and its corresponding fault type as training data described preparatory The process of the electric network failure diagnosis model of design includes:

Obtain corresponding 3 machine, 9 node system figure of power grid historical trend data；

By the corresponding Matrix Power value of component and 28 in corresponding 3 machine, 9 node system figure of the power grid historical trend data The element replacement of the corresponding predetermined location of component, obtains historical trend data matrix in × 28 null matrix；

The historical trend data matrix is converted into its corresponding computer vision by the images function of MATLAB Tidal current chart sample；

Using the computer vision tidal current chart sample as the input of initial convolution neural network model, the computer view Feel that the corresponding fault type of tidal current chart sample is trained as the output of initial convolution neural network model, obtains described preparatory The electric network failure diagnosis model of design；

Wherein, when extracting input sample feature during the initial convolution neural network model of training, the spy of extraction is obtained Similarity between sign, similarity is not less than two features of the similarity threshold threshold of setting if it exists, then described in reservation A feature in two features.

Further, described by the corresponding square of component in corresponding 3 machine, 9 node system figure of the power grid historical trend data The element replacement of battle array performance number predetermined location corresponding with component in 28 × 28 null matrix, obtains historical trend data square Battle array, comprising:

By the Matrix Power value of the first generator in corresponding 3 machine, 9 node system figure of the power grid historical trend data point The element not arranged with the 21st row the 14th column, the 21st row the 15th column, the 22nd row the 14th column and the 22nd row the 15th in the null matrix Value replacement；

By the Matrix Power value of the second generator in corresponding 3 machine, 9 node system figure of the power grid historical trend data with The element value replacement of the column of eighth row the 4th, the column of eighth row the 5th, the 9th row the 4th column and the 9th row the 5th column in the null matrix；

By the Matrix Power value of third generator in corresponding 3 machine, 9 node system figure of the power grid historical trend data with The element value replacement of the column of eighth row the 24th, the column of eighth row the 25th, the 9th row the 24th column and the 9th row the 25th column in the null matrix；

By the phase of the Matrix Power value of the first load in corresponding 3 machine, 9 node system figure of the power grid historical trend data Anti- number is arranged with the 13rd row the 7th in the null matrix, the 13rd row the 8th column, the 14th row the 7th arrange and the element value of the 14th row the 8th column replaces It changes；

By the phase of the Matrix Power value of the second load in corresponding 3 machine, 9 node system figure of the power grid historical trend data Anti- several elements with the 13rd row the 21st column, the 13rd row the 22nd column, the 14th row the 21st column and the 14th row the 22nd column in the null matrix Value replacement；

By the phase of the Matrix Power value of third load in corresponding 3 machine, 9 node system figure of the power grid historical trend data Anti- several element values with the 9th row the 14th column, the 9th row the 15th column, the 10th row the 14th column and the 10th row the 15th column in the null matrix Replacement；

By the Matrix Power value of first line and institute in corresponding 3 machine, 9 node system figure of the power grid historical trend data State the element value replacement of the 13rd row the 10th column, the 14th row the 10th column and the 15th row the 10th column in null matrix；

By the Matrix Power value of the second route and institute in corresponding 3 machine, 9 node system figure of the power grid historical trend data State the element value replacement of the 13rd row the 19th column, the 14th row the 19th column and the 15th row the 19th column in null matrix；

By the phase of the Matrix Power value of tertiary circuit in corresponding 3 machine, 9 node system figure of the power grid historical trend data Anti- several element values with the 9th row the 10th column, the 10th row the 10th column and the 11st row the 10th column in the null matrix are replaced；

By the phase of the Matrix Power value of the 4th route in corresponding 3 machine, 9 node system figure of the power grid historical trend data Anti- several element values with the 9th row the 19th column, the 10th row the 19th column and the 11st row the 19th column in the null matrix are replaced；

By the Matrix Power value of the 5th route and institute in corresponding 3 machine, 9 node system figure of the power grid historical trend data State the column of eighth row the 11st in null matrix, the element value replacement that the column of eighth row the 12nd and eighth row the 13rd arrange；

By the phase of the Matrix Power value of the 6th route in corresponding 3 machine, 9 node system figure of the power grid historical trend data The anti-element value replacement for counting the column of eighth row the 16th, the column of eighth row the 17th and the column of eighth row the 18th in the null matrix；

By the Matrix Power value of the first transformer in corresponding 3 machine, 9 node system figure of the power grid historical trend data with The element value replacement of 17th row the 14th column, the 18th row the 14th column and the 19th row the 14th column in the null matrix；

By the Matrix Power value of the second transformer in corresponding 3 machine, 9 node system figure of the power grid historical trend data with The element value replacement that the column of eighth row the 7th, the column of eighth row the 8th and eighth row the 9th arrange in the null matrix；

By the Matrix Power value of third transformer in corresponding 3 machine, 9 node system figure of the power grid historical trend data with The element value replacement that the column of eighth row the 20th, the column of eighth row the 21st and eighth row the 22nd arrange in the null matrix.

Further, if the not event of branch where component in corresponding 3 machine, 9 node system figure of the power grid historical trend data Barrier, then using the random active power of the component as its corresponding Matrix Power value；

If branch trouble where component in corresponding 3 machine, 9 node system figure of the power grid historical trend data, by the portion The random noise power value P of part_lAs its corresponding Matrix Power value, wherein determine the random noise function of the component as the following formula Rate value P_l:

P_l=rand [- P_dis,P_dis]

Further, if the component is the second hair in corresponding 3 machine, 9 node system figure of the power grid historical trend data Motor or third generator determine its random active power as the following formula:

If the component is the first load in current corresponding 3 machine, 9 node system figure of flow data of the power grid, the Two loads or third load determine its random active power as the following formula:

If the component be the first generator in current corresponding 3 machine, 9 node system figure of flow data of the power grid, First transformer, the second transformer, third transformer, first line, the second route, tertiary circuit, the 4th route, the 5th route Or the 6th route, then its random active power is determined as the following formula

In formula, P_GiThe injecting power of the second generator or third generator in standard example is calculated for electric power system tide, P_LjThe injecting power of the first load, the second load or third load in standard example, K are calculated for electric power system tide_GiAnd K_LjFor Preset to be randomly provided parameter, rand [] is random number functions；P_TIt is calculated first in standard example for electric power system tide Generator, the first transformer, the second transformer, third transformer, first line, the second route, tertiary circuit, the 4th route, The active power of 5th route or the 6th route.

Further, the coefficient of similarity r between feature A and feature B is determined as the following formula_c:

Wherein,A_mnIt is characterized figure A m row the n-th column picture The pixel value of plain block, B_mnIt is characterized the pixel value of figure B m row the n-th column block of pixels, m ∈ [1, N_c1], n ∈ [1, N_c2],For spy The pixel average of sign figure A,It is characterized the pixel average of figure B, N_c1It is characterized total line number of figure, N_c2It is characterized total column of figure Number, N_c1=N_c2。

Further, which is characterized in that 3 machine, 9 node system includes:

Second generator, bus 2, the second transformer, bus 7, the 5th route, bus 8, third load, bus the 9, the 6th Route, third transformer, bus 3, third generator, tertiary circuit, bus 5, the first load, first line, bus 4, first Transformer, bus 1, the first generator, the second load, the second route, bus 6 and the 6th route；

Second generator, bus 2, the second transformer, bus 7, the 5th route, bus 8, third load, bus 9, 6th route, third transformer, bus 3 and third generator are sequentially connected；

5th route, tertiary circuit, bus 5, first line, bus 4, the first transformer, bus 1 and the first power generation Machine is sequentially connected；

6th route, the 4th route, bus 6, the second route, bus 4, the first transformer, bus 1 and the first power generation Machine is sequentially connected；

First load is connect with bus 5；Second load is connect with bus 6.

A kind of electric network failure diagnosis system based on computer vision tidal current chart, it is improved in that the system packet It includes:

Tidal current chart generation module, for generating the current computer vision trend of power grid according to the current flow data of power grid Figure；

Diagnostic module utilizes the power grid event being pre-designed for the computer vision tidal current chart current according to the power grid Hinder diagnostic model and fault diagnosis is carried out to power grid；

Preferably, the tidal current chart generation module, comprising:

Acquiring unit, for obtaining corresponding 3 machine, 9 node system figure of the current flow data of power grid；

Replacement unit, for component in corresponding 3 machine, 9 node system figure of the current flow data of the power grid is corresponding The element of Matrix Power value predetermined location corresponding with component in 28 × 28 null matrix is replaced, and current flow data is obtained Matrix；

Converting unit, for being corresponded to the current flow data matrix conversion for it by the images function of MATLAB Computer vision tidal current chart.

P_l=rand [- P_dis,P_dis]

Further, the replacement unit, is used for:

P_l=rand [- P_dis,P_dis]

Further, 3 machine, 9 node system includes:

First load is connect with bus 5；Second load is connect with bus 6.

Compared with the immediate prior art, the present invention is also had the following beneficial effects:

Using technical solution of the present invention, the current computer vision tide of power grid is generated according to the current flow data of power grid The flow data of power grid, is converted to the tidal current chart of computer visualization by flow graph, has not only remained the electric data of electric network swim, but also The topological structure of power grid is remained, provides strong support for the accuracy of fault diagnosis result；Current according to the power grid Computer vision tidal current chart carries out fault diagnosis to power grid using the electric network failure diagnosis model being pre-designed, wherein described pre- The electric network failure diagnosis model first designed is obtained using power grid historical trend data and its corresponding fault type as training data It takes, using power grid historical data, that is, fault type Training diagnosis model, ensure that the reliability of diagnostic model, be based on the technology Scheme effectively reduces the scheduling pressure of power grid, overcomes simultaneously it is not necessary that large-scale fault data is sent to control centre There is interfering result mistake in operating condition in the prior art, improve robustness and diagnosis efficiency.

Detailed description of the invention

Fig. 1 is the flow chart of electric network failure diagnosis method of the embodiment of the present invention based on computer vision tidal current chart；

Fig. 2 is the flow data matrix of electric network failure diagnosis method of the embodiment of the present invention based on computer vision tidal current chart Schematic diagram figure；

Fig. 3 is the computer visualization of electric network failure diagnosis method of the embodiment of the present invention based on computer vision tidal current chart Tidal current chart；

Fig. 4 is the electric network failure diagnosis of electric network failure diagnosis method of the embodiment of the present invention based on computer vision tidal current chart Modulus principle figure；

Fig. 5 is that the embodiment of the present invention deletes the spy after weight based on the electric network failure diagnosis method characteristic of computer vision tidal current chart Levy schematic diagram；

Fig. 6 is electric network failure diagnosis method electric network swim data of the embodiment of the present invention based on computer vision tidal current chart 3 machine, 9 node system figure；

Fig. 7 is the structural schematic diagram of electric network failure diagnosis system of the embodiment of the present invention based on computer vision tidal current chart；

3 machine, 9 node when Fig. 8 is electric network failure diagnosis method of the embodiment of the present invention by taking IEEE9 system as an example non-failure System diagram；

3 machine, 9 node system when Fig. 9 is electric network failure diagnosis method failure of the embodiment of the present invention by taking IEEE9 system as an example System figure.

Specific embodiment

It elaborates with reference to the accompanying drawing to a specific embodiment of the invention.

In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art All other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.

Power grid measurement equipment is likely to the case where by noise jamming in actual motion environment, even if at this moment route has broken Open, on measurement equipment still can display interference wrong measuring value, this makes operations staff's failure point data obtained be complete Mistake；Therefore, it is necessary to develop a kind of method that the power grid that can be effectively applicable in interference operating condition carries out fault diagnosis； In recent years, deep learning development is very rapid, and related researcher proposes such as stacking autocoder, recurrent neural net Network, deepness belief network and convolutional neural networks (Convolutional Neural Network, CNN) even depth learn mould Type, in these types of model, CNN is a kind of using relatively broad and mature new technology, can be automatically from initial data Validity feature information is extracted in (especially image), meanwhile, compared to traditional ANN method, partially connected and weight are shared Two technologies can greatly simplify network, accelerate the training speed of neural network.In addition, CNN can be with a kind of more macroscopical Mode catch the feature of image, even if image by noise jamming or it is imperfect when CNN still be able to well carry out it Identification, the above technical advantage make CNN have vast potential for future development in power system failure diagnostic field.

The electric network failure diagnosis method based on deep learning that the invention proposes a kind of.This method is first mentioned SCADA The electric network swim data of confession are converted into electric network current diagram (the Computer Visualized Power Flow of computer vision Image, CVPFI), convolutional neural networks are then inputted, network can carry out the position of grid collapses accurate and fast later The judgement of speed.

The present invention provides a kind of accumulation energy type photovoltaic plants for supporting power grid"black-start" to start method and system, carries out below Explanation.

Fig. 1 shows the flow chart of electric network failure diagnosis method of the embodiment of the present invention based on computer vision tidal current chart, As shown in Figure 1, the method may include:

101. generating the current computer vision tidal current chart of power grid according to the current flow data of power grid；

102. utilizing the electric network failure diagnosis model being pre-designed according to the current computer vision tidal current chart of the power grid Fault diagnosis is carried out to power grid；

Wherein, the electric network failure diagnosis model being pre-designed utilizes power grid historical trend data and its corresponding failure Type is obtained as training data；

In view of it influences distribution presentation " from the near to the distant " feature on plane space after grid collapses, use The electric network swim data of CVPFI substitution numeric form preferably extract the spatial information contained in trend as the input energy of CNN And then the position of system jam is judged；

Wherein, the flow data current according to power grid generates the current computer vision tidal current chart of power grid, can wrap It includes:

Once open circuit fault occurs for power grid, the power distribution of network will change.Meanwhile changed wattful power Rate can be gathered in around the region broken down, and changing value can successively decrease in region around.Power after failure, which is distributed, to be presented Such a complex form explanation, a kind of specific power be distributed with and only a kind of system failure (generator, load are removed Or open circuit) be corresponding to it.In view of the situation, although the monitoring device in region of breaking down may power off either by noise Interference, the but because of one-to-one relationship that power is distributed after fault type and failure, when troubleshooting using fault zone it Outer flow data is theoretically feasible, if therefore current corresponding 3 machine, 9 node system figure of flow data of the power grid The non-failure of branch where middle component, then using the active power of the component as its corresponding Matrix Power value；

P_l=rand [- P_dis,P_dis]

The flow data of electric network failure diagnosis method Fig. 2 shows the embodiment of the present invention based on computer vision tidal current chart Matrix schematic diagram figure, as shown in Fig. 2, generator and load are plotted as 2 × 2 sizes in figure, the absolute value of four points is that its is active Performance number, line and transformer are plotted as 1 × 3 size, and the absolute value of three points is its active power value, and white space owns Pixel value is set as zero,

The corresponding Matrix Power of component in corresponding 3 machine, 9 node system figure of the flow data that the power grid is current It is worth the element replacement of predetermined location corresponding with component in 28 × 28 null matrix, obtains current flow data matrix, it can To include:

By the Matrix Power value of third transformer in corresponding 3 machine, 9 node system figure of the current flow data of the power grid With the element value replacement that the column of eighth row the 20th, the column of eighth row the 21st and eighth row the 22nd arrange in the null matrix；

Each electrical equipment in CVPFI matrix position and its position in network system figure correspond, CVPFI square The absolute value of each element is used to indicate that the numerical value of power in flow data, symbol are used to indicate the direction of trend in battle array；

Generator and the element value symbol of load are that positive sign is indicated to power grid injecting power, indicate to absorb from power grid for negative sign Power；The element value symbol of transformer and route is that positive sign indicates that flow of power direction is consistent with positive direction, indicates function for negative sign Rate flow direction is opposite with positive direction.

The computer that Fig. 3 shows electric network failure diagnosis method of the embodiment of the present invention based on computer vision tidal current chart can Depending on changing tidal current chart, shown in figure Fig. 3, the current flow data matrix conversion is corresponded to for it by the images function of MATLAB Computer vision tidal current chart；Wherein the element value in the pixel value of block of pixels and the flow data matrix corresponds, net The size and Orientation of the electrical equipment of network, topological structure and trend is just fully retained in CVPFI.

It needs to input great amount of samples when training CNN, self-optimization is carried out to improve identification by the continuous iteration of network Different network failure types (including normal condition) can be generated in accuracy rate, the Load flow calculation standard example based on network system The corresponding sample of CVPFI containing noise and its label, and then the sample as training CNN:

Using power grid historical trend data and its corresponding fault type as training data obtain described in be pre-designed The process of electric network failure diagnosis model may include:

Fig. 4 shows the electric network fault of electric network failure diagnosis method of the embodiment of the present invention based on computer vision tidal current chart Modulus principle figure is diagnosed, as shown in figure 4, here using classical CNN structure；The neuron of input layer is in 28 × 28 square arrangement (its 28 × 28 pixel for corresponding CVPFI), extract the spatial structure characteristic of CVPFI, most using convolution twice later 7 longitudinal arrangement neurons of whole network output layer respectively correspond seven kinds of fault types (including normal condition) of network；At CNN The process for managing CVPFI is as described below: after a CVPFI sample inputs CNN, by the 1st convolution sum pond, CNN can be extracted CVPFI shallow-layer spatial structure characteristic out, and it is expressed as the characteristic pattern that size is 12 × 12, as shown in P1 layers；Later, pass through Behind 2nd convolution sum pond, CNN has just extracted the depth local spatial feature of CVPFI, as shown in P2 layers in Fig. 2, size It is more abstract for the characteristic pattern ratio P1 in 4 × 4, P2；Finally, the 12 width characteristic patterns obtained behind the 2nd pond be converted into column to Measure and then input full articulamentum F1；A large amount of CVPFI and its label (fault type, including normal condition) are inputted into CNN, CNN Constantly can voluntarily adjust the parameter of network so that when CVPFI is inputted the output of network constantly close to the CVPFI's therefore Hinder type label；When the discrimination of network is sufficiently high, CNN also just has the overall recognition ability to CVPFI, and then can Fault diagnosis is carried out to power grid using the metric data that SCADA under actual state is provided.

Fig. 5 show the embodiment of the present invention based on the electric network failure diagnosis method characteristic of computer vision tidal current chart delete weight after Feature schematic diagram, as shown in figure 5, input CVPFI tidal current chart, check the similitude of each layer characteristic pattern and delete the part of redundancy To optimize network structure, improve training speed, steps are as follows:

(1) the likeness coefficient r between characteristic pattern two-by-two is calculated in C1 layers_c, and corresponding threshold value is set, it is tied according to calculating Fruit can consider f1 and f3, and f1 and f5 and f1 and f6 are highly relevant, therefore f3, f5 and f6 can be deleted from C1 layers. In this way, only retaining three characteristic patterns in C1.

(2) equally to C2 layers of progress similitude inspection.Calculated result illustrates that f3 and f4, f6 and f10 are highly relevant, therefore from F4 and f10 is deleted in C2 layers.

It (3) so far, is all not in the characteristic pattern of redundancy in C1 layers or C2 layers.In this way, 3 characteristic patterns and C2 in C1 layers 10 characteristic patterns in layer constitute the final structure of CNN.It can be seen that in two convolutional layers from the reconstruction result in Fig. 6 There are larger differences each other for characteristic pattern, and the characteristic pattern of redundancy is not present in the CNN after reconstruct.

After being trained initial network using the magnanimity computer visualization tidal current chart training sample generated, need to delete There is redundancy feature figure no longer except redundancy feature figure existing for convolutional layer enables in each convolutional layer of CNN to optimize network structure, because This, when extracting input sample feature during the initial convolution neural network model of training, obtains similar between the feature of extraction Degree, similarity then retains in described two features not less than two features of the similarity threshold threshold of setting if it exists A feature, in actual condition, similarity threshold threshold >=96% is usually set, and value is bigger, the instruction of network It is faster to practice speed, but accuracy of identification can be declined；Conversely, the accuracy of identification of network is higher, but training speed can decrease； It can be examined according to the scale of power grid and the performance of training neural network equipment, comprehensive both sides demand in actual condition Amount.

It is described by the corresponding Matrix Power value of component in corresponding 3 machine, 9 node system figure of the power grid historical trend data The element of predetermined location corresponding with component in 28 × 28 null matrix is replaced, and historical trend data matrix is obtained, can be with Include:

By the Matrix Power value of third transformer in corresponding 3 machine, 9 node system figure of the power grid historical trend data with The element value replacement that the column of eighth row the 20th, the column of eighth row the 21st and eighth row the 22nd arrange in the null matrix；

The random injecting power for first calculating specified generator and load, then according to network system Load flow calculation standard example The active power distribution that power grid is obtained according to the random injecting power of the specified generator and load, particularly, for occurring The case where failure, carry out Load flow calculation when will be deleted the branch to break down in network topology, for corresponding component setting with Machine noise power-value；If the non-failure of branch where component in corresponding 3 machine, 9 node system figure of the power grid historical trend data, Then using the random active power of the component as its corresponding Matrix Power value；

For simulating grid existing harsh electromagnetic interference in actual operation, if the power grid historical trend data are corresponding 3 machine, 9 node system figure in branch trouble where component, then by the random noise power value P of the component_lAs its corresponding square Battle array performance number, wherein determine the random noise power value P of the component as the following formula_l:

P_l=rand [- P_dis,P_dis]

If the component is the second generator or the in corresponding 3 machine, 9 node system figure of the power grid historical trend data Three generators determine its random active power as the following formula:

In formula, P_GiThe injecting power of the second generator or third generator in standard example is calculated for electric power system tide, P_LjThe injecting power of the first load, the second load or third load in standard example, K are calculated for electric power system tide_GiAnd K_LjFor It is preset to be randomly provided parameter, it can be configured according to power grid history scheduling method, rand [] is random number functions；P_TFor Electric power system tide calculates the first generator, the first transformer, the second transformer, third transformer, First Line in standard example Road, the second route, tertiary circuit, the 4th route, the 5th route or the 6th route active power；

Assuming that fault type one shares N_{fault_type}Kind, for a kind of random injecting power, it is contemplated that the whole of system is opened up Structure is flutterred, (N can be generated_{fault_type}+ 1) width CVPFI；Repeatedly generate N_sSecondary random injecting power, so that it may obtain N_s× (N_{fault_type}+ 1) width CVPFI.In this manner it is possible to be carried out using the technology Massive Sample of CVPFI containing noise generated to CNN It is trained.

Specifically, the coefficient of similarity r between feature A and feature B is determined as the following formula_c:

Wherein,A_mnIt is characterized figure A m row the n-th column picture The pixel value of plain block, B_mnIt is characterized the pixel value of figure B m row the n-th column block of pixels, m ∈ [1, N_c1], n ∈ [1, N_c2],For spy The pixel average of sign figure A,The pixel average of characteristic pattern B, N_c1It is characterized total line number of figure, N_c2It is characterized total column of figure Number, N_c1=N_c2。

Fig. 6 shows electric network failure diagnosis method electric network swim number of the embodiment of the present invention based on computer vision tidal current chart According to 3 machine, 9 node system figure, as shown in fig. 6,3 machine, 9 node system may include:

First load is connect with bus 5；Second load is connect with bus 6.

Show that CNN network reconfiguration technology that the invention is proposed effectively raises the training of network by emulation experiment Efficiency；Meanwhile being had using the CNN of Noise CVPFI sample training for the interference signal that usually will appear under complex working condition Good robustness.

Embodiment two

Fig. 7 shows the structural representation of electric network failure diagnosis system of the embodiment of the present invention based on computer vision tidal current chart Figure, as shown in fig. 7, the system may include:

Wherein, the tidal current chart generation module may include:

If the non-failure of branch where component in corresponding 3 machine, 9 node system figure of the current flow data of the power grid, will The active power of the component is as its corresponding Matrix Power value；

P_l=rand [- P_dis,P_dis]

Specifically, the replacement unit, is used for: by corresponding 3 machine, 9 node system figure of the current flow data of the power grid In the first generator Matrix Power value respectively in the null matrix the 21st row the 14th column, the 21st row the 15th column, the 22nd row The element value replacement of 14th column and the 22nd row the 15th column；

Specifically, if the not event of branch where component in corresponding 3 machine, 9 node system figure of the power grid historical trend data Barrier, then using the random active power of the component as its corresponding Matrix Power value；

P_l=rand [- P_dis,P_dis]

Wherein, the coefficient of similarity r between feature A and feature B is determined as the following formula_c:

3 machine, 9 node system may include: the second generator, bus 2, the second transformer, bus 7, the 5th route, Bus 8, third load, bus 9, the 6th route, third transformer, bus 3, third generator, tertiary circuit, bus 5, first Load, first line, bus 4, the first transformer, bus 1, the first generator, the second load, the second route, bus 6 and the 6th Route；

First load is connect with bus 5；Second load is connect with bus 6.

Embodiment three

Magnanimity CVPFI sample needed for statement generates training CNN by taking IEEE9 system as an example, and sent out for finding out in system The process of the transmission line of electricity of raw open circuit fault.

(1) data that obtains Load flow calculation are converted into CVPFI

It, can be by the flow data (wattful power of network by calculating below with the IEEE9 system example explanation of a standard Rate) it is expressed as the form of 3 machine, 9 node diagram, it is exactly " the electric network swim datagram of human viewableization ", but it cannot be directly by CNN It is identified, needs to design certain transformation rule, be translated into the electric network current diagram (CVPFI) of computer vision.

Convert 3 machine, 9 node system figure of human viewableization in the electric network swim matrix of computer vision.Rule is such as Under:

1) is 28*28 pixel in view of CVPFI size is arranged in the scale of network, i.e. the line number of CVPFI matrix and columns is equal It is 28.

2) load and transformer are plotted as 2*2 size, and line and transformer is plotted as 1*3 size, in CVPFI square Position in battle array is approximate with its position in figure one, specifically:

(21,14), (21,15), (22,14) and (22,15) indicate generator G in matrix₁；(8,4) in matrix, (8,5), (9,4) and (9,5) indicate generator G₂；(8,24), (8,25), (9,24) and (9,25) indicate generator G in matrix₃。

(13,7), (13,8), (14,7) and (14,8) indicate load L in matrix₁；(13,21) in matrix, (13,22), (14,21) and (14,22) indicate load L₂；(9,14), (9,15), (10,14) and (10,15) indicate load L in matrix₃。

(13,10), (14,10) and (15,10) indicate route 4-5 in matrix；(13,19) in matrix, (14,19) and (15, 19) route 4-6 is indicated；(9,10), (10,10) and (11,10) indicate route 5-7 in matrix；(9,19), (10,19) in matrix (11,19) route 6-9 is indicated；(8,11), (8,12) and (8,13) indicate route 7-8 in matrix；(8,16) in matrix, (8, 17) and (8,18) indicate route 8-9.

(17,14), (18,14) and (19,14) indication transformer T in matrix₁(route 1-4)；(8,7), (8,8) in matrix (8,9) indication transformer T₂(route 2-7)；(8,20), (8,21) and (8,22) indication transformer T in matrix₃(route 3-9).

3) a shared 28*28=784 position in .CVPFI matrix, in the range of -1~1.Generator, load, power transmission line The numerical value of road and transformer in a matrix is provided that

For generator and load, numerical value, which is positive, indicates it to system injecting power, and the expression system that is negative injects function to it Rate, the watt level that the absolute value representation system of numerical value is injected or injected by system.

For transmission line of electricity and transformer, numerical value, which is positive, indicates that the flow direction of power is from the lesser bus flow direction of number Biggish bus is numbered, being negative indicates that the flow direction of power is to count from the lesser bus of biggish bus flow direction number is numbered The size of the absolute value representation active power of value.

4) element value of all white spaces of is set as zero.

By above four steps corresponding trend can be converted by the electric network swim data of 3 machine, 9 node system figure Data matrix can be translated into CVPFI tidal current chart from tape function imagesc using MATLAB, and such tidal current chart can To input CNN, and then it is identified by it.

(2) random network initial condition is obtained based on IEEE9 system parameter

1) is for PV node (generator 2, generator 3):

A. the active power of injected system is set as random value.In formula, P_GiIt is generator in IEEE9 Load flow calculation standard example G_iActive power, k takes a random number between 0 and 1:

B. the voltage magnitude of node is constant, is still the value in IEEE9 Load flow calculation standard example.

2) is for PQ node (load 1, load 2 and load 3):

The active power that system injects it is set as random value.In formula, P_LiIt is load L in IEEE9 Load flow calculation standard example_i Active power, k takes a random number between 0 and 1:

The reactive power that system injects it is set as random value.In formula, Q_LiIt is load L in IEEE9 Load flow calculation standard example_i Reactive power, k takes a random number between 0 and 1:

3) does not change the parameter of SL node (generator 1).

4) does not change the parameter of line and transformer.

IEEE9 Load flow calculation standard example (can be based on) by above four rule and obtain one group of random initial strip Part.

(3) active power under heterogeneous networks state is obtained according to the primary condition of network to be distributed

Here heterogeneous networks state refers to different fault conditions: fault-free, route 4-5 are breaking, route 4-6 is breaking, Route 5-7 open circuit, route 6-9 open circuit, route 7-8 open circuit and route 8-9 open circuit, seven kinds altogether, transformer open circuit is ignored.

For non-failure conditions: carrying out Load flow calculation active power point according to the Random Initial Condition of network and topological structure Cloth can be obtained required result.

The case where for breaking down, then the explanation for needing to modify to the topological structure of network, below is with route 4-5 For open circuit.

Fig. 8 shows 3 machines 9 when the non-failure of electric network failure diagnosis method of the embodiment of the present invention by taking IEEE9 system as an example Node system figure, as shown in figure 8, topological structure of electric is completely when IEEE9 system does not break down, route 4-5 normally connects It connects, Fig. 9 shows 3 machine, 9 node system when electric network failure diagnosis method failure of the embodiment of the present invention by taking IEEE9 system as an example Figure, as shown in figure 9, needing to leave out route 4-5 before carrying out Load flow calculation, then according to the network topology structure after leaving out The active power distribution for calculating current primary condition lower network, does not include the wattful power on route 4-5 naturally, in calculated result Rate.The active power for enabling route 4-5 is a random value:

P_4-5=rand [- P_dis,P_dis] (4)

In formula, P_disIt can be set to the maximum value of the branch measurement equipment noise, for simplification, be set as 1, i.e. P here_4-5 It is a random number between -1 and 1.

In this way, the active power on route 4-5 that formula 4 generates at random is added to topological structure meter as shown in Figure 5 Calculate in resulting result, can be obtained one it is complete, when route 4-5 breaks down, in view of measurement equipment is done by noise Network active power distribution when disturbing.Process when All other routes break down is identical.

It should be understood by those skilled in the art that, embodiments herein can provide as method, system or computer program Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the application Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the application, which can be used in one or more, The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces The form of product.

The application is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present application Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates, Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or The function of being specified in multiple boxes.

These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one The step of function of being specified in a box or multiple boxes.

Finally it should be noted that: the above embodiments are merely illustrative of the technical scheme of the present invention and are not intended to be limiting thereof, to the greatest extent Invention is explained in detail referring to above-described embodiment for pipe, it should be understood by those ordinary skilled in the art that: still It can be with modifications or equivalent substitutions are made to specific embodiments of the invention, and without departing from any of spirit and scope of the invention Modification or equivalent replacement, should all cover within the scope of the claims of the present invention.

Claims

1. a kind of electric network failure diagnosis method based on computer vision tidal current chart, which is characterized in that the described method includes:

According to the current computer vision tidal current chart of the power grid, using the electric network failure diagnosis model being pre-designed to power grid into Row fault diagnosis；

Wherein, the electric network failure diagnosis model being pre-designed utilizes power grid historical trend data and its corresponding fault type It is obtained as training data.

2. the method as described in claim 1, which is characterized in that it is current that the flow data current according to power grid generates power grid Computer vision tidal current chart, comprising:

By the corresponding Matrix Power value of component in corresponding 3 machine, 9 node system figure of the current flow data of the power grid with 28 × The element replacement of the corresponding predetermined location of component, obtains current flow data matrix in 28 null matrix；

It by the current flow data matrix conversion is its corresponding computer vision trend by the images function of MATLAB Figure.

3. method according to claim 2, which is characterized in that if corresponding 3 machine, 9 node of flow data that the power grid is current The non-failure of branch where component in system diagram, then using the active power of the component as its corresponding Matrix Power value；

If branch trouble where component in corresponding 3 machine, 9 node system figure of the current flow data of the power grid, by the component Random noise power value P_lAs its corresponding Matrix Power value, wherein determine the random noise power of the component as the following formula Value P_l:

P_l=rand [- P_dis,P_dis]

4. method according to claim 2, which is characterized in that corresponding 3 machine 9 of the flow data that the power grid is current The member of the corresponding Matrix Power value of component predetermined location corresponding with component in 28 × 28 null matrix in node system figure Element replacement, obtains current flow data matrix, comprising:

By the Matrix Power value difference of the first generator in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The element value arranged with the 21st row the 14th column, the 21st row the 15th column, the 22nd row the 14th column and the 22nd row the 15th in the null matrix Replacement；

By the Matrix Power value of the second generator and institute in corresponding 3 machine, 9 node system figure of the current flow data of the power grid State the element value replacement of the column of eighth row the 4th in null matrix, the column of eighth row the 5th, the 9th row the 4th column and the 9th row the 5th column；

By the Matrix Power value of third generator and institute in corresponding 3 machine, 9 node system figure of the current flow data of the power grid State the element value replacement of the column of eighth row the 24th in null matrix, the column of eighth row the 25th, the 9th row the 24th column and the 9th row the 25th column；

By in corresponding 3 machine, 9 node system figure of the current flow data of the power grid Matrix Power value of the first load it is opposite The element value replacement of 13rd row the 7th column, the 13rd row the 8th column, the 14th row the 7th column and the 14th row the 8th column in the several and null matrix；

By in corresponding 3 machine, 9 node system figure of the current flow data of the power grid Matrix Power value of the second load it is opposite Several element values with the 13rd row the 21st column, the 13rd row the 22nd column, the 14th row the 21st column and the 14th row the 22nd column in the null matrix Replacement；

By in corresponding 3 machine, 9 node system figure of the current flow data of the power grid Matrix Power value of third load it is opposite Number is replaced with the element value that the 9th row the 14th column, the 9th row the 15th column, the 10th row the 14th column and the 10th row the 15th arrange in the null matrix It changes；

By the Matrix Power value of first line in corresponding 3 machine, 9 node system figure of the current flow data of the power grid with it is described The element value replacement of 13rd row the 10th column, the 14th row the 10th column and the 15th row the 10th column in null matrix；

By the Matrix Power value of the second route in corresponding 3 machine, 9 node system figure of the current flow data of the power grid with it is described The element value replacement of 13rd row the 19th column, the 14th row the 19th column and the 15th row the 19th column in null matrix；

By in corresponding 3 machine, 9 node system figure of the current flow data of the power grid Matrix Power value of tertiary circuit it is opposite Several element values with the 9th row the 10th column, the 10th row the 10th column and the 11st row the 10th column in the null matrix are replaced；

By in corresponding 3 machine, 9 node system figure of the current flow data of the power grid Matrix Power value of the 4th route it is opposite Several element values with the 9th row the 19th column, the 10th row the 19th column and the 11st row the 19th column in the null matrix are replaced；

By the Matrix Power value of the 5th route in corresponding 3 machine, 9 node system figure of the current flow data of the power grid with it is described The element value replacement that the column of eighth row the 11st, the column of eighth row the 12nd and eighth row the 13rd arrange in null matrix；

By in corresponding 3 machine, 9 node system figure of the current flow data of the power grid Matrix Power value of the 6th route it is opposite The element value replacement that the column of eighth row the 16th, the column of eighth row the 17th and eighth row the 18th arrange in the number null matrix；

By the Matrix Power value of the first transformer and institute in corresponding 3 machine, 9 node system figure of the current flow data of the power grid State the element value replacement of the 17th row the 14th column, the 18th row the 14th column and the 19th row the 14th column in null matrix；

By the Matrix Power value of the second transformer and institute in corresponding 3 machine, 9 node system figure of the current flow data of the power grid State the column of eighth row the 7th in null matrix, the element value replacement that the column of eighth row the 8th and eighth row the 9th arrange；

By the Matrix Power value of third transformer and institute in corresponding 3 machine, 9 node system figure of the current flow data of the power grid State the column of eighth row the 20th in null matrix, the element value replacement that the column of eighth row the 21st and eighth row the 22nd arrange.

5. the method as described in claim 1, which is characterized in that utilize power grid historical trend data and its corresponding fault type As training data obtain described in the process of electric network failure diagnosis model that is pre-designed include:

By the corresponding Matrix Power value of component and 28 × 28 in corresponding 3 machine, 9 node system figure of the power grid historical trend data Null matrix in the corresponding predetermined location of component element replacement, obtain historical trend data matrix；

The historical trend data matrix is converted into its corresponding computer vision trend by the images function of MATLAB Pattern sheet；

Using the computer vision tidal current chart sample as the input of initial convolution neural network model, the computer vision tide The corresponding fault type of flow graph sample is trained as the output of initial convolution neural network model, is pre-designed described in acquisition Electric network failure diagnosis model；

Wherein, when extracting input sample feature during the initial convolution neural network model of training, between the feature for obtaining extraction Similarity, if it exists similarity not less than setting similarity threshold threshold two features, then retain described two A feature in feature.

6. method as claimed in claim 5, which is characterized in that described by corresponding 3 machine, 9 section of the power grid historical trend data The element of the corresponding Matrix Power value of component predetermined location corresponding with component in 28 × 28 null matrix in dot system figure Replacement obtains historical trend data matrix, comprising:

By the Matrix Power value of the first generator in corresponding 3 machine, 9 node system figure of the power grid historical trend data respectively with The element value of the 21st row the 14th column, the 21st row the 15th column, the 22nd row the 14th column and the 22nd row the 15th column in the null matrix replaces It changes；

By the Matrix Power value of the second generator in corresponding 3 machine, 9 node system figure of the power grid historical trend data with it is described The element value replacement of the column of eighth row the 4th, the column of eighth row the 5th, the 9th row the 4th column and the 9th row the 5th column in null matrix；

By the Matrix Power value of third generator in corresponding 3 machine, 9 node system figure of the power grid historical trend data with it is described The element value replacement of the column of eighth row the 24th, the column of eighth row the 25th, the 9th row the 24th column and the 9th row the 25th column in null matrix；

By the opposite number of the Matrix Power value of the first load in corresponding 3 machine, 9 node system figure of the power grid historical trend data With the element value replacement of the 13rd row the 7th column, the 13rd row the 8th column, the 14th row the 7th column and the 14th row the 8th column in the null matrix；

By the opposite number of the Matrix Power value of the second load in corresponding 3 machine, 9 node system figure of the power grid historical trend data It is replaced with the 13rd row the 21st column, the 13rd row the 22nd column, the element value that the 14th row the 21st arranges and the 14th row the 22nd arranges in the null matrix It changes；

By the opposite number of the Matrix Power value of third load in corresponding 3 machine, 9 node system figure of the power grid historical trend data With the element value replacement of the 9th row the 14th column, the 9th row the 15th column, the 10th row the 14th column and the 10th row the 15th column in the null matrix；

By the Matrix Power value and described zero of first line in corresponding 3 machine, 9 node system figure of the power grid historical trend data The element value replacement of 13rd row the 10th column, the 14th row the 10th column and the 15th row the 10th column in matrix；

By the Matrix Power value and described zero of the second route in corresponding 3 machine, 9 node system figure of the power grid historical trend data The element value replacement of 13rd row the 19th column, the 14th row the 19th column and the 15th row the 19th column in matrix；

By the opposite number of the Matrix Power value of tertiary circuit in corresponding 3 machine, 9 node system figure of the power grid historical trend data It is replaced with the element value of the 9th row the 10th column, the 10th row the 10th column and the 11st row the 10th column in the null matrix；

By the opposite number of the Matrix Power value of the 4th route in corresponding 3 machine, 9 node system figure of the power grid historical trend data It is replaced with the element value of the 9th row the 19th column, the 10th row the 19th column and the 11st row the 19th column in the null matrix；

By the Matrix Power value and described zero of the 5th route in corresponding 3 machine, 9 node system figure of the power grid historical trend data The element value replacement that the column of eighth row the 11st, the column of eighth row the 12nd and eighth row the 13rd arrange in matrix；

By the opposite number of the Matrix Power value of the 6th route in corresponding 3 machine, 9 node system figure of the power grid historical trend data The element value replacement that the column of eighth row the 16th, the column of eighth row the 17th and eighth row the 18th arrange in the null matrix；

By the Matrix Power value of the first transformer in corresponding 3 machine, 9 node system figure of the power grid historical trend data with it is described The element value replacement of 17th row the 14th column, the 18th row the 14th column and the 19th row the 14th column in null matrix；

By the Matrix Power value of the second transformer in corresponding 3 machine, 9 node system figure of the power grid historical trend data with it is described The element value replacement that the column of eighth row the 7th, the column of eighth row the 8th and eighth row the 9th arrange in null matrix；

By the Matrix Power value of third transformer in corresponding 3 machine, 9 node system figure of the power grid historical trend data with it is described The element value replacement that the column of eighth row the 20th, the column of eighth row the 21st and eighth row the 22nd arrange in null matrix.

7. method as claimed in claim 6, which is characterized in that if corresponding 3 machine, the 9 node system of the power grid historical trend data The non-failure of branch where component in system figure, then using the random active power of the component as its corresponding Matrix Power value；

If branch trouble where component in corresponding 3 machine, 9 node system figure of the power grid historical trend data, by the component Random noise power value P_lAs its corresponding Matrix Power value, wherein determine the random noise power value of the component as the following formula P_l:

P_l=rand [- P_dis,P_dis]

8. the method for claim 7, which is characterized in that if the component is that the power grid historical trend data are corresponding Second generator or third generator in 3 machine, 9 node system figure, determine its random active power as the following formula:

If the component is the first load in current corresponding 3 machine, 9 node system figure of flow data of the power grid, second negative Lotus or third load determine its random active power as the following formula:

If the component is the first generator in current corresponding 3 machine, 9 node system figure of flow data of the power grid, first Transformer, the second transformer, third transformer, first line, the second route, tertiary circuit, the 4th route, the 5th route or Six routes then determine its random active power as the following formula

In formula, P_GiThe injecting power of the second generator or third generator in standard example, P are calculated for electric power system tide_LjFor Electric power system tide calculates the injecting power of the first load, the second load or third load in standard example, K_GiAnd K_LjIt is preparatory Setting is randomly provided parameter, and rand [] is random number functions；P_TThe first power generation in standard example is calculated for electric power system tide Machine, the first transformer, the second transformer, third transformer, first line, the second route, tertiary circuit, the 4th route, the 5th The active power of route or the 6th route.

9. method as claimed in claim 5, which is characterized in that determine the coefficient of similarity r between feature A and feature B as the following formula_c:

Wherein,A_mnIt is characterized figure A m row the n-th column block of pixels Pixel value, B_mnIt is characterized the pixel value of figure B m row the n-th column block of pixels, m ∈ [1, N_c1], n ∈ [1, N_c2],It is characterized figure A's Pixel average,It is characterized the pixel average of figure B, N_c1It is characterized total line number of figure, N_c2It is characterized total columns of figure, N_c1 =N_c2。

10. the method as described in claim 2 or 5, which is characterized in that 3 machine, 9 node system includes:

Second generator, bus 2, the second transformer, bus 7, the 5th route, bus 8, third load, bus 9, the 6th route, Third transformer, bus 3, third generator, tertiary circuit, bus 5, the first load, first line, bus 4, the first transformation Device, bus 1, the first generator, the second load, the second route, bus 6 and the 6th route；

Second generator, bus 2, the second transformer, bus 7, the 5th route, bus 8, third load, bus the 9, the 6th Route, third transformer, bus 3 and third generator are sequentially connected；

5th route, tertiary circuit, bus 5, first line, bus 4, the first transformer, bus 1 and the first generator according to Secondary connection；

6th route, the 4th route, bus 6, the second route, bus 4, the first transformer, bus 1 and the first generator according to Secondary connection；

First load is connect with bus 5；Second load is connect with bus 6.

11. a kind of electric network failure diagnosis system based on computer vision tidal current chart, which is characterized in that the system comprises:

Tidal current chart generation module, for generating the current computer vision tidal current chart of power grid according to the current flow data of power grid；

Diagnostic module is examined for the computer vision tidal current chart current according to the power grid using the electric network fault being pre-designed Disconnected model carries out fault diagnosis to power grid；

12. system as claimed in claim 11, which is characterized in that the tidal current chart generation module, comprising:

Replacement unit, for by the corresponding matrix of component in corresponding 3 machine, 9 node system figure of the current flow data of the power grid The element of performance number predetermined location corresponding with component in 28 × 28 null matrix is replaced, and current flow data square is obtained Battle array；

Converting unit, by by the images function of MATLAB will the current flow data matrix conversion for its it is corresponding based on Calculation machine vision tidal current chart.

13. system as claimed in claim 12, which is characterized in that if corresponding 3 machine, 9 section of flow data that the power grid is current The non-failure of branch where component in dot system figure, then using the active power of the component as its corresponding Matrix Power value；

P_l=rand [- P_dis,P_dis]

14. system as claimed in claim 12, which is characterized in that the replacement unit is used for:

15. system as claimed in claim 11, which is characterized in that utilize power grid historical trend data and its corresponding failure classes Type as training data obtain described in the process of electric network failure diagnosis model that is pre-designed include:

16. system as claimed in claim 15, which is characterized in that described by corresponding 3 machine 9 of the power grid historical trend data The member of the corresponding Matrix Power value of component predetermined location corresponding with component in 28 × 28 null matrix in node system figure Element replacement, obtains historical trend data matrix, comprising:

17. system as claimed in claim 16, which is characterized in that if corresponding 3 machine, 9 node of the power grid historical trend data The non-failure of branch where component in system diagram, then using the random active power of the component as its corresponding Matrix Power value；

P_l=rand [- P_dis,P_dis]

18. system as claimed in claim 17, which is characterized in that if the component is corresponding for the power grid historical trend data 3 machine, 9 node system figure in the second generator or third generator, determine its random active power as the following formula:

19. system as claimed in claim 15, which is characterized in that determine the coefficient of similarity between feature A and feature B as the following formula r_c:

20. the system as described in claim 12 or 15, which is characterized in that 3 machine, 9 node system includes:

First load is connect with bus 5；Second load is connect with bus 6.