CN113554010A - Power grid line fault recognition model training method - Google Patents
Power grid line fault recognition model training method Download PDFInfo
- Publication number
- CN113554010A CN113554010A CN202111103885.4A CN202111103885A CN113554010A CN 113554010 A CN113554010 A CN 113554010A CN 202111103885 A CN202111103885 A CN 202111103885A CN 113554010 A CN113554010 A CN 113554010A
- Authority
- CN
- China
- Prior art keywords
- fault
- graph
- line
- target line
- identification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012549 training Methods 0.000 title claims abstract description 39
- 238000013528 artificial neural network Methods 0.000 claims abstract description 25
- 238000010586 diagram Methods 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 76
- 238000012544 monitoring process Methods 0.000 claims description 67
- 230000008859 change Effects 0.000 claims description 17
- 238000004088 simulation Methods 0.000 claims description 14
- 230000005540 biological transmission Effects 0.000 claims description 10
- 238000007621 cluster analysis Methods 0.000 claims description 9
- 230000010354 integration Effects 0.000 claims description 6
- 230000001537 neural effect Effects 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000012512 characterization method Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000012545 processing Methods 0.000 abstract description 7
- 230000004927 fusion Effects 0.000 abstract description 5
- 238000003745 diagnosis Methods 0.000 description 9
- 238000004422 calculation algorithm Methods 0.000 description 4
- 239000002699 waste material Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses a power grid line fault recognition model training method, which comprises the following steps: s1, collecting the electrical parameters of the target line, and converting the electrical parameters of the target line into a fault recognition diagram of the target line; step S2, carrying out image recognition training on the deep neural network based on the fault recognition graph to obtain a line fault recognition model; and step S3, updating the line fault identification model by analyzing and updating the fault identification diagram according to the feature similarity. The invention clusters the electrical parameters with the same attribute, performs intra-cluster fusion to quantize the discrete electrical parameters into three identification parameters consistent with the R pixel value, the G pixel value and the B pixel value of the graph, and constructs a fault identification graph consistent with the deep neural network adept processing field by using the three identification parameters, thereby improving the identification matching degree of the deep neural network on the electric power system fault based on the electrical parameters and improving the fault identification precision.
Description
Technical Field
The invention relates to the technical field of fault recognition, in particular to a power grid line fault recognition model training method.
Background
The fault identification is used as an important link of fault diagnosis of the power system, and the establishment of the power grid data acquisition system and the fault information system can provide event information and wave recording data during the fault period, thereby laying a data foundation for the application of an artificial intelligent algorithm. Related researchers have proposed that the traditional artificial intelligence algorithm is applied to fault type identification, however, the traditional shallow learning algorithm comprises two parts, namely a characteristic parameter extraction link and a classification identification link, the identification effect of the traditional shallow learning algorithm depends on the artificially designed characteristic parameter extraction link to a great extent, and the quality of characteristic parameter selection can directly influence the identification effect of the whole model. Deep learning does not depend on a manual design feature extraction link, a deep neural network is constructed through mass data training, input data features are automatically extracted, induction and classification are carried out, and the method has great application potential in the aspect of fault type identification.
In the prior art, the adept scene of a deep neural network for fault diagnosis is image processing identification, the matching degree is easy to reduce when the deep neural network is directly applied to an electric parameter, so that the fault diagnosis precision is reduced, a fault identification model is updated regularly to adapt to the topological change of a target line, in practice, the topology of the target line does not occur all the time, and due to the fact that the target line quantity and the characteristic data quantity are huge, huge operation pressure can be caused at the time of model updating, if the target line is updated indiscriminately, only the waste of operation resources and the prolonging of operation time can be caused, and the operation efficiency of diagnosis is reduced.
Disclosure of Invention
The invention aims to provide a power grid line fault recognition model training method, which aims to solve the technical problems that in the prior art, the adept scene of a deep neural network used for fault diagnosis is image processing recognition, the matching degree is easy to reduce when the deep neural network is directly applied to power parameters, the fault diagnosis precision is reduced, and the indiscriminate update of a target line only causes the waste of operation resources and the extension of operation time, so that the operation efficiency of diagnosis is reduced.
In order to solve the technical problems, the invention specifically provides the following technical scheme:
a power grid line fault recognition model training method comprises the following steps:
s1, collecting the electrical parameters of the target line, and converting the electrical parameters of the target line into a fault recognition diagram of the target line;
step S2, carrying out image recognition training on the deep neural network based on the fault recognition graph to obtain a line fault recognition model;
and step S3, updating the line fault identification model by analyzing and updating the fault identification diagram according to the feature similarity.
As a preferable aspect of the present invention, in step S1, the specific method for converting the electrical parameter of the target line into the fault identification map of the target line includes:
step S101, performing cluster analysis on the electrical parameters to divide all the electrical parameters into three parameter clusters;
s102, fusing and normalizing all electrical parameters in each parameter cluster to integrate the three parameter clusters into three identification parameters;
step S103, converting the three identification parameters into R pixel values, G pixel values and B pixel values of the graphs respectively, and taking the line number of the target line and the acquisition time sequence of the three identification parameters of the target line as graph coordinate points of the R pixel values, the G pixel values and the B pixel values of the graphs to form the fault identification graph.
As a preferable aspect of the present invention, in step S101, the specific method for performing cluster analysis on the electrical parameter includes:
sequentially converging and calibrating the electrical parameters of all target lines according to the same category to be a parameter vector set
Wherein
characterized by the ith parametric vector term,
characterized by the ith electrical parameter of the target line n;
selecting three vector items with the maximum difference degree from the parameter vector set I as three cluster centers, and carrying out K-means cluster analysis on the parameter vector set I based on the three cluster centers to obtain three parameter clusters, wherein the three parameter clusters are respectively
,
,
Wherein
，
，
、
、
respectively characterized as a parametric cluster
、
、
To (1) a
、
、
The term of the parameter vector is defined as,
、
、
respectively characterized as the second of the target line n
、
、
An electrical parameter.
As a preferred embodiment of the present invention, in step S102, a specific method for integrating three parameter clusters into three identified parameters includes:
converting the parametric vector form in the parametric cluster into the electrical parametric vector form of the target line
Wherein
，
，
，
，
、
、
characterised by the target line k according to said parametric cluster
、
、
The electrical parameter arranged in a cluster format,
the representation is the final value of the line number of the target line;
transforming the electric parameter vector form of the target line
According to the parameter cluster
、
、
Performing intra-cluster linear integration and converting the intra-cluster linear integration into an identification parameter vector form of a target line, wherein the identification parameter vector form of the target line is as follows:
wherein,
characterized by the fact that the product operator is a product operator,
、
、
respectively characterized as the second of the target line k
、
、
An electrical parameter.
In a preferred embodiment of the present invention, the step S103 identifies the parameters
Performing normalization treatment to [0,255%]The value ranges are respectively used as R pixel value, G pixel value and B pixel value of the graph.
As a preferable aspect of the present invention, in step S2, the specific method for training the line fault recognition model includes:
step S201, model training samples in a fault recognition graph format are obtained and converted from a power grid operation log, wherein positive labels are given to the fault recognition graph samples corresponding to normal power grid operation, and negative labels are given to the fault recognition graph samples corresponding to abnormal power grid operation;
step S202, preprocessing the fault recognition pattern book into three-channel pattern vectors, using the pattern vectors as the input of a deep neural model, using the labels of the fault recognition pattern samples as the output of the deep neural model to perform image recognition training, and realizing learning of a line fault recognition model for judging the operation condition of the power grid.
As a preferable aspect of the present invention, in step S3, the specific method for updating the line fault identification model includes:
setting a monitoring interval, and carrying out similarity monitoring on a fault identification graph of a target line once after the fault identification graph passes through the monitoring interval, wherein the similarity monitoring is used for monitoring the topology change attribute of the target line;
setting the overall similarity between the fault identification graph of the monitored target line and the fault identification graph of the target line before monitoring as a monitoring coefficient, wherein the operational formula of the monitoring coefficient is as follows:
wherein,
the characterization is that the listening coefficient is,
total number of graph vectors characterized as fault identification graph
、
Respectively characterized as post-snoop and pre-snoop fault identificationThe jth graph vector of the graph;
as a preferable aspect of the present invention, in step S3, the specific method for updating the line fault identification model further includes:
setting a monitoring threshold, and comparing the monitoring coefficient with the monitoring threshold, specifically:
if the monitoring coefficient is higher than the monitoring threshold value, the line fault identification model before monitoring does not need to be updated;
if the monitoring coefficient is lower than the monitoring threshold, the line fault identification model before monitoring needs to be updated, specifically:
calculating the single similarity of each graph vector in the fault identification graph of the monitored target line and each graph vector in the fault identification graph of the target line before monitoring, and selecting all graph vectors with the single similarity higher than a monitoring threshold value from the fault identification graph of the monitored target line as a graph vector updating chain, wherein the calculation formula of the single similarity is as follows:
wherein,
the j graph vector in the fault identification graph characterized as the monitored target line
The jth graph vector in the fault identification graph of the target line before monitoring
Similarity of single items between the two;
and replacing the graph vector updating chain to the corresponding graph vector in the fault recognition graph of the target line before monitoring, completing updating of the fault recognition graph adapting to the topological variation attribute of the target line, bringing the updated fault recognition graph into the line fault recognition model for training, and completing updating of the line fault recognition model.
As a preferable scheme of the invention, the line fault identification model takes the identification rate and the training time as optimization targets to determine the optimal parameters of the deep neural network.
As a preferred scheme of the present invention, the power grid operation log is generated by a power transmission line fault simulation model to generate fault time sequence data of a target line, the topology variation attribute of the target line is generated by the power transmission line fault simulation model, and the electrical parameter of the target line is generated by the power transmission line fault simulation model.
Compared with the prior art, the invention has the following beneficial effects:
the invention clusters the electrical parameters with the same attribute, performs intra-cluster fusion to quantize the discrete electrical parameters into three identification parameters consistent with the R pixel value, the G pixel value and the B pixel value of the graph, constructs a fault identification graph consistent with the deep neural network adept processing field by using the three identification parameters, can improve the identification matching degree of the deep neural network to the power system fault based on the electrical parameters, improves the fault identification precision, and by setting the monitoring coefficient, the topology change information of the target line is obtained, the fault identification model is updated for the target line with topology change, the fault identification model is not required to be updated for the target line without topology change, the original fault identification model is used, the method can avoid the waste of operation resources and the extension of operation time caused by indiscriminate update of the target line, and improve the operation efficiency of diagnosis.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It should be apparent that the drawings in the following description are merely exemplary, and that other embodiments can be derived from the drawings provided by those of ordinary skill in the art without inventive effort.
Fig. 1 is a flowchart of a power grid line fault recognition model training method provided in an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the invention provides a power grid line fault recognition model training method, which includes the following steps:
the deep neural network is mainly used for recognition models of power grid lines, but the deep neural network is adept to process images in the field of processing, namely, the input is a graph vector of the images, the output is a label of the images, if the electrical parameters of the power grid lines are directly used as the input of the deep neural network, the input items are not matched with the deep neural network, and the precision of the trained fault recognition model is reduced, so that the embodiment provides a method for converting the electrical parameters into a graph format, the input of the deep neural network is matched, the characteristic processing advantages of the deep neural network are fully exerted, and the specific steps of converting the electrical parameters into the graph format are as follows:
s1, collecting the electrical parameters of the target line, and converting the electrical parameters of the target line into a fault recognition diagram of the target line;
in step S1, the specific method for converting the electrical parameter of the target line into the fault identification map of the target line includes:
step S101, performing cluster analysis on the electrical parameters to divide all the electrical parameters into three parameter clusters;
in step S101, the specific method for performing cluster analysis on the electrical parameter includes:
sequentially converging and calibrating the electrical parameters of all target lines according to the same category to be a parameter vector set
Wherein
characterized by the ith parametric vector term,
characterized by the ith electrical parameter of the target line n;
selecting three vector items with the maximum difference degree from the parameter vector set I as three cluster centers, and carrying out K-means cluster analysis on the parameter vector set I based on the three cluster centers to obtain three parameter clusters which are respectively
, 
,
Wherein,
，
，
、
、
respectively characterized as a parametric cluster
、
、
To (1) a
、
、
The term of the parameter vector is defined as,
、
、
respectively characterized as the second of the target line n
、
、
An electrical parameter.
Specifically, in order to increase the recognition accuracy of the fault recognition model, the number of the required electrical parameters is increased, and the electrical parameters with large number are converted into a graph format, the number of the electrical parameters needs to be compressed to three, so that the electrical parameters are consistent with R, G, B pixels of the graph.
S102, fusing and normalizing all electrical parameters in each parameter cluster to integrate the three parameter clusters into three identification parameters;
in step S102, a specific method for integrating the three parameter clusters into three identification parameters includes:
converting the parametric vector form in the parametric cluster into the electrical parametric vector form of the target line
Wherein
，
，
，
，
、
、
characterized in that the target line k is clustered according to parameters
、
、
The electrical parameter arranged in a cluster format,
the representation is the final value of the line number of the target line;
the vector form of the electric parameter of the target line
According to parameter cluster
、
、
And performing intra-cluster linear integration to convert the intra-cluster linear integration into an identification parameter vector form of the target line, wherein the identification parameter vector form of the target line is as follows:
wherein,
characterized by the fact that the product operator is a product operator,
、
、
respectively characterized as the second of the target line k
、
、
An electrical parameter.
Specifically, all the electrical parameters in the same parameter cluster are linearly multiplied to unify into a single identification parameter, so that the electrical parameter vector display of the target line is compressed into three fusion items by a plurality of discrete items, namely, the three fusion items matched into the R, G, B pixel form are formed.
And S103, converting the three identification parameters into R pixel values, G pixel values and B pixel values of the graphs respectively, and taking the line number of the target line and the acquisition time sequence of the three identification parameters of the target line as graph coordinate points of the R pixel values, the G pixel values and the B pixel values of the graphs to form a fault identification graph.
Specifically, a graph coordinate system is constructed by using a line number as a vertical coordinate and a time sequence as a horizontal coordinate, a target line is determined as a graph pixel point by using the line number of each target line and the acquisition time sequence of three identification parameters of the target line, the three identification parameters of the target line are respectively converted into a graph R pixel value, a graph G pixel value and a graph B pixel value of the corresponding graph pixel point, and finally the parameters of the target line are converted into a graph format to be used as the input of the deep neural network.
Step S103, identifying the parameters
Performing normalization treatment to [0,255%]The value ranges are respectively used as R pixel value, G pixel value and B pixel value of the graph.
Step S2, carrying out image recognition training on the deep neural network based on the fault recognition graph to obtain a line fault recognition model;
in step S2, the specific method for training the line fault recognition model includes:
step S201, model training samples in a fault recognition graph format are obtained and converted from a power grid operation log, wherein positive labels are given to the fault recognition graph samples corresponding to normal power grid operation, and negative labels are given to the fault recognition graph samples corresponding to abnormal power grid operation;
step S202, preprocessing the fault recognition image sample into a three-channel image vector, using the image vector as the input of the deep neural model, using the label of the fault recognition image sample as the output of the deep neural model, and performing image recognition training to realize learning of the line fault recognition model for judging the operation state of the power grid.
And step S3, updating the line fault identification model by analyzing and updating the fault identification diagram according to the feature similarity.
In step S3, the specific method for updating the line fault identification model includes:
setting a monitoring interval, and carrying out similarity monitoring on the fault identification graph of the target line once after every monitoring interval, wherein the similarity monitoring is used for monitoring the topology change attribute of the target line;
setting the overall similarity between the fault identification graph of the monitored target line and the fault identification graph of the monitored target line as a monitoring coefficient, wherein the operational formula of the monitoring coefficient is as follows:
wherein,
the characterization is that the listening coefficient is,
the total number of pattern vectors characterized as a fault identification pattern,
、
j-th graph vectors respectively characterized as a post-monitoring and pre-monitoring fault identification graph;
in step S3, the specific method for updating the line fault identification model further includes:
setting a monitoring threshold value, and comparing a monitoring coefficient with the monitoring threshold value, specifically:
if the monitoring coefficient is higher than the monitoring threshold value, the line fault identification model before monitoring does not need to be updated;
if the monitoring coefficient is lower than the monitoring threshold, the line fault identification model before monitoring needs to be updated, specifically:
calculating the single similarity of each graph vector in the fault identification graph of the monitored target line and each graph vector in the fault identification graph of the target line before monitoring, and selecting all graph vectors with the single similarity higher than a monitoring threshold value from the fault identification graph of the monitored target line as a graph vector updating chain, wherein the calculation formula of the single similarity is as follows:
wherein,
the j graph vector in the fault identification graph characterized as the monitored target line
The jth graph vector in the fault identification graph of the target line before monitoring
Similarity of single items between the two;
and replacing the graph vector updating chain to the corresponding graph vector in the fault recognition graph of the target line before monitoring, completing updating of the fault recognition graph which is adaptive to the topological variation attribute of the target line, and bringing the updated fault recognition graph into a line fault recognition model for training to complete updating of the line fault recognition model.
Setting a monitoring coefficient, rapidly identifying whether the target line has topology change, if the target line has topology change by triggering an updating mechanism, representing that the graph vector changes on a fault identification graph, changing and replacing the graph vector with larger graph vector change in the fault identification graph of the target line, and reserving the graph vector with smaller original change, at the moment, updating the topology structure of the target line is realized, the updated graph vector not only reflects the changed topology attribute, but also has the original topology attribute, and the updated fault identification graph is brought into a fault identification model to obtain a new fault identification model, so that the fault identification model not only has a function of identifying the topology change of the target line, but also has a function of identifying the original topology of the target line, and can realize that only the changed topology is directionally updated on the basis of the original fault identification model, model learning for the changed topology is not needed, and due to the fact that graphs with large changes are connected, the representation of the changed attribute can be well met, data operation can be further reduced, system operation pressure is relieved, response speed is improved, and a fault recognition model adaptive to the new topology of the target line is quickly generated.
And the line fault identification model determines the optimal parameters of the deep neural network by taking the identification rate and the training time as optimization targets.
The power grid operation log is used for generating fault time sequence data of a target line through a power transmission line fault simulation model, the topological variation attribute of the target line is generated through simulation of the power transmission line fault simulation model, and the electrical parameter of the target line is generated through simulation of the power transmission line fault simulation model.
In an actual situation, the fault recognition model cannot be established on the actual power grid line, so that the power transmission line fault simulation model can be established on the actual power grid line, a large amount of power grid operation logs are obtained, real-time collection of electrical parameters is facilitated, the change convenience of the target line topology change attribute is improved, collection of model training samples can be more conveniently achieved, and the actual application effect of the fault recognition model can be evaluated.
The invention clusters the electrical parameters with the same attribute, performs intra-cluster fusion to quantize the discrete electrical parameters into three identification parameters consistent with the R pixel value, the G pixel value and the B pixel value of the graph, constructs a fault identification graph consistent with the deep neural network adept processing field by using the three identification parameters, can improve the identification matching degree of the deep neural network to the power system fault based on the electrical parameters, improves the fault identification precision, and by setting the monitoring coefficient, the topology change information of the target line is obtained, the fault identification model is updated for the target line with topology change, the fault identification model is not required to be updated for the target line without topology change, the original fault identification model is used, the method can avoid the waste of operation resources and the extension of operation time caused by indiscriminate update of the target line, and improve the operation efficiency of diagnosis.
The above embodiments are only exemplary embodiments of the present application, and are not intended to limit the present application, and the protection scope of the present application is defined by the claims. Various modifications and equivalents may be made by those skilled in the art within the spirit and scope of the present application and such modifications and equivalents should also be considered to be within the scope of the present application.
Claims (10)
1. A power grid line fault recognition model training method is characterized by comprising the following steps:
s1, collecting the electrical parameters of the target line, and converting the electrical parameters of the target line into a fault recognition diagram of the target line;
step S2, carrying out image recognition training on the deep neural network based on the fault recognition graph to obtain a line fault recognition model;
and step S3, updating the line fault identification model by analyzing and updating the fault identification diagram according to the feature similarity.
2. The power grid line fault recognition model training method according to claim 1, characterized in that: in step S1, the specific method for converting the electrical parameter of the target line into the fault identification map of the target line includes:
step S101, performing cluster analysis on the electrical parameters to divide all the electrical parameters into three parameter clusters;
s102, fusing and normalizing all electrical parameters in each parameter cluster to integrate the three parameter clusters into three identification parameters;
step S103, converting the three identification parameters into R pixel values, G pixel values and B pixel values of the graphs respectively, and taking the line number of the target line and the acquisition time sequence of the three identification parameters of the target line as graph coordinate points of the R pixel values, the G pixel values and the B pixel values of the graphs to form the fault identification graph.
3. The power grid line fault recognition model training method according to claim 2, characterized in that: in step S101, the specific method for performing cluster analysis on the electrical parameter includes:
sequentially converging and calibrating the electrical parameters of all target lines according to the same category to be a parameter vector set
Wherein
characterized by the ith parametric vector term,
characterized by the ith electrical parameter of the target line n;
selecting three vector items with the maximum difference degree from the parameter vector set I as three cluster centers, and carrying out K-means cluster analysis on the parameter vector set I based on the three cluster centers to obtain three parameter clusters, wherein the three parameter clusters are respectively
,
, 
Wherein
，
，
、
、
respectively characterized as a parametric cluster
、
、
To (1) a
、
、
The term of the parameter vector is defined as,
、
、
respectively characterized as the second of the target line n
、
、
An electrical parameter.
4. The power grid line fault recognition model training method according to claim 3, characterized in that: in step S102, a specific method for integrating the three parameter clusters into three identification parameters includes:
converting the parametric vector form in the parametric cluster into the electrical parametric vector form of the target line
Wherein
，
，
，
，
、
、
characterised by the target line k according to said parametric cluster
、
、
The electrical parameter arranged in a cluster format,
the representation is the final value of the line number of the target line;
transforming the electric parameter vector form of the target line
According to the parameter cluster
、
、
Performing intra-cluster linear integration and converting the intra-cluster linear integration into an identification parameter vector form of a target line, wherein the identification parameter vector form of the target line is as follows:
wherein,
characterized by the fact that the product operator is a product operator,
、
、
respectively characterized as the second of the target line k
、
、
An electrical parameter.
5. The power grid line fault recognition model training method according to claim 4, characterized in that: the step S103 is to identify the parameters
Performing normalization treatment to [0,255%]The value ranges are respectively used as R pixel value, G pixel value and B pixel value of the graph.
6. The power grid line fault recognition model training method according to claim 5, characterized in that: in step S2, the specific method for training the line fault recognition model includes:
step S201, model training samples in a fault recognition graph format are obtained and converted from a power grid operation log, wherein positive labels are given to the fault recognition graph samples corresponding to normal power grid operation, and negative labels are given to the fault recognition graph samples corresponding to abnormal power grid operation;
step S202, preprocessing the fault recognition pattern book into three-channel pattern vectors, using the pattern vectors as the input of a deep neural model, using the labels of the fault recognition pattern samples as the output of the deep neural model to perform image recognition training, and realizing learning of a line fault recognition model for judging the operation condition of the power grid.
7. The power grid line fault recognition model training method according to claim 6, characterized in that: in step S3, the specific method for updating the line fault identification model includes:
setting a monitoring interval, and carrying out similarity monitoring on a fault identification graph of a target line once after the fault identification graph passes through the monitoring interval, wherein the similarity monitoring is used for monitoring the topology change attribute of the target line;
setting the overall similarity between the fault identification graph of the monitored target line and the fault identification graph of the target line before monitoring as a monitoring coefficient, wherein the operational formula of the monitoring coefficient is as follows:
wherein,
the characterization is that the listening coefficient is,
the total number of pattern vectors characterized as a fault identification pattern,
、
the jth graph vector is characterized as a post-snoop and pre-snoop fault identification graph, respectively.
8. The power grid line fault recognition model training method according to claim 7, wherein in step S3, the specific method for updating the line fault recognition model further includes:
setting a monitoring threshold, and comparing the monitoring coefficient with the monitoring threshold, specifically:
if the monitoring coefficient is higher than the monitoring threshold value, the line fault identification model before monitoring does not need to be updated;
if the monitoring coefficient is lower than the monitoring threshold, the line fault identification model before monitoring needs to be updated, specifically:
calculating the single similarity of each graph vector in the fault identification graph of the monitored target line and each graph vector in the fault identification graph of the target line before monitoring, and selecting all graph vectors with the single similarity higher than a monitoring threshold value from the fault identification graph of the monitored target line as a graph vector updating chain, wherein the calculation formula of the single similarity is as follows:
wherein,
the j graph vector in the fault identification graph characterized as the monitored target line
The jth graph vector in the fault identification graph of the target line before monitoring
Similarity of single items between the two;
and replacing the graph vector updating chain to the corresponding graph vector in the fault recognition graph of the target line before monitoring, completing updating of the fault recognition graph adapting to the topological variation attribute of the target line, bringing the updated fault recognition graph into the line fault recognition model for training, and completing updating of the line fault recognition model.
9. The power grid line fault recognition model training method according to claim 8, wherein the line fault recognition model determines optimal parameters of a deep neural network by taking a recognition rate and training time as optimization targets.
10. The power grid line fault recognition model training method according to claim 9, wherein the power grid operation log is generated by a power transmission line fault simulation model to generate fault timing sequence data of a target line, the topology variation attribute of the target line is generated by the power transmission line fault simulation model in a simulation manner, and the electrical parameter of the target line is generated by the power transmission line fault simulation model in a simulation manner.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111103885.4A CN113554010B (en) | 2021-09-22 | 2021-09-22 | Power grid line fault recognition model training method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111103885.4A CN113554010B (en) | 2021-09-22 | 2021-09-22 | Power grid line fault recognition model training method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113554010A true CN113554010A (en) | 2021-10-26 |
| CN113554010B CN113554010B (en) | 2022-01-21 |
Family
ID=78106612
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111103885.4A Active CN113554010B (en) | 2021-09-22 | 2021-09-22 | Power grid line fault recognition model training method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113554010B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117169657A (en) * | 2023-11-03 | 2023-12-05 | 山东和兑智能科技有限公司 | High-voltage cable state monitoring method and system based on artificial intelligence |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120304008A1 (en) * | 2009-09-17 | 2012-11-29 | Holger Hackstein | Supervised fault learning using rule-generated samples for machine condition monitoring |
| US20160054372A1 (en) * | 2013-04-09 | 2016-02-25 | Companhia Hidro Elétrica Do São Francisco | Real time, automatic diagnostic system and method for electric networks |
| CN107192939A (en) * | 2017-06-08 | 2017-09-22 | 京东方科技集团股份有限公司 | A kind of detection method and detecting system of the working condition of circuit |
| CN108304931A (en) * | 2018-02-06 | 2018-07-20 | 国网天津市电力公司电力科学研究院 | A kind of Condition-based Maintenance of Substation Equipment method for diagnosing faults |
| EP3460494A1 (en) * | 2017-09-26 | 2019-03-27 | Siemens Aktiengesellschaft | A method and apparatus for automatic detection of a fault type |
| CN109670584A (en) * | 2018-12-27 | 2019-04-23 | 张婧 | A kind of fault diagnosis method and system based on big data |
| CN111046581A (en) * | 2019-12-27 | 2020-04-21 | 国网江苏省电力有限公司电力科学研究院 | Power transmission line fault type identification method and system |
| CN112149554A (en) * | 2020-09-21 | 2020-12-29 | 广东电网有限责任公司清远供电局 | Fault classification model training method, fault detection method and related device |
| CN112180204A (en) * | 2020-07-10 | 2021-01-05 | 国网河北省电力有限公司雄安新区供电公司 | Power grid line fault diagnosis method based on electric quantity information |
| CN112630562A (en) * | 2020-12-04 | 2021-04-09 | 国网四川省电力公司广安供电公司 | Switch cabinet fault identification method and device based on deep neural network |
-
2021
- 2021-09-22 CN CN202111103885.4A patent/CN113554010B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120304008A1 (en) * | 2009-09-17 | 2012-11-29 | Holger Hackstein | Supervised fault learning using rule-generated samples for machine condition monitoring |
| US20160054372A1 (en) * | 2013-04-09 | 2016-02-25 | Companhia Hidro Elétrica Do São Francisco | Real time, automatic diagnostic system and method for electric networks |
| CN107192939A (en) * | 2017-06-08 | 2017-09-22 | 京东方科技集团股份有限公司 | A kind of detection method and detecting system of the working condition of circuit |
| EP3460494A1 (en) * | 2017-09-26 | 2019-03-27 | Siemens Aktiengesellschaft | A method and apparatus for automatic detection of a fault type |
| CN108304931A (en) * | 2018-02-06 | 2018-07-20 | 国网天津市电力公司电力科学研究院 | A kind of Condition-based Maintenance of Substation Equipment method for diagnosing faults |
| CN109670584A (en) * | 2018-12-27 | 2019-04-23 | 张婧 | A kind of fault diagnosis method and system based on big data |
| CN111046581A (en) * | 2019-12-27 | 2020-04-21 | 国网江苏省电力有限公司电力科学研究院 | Power transmission line fault type identification method and system |
| CN112180204A (en) * | 2020-07-10 | 2021-01-05 | 国网河北省电力有限公司雄安新区供电公司 | Power grid line fault diagnosis method based on electric quantity information |
| CN112149554A (en) * | 2020-09-21 | 2020-12-29 | 广东电网有限责任公司清远供电局 | Fault classification model training method, fault detection method and related device |
| CN112630562A (en) * | 2020-12-04 | 2021-04-09 | 国网四川省电力公司广安供电公司 | Switch cabinet fault identification method and device based on deep neural network |
Non-Patent Citations (1)
| Title |
|---|
| 杨光亮等: "基于小波神经网络和故障录波数据的电网故障类型识别", 《中国电机工程学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117169657A (en) * | 2023-11-03 | 2023-12-05 | 山东和兑智能科技有限公司 | High-voltage cable state monitoring method and system based on artificial intelligence |
| CN117169657B (en) * | 2023-11-03 | 2024-01-12 | 山东和兑智能科技有限公司 | High-voltage cable state monitoring method and system based on artificial intelligence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113554010B (en) | 2022-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2020102885A4 (en) | Disease recognition method of winter jujube based on deep convolutional neural network and disease image | |
| CN112859822B (en) | Equipment health analysis and fault diagnosis method and system based on artificial intelligence | |
| CN110995475A (en) | Power communication network fault detection method based on transfer learning | |
| CN113408087B (en) | Substation inspection method based on cloud side system and video intelligent analysis | |
| CN104346503A (en) | Human face image based emotional health monitoring method and mobile phone | |
| CN110532936A (en) | A kind of method and system identifying field crop growing way monitoring image Green plant | |
| CN114553475A (en) | Network attack detection method based on network flow attribute directed topology | |
| CN116760772B (en) | Control system and method for converging flow divider | |
| CN117540908B (en) | Agricultural resource integration method and system based on big data | |
| CN111275165A (en) | Network intrusion detection method based on improved convolutional neural network | |
| CN117315380B (en) | Deep learning-based pneumonia CT image classification method and system | |
| CN111461921A (en) | Load modeling typical user database updating method based on machine learning | |
| CN113567159A (en) | Scraper conveyor state monitoring and fault diagnosis method based on edge cloud cooperation | |
| CN113554010B (en) | Power grid line fault recognition model training method | |
| CN110321555A (en) | A kind of power network signal classification method based on Recognition with Recurrent Neural Network model | |
| CN114062812B (en) | Metering cabinet fault diagnosis method and system | |
| CN113935413A (en) | Distribution network wave recording file waveform identification method based on convolutional neural network | |
| CN117993282A (en) | Domain adaptability information bottleneck federal learning method for intelligent manufacturing fault diagnosis | |
| CN117809107A (en) | Tobacco leaf state identification method and device in baking stage | |
| CN117252459A (en) | Fruit quality evaluation system based on deep learning | |
| CN116304112A (en) | Intelligent monitoring method based on big data technology | |
| CN112529637B (en) | Service demand dynamic prediction method and system based on context awareness | |
| CN114064754B (en) | Industrial time sequence data characteristic learning method | |
| Zhang et al. | Load prediction based on depthwise separable convolution model | |
| CN117370870B (en) | Knowledge and data compound driven equipment multi-working condition identification and performance prediction method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |