CN113899987A - Power grid fault diagnosis method based on deep pyramid convolutional neural network - Google Patents
Power grid fault diagnosis method based on deep pyramid convolutional neural network Download PDFInfo
- Publication number
- CN113899987A CN113899987A CN202111225428.2A CN202111225428A CN113899987A CN 113899987 A CN113899987 A CN 113899987A CN 202111225428 A CN202111225428 A CN 202111225428A CN 113899987 A CN113899987 A CN 113899987A
- Authority
- CN
- China
- Prior art keywords
- fault
- alarm information
- model
- equipment
- neural network
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention provides a power grid fault diagnosis method based on a deep pyramid convolutional neural network. The method comprises the following steps: a depth pyramid convolution neural network model for extracting the overall characteristics of a fault event is constructed by a fault-oriented alarm information set, and the fault type is judged; constructing a deep pyramid convolutional neural network model facing to single alarm information, extracting text description characteristics of the single alarm information, determining key information corresponding to faults in an alarm information set, and screening suspicious fault equipment; and providing a fault equipment identification strategy integrating fault types and time sequence priority to identify fault equipment.
Description
Technical Field
The invention provides a grid fault diagnosis method based on a deep pyramid convolutional neural network, which utilizes the deep feature extraction capability of the deep pyramid convolutional neural network to extract fault feature knowledge of the text connotation of alarm information and realize end-to-end fault classification and fault equipment identification.
Background
The rapid and accurate power grid fault diagnosis plays an important role in accelerating the accident handling and system recovery process and ensuring the safe and stable operation of the system. With the continuous expansion of the power grid scale, the real-time alarm information of each station acquired by an Energy Management System (EMS) is larger and larger, and the logical relationship between the alarm information is more and more complex. When a power grid fails, particularly under the conditions of protection and failure and misoperation of a breaker, a dispatcher is difficult to determine key alarm information in a short time, accurately analyze a fault process and identify fault equipment. The power grid fault diagnosis method capable of rapidly and accurately identifying the fault equipment in the complex data environment is researched, and has important significance for assisting a dispatcher in judging the fault and realizing dispatching intellectualization.
Disclosure of Invention
In order to solve the above problems or at least partially solve the above problems, the present invention provides a grid fault diagnosis method based on a deep pyramid convolutional neural network. The method utilizes the deep knowledge extraction capability of a Deep Pyramid Convolutional Neural Network (DPCNN) on alarm information, establishes a fault classification model and a key information extraction model based on the DPCNN respectively facing an alarm information set and a single alarm information, and realizes fault classification and key information extraction. On the basis, a suspicious fault equipment set is determined according to the output results of the two models, and fault equipment is identified by adopting a fault equipment identification strategy integrating fault types and time sequence priority. The following scheme is adopted specifically:
s1, acquiring an alarm information set of historical fault events acquired by the SCADA system as a sample set, and performing vectorization representation on the alarm information sample set and a single piece of alarm information by using a word2vec model;
s2, constructing a fault classification model based on a Deep Pyramid Convolutional Neural Network (DPCNN), dividing sample sets based on the fault classification model to form two groups of fault sample sets, attaching corresponding fault event labels to each group of sample sets, and constructing a fault sample set for model learning;
s3, constructing a key information extraction model based on a Deep Pyramid Convolutional Neural Network (DPCNN), and setting a model hyper-parameter;
s4, determining a suspicious fault equipment set, and integrating fault types and fault equipment identification strategies with time sequence priority;
preferably, the step S1 includes:
and after the power grid fails, the protection and breaker action information related to the fault equipment is uploaded to the SCADA system in real time. The SCADA system also receives some alarm information such as device faults, non-stored energy of a spring of a switching mechanism and the like, and the information is recorded according to the time sequence to generate an alarm information set when the fault occurs. In order to facilitate computer processing, a text formed by an alarm information set is represented in a numerical mode, and the method adopts a word2vec model to carry out vectorization representation on the alarm information text.
Preferably, the step S2 includes:
and constructing a DPCNN fault classification model, and performing fault classification of different complexity and different fault equipment types on the alarm information set generated in the fault. Distinguishing between non-faults, simple faults, full protection refusal, incomplete protection refusal, and developmental faults. The complete protection refusal refers to the refusal of all the multiple sets of protection of a certain device, and the incomplete refusal refers to at least one set of protection action. Line faults, bus faults and transformer faults are distinguished.
The input layer of the model is a sentence vector matrix corresponding to the alarm information set to be classified, wherein the sizes of convolution kernels of region embedding and convolution layers are set to be 3, the correlation characteristics between three adjacent alarm information sentences in the alarm information text are extracted, and 200 matrix windows with different weight values are set simultaneously to obtain more complete characteristic expression. After the alarm information text is subjected to feature extraction, the probability of each classification label is calculated by the full connection layer, and the label with the maximum probability is a fault classification result corresponding to the alarm information text.
Preferably, the step S3 includes:
and designing a DPCNN-based key information extraction model facing to the single alarm information in the alarm information set, and extracting key information in the alarm information set, namely alarm information of protection and breaker action. The input layer of the key information extraction model is a word vector matrix corresponding to the alarm information sentences to be classified, m is the number of words contained in the sentences, and k is the dimension of the word vector. The output of the model is the classification result of the alarm information statement, namely the key information or the non-key information.
Preferably, the step S4 includes:
and inputting the alarm information set into the key information extraction model, and outputting the classification result of each alarm information statement. If the alarm information statement is key information, the equipment related to the statement is defined as suspicious fault equipment. And all devices related to the key information in the alarm information set are represented by a suspected fault device set D.
D＝{d1,d2,…,di,…,dn}
In the formula (d)iThe device related to the ith piece of key information in the alarm information set is n, and the number of the devices with suspicious faults is n.
After the suspicious faulty equipment set is determined by the key information and the fault classification model, faulty equipment needs to be identified. Aiming at the time sequence distribution characteristics of alarm information of different fault types, the method provides a fault equipment identification strategy integrating fault types and time sequences with priority:
1) and when the classification result of the fault classification model is a simple fault, identifying the equipment in the suspicious fault equipment set as fault equipment.
2) And when the classification results of the fault classification model are switch failure, developmental fault and incomplete protection failure, identifying the first equipment in the suspicious fault set as fault equipment.
3) And when the classification result of the fault classification model is complete protection refusal, determining fault equipment by combining a network topology structure.
Drawings
Fig. 1 is a flow chart of the grid fault diagnosis according to a preferred embodiment of the present invention;
FIG. 2 is a diagram illustrating the distribution of the faulty device information in the alarm information set of a simple fault event according to a preferred embodiment of the present invention;
FIG. 3 illustrates the fault identification strategy in accordance with a preferred embodiment of the present invention;
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
The invention provides a power grid fault diagnosis method based on a deep pyramid convolutional neural network. The method comprises the following steps: a depth pyramid convolution neural network model for extracting the overall characteristics of a fault event is constructed by a fault-oriented alarm information set, and the fault type is judged; constructing a deep pyramid convolutional neural network model facing to single alarm information, extracting text description characteristics of the single alarm information, determining key information corresponding to faults in an alarm information set, and screening suspicious fault equipment; and providing a fault equipment identification strategy integrating fault types and time sequence priority to identify fault equipment.
Referring to fig. 1, the method specifically includes the following steps:
s1, acquiring an alarm information set of historical fault events acquired by the SCADA system as a sample set, and performing vectorization representation on the alarm information sample set and a single piece of alarm information by using a word2vec model.
S2, constructing a DPCNN fault classification model, and performing fault classification of different complexity and different fault equipment types on the alarm information set generated in the fault. Distinguishing between non-faults, simple faults, full protection refusal, incomplete protection refusal, and developmental faults. The complete protection refusal refers to the refusal of all the multiple sets of protection of a certain device, and the incomplete refusal refers to at least one set of protection action. Line faults, bus faults and transformer faults are distinguished.
The input layer of the model is a sentence vector matrix corresponding to the alarm information set to be classified, wherein the sizes of convolution kernels of region embedding and convolution layers are set to be 3, the correlation characteristics between three adjacent alarm information sentences in the alarm information text are extracted, and 200 matrix windows with different weight values are set simultaneously to obtain more complete characteristic expression. After the alarm information text is subjected to feature extraction, the probability of each classification label is calculated by the full connection layer, and the label with the maximum probability is a fault classification result corresponding to the alarm information text.
S3, constructing a key information extraction model based on the Depth Pyramid Convolutional Neural Network (DPCNN), and setting a model hyper-parameter.
S4, aiming at the time sequence distribution characteristics of the alarm information of different fault types, a fault equipment identification strategy integrating the fault types and time sequence priority is provided. Referring to fig. 2, the distribution of the fault device information in the alarm information set of the simple fault event is shown. Referring to fig. 3, when the classification result of the fault classification model 1 is a simple fault, devices in the suspected fault device set are identified as fault devices; when the classification result of the fault classification model 1 is a switch failure, a developing fault and an incomplete protection failure, identifying the first equipment in the suspicious fault set as fault equipment; and when the classification result of the fault classification model 1 is complete protection refusal, determining fault equipment by combining a network topology result.
Claims (5)
1. A power grid fault diagnosis method based on a deep pyramid convolutional neural network is characterized by comprising the following steps:
s1, acquiring an alarm information set of historical fault events acquired by the SCADA system as a sample set, and performing vectorization representation on the alarm information sample set and a single piece of alarm information by using a word2vec model;
s2, constructing a fault classification model based on a Deep Pyramid Convolutional Neural Network (DPCNN), dividing sample sets based on the fault classification model to form two groups of fault sample sets, attaching corresponding fault event labels to each group of sample sets, and constructing a fault sample set for model learning;
s3, constructing a key information extraction model based on a Deep Pyramid Convolutional Neural Network (DPCNN), and setting a model hyper-parameter;
and S4, determining a suspicious fault equipment set, and integrating fault types and fault equipment identification strategies with time sequence priority.
2. The grid fault diagnosis method based on the deep pyramid convolutional neural network of claim 1, wherein the step S1 includes:
and after the power grid fails, the protection and breaker action information related to the fault equipment is uploaded to the SCADA system in real time. The SCADA system also receives some alarm information such as device faults, non-stored energy of a spring of a switching mechanism and the like, and the information is recorded according to the time sequence to generate an alarm information set when the fault occurs. In order to facilitate computer processing, a text formed by an alarm information set is represented in a numerical mode, and the method adopts a word2vec model to carry out vectorization representation on the alarm information text.
3. The grid fault diagnosis method based on the deep pyramid convolutional neural network of claim 1, wherein the step S2 includes:
and constructing a DPCNN fault classification model, and performing fault classification of different complexity and different fault equipment types on the alarm information set generated in the fault. Distinguishing between non-faults, simple faults, full protection refusal, incomplete protection refusal, and developmental faults. The complete protection refusal refers to the refusal of all the multiple sets of protection of a certain device, and the incomplete refusal refers to at least one set of protection action. Line faults, bus faults and transformer faults are distinguished.
The input layer of the model is a sentence vector matrix corresponding to the alarm information set to be classified, wherein the sizes of convolution kernels of region embedding and convolution layers are set to be 3, the correlation characteristics between three adjacent alarm information sentences in the alarm information text are extracted, and 200 matrix windows with different weight values are set simultaneously to obtain more complete characteristic expression. After the alarm information text is subjected to feature extraction, the probability of each classification label is calculated by the full connection layer, and the label with the maximum probability is a fault classification result corresponding to the alarm information text. Firstly, forming a subset min _ batch by every 32 training set samples, inputting a fault classification model for training, calculating a network loss function by adopting a cross entropy loss function, and optimizing an algorithm by an Adam optimizer. And forming a generation epoch after the training samples are all input into the model and trained once, and recording the model classification result once every 10 generation epochs. The method carries out 100 epoch training models, and judges the classification result by adopting the accuracy, the recall rate and the F1 value.
4. The grid fault diagnosis method based on the deep pyramid convolutional neural network of claim 1, wherein the step S3 includes:
and designing a DPCNN-based key information extraction model facing to the single alarm information in the alarm information set, and extracting key information in the alarm information set, namely alarm information of protection and breaker action. The input layer of the key information extraction model is a word vector matrix corresponding to the alarm information sentences to be classified, m is the number of words contained in the sentences, and k is the dimension of the word vector. The output of the model is the classification result of the alarm information statement, namely the key information or the non-key information.
5. The grid fault diagnosis method based on the deep pyramid convolutional neural network of claim 1, wherein the step S4 includes:
and inputting the alarm information set into the key information extraction model, and outputting the classification result of each alarm information statement. If the alarm information statement is key information, the equipment related to the statement is defined as suspicious fault equipment. And all devices related to the key information in the alarm information set are represented by a suspected fault device set D.
D＝{d1,d2,…,di,…,dn}
In the formula (d)iThe device related to the ith piece of key information in the alarm information set is n, and the number of the devices with suspicious faults is n.
After the suspicious faulty equipment set is determined by the key information and the fault classification model, faulty equipment needs to be identified. Aiming at the time sequence distribution characteristics of alarm information of different fault types, the method provides a fault equipment identification strategy integrating fault types and time sequences with priority:
1) and when the classification result of the fault classification model is a simple fault, identifying the equipment in the suspicious fault equipment set as fault equipment.
2) And when the classification results of the fault classification model are switch failure, developmental fault and incomplete protection failure, identifying the first equipment in the suspicious fault set as fault equipment.
3) And when the classification result of the fault classification model is complete protection refusal, determining fault equipment by combining a network topology structure.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111225428.2A CN113899987A (en) | 2021-10-21 | 2021-10-21 | Power grid fault diagnosis method based on deep pyramid convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111225428.2A CN113899987A (en) | 2021-10-21 | 2021-10-21 | Power grid fault diagnosis method based on deep pyramid convolutional neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113899987A true CN113899987A (en) | 2022-01-07 |
Family
ID=79193030
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111225428.2A Pending CN113899987A (en) | 2021-10-21 | 2021-10-21 | Power grid fault diagnosis method based on deep pyramid convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113899987A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114996461A (en) * | 2022-07-18 | 2022-09-02 | 北京大学 | Method, device, electronic equipment and medium for classifying text of medical adverse event |
| CN117521662A (en) * | 2023-10-19 | 2024-02-06 | 湖北华中电力科技开发有限责任公司 | Power dispatching semantic analysis method based on deep learning |
-
2021
- 2021-10-21 CN CN202111225428.2A patent/CN113899987A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114996461A (en) * | 2022-07-18 | 2022-09-02 | 北京大学 | Method, device, electronic equipment and medium for classifying text of medical adverse event |
| CN117521662A (en) * | 2023-10-19 | 2024-02-06 | 湖北华中电力科技开发有限责任公司 | Power dispatching semantic analysis method based on deep learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111274395B (en) | Power grid monitoring alarm event identification method based on convolution and long-short term memory network | |
| CN111914873B (en) | Two-stage cloud server unsupervised anomaly prediction method | |
| CN106570513A (en) | Fault diagnosis method and apparatus for big data network system | |
| CN113899987A (en) | Power grid fault diagnosis method based on deep pyramid convolutional neural network | |
| CN109633369B (en) | Power grid fault diagnosis method based on multi-dimensional data similarity matching | |
| CN105974265A (en) | SVM (support vector machine) classification technology-based power grid fault cause diagnosis method | |
| CN109409444B (en) | Multivariate power grid fault type discrimination method based on prior probability | |
| Chow et al. | Recognizing animal-caused faults in power distribution systems using artificial neural networks | |
| CN108020781A (en) | A kind of circuit breaker failure diagnostic method | |
| CN113676343B (en) | Fault source positioning method and device for power communication network | |
| CN114661905A (en) | Power grid fault diagnosis method based on BERT | |
| CN116205265A (en) | Power grid fault diagnosis method and device based on deep neural network | |
| CN108304567A (en) | High-tension transformer regime mode identifies and data classification method and system | |
| CN114091549A (en) | Equipment fault diagnosis method based on deep residual error network | |
| CN110794254A (en) | Power distribution network fault prediction method and system based on reinforcement learning | |
| CN113740666B (en) | Method for positioning root fault of storm alarm in power system of data center | |
| CN116593883A (en) | Breaker body fault diagnosis method, device and equipment of intelligent high-voltage switch and storage medium | |
| CN114638318A (en) | Power grid fault diagnosis method and system based on SE-inclusion network | |
| CN115017828A (en) | Power cable fault identification method and system based on bidirectional long-short-time memory network | |
| CN113722140B (en) | Industrial alarm flooding source diagnosis method based on small sample learning and storage medium | |
| Xu et al. | Fault diagnosis and identification of malfunctioning protection devices in a power system via time series similarity matching | |
| CN116032726A (en) | Fault root cause positioning model training method, device, equipment and readable storage medium | |
| CN113076217B (en) | Disk fault prediction method based on domestic platform | |
| CN113825162B (en) | Method and device for positioning fault reasons of telecommunication network | |
| CN113807462A (en) | AI-based network equipment fault reason positioning method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |