CN116894115A - Automatic archiving method for power grid infrastructure files - Google Patents
Automatic archiving method for power grid infrastructure files Download PDFInfo
- Publication number
- CN116894115A CN116894115A CN202310690485.0A CN202310690485A CN116894115A CN 116894115 A CN116894115 A CN 116894115A CN 202310690485 A CN202310690485 A CN 202310690485A CN 116894115 A CN116894115 A CN 116894115A
- Authority
- CN
- China
- Prior art keywords
- power grid
- grid infrastructure
- gru
- archiving
- gru model
- 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 34
- 230000004927 fusion Effects 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims abstract description 9
- 238000013528 artificial neural network Methods 0.000 claims abstract description 8
- 230000000306 recurrent effect Effects 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 41
- 230000004913 activation Effects 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 230000002457 bidirectional effect Effects 0.000 claims description 6
- 239000006185 dispersion Substances 0.000 claims description 6
- 230000003213 activating effect Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000013215 result calculation Methods 0.000 claims description 2
- 238000007726 management method Methods 0.000 abstract description 3
- 238000013461 design Methods 0.000 abstract description 2
- 239000000463 material Substances 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 12
- 238000010276 construction Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 241000283153 Cetacea Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000011056 performance test Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/93—Document management systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/906—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Business, Economics & Management (AREA)
- Business, Economics & Management (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
A method for automatically archiving a power grid infrastructure archive includes the steps of firstly, carrying out feature extraction on power grid infrastructure archive information by using a Bi-directional gating circulating unit Bi-GRU, then carrying out feature fusion on the extracted power grid infrastructure archive information features to obtain a feature data set, finally, configuring a feature tag according to archiving requirements, automatically archiving the feature data set by using a Bi-directional recurrent neural network BRNN, and outputting archiving results. The design realizes the automatic archiving of various complicated and various power grid infrastructure files, provides materials for building power grid infrastructure archives according to power grid infrastructure planning, is convenient for power grid infrastructure management staff to inquire engineering data in the power grid infrastructure process, and provides data support for the follow-up infrastructure process.
Description
Technical Field
The invention belongs to the technical field of electronic archiving, and particularly relates to an automatic archiving method for a power grid infrastructure file.
Background
As the project size of the power grid infrastructure project is larger and larger, mass data will be generated in the corresponding project, however, the problem of huge scale of the power grid infrastructure files is generated, and the problem of low efficiency exists in manual archiving. On the basis of project file management of a power grid enterprise, documents Zhang Lan and the like try to establish an informationized project file utilization method from the aspects of flow of project files, informationization of a system, concept construction of related personnel used and the like, documents Ouyang Suyu and the like are used for solving a scientific text extraction text drawing, documents Sun Erhua and the like are used for solving the problems of low classification precision and easy sinking into a local optimal state in the current unbalanced data classification algorithm, such as a scientific text extraction method based on word mixing and GRU, a scientific text knowledge extraction method based on a word mixing and gate control unit is provided, and the like are used for effectively solving the problem of unbalanced data set classification based on a whale optimizing and deep learning. However, most of the above methods are text classification, and no research is conducted on an electronic automatic archiving method and system for the grid-based files.
Disclosure of Invention
The invention aims to overcome the problems in the prior art and provide an automatic archiving method for the power grid infrastructure files.
In order to achieve the above object, the present invention provides the following technical solutions:
an automatic archiving method for a power grid infrastructure archive, which comprises the following steps in sequence:
s1, carrying out feature extraction on information of a power grid infrastructure file by utilizing a Bi-directional gating circulating unit Bi-GRU;
s2, carrying out feature fusion on the information features of the power grid infrastructure files extracted in the step S1 to obtain a feature data set;
and S3, configuring a feature tag according to the archiving requirement, automatically archiving the feature data set by using the bidirectional recurrent neural network BRNN, and outputting an archiving result.
In step S2, the feature fusion includes:
taking the information characteristic data of each power grid infrastructure archive obtained in the step S1 as a cluster respectively, calculating a merging value between any two clusters according to the following formula, merging two clusters with the minimum merging value into a new cluster, and eliminating redundancy of the information characteristic of the power grid infrastructure archive:
in the above formula, C is a combined value,for the new cluster a' x ∪a′ y The sum of the squares of the deviations of the new cluster a' x ∪a′ y From the x-th cluster a' x With cluster y a' y Combining to obtain (I) herba Cistanchis>For the x th cluster a' x Sum of squares of dispersion of>For cluster y a' y Sum of squares of dispersion of>For cluster a't' t The sum of the squares of the deviations of (a); a' t 、a′ x 、a′ y The characteristic data of the information of the power grid infrastructure files are respectively at the time t, the time x and the time y; t=1, 2, …, m, m is the total number of the grid infrastructure archive information characteristic data extracted in the step S1.
The Bi-GRU comprises an output layer and a hidden layer formed by two GRU models which are respectively arranged in the forward direction and the reverse direction, wherein the information of the power grid infrastructure archive is firstly input into the two GRU models of the hidden layer to respectively perform matrix calculation, and then the output vectors from the two GRU models are combined through the output layer, so that the extracted characteristic data of the information of the power grid infrastructure archive is output.
The GRU model is as follows:
in the above, z t Update gate for GRU model, r t For the reset gate of the GRU model,activating a function for sigmoid, h t-1 Outputting a vector for the GRU model at the time t-1, and h t The output vector of the GRU model at the moment t; w (W) z 、U z The weight of the GRU model output vector at the t-1 moment in the updating gate and the weight of the GRU model input vector at the t moment in the updating gate are respectively obtained; w (W) r 、U r The weight of the GRU model output vector at the t moment in the reset gate and the weight of the GRU model input vector at the t moment in the reset gate are respectively; />Grid infrastructure archive for GRU model at time tInformation, W is the weight of GRU model input vector, x t And (3) inputting a vector for the GRU model at the time t, namely inputting the power grid infrastructure file information, wherein U is the weight of the forgotten power grid infrastructure file information in the reset gate.
The output layer merges output vectors from the forward GRU model, the reverse GRU model according to the following formula:
in the above, a' t For the information characteristic data of the power grid infrastructure archive outputted by the Bi-gate control circulation unit Bi-GRU at the t moment,respectively outputting vectors of a forward GRU model and a reverse GRU model at the moment t; x is x t Inputting a vector for the GRU model at the time t, namely inputting information of the power grid infrastructure files; />The output of the hidden layer at the moment of the forward GRU model and the reverse GRU model t-1 is respectively; n (N) i 、M i And respectively obtaining weights corresponding to the hidden layers of the forward GRU model and the reverse GRU model at the moment t.
The bidirectional recurrent neural network BRNN comprises an input layer, two hidden layers arranged in opposite directions and an output layer, wherein the output result calculation formula of the output layer is as follows:
y t =α(W fy h ft +W by h bt +b y );
h ft =tanh(W fh h ft-1 +W fx a t +b f h);
h bt =tanh(W by h bt+1 +W bx a t +b bh );
in the above, a t For the input information of the input layer at time t, i.e. the feature data set a= [ a ] obtained by feature fusion 1 ,a 2 ,…,a t ,…,a k ]Feature data at time t in (1), wherein t=1, 2, …, k, k is the total number of the remaining feature data after feature fusion, y t Outputting the result for the output layer at the moment t, wherein alpha is an archiving activation function, W fy Weights connecting hidden layer to output layer for forward direction, W by Weights connecting hidden layer to output layer for backward direction, h ft Outputting a result for a hidden layer in the forward direction at the time t, and h bt B, outputting a result for a hidden layer in a backward direction at the time t y B is the bias vector of the hidden layer fh A hidden layer deviation amount of the forward direction b bh A hidden layer deviation amount of backward direction, h bt+1 Outputting a result h for a hidden layer in a backward direction at a time t+1 ft-1 Outputting a result for a hidden layer in a forward direction at a time t-1, W fh Weights connecting hidden layer to output layer for forward direction, W fx Weights connecting hidden layer to output layer for forward direction, W bx The hidden layer, which is the backward direction, is connected to the weights of the output layer.
The expression of the archiving activation function alpha is:
in the above formula, x is the input of the nonlinear activation function T-ReLU, and k' is a constant coefficient.
The archive requirements are handover requirements, engineering type or data source.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention relates to an automatic archiving method for a power grid infrastructure file, which comprises the steps of firstly carrying out feature extraction on power grid infrastructure file information by utilizing a Bi-directional gating circulating unit Bi-GRU, then carrying out feature fusion on the extracted power grid infrastructure file information features to obtain a feature data set, finally configuring a feature tag according to archiving requirements, automatically archiving the feature data set by utilizing a Bi-directional recurrent neural network BRNN, and outputting an archiving result; the design realizes the automatic archiving of various complicated and various power grid infrastructure files, provides materials for building power grid infrastructure archives according to power grid infrastructure planning, is convenient for power grid infrastructure management staff to inquire engineering data in the power grid infrastructure process, and provides data support for the follow-up infrastructure process. Therefore, the invention can realize the automatic archiving of the power grid infrastructure files and provide data support for the subsequent infrastructure process.
2. In the automatic archiving method of the power grid infrastructure archives, the obtained characteristic data of each power grid infrastructure archives are respectively used as one cluster, the merging value between any two clusters is calculated, the merging degree between the two clusters is evaluated by using the merging value, if the merging value is smaller, the similarity between the characteristics contained in the two clusters is higher, and the two clusters can be merged into one cluster, so that redundancy of the power grid infrastructure archives information characteristics is eliminated, the characteristic dimension reduction is facilitated, the calculation complexity of a model is reduced, the archiving accuracy is improved, otherwise, if the merging value is larger, the two clusters cannot be merged; the method can process the abnormal values in the feature data set by adopting the mode of realizing feature fusion, so as to reduce the influence on the feature fusion result. Therefore, the feature fusion is realized through the combined value, the calculation complexity of the model can be reduced, and the archiving accuracy is improved.
Drawings
FIG. 1 is a flow chart of the present invention.
FIG. 2 shows the result of the archival effect evaluation of the method of the present invention.
Detailed Description
The invention is further described below with reference to the drawings and the detailed description.
Referring to fig. 1, an automatic archiving method for a power grid infrastructure file sequentially comprises the following steps:
s1, performing feature extraction on an electronic file of a grid infrastructure project collected under a line by utilizing a Bi-directional gating circulating unit Bi-GRU to obtain an initial feature data set A '= [ a ]' 1 ,a′ 2 ,…,a′ t ,…,a′ m ]T=1, 2, …, m, m is the total number of feature data in the initial feature data set;
the Bi-GRU is a model simplified by an LSTM deep network model, only an update gate and a reset gate exist, the update gate is helpful to determine how much of previous grid construction archive information needs to be transferred to the model, all information in the past is prevented from being copied by the model, the reset gate is used to determine how much of previous grid construction archive information is forgotten and determined in the model, the Bi-GRU comprises an output layer and a hidden layer formed by two GRU models which are respectively arranged in a forward direction and a reverse direction, the feature extraction step is that the grid construction archive information is firstly input into the two GRU models of the hidden layer to respectively perform matrix calculation, and the GRU models are as follows:
in the above, z t Update gate for GRU model, r t For the reset gate of the GRU model,activating a function for sigmoid, h t-1 Outputting a vector for the GRU model at the time t-1, and h t The output vector of the GRU model at the moment t; w (W) z 、U z The weight of the GRU model output vector at the t-1 moment in the updating gate and the weight of the GRU model input vector at the t moment in the updating gate are respectively obtained; w (W) r 、U r The weight of the GRU model output vector at the t moment in the reset gate and the weight of the GRU model input vector at the t moment in the reset gate are respectively; />For building archive information of a GRU model power grid at time t, W is weight of input vector of the GRU model, and x is the weight of the GRU model input vector t Inputting a vector for the GRU model at the time t, namely inputting power grid basic building archive information, wherein U is the weight of the forgotten power grid basic building archive information in a reset gate;
and then combining output vectors from the forward GRU model and the reverse GRU model through an output layer according to the following formula, and outputting the extracted information characteristic data of the power grid infrastructure file:
in the above, a' t For the information characteristic data of the power grid infrastructure archive outputted by the Bi-gate control circulation unit Bi-GRU at the t moment,respectively outputting vectors of a forward GRU model and a reverse GRU model at the moment t; x is x t Inputting a vector for the GRU model at the time t, namely inputting information of the power grid infrastructure files; />The output of the hidden layer at the moment of the forward GRU model and the reverse GRU model t-1 is respectively; n (N) i 、M i Respectively corresponding weights of the hidden layers of the forward GRU model and the reverse GRU model at the moment t;
s2, performing feature fusion on the initial feature data set to obtain a feature data set A= [ a ] 1 ,a 2 ,…,a t ,…,a k ]T=1, 2, …, k, k is the total number of feature data remaining after feature fusion, and the feature fusion step specifically includes:
each feature data in the initial feature data set is respectively used as a cluster, the merging value between any two clusters is calculated, whether feature fusion can be achieved is determined by comparing the sizes of the merging values, if the merging value of the two clusters is minimum, the similarity between features contained in the two clusters is higher, and if the merging value of a plurality of merging values for comparison is minimum, the two clusters with the minimum merging value are merged into a new cluster, so that redundancy of the information features of the power grid foundation files is eliminated, and the calculation formula of the merging value is as follows:
in the above formula, C is a combined value,for the new cluster a' x ∪a′ y The sum of the squares of the deviations of the new cluster a' x ∪a′ y From the x-th cluster a' x With cluster y a' y Combining to obtain (I) herba Cistanchis>For the x th cluster a' x Sum of squares of dispersion of>For cluster y a' y Sum of squares of dispersion of>For cluster a't' t The sum of the squares of the deviations of (a); a' t 、a′ x 、a′ y The characteristic data of the information of the power grid infrastructure files are respectively at the time t, the time x and the time y;
s3, configuring a feature tag according to archiving requirements, and identifying and classifying a feature data set by utilizing a bidirectional recurrent neural network BRNN, wherein the bidirectional recurrent neural network BRNN comprises an input layer, two hidden layers which are arranged in opposite directions and an output layer, and the input feature data a t The input BRNN carries out forward and backward transfer, meanwhile acquires forward and backward hidden states, then splices the forward and backward hidden states to obtain a characteristic representation at the time t, and then transmits the characteristic representation at the time t into an output layer (such as a full connection layer) and maps the characteristic representation to a corresponding output result y through an activation function t Because the k features are integrated, the output of the BRNN can be set as a k-dimensional vector, each dimension corresponds to one feature classification, a feature label is marked, and files are archived in the classification with the corresponding feature label according to archiving requirements, so that a user can quickly locate and search when searching the files;
output result y of the output layer t The calculation formula of (2) is as follows:
y t =α(W fy h ft +W by h bt +b y );
h ft =tanh(W fh h ft-1 +W fx a t +b fh );
h bt =tanh(W by h bt+1 +W bx a t +b bh );
in the above, a t For the input information of the input layer at time t, i.e. the feature data at time t in the feature data set obtained by feature fusion, t=1, 2, …, kIs the total number of the residual characteristic data after the characteristic fusion, y t Output result of output layer at time t, W fy Weights connecting hidden layer to output layer for forward direction, W by Weights connecting hidden layer to output layer for backward direction, h ft Outputting a result for a hidden layer in the forward direction at the time t, and h bt B, outputting a result for a hidden layer in a backward direction at the time t y B is the bias vector of the hidden layer fh A hidden layer deviation amount of the forward direction b bh A hidden layer deviation amount of backward direction, h bt+1 Outputting a result h for a hidden layer in a backward direction at a time t+1 ft-1 Outputting a result for a hidden layer in a forward direction at a time t-1, W fh Weights connecting hidden layer to output layer for forward direction, W fx Weights connecting hidden layer to output layer for forward direction, W bx For the weight of the hidden layer connected to the output layer in the backward direction, alpha is an archiving activation function, the archiving activation function combines the linear characteristic of the active region of the ReLU function and the soft saturation characteristic of the inactive region of the Tanh function, the defects of the original two functions are overcome, and the expression is as follows:
in the above formula, x is the input of a nonlinear activation function T-ReLU, k' is a constant coefficient, and the value range is [0.1,0.5].
Performance test:
according to different handover requirements, different engineering types and different data sources, three electronic repositories are respectively constructed, corresponding characteristic labels are configured according to the file content of the power grid infrastructure projects, the electronic files of the plurality of power grid infrastructure projects are collected and obtained under the line by adopting the automatic filing method to carry out automatic classified filing, the automatic classified filing result is evaluated according to three performance indexes of accuracy, precision and recall, the calculation results of the three performance indexes are shown in figure 2, and the calculation formula is as follows:
in the above description, TP, TN, FP, FN is the number of files with correct true archive, correct false archive and incorrect false archive respectively;
as can be seen from fig. 2, the automatic archiving method of the present invention has excellent performance in terms of three performance indexes, namely, accuracy, precision and recall, wherein the archiving effect on engineering types and data sources is excellent, the three performance indexes are over 90%, and the archiving effect on handover requirements is poor. In summary, the automatic archiving method has good archiving effect on the power grid infrastructure files, and can be used for inquiring engineering data in the power grid infrastructure process by power grid infrastructure manager to provide data support for the follow-up infrastructure process.
Claims (8)
1. An automatic archiving method for a power grid infrastructure file is characterized by comprising the following steps of:
the archiving method sequentially comprises the following steps:
s1, carrying out feature extraction on information of a power grid infrastructure file by utilizing a Bi-directional gating circulating unit Bi-GRU;
s2, carrying out feature fusion on the information features of the power grid infrastructure files extracted in the step S1 to obtain a feature data set;
and S3, configuring a feature tag according to the archiving requirement, automatically archiving the feature data set by using the bidirectional recurrent neural network BRNN, and outputting an archiving result.
2. The method for automatically archiving a power grid infrastructure file according to claim 1, wherein:
in step S2, the feature fusion includes:
taking the information characteristic data of each power grid infrastructure archive obtained in the step S1 as a cluster respectively, calculating a merging value between any two clusters according to the following formula, merging two clusters with the minimum merging value into a new cluster, and eliminating redundancy of the information characteristic of the power grid infrastructure archive:
in the above formula, C is a combined value,for the new cluster a' x ∪a′ y The sum of the squares of the deviations of the new cluster a' x ∪a′ y From the x-th cluster a' x With cluster y a' y Combining to obtain (I) herba Cistanchis>For the x th cluster a' x Sum of squares of dispersion of>For cluster y a' y Sum of squares of dispersion of>For cluster a't' t The sum of the squares of the deviations of (a); a' t 、a′ x 、a′ y The characteristic data of the information of the power grid infrastructure files are respectively at the time t, the time x and the time y; t=1, 2, …, m, m is the total number of the grid infrastructure archive information characteristic data extracted in the step S1.
3. An automatic archiving method for electric network infrastructure files according to claim 1 or 2, characterized in that:
the Bi-GRU comprises an output layer and a hidden layer formed by two GRU models which are respectively arranged in the forward direction and the reverse direction, wherein the information of the power grid infrastructure archive is firstly input into the two GRU models of the hidden layer to respectively perform matrix calculation, and then the output vectors from the two GRU models are combined through the output layer, so that the extracted characteristic data of the information of the power grid infrastructure archive is output.
4. A method for automatically archiving a grid-based archive in accordance with claim 3, wherein:
the GRU model is as follows:
in the above, z t Update gate for GRU model, r t For the reset gate of the GRU model,activating a function for sigmoid, h t-1 Outputting a vector for the GRU model at the time t-1, and h t The output vector of the GRU model at the moment t; w (W) z 、U z The weight of the GRU model output vector at the t-1 moment in the updating gate and the weight of the GRU model input vector at the t moment in the updating gate are respectively obtained; w (W) r 、U r The weight of the GRU model output vector at the t moment in the reset gate and the weight of the GRU model input vector at the t moment in the reset gate are respectively; />For building archive information of a GRU model power grid at time t, W is weight of input vector of the GRU model, and x is the weight of the GRU model input vector t And (3) inputting a vector for the GRU model at the time t, namely inputting the power grid infrastructure file information, wherein U is the weight of the forgotten power grid infrastructure file information in the reset gate.
5. The method for automatically archiving a power grid infrastructure archive of claim 4, wherein:
the output layer merges output vectors from the forward GRU model, the reverse GRU model according to the following formula:
in the above, a' t For the information characteristic data of the power grid infrastructure archive outputted by the Bi-gate control circulation unit Bi-GRU at the t moment,respectively outputting vectors of a forward GRU model and a reverse GRU model at the moment t; x is x t Inputting a vector for the GRU model at the time t, namely inputting information of the power grid infrastructure files; />The output of the hidden layer at the moment of the forward GRU model and the reverse GRU model t-1 is respectively; n (N) i 、M i Hidden layer pairs of a forward GRU model and a reverse GRU model at the moment t respectivelyWeight of the weight.
6. The method for automatically archiving a power grid infrastructure file according to claim 1, wherein:
the bidirectional recurrent neural network BRNN comprises an input layer, two hidden layers arranged in opposite directions and an output layer, wherein the output result calculation formula of the output layer is as follows:
y t =α(W fy h ft +W by h bt +b y );
h ft =tanh(W fh h ft-1 +W fx a t +b fh );
h nt =tanh(W by h bt+1 +W bx a t +b bh );
in the above, a t For the input information of the input layer at time t, i.e. the feature data set a= [ a ] obtained by feature fusion 1 ,a 2 ,…,a t ,…,a k ]Feature data at time t in (1), wherein t=1, 2, …, k, k is the total number of the remaining feature data after feature fusion, y t Outputting the result for the output layer at the moment t, wherein alpha is an archiving activation function, W fy Weights connecting hidden layer to output layer for forward direction, W by Weights connecting hidden layer to output layer for backward direction, h ft Outputting a result for a hidden layer in the forward direction at the time t, and h bt B, outputting a result for a hidden layer in a backward direction at the time t y B is the bias vector of the hidden layer fh A hidden layer deviation amount of the forward direction b bh A hidden layer deviation amount of backward direction, h bt+1 Outputting a result h for a hidden layer in a backward direction at a time t+1 ft-1 Outputting a result for a hidden layer in a forward direction at a time t-1, W fh Weights connecting hidden layer to output layer for forward direction, W fx Weights connecting hidden layer to output layer for forward direction, W bx The hidden layer, which is the backward direction, is connected to the weights of the output layer.
7. The method for automatically archiving a power grid infrastructure archive of claim 6, wherein:
the expression of the archiving activation function alpha is:
in the above formula, x is the input of the nonlinear activation function T-ReLU, and k' is a constant coefficient.
8. An automatic archiving method for electric network infrastructure files according to claim 1 or 2, characterized in that:
the archive requirements are handover requirements, engineering type or data source.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310690485.0A CN116894115B (en) | 2023-06-12 | 2023-06-12 | Automatic archiving method for power grid infrastructure files |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310690485.0A CN116894115B (en) | 2023-06-12 | 2023-06-12 | Automatic archiving method for power grid infrastructure files |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116894115A true CN116894115A (en) | 2023-10-17 |
CN116894115B CN116894115B (en) | 2024-05-24 |
Family
ID=88312781
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310690485.0A Active CN116894115B (en) | 2023-06-12 | 2023-06-12 | Automatic archiving method for power grid infrastructure files |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116894115B (en) |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108846017A (en) * | 2018-05-07 | 2018-11-20 | 国家计算机网络与信息安全管理中心 | The end-to-end classification method of extensive newsletter archive based on Bi-GRU and word vector |
CN112527966A (en) * | 2020-12-18 | 2021-03-19 | 重庆邮电大学 | Network text emotion analysis method based on Bi-GRU neural network and self-attention mechanism |
CN113035348A (en) * | 2021-03-22 | 2021-06-25 | 山西三友和智慧信息技术股份有限公司 | Diabetes diagnosis method based on GRU feature fusion |
WO2021155699A1 (en) * | 2020-02-03 | 2021-08-12 | 苏州科技大学 | Global encoding method for automatic abstract of chinese long text |
CN113869054A (en) * | 2021-10-13 | 2021-12-31 | 天津大学 | Deep learning-based electric power field project feature identification method |
CN113869055A (en) * | 2021-10-13 | 2021-12-31 | 天津大学 | Power grid project characteristic attribute identification method based on deep learning |
WO2022011977A1 (en) * | 2020-07-15 | 2022-01-20 | 中国科学院深圳先进技术研究院 | Network anomaly detection method and system, terminal and storage medium |
CN114020917A (en) * | 2021-11-05 | 2022-02-08 | 南京理工大学 | News text classification method based on self-attention mechanism and convolutional neural network |
CN114020964A (en) * | 2021-11-15 | 2022-02-08 | 上海大学 | Method for realizing video abstraction by using memory network and gated cyclic unit |
CN114443923A (en) * | 2022-01-14 | 2022-05-06 | 国网四川省电力公司天府新区供电公司 | Electronic archiving method and system for power grid operation and inspection project archives |
CN114760098A (en) * | 2022-03-16 | 2022-07-15 | 南京邮电大学 | CNN-GRU-based power grid false data injection detection method and device |
CN117373116A (en) * | 2023-10-10 | 2024-01-09 | 合肥工业大学 | Human body action detection method based on lightweight characteristic reservation of graph neural network |
-
2023
- 2023-06-12 CN CN202310690485.0A patent/CN116894115B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108846017A (en) * | 2018-05-07 | 2018-11-20 | 国家计算机网络与信息安全管理中心 | The end-to-end classification method of extensive newsletter archive based on Bi-GRU and word vector |
WO2021155699A1 (en) * | 2020-02-03 | 2021-08-12 | 苏州科技大学 | Global encoding method for automatic abstract of chinese long text |
WO2022011977A1 (en) * | 2020-07-15 | 2022-01-20 | 中国科学院深圳先进技术研究院 | Network anomaly detection method and system, terminal and storage medium |
CN112527966A (en) * | 2020-12-18 | 2021-03-19 | 重庆邮电大学 | Network text emotion analysis method based on Bi-GRU neural network and self-attention mechanism |
CN113035348A (en) * | 2021-03-22 | 2021-06-25 | 山西三友和智慧信息技术股份有限公司 | Diabetes diagnosis method based on GRU feature fusion |
CN113869054A (en) * | 2021-10-13 | 2021-12-31 | 天津大学 | Deep learning-based electric power field project feature identification method |
CN113869055A (en) * | 2021-10-13 | 2021-12-31 | 天津大学 | Power grid project characteristic attribute identification method based on deep learning |
CN114020917A (en) * | 2021-11-05 | 2022-02-08 | 南京理工大学 | News text classification method based on self-attention mechanism and convolutional neural network |
CN114020964A (en) * | 2021-11-15 | 2022-02-08 | 上海大学 | Method for realizing video abstraction by using memory network and gated cyclic unit |
CN114443923A (en) * | 2022-01-14 | 2022-05-06 | 国网四川省电力公司天府新区供电公司 | Electronic archiving method and system for power grid operation and inspection project archives |
CN114760098A (en) * | 2022-03-16 | 2022-07-15 | 南京邮电大学 | CNN-GRU-based power grid false data injection detection method and device |
CN117373116A (en) * | 2023-10-10 | 2024-01-09 | 合肥工业大学 | Human body action detection method based on lightweight characteristic reservation of graph neural network |
Non-Patent Citations (1)
Title |
---|
黄存强等: ""基于并行数据计算的网格化电网空间负荷预测方法研究"", 《电子设计工程》, 31 May 2023 (2023-05-31), pages 123 - 127 * |
Also Published As
Publication number | Publication date |
---|---|
CN116894115B (en) | 2024-05-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ding et al. | Multi-scale relation network for few-shot learning based on meta-learning | |
CN109902714B (en) | Multi-modal medical image retrieval method based on multi-graph regularization depth hashing | |
CN113254659A (en) | File studying and judging method and system based on knowledge graph technology | |
CN109902192A (en) | Remote sensing image retrieval method, system, equipment and the medium returned based on unsupervised depth | |
CN113821670A (en) | Image retrieval method, device, equipment and computer readable storage medium | |
AU2019200371A1 (en) | Utilizing artificial intelligence to integrate data from multiple diverse sources into a data structure | |
CN117236677A (en) | RPA process mining method and device based on event extraction | |
CN113742396B (en) | Mining method and device for object learning behavior mode | |
CN116894115B (en) | Automatic archiving method for power grid infrastructure files | |
De et al. | Engineering a machine learning pipeline for automating metadata extraction from longitudinal survey questionnaires | |
Yang et al. | Multi-Label Learning Based on Transfer Learning and Label Correlation. | |
WO2023168818A1 (en) | Method and apparatus for determining similarity between video and text, electronic device, and storage medium | |
Yang et al. | Named entity recognition of power substation knowledge based on transformer-BiLSTM-CRF network | |
CN114757097A (en) | Line fault diagnosis method and device | |
Chukhray et al. | Proximate objects probabilistic searching method | |
Xu et al. | Ensemble learning based multi-source information fusion | |
Yan et al. | Multi-sql: An automatic multi-model data management system | |
Liu et al. | Lekan: Extracting long-tail relations via layer-enhanced knowledge-aggregation networks | |
Li et al. | Improved YOLOV3 Surveillance Device Object Detection Method Based on Federated Learning | |
Han et al. | A review on financial robot process auto-mining based on reinforcement learning | |
CN117151247B (en) | Method, apparatus, computer device and storage medium for modeling machine learning task | |
Feng et al. | Construction of Legal Reporting Information Platform Based on Natural Optimization Algorithm | |
Zhang et al. | Subgraph Isomorphism Building on A Hierarchical Query Graph | |
Zhang et al. | Application of a multimodal model optimized by multi-head-attention mechanism in the classification of fishery resource remote sensing data files | |
Lawabni et al. | A novel update propagation module for the data provenance problem: A contemplating vision on realizing data provenance from models to storage |
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 |