CN114386524A - Power equipment identification method for dynamic self-adaptive graph layering simulation learning - Google Patents
Power equipment identification method for dynamic self-adaptive graph layering simulation learning Download PDFInfo
- Publication number
- CN114386524A CN114386524A CN202210048517.2A CN202210048517A CN114386524A CN 114386524 A CN114386524 A CN 114386524A CN 202210048517 A CN202210048517 A CN 202210048517A CN 114386524 A CN114386524 A CN 114386524A
- Authority
- CN
- China
- Prior art keywords
- learning
- feature set
- power equipment
- graph
- historical
- 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
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000004088 simulation Methods 0.000 title claims abstract description 31
- 230000007246 mechanism Effects 0.000 claims abstract description 19
- 238000010606 normalization Methods 0.000 claims abstract description 12
- 239000012212 insulator Substances 0.000 claims abstract description 8
- 230000002776 aggregation Effects 0.000 claims abstract description 7
- 238000004220 aggregation Methods 0.000 claims abstract description 7
- 230000008569 process Effects 0.000 claims description 21
- 230000003044 adaptive effect Effects 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 15
- 230000004913 activation Effects 0.000 claims description 7
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 4
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 3
- 238000005314 correlation function Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000003278 mimic effect Effects 0.000 claims description 2
- 238000012512 characterization method Methods 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 3
- 230000009471 action Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 239000000654 additive Substances 0.000 description 2
- 230000000996 additive effect Effects 0.000 description 2
- 230000004931 aggregating effect Effects 0.000 description 2
- 239000003795 chemical substances by application Substances 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000006116 polymerization reaction Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000007704 transition 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
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- 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)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Probability & Statistics with Applications (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a power equipment identification method for dynamic self-adaptive graph layered simulation learning, which comprises the following steps: acquiring characteristic information based on a multi-head attention mechanism to obtain real-time characteristic information of the power equipment; performing aggregation processing and normalization processing on the real-time characteristic information based on a graph attention network to obtain a characteristic set; based on a layered simulation learning mechanism, learning processing is carried out on the feature set, and KL divergence and the historical feature set are adopted for judgment, so that a strategy model of the historical feature set is obtained, and the strategy model of the historical feature set is optimized; and learning the strategy model based on a dynamic self-adaptive optimizer and classifying, distinguishing and processing the transformer, the insulator, the breaker and the wire power equipment by adopting a classifier so as to achieve the aim of image recognition. The invention can solve the problem of non-grid data among different devices; the collected image data can be intelligently analyzed; the automatic analysis and identification can be carried out on the power equipment; the recognition efficiency and accuracy can be improved.
Description
Technical Field
The invention belongs to the technical field of power equipment identification, and particularly relates to a power equipment identification method for dynamic self-adaptive graph layered simulation learning.
Background
At present, the introduction robot is patrolled and examined to electric power, and unmanned aerial vehicle, technologies such as high definition camera detect the circuit through visible light or infrared ray, can reduce the cost of labor, improve work efficiency. However, since the variety of power devices, such as transformers, insulators, circuit breakers, wires, etc., is large, and it is easily affected by weather and unknown objects, it provides challenges for collecting information of the power devices. The collected images are often selected and identified manually, which is limited by human subjectivity and is easy to cause visual fatigue and fuzziness. In recent years, artificial intelligence, especially deep learning, has achieved remarkable results in image processing, and the deep learning can process high-dimensional and complex data by means of layer-by-layer training. However, different types of devices in the power network are not regular grid data, and the traditional neural network cannot process non-grid data.
The graph neural network can treat each electric power device as connected nodes, and solves the difficult problem brought by non-Euclidean data in reality by modeling the connection relation among the nodes. The attention mechanism is added on the basis of the graph neural network, so that more important node information (such as color, shape, additive, texture, brightness, direction and other characteristics) in the graph structure can be effectively selected, and further, the appropriate weight is distributed to each node. However, the pure dependency graph attention network outputs a new feature set with certain deviation. Therefore, the invention provides a method for combining the graph attention network with the simulation learning, so as to correct the deviation caused by the graph attention network.
The simulation learning does not need interaction between an intelligent agent and the environment, the strategy is directly obtained from the historical characteristic set and executed, and finally the expected strategy is obtained through continuous interaction between the initial strategy and the environment. However, the mimic learning has a limitation on complex data and is slow in calculation speed.
Disclosure of Invention
The invention aims to provide a power equipment identification method for dynamic self-adaptive graph layering simulation learning, which can solve the problems of subjectivity and inaccuracy caused by a manual method, can avoid a large number of redundant and repeated images, can solve non-grid data among different equipment, and can perform intelligent analysis on image data.
In order to achieve the above object, the present invention provides a power equipment identification method for dynamic adaptive graph layered simulation learning, including:
acquiring characteristic information based on a multi-head attention mechanism to obtain real-time characteristic information of the power equipment;
performing aggregation processing and normalization processing on the real-time characteristic information based on a graph attention network to obtain a characteristic set;
based on a layered simulation learning mechanism, learning processing is carried out on the feature set, and KL divergence and the historical feature set are adopted for judgment, so that a strategy model of the historical feature set is obtained, and the strategy model of the historical feature set is optimized;
and learning the strategy model based on a dynamic self-adaptive optimizer and classifying, distinguishing and processing the transformer, the insulator, the breaker and the wire power equipment by adopting a classifier so as to achieve the aim of image recognition.
Optionally, the real-time feature information includes: color, shape, add-on, texture, brightness and orientation.
Optionally, the process of the aggregation treatment includes:
firstly, the graph attention network aggregates feature sets scattered nearby and sets the weight of each node, and the expression is as follows:
Y＝{y1，y2，...，yi}，Yi∈RF
Y′＝{y′1，y′2，...，y′i}，Y′i∈RF
wherein Y and Y' represent feature sets of the input image and the output image of the graph attention network, respectively;
the aggregated and weighted set of features is represented by the attention coefficient of node j relative to node i through the activation function softmax as follows:
in the formula, the correlation function between the node i and the node j is represented by α, and W represents affine transformation.
Optionally, the normalization process includes: obtaining the attention coefficient a after normalization by softmaxijAnd outputting a new feature set by means of average integration, wherein the expression is as follows:
wherein K is greater than or equal to 2; a isij，kAnd WYj，kWeight coefficients and learning parameters respectively representing the kth group attention mechanism; y'iA new feature set representing layer node i; σ denotes the activation function softmax function.
Optionally, the new feature set satisfies the historical features, and if the new feature set satisfies the historical features, the feature set is optimized; otherwise, the learning process of the layered simulation learning mechanism is carried out.
Optionally, the learning process includes:
combining the thought of the layered learning on the basis of the simulated learning, and taking a new data characteristic set as an expected return of the layered simulated learning:
wherein γ is a discount factor;
and automatically generating a characteristic set which is in accordance with the historical characteristic distribution by learning the parameters of the reward function through the adaptive optimizer.
Optionally, the KL divergence determining process includes:
and judging the feature set by adopting KL divergence in graph layering simulation learning, wherein the smaller the KL divergence, the higher the similarity between the feature set representing dynamic input and the historical feature set is.
Optionally, the power device identification process includes: automatically learning the dynamically input feature set by adopting an Adam self-adaptive optimizer, and judging the input feature set through KL divergence; then distinguishing input images by adopting a softmax classifier; and finally, confirming that the input image is a transformer, an insulator, a breaker and a wire power device so as to achieve the purpose of image identification.
The invention has the beneficial effects that:
the invention provides a power equipment identification method for dynamic adaptive graph layered simulation learning, which adopts a multi-head attention mechanism to extract key characteristic information of power equipment, and solves the problems of subjectivity, ambiguity and inaccuracy caused by a manual method; the real-time characteristic information is subjected to aggregation processing and normalization processing through the graph attention network, so that the problem that non-grid data among different devices cannot be modeled can be solved; the feature set is learned through a layered simulation learning mechanism to obtain an optimal strategy model, intelligent analysis can be performed on the acquired image data, and the identification efficiency and accuracy can be improved; on the other hand, the dynamic self-adaptive graph layering simulation learning evaluates the approximate repeated images through KL divergence, the higher the similarity between the input feature set and the historical feature set is, the higher the similarity is, other images are deleted, and therefore the occurrence of a large number of repeated images is avoided.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the application and, together with the description, serve to explain the application and are not intended to limit the application. In the drawings:
fig. 1 is a schematic flow chart of a power equipment identification method for dynamic adaptive graph layered simulation learning according to a first embodiment of the present invention;
FIG. 2 is a schematic diagram illustrating an attention network structure according to a second embodiment of the present invention;
fig. 3 is a schematic diagram of a hierarchical learning simulation structure according to a third embodiment of the present invention.
Detailed Description
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.
It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer-executable instructions and that, although a logical order is illustrated in the flowcharts, in some cases, the steps illustrated or described may be performed in an order different than presented herein.
Example one
As shown in fig. 1, the present embodiment provides a method for identifying a power device by dynamic adaptive graph hierarchical simulation learning, including:
step 1: extracting characteristic information through a graph attention network;
step 2: aggregating the feature sets of the adjacent dispersion through an attention mechanism and setting the weight among the nodes;
and step 3: normalizing the aggregated feature set by softmax to obtain the attention coefficient aij；
And 4, step 4: outputting a new feature set by adopting a multi-head attention mechanism and adopting an average integration mode;
and 5: judging whether the input feature set is consistent with the historical data feature set or not;
step 6: if the images are consistent, an Adam self-adaptive optimizer and a softmax classifier are adopted to achieve the purpose of image recognition;
and 7: if not, taking the new data feature set as an initial strategy of the simulated learning;
and 8: calculating a reward function simulating learning and optimizing a strategy of the reward function;
and step 9: in order to improve the accuracy of solving, the simulation learning is decomposed into a plurality of levels;
step 10: judging by adopting KL divergence, wherein the smaller the KL divergence, the higher the similarity between the distribution of the characteristic samples representing dynamic input and the historical characteristic distribution;
step 11: performing dynamic adaptive learning by adopting an Adam adaptive optimizer;
step 12: and distinguishing by adopting a softmax classifier so as to achieve the purpose of image recognition.
Acquiring characteristic information based on a multi-head attention mechanism to obtain real-time characteristic information of the power equipment;
performing aggregation processing and normalization processing on the real-time characteristic information based on a graph attention network to obtain a characteristic set;
based on a layered simulation learning mechanism, learning processing is carried out on the feature set, and KL divergence and the historical feature set are adopted for judgment, so that a strategy model of the historical feature set is obtained, and the strategy model of the historical feature set is optimized;
and learning the strategy model based on a dynamic self-adaptive optimizer and classifying, distinguishing and processing the transformer, the insulator, the breaker and the wire power equipment by adopting a classifier so as to achieve the aim of image recognition.
In a further optimization scheme, the real-time feature information includes: color, shape, add-on, texture, brightness and orientation.
In a further optimization scheme, the process of the polymerization treatment comprises the following steps:
firstly, the graph attention network aggregates feature sets scattered nearby and sets the weight of each node, and the expression is as follows:
Y＝{y1，y2，...，yi}，Yi∈RF
Y′＝{y′1，y′2，...，y′i}，Yi∈RF′
wherein Y and Y' represent feature sets of the input image and the output image of the graph attention network, respectively;
the aggregated and weighted set of features is represented by the attention coefficient of node j relative to node i through the activation function softmax as follows:
in the formula, the correlation function between the node i and the node j is represented by α, and W represents affine transformation.
Further optimizing the scheme, the normalization process comprises: obtaining the attention coefficient a after normalization by softmaxijAnd outputting a new feature set by means of average integration, wherein the expression is as follows:
wherein K is greater than or equal to 2; a isij，kAnd WYj，kWeight coefficients and learning parameters respectively representing the kth group attention mechanism; y'iA new feature set representing layer node i; σ denotes the activation function softmax function.
Further optimizing the scheme, wherein the new feature set meets the historical features, and if the new feature set meets the historical features, the feature set is optimized; otherwise, the learning process of the layered simulation learning mechanism is carried out.
Further optimizing the scheme, the learning process comprises:
combining the thought of the layered learning on the basis of the simulated learning, and taking a new data characteristic set as an expected return of the layered simulated learning:
wherein γ is a discount factor;
and automatically generating a characteristic set which is in accordance with the historical characteristic distribution by learning the parameters of the reward function through the adaptive optimizer.
Further optimizing the scheme, the KL divergence judgment processing process includes:
and judging the feature set by adopting KL divergence in graph layering simulation learning, wherein the smaller the KL divergence, the higher the similarity between the feature set representing dynamic input and the historical feature set is.
In a further optimization scheme, the power equipment identification process comprises: automatically learning the dynamically input feature set by adopting an Adam self-adaptive optimizer, and judging the input feature set through KL divergence; then distinguishing input images by adopting a softmax classifier; and finally, confirming that the input image is a transformer, an insulator, a breaker and a wire power device so as to achieve the purpose of image identification.
Example two
As shown in fig. 2, the present embodiment provides a method for identifying a power device by dynamic adaptive graph hierarchical simulation learning, including: the graph attention network collects various dynamic information of power equipment in a power system and is represented by graph data of G (Y, A), wherein Y is characteristic information of various equipment, such as color, shape, additive, texture, brightness, direction and other characteristics, and A is a proximity matrix describing nodes of the graph data; firstly, the attention network aggregates the feature sets which are adjacent and dispersed and sets the weight among all nodes; then, in order to improve the extraction capability of the graph attention network, a multi-head attention system is adopted to obtain a multi-head attention coefficient of the node j relative to the node i through the normalization processing of an activation function softmax; and finally, solving a new data feature set by taking an average mode.
EXAMPLE III
As shown in fig. 3, the present embodiment provides a method for identifying a power device by dynamic adaptive graph hierarchical simulation learning, including: simulating learning to generate an initialization strategy according to the historical characteristic data, wherein the initialization strategy is a top-level strategy; decomposing a hierarchy in the current state s, and determining the selection action a by the agent through similarity measurement according to the strategy of the hierarchy1Or to solve a hierarchy; if the action a is selected1Then to understandThe state s of the energy body directly executing the transition to the next momentt+1(ii) a If another level is selected, the selection continues with the new level selected until an action similar to the duration feature is finally derived. And finally, continuously optimizing the strategy through interaction of the underlying strategy and the environment so as to obtain the optimal expected strategy.
Firstly, extracting key characteristic information of the power equipment by adopting an image attention network; then, aggregating the collected characteristic information and outputting a new characteristic set; thirdly, learning by taking the new feature set as an initial strategy through layered simulation learning to obtain a strategy model close to the historical feature set; and finally, performing dynamic self-adaptive learning by adopting an Adam self-adaptive optimizer and classifying by adopting a softmax classifier so as to achieve the aim of image recognition. The invention provides a power equipment identification method for graph layering simulation learning, which can automatically analyze and identify power equipment, improve identification efficiency and accuracy and ensure safe and stable operation of a power system.
The above description is only for the preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (8)
1. A power equipment identification method for dynamic adaptive graph layering simulation learning is characterized by comprising the following steps:
acquiring characteristic information based on a multi-head attention mechanism to obtain real-time characteristic information of the power equipment;
performing aggregation processing and normalization processing on the real-time characteristic information based on a graph attention network to obtain a characteristic set;
based on a layered simulation learning mechanism, learning processing is carried out on the feature set, and KL divergence and the historical feature set are adopted for judgment, so that a strategy model of the historical feature set is obtained, and the strategy model of the historical feature set is optimized;
and learning the strategy model based on a dynamic self-adaptive optimizer and classifying, distinguishing and processing the transformer, the insulator, the breaker and the wire power equipment by adopting a classifier so as to achieve the aim of image recognition.
2. The method of power device identification for dynamic adaptive graph layering mimic learning as claimed in claim 1, wherein the real-time characterization information comprises: color, shape, add-on, texture, brightness and orientation.
3. The method of power device identification for dynamic adaptive graph layering emulation learning of claim 1, wherein the process of the aggregation process comprises:
firstly, the graph attention network aggregates feature sets scattered nearby and sets the weight of each node, and the expression is as follows:
Y＝{y1，y2，...，yi}，Yi∈RF
Y′＝{y′1，y′2，...，y′i}，Y′i∈RF
wherein Y and Y' represent feature sets of the input image and the output image of the graph attention network, respectively;
the aggregated and weighted set of features is represented by the attention coefficient of node j relative to node i through the activation function softmax as follows:
in the formula, the correlation function between the node i and the node j is represented by α, and W represents affine transformation.
4. The method of power device identification for dynamic adaptive graph layering emulation learning of claim 1, wherein the normalization process comprises: obtaining the attention coefficient a after normalization by softmaxijAnd integrated by averagingThe new feature set is output, which is expressed as follows:
wherein K is greater than or equal to 2; a isij，kAnd WYj，kWeight coefficients and learning parameters respectively representing the kth group attention mechanism; y isi' New feature set representing layer node i; σ denotes the activation function softmax function.
5. The method for identifying electric power equipment through dynamic adaptive graph layering simulation learning according to claim 1, wherein the new feature set meets historical features, and if the new feature set meets the historical features, feature set optimization processing is carried out; otherwise, the learning process of the layered simulation learning mechanism is carried out.
6. The method for power equipment identification by dynamic adaptive graph layering emulation learning of claim 1, wherein the learning process comprises:
combining the thought of the layered learning on the basis of the simulated learning, and taking a new data characteristic set as an expected return of the layered simulated learning:
wherein γ is a discount factor;
and automatically generating a characteristic set which is in accordance with the historical characteristic distribution by learning the parameters of the reward function through the adaptive optimizer.
7. The method according to claim 1, wherein the KL divergence determination process includes:
and judging the feature set by adopting KL divergence in graph layering simulation learning, wherein the smaller the KL divergence, the higher the similarity between the feature set representing dynamic input and the historical feature set is.
8. The method of power device identification for dynamic adaptive graph layering emulation learning of claim 1, in which the power device identification process comprises: automatically learning the dynamically input feature set by adopting an Adam self-adaptive optimizer, and judging the input feature set through KL divergence; then distinguishing input images by adopting a softmax classifier; and finally, confirming that the input image is a transformer, an insulator, a breaker and a wire power device so as to achieve the purpose of image identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210048517.2A CN114386524A (en) | 2022-01-17 | 2022-01-17 | Power equipment identification method for dynamic self-adaptive graph layering simulation learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210048517.2A CN114386524A (en) | 2022-01-17 | 2022-01-17 | Power equipment identification method for dynamic self-adaptive graph layering simulation learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114386524A true CN114386524A (en) | 2022-04-22 |
Family
ID=81201732
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210048517.2A Pending CN114386524A (en) | 2022-01-17 | 2022-01-17 | Power equipment identification method for dynamic self-adaptive graph layering simulation learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114386524A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105677852A (en) * | 2016-01-07 | 2016-06-15 | 陕西师范大学 | Personalized healthy diet recommendation service method |
| CN105930868A (en) * | 2016-04-20 | 2016-09-07 | 北京航空航天大学 | Low-resolution airport target detection method based on hierarchical reinforcement learning |
| CN110197317A (en) * | 2018-08-31 | 2019-09-03 | 腾讯科技（深圳）有限公司 | Target user determines method and device, electronic equipment and storage medium |
| CN110378408A (en) * | 2019-07-12 | 2019-10-25 | 台州宏创电力集团有限公司 | Power equipment image-recognizing method and device based on transfer learning and neural network |
| CN111144580A (en) * | 2019-12-31 | 2020-05-12 | 中国电子科技集团公司信息科学研究院 | Hierarchical reinforcement learning training method and device based on simulation learning |
| CN111461038A (en) * | 2020-04-07 | 2020-07-28 | 中北大学 | Pedestrian re-identification method based on layered multi-mode attention mechanism |
| CN112365005A (en) * | 2020-12-11 | 2021-02-12 | 浙江工业大学 | Neuron distribution characteristic-based federal learning poisoning detection method |
| CN112906720A (en) * | 2021-03-19 | 2021-06-04 | 河北工业大学 | Multi-label image identification method based on graph attention network |
| CN113255936A (en) * | 2021-05-28 | 2021-08-13 | 浙江工业大学 | Deep reinforcement learning strategy protection defense method and device based on simulation learning and attention mechanism |
| CN113628059A (en) * | 2021-07-14 | 2021-11-09 | 武汉大学 | Associated user identification method and device based on multilayer graph attention network |
| CN113693563A (en) * | 2021-08-27 | 2021-11-26 | 北京工业大学 | Brain function network classification method based on hypergraph attention network |
-
2022
- 2022-01-17 CN CN202210048517.2A patent/CN114386524A/en active Pending
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105677852A (en) * | 2016-01-07 | 2016-06-15 | 陕西师范大学 | Personalized healthy diet recommendation service method |
| CN105930868A (en) * | 2016-04-20 | 2016-09-07 | 北京航空航天大学 | Low-resolution airport target detection method based on hierarchical reinforcement learning |
| CN110197317A (en) * | 2018-08-31 | 2019-09-03 | 腾讯科技（深圳）有限公司 | Target user determines method and device, electronic equipment and storage medium |
| CN110378408A (en) * | 2019-07-12 | 2019-10-25 | 台州宏创电力集团有限公司 | Power equipment image-recognizing method and device based on transfer learning and neural network |
| CN111144580A (en) * | 2019-12-31 | 2020-05-12 | 中国电子科技集团公司信息科学研究院 | Hierarchical reinforcement learning training method and device based on simulation learning |
| CN111461038A (en) * | 2020-04-07 | 2020-07-28 | 中北大学 | Pedestrian re-identification method based on layered multi-mode attention mechanism |
| CN112365005A (en) * | 2020-12-11 | 2021-02-12 | 浙江工业大学 | Neuron distribution characteristic-based federal learning poisoning detection method |
| CN112906720A (en) * | 2021-03-19 | 2021-06-04 | 河北工业大学 | Multi-label image identification method based on graph attention network |
| CN113255936A (en) * | 2021-05-28 | 2021-08-13 | 浙江工业大学 | Deep reinforcement learning strategy protection defense method and device based on simulation learning and attention mechanism |
| CN113628059A (en) * | 2021-07-14 | 2021-11-09 | 武汉大学 | Associated user identification method and device based on multilayer graph attention network |
| CN113693563A (en) * | 2021-08-27 | 2021-11-26 | 北京工业大学 | Brain function network classification method based on hypergraph attention network |
Non-Patent Citations (5)
| Title |
|---|
| 李佳玮等: "基于图注意力网络的配电网故障定位方法", 《电网技术》 * |
| 李帅龙等: "模仿学习方法综述及其在机器人领域的应用", 《计算机工程与应用》 * |
| 牛斌等: "一种基于注意力机制与多模态的图像描述方法", 《辽宁大学学报(自然科学版)》 * |
| 钟智等: "基于图注意力深度网络的电力系统暂态稳定评估", 《电网技术》 * |
| 高明柯等: "基于注意力机制和特征融合的手势识别方法", 《计算机应用与软件》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110598736B (en) | Power equipment infrared image fault positioning, identifying and predicting method | |
| CN110188635B (en) | Plant disease and insect pest identification method based on attention mechanism and multi-level convolution characteristics | |
| CN112163465B (en) | Fine-grained image classification method, fine-grained image classification system, computer equipment and storage medium | |
| CN107944396B (en) | Knife switch state identification method based on improved deep learning | |
| CN109215034B (en) | Weak supervision image semantic segmentation method based on spatial pyramid covering pooling | |
| CN104268593A (en) | Multiple-sparse-representation face recognition method for solving small sample size problem | |
| CN109614488B (en) | Text classification and image recognition-based distribution network live working condition judgment method | |
| CN109389171B (en) | Medical image classification method based on multi-granularity convolution noise reduction automatic encoder technology | |
| CN110212528A (en) | Reconstructing method is lacked based on the power distribution network metric data for generating confrontation and dual Semantic Aware | |
| She et al. | Insulator breakage detection utilizing a convolutional neural network ensemble implemented with small sample data augmentation and transfer learning | |
| CN114169445A (en) | Day-ahead photovoltaic power prediction method, device and system based on CAE and GAN hybrid network | |
| CN111540203B (en) | Method for adjusting green light passing time based on fast-RCNN | |
| CN110334584A (en) | A kind of gesture identification method based on the full convolutional network in region | |
| CN115564194A (en) | Method and system for constructing metering abnormality diagnosis information generation model of smart power grid | |
| CN108985385A (en) | Based on the quick Weakly supervised object detection method for generating confrontation study | |
| Hu et al. | Multiobjective image color quantization algorithm based on self-adaptive hybrid differential evolution | |
| CN113221614A (en) | Power grid insulator damage image identification method based on hybrid neural network | |
| CN114386524A (en) | Power equipment identification method for dynamic self-adaptive graph layering simulation learning | |
| Ling et al. | A facial expression recognition system for smart learning based on YOLO and vision transformer | |
| CN110610203A (en) | Electric energy quality disturbance classification method based on DWT and extreme learning machine | |
| CN113554010B (en) | Power grid line fault recognition model training method | |
| CN112529637B (en) | Service demand dynamic prediction method and system based on context awareness | |
| CN115144211A (en) | Fault detection method and device based on discrete wavelet transform and gated cyclic unit | |
| CN114372418A (en) | Wind power space-time situation description model establishing method | |
| CN108154175B (en) | Method for accurately identifying wave recording multiple working conditions of power distribution network |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220422 |