CN113721162A - Fusion magnet power failure intelligent diagnosis method based on deep learning - Google Patents
Fusion magnet power failure intelligent diagnosis method based on deep learning Download PDFInfo
- Publication number
- CN113721162A CN113721162A CN202110998405.9A CN202110998405A CN113721162A CN 113721162 A CN113721162 A CN 113721162A CN 202110998405 A CN202110998405 A CN 202110998405A CN 113721162 A CN113721162 A CN 113721162A
- Authority
- CN
- China
- Prior art keywords
- fusion
- layer
- network
- power supply
- noise reduction
- 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
- 230000004927 fusion Effects 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000003745 diagnosis Methods 0.000 title claims abstract description 29
- 238000013135 deep learning Methods 0.000 title claims abstract description 11
- 238000012549 training Methods 0.000 claims abstract description 34
- 230000009467 reduction Effects 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 11
- 238000013528 artificial neural network Methods 0.000 claims abstract description 3
- 239000004576 sand Substances 0.000 claims description 11
- 238000013527 convolutional neural network Methods 0.000 claims description 10
- 230000006870 function Effects 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 2
- 210000002569 neuron Anatomy 0.000 claims description 2
- 238000010606 normalization Methods 0.000 claims description 2
- 238000011176 pooling Methods 0.000 claims description 2
- 238000013461 design Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 5
- 230000005012 migration Effects 0.000 description 3
- 238000013508 migration Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 2
- 238000010168 coupling process Methods 0.000 description 2
- 238000005859 coupling reaction Methods 0.000 description 2
- 238000005096 rolling process Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000010791 quenching Methods 0.000 description 1
- 230000000171 quenching effect Effects 0.000 description 1
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
- G01R31/40—Testing power supplies
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention discloses a fusion magnet power failure intelligent diagnosis method based on deep learning, which uses thyristor three-phase full-control bridge converter unit to output voltage UdAnd an output IdThe current is the input signal of fault diagnosis, and passes through a depth noise reduction self-encoder F comprising seven fully-connected layersDAfter noise reduction processing is carried out, real-time diagnosis is carried out on inversion failure and pulse loss fault types of the fusion superconducting magnet high-power supply system through a double-path heterogeneous deep convolution neural network, and intelligent fault diagnosis of the fusion superconducting magnet high-power supply system under the condition of small samples is achieved. By using the method, two controllable fault types of inversion failure and pulse loss of the fusion superconducting magnet power supply system can be diagnosed in real time based on a small amount of training data, so that high recognition rate is achieved, controllable faults can be eliminated by adopting a control means, and unnecessary shutdown of the fusion superconducting magnet power supply is avoided.
Description
Technical Field
The invention belongs to the technical field of fusion magnet power supply control, and particularly relates to an intelligent fault diagnosis method for a high-power supply of a fusion superconducting magnet.
Background
A fusion superconducting magnet power supply system is a key subsystem of a magnetic confinement nuclear fusion device, and the fusion superconducting magnet power supply system has the function of providing indispensable support for the generation, confinement, maintenance, heating and control of plasma. The fusion superconducting magnet power supply system consists of a plurality of groups of super-power thyristor converters, the total installed capacity is up to thousands of megavolt-ampere, and at present, the fault protection of the fusion superconducting magnet power supply system adopts the conventional overvoltage and overcurrent detection in engineering. And when the current or the voltage exceeds a certain value, the protection is carried out by adopting an emergency shutdown mode. And the plasma hard fracture can be caused by the emergency shutdown of the power supply system, so that the circulation current of the vacuum chamber is seriously damaged due to the coupling of tens of thousands of amperes, and meanwhile, the current rapid change caused by the plasma hard fracture can be coupled to the poloidal field superconducting magnet through mutual inductance to induce the quenching of the poloidal field superconducting magnet, so that the safety of the device is seriously threatened, and meanwhile, a large amount of impurities are brought into the vacuum chamber by the plasma hard fracture, so that the plasma quality and the experimental efficiency are seriously influenced. Therefore, the power supply fault type is diagnosed in real time, the controllable fault is eliminated through a control means, and the uncontrollable fault is intervened in advance, so that the method is not only the key of the stable operation of the power supply, but also the powerful guarantee of the safe and stable operation of the device.
Disclosure of Invention
In order to solve the technical problem, the invention provides an intelligent fusion magnet power failure diagnosis method based on deep learning. According to the method, deep network knowledge migration which is constructed in other fields and trained under the support of a large amount of data is applied to fault diagnosis through migration learning, the problem of difference of data characteristics in different fields is solved through two-way heterogeneous network design, and a designed target domain network is more applicable due to less target domain samples. A method for carrying out advanced denoising on sample data by a denoising encoder is designed, and the identification stability and the denoising performance of the method are improved. Through testing, the method can be used for diagnosing two controllable fault types, namely inversion failure and pulse loss, of the fusion superconducting magnet power supply system in real time based on a small amount of training data, and achieves high recognition rate, so that the controllable faults can be eliminated by adopting a control means, and unnecessary shutdown of the fusion superconducting magnet power supply is avoided.
The technical scheme adopted by the invention is as follows:
a fusion magnet power supply fault intelligent diagnosis method based on deep learning is characterized in that a fusion magnet power supply is composed of thyristor three-phase full-controlled bridge converter units which are connected in series or in parallel, and each converter unit outputs 6 pulse wave voltage UdAnd output IdCurrent following a given current I in the rated current rangerefThe method comprises the following steps:
Step 2, signal U after noise reduction processingd' and IdThe method comprises the steps of performing real-time diagnosis on inversion failure and pulse loss fault types of a fusion superconducting magnet high-power supply system through a dual-path heterogeneous deep convolution neural network, and realizing intelligent fault diagnosis of the fusion superconducting magnet high-power supply system under the condition of small samples.
The invention has the beneficial effects that:
(1) according to the method, a large amount of historical data of the existing similar system in normal operation is used as support through transfer learning, deep network knowledge transfer model training is carried out, and the deep network knowledge transfer model training is applied to fault diagnosis; the method is embodied in the training process of step four, namely, firstly, the existing fault diagnosis data of the rolling bearing are utilized to form a training data set, and a network F is subjected tosTraining, combining the source domain model and the target domain model into a two-way heterogeneous deep convolutional neural network, and performing real data-based alignment on the target network FDAnd (5) training.
(2) Through the design of a double-path heterogeneous network, the problem of difference of data characteristics in different fields is avoided. Due to the fact that the target domain samples are few, the designed target domain network is more applicable.
(3) Aiming at multiple interference sources, strong noise and multiple fault coupling signals of a fusion magnet power supply, the invention designs a method for carrying out early denoising on sample data by a denoising encoder, and improves the identification stability and the denoising performance of the method. After testing, the method can be used for diagnosing two controllable fault types, namely inversion failure and pulse loss, of the fusion superconducting magnet power supply system in real time based on a small amount of training data, and the recognition rate is more than 95%, so that the controllable faults can be eliminated by adopting a control means, unnecessary shutdown of the fusion superconducting magnet power supply is avoided, and the safety of the fusion magnet power supply system and the fusion superconducting magnet power supply device is improved.
Drawings
FIG. 1 fusion magnet power topology;
FIG. 2 noise reduction self-encoder network FDAnd (5) structure.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments, and all other embodiments obtained by a person skilled in the art based on the embodiments of the present invention belong to the protection scope of the present invention without creative efforts.
According to the embodiment of the invention, the fusion magnet power supply fault intelligent diagnosis method based on deep learning is provided, the fusion superconducting magnet high-power supply system is formed by connecting converter units of a thyristor three-phase full-control bridge in series or in parallel, and each converter unit outputs 6 pulse wave voltage UdAnd output IdThe current can follow a given current I in the rated current rangeref. As shown in fig. 1, a fusion magnet power supply topology;
the invention relates to an intelligent fault diagnosis method, which uses the voltage U of a converter unitdAnd current IdAs input signal for fault diagnosis, via a deep noise reduction self-encoder F comprising seven fully connected layersDTo UdAnd IdNoise reduction processing is carried out, and the signal U after the noise reduction processing is carried outd' and IdThe inverse loss of the fusion superconducting magnet high-power supply system is realized through a two-way heterogeneous deep convolution neural networkFailure and pulse loss fault types are diagnosed in real time, and intelligent fault diagnosis of the fusion superconducting magnet high-power supply system is achieved under the condition of small samples. The method specifically comprises the following steps:
the method comprises the following steps: designing a noise reduction self-encoder
Firstly, designing a depth noise reduction self-encoder F containing seven fully-connected layersDThe structure is shown in FIG. 2, which is Input Layer (None,3000) -)>Dense(None,128)—>Dense(None,64)—>Dense(None,32)—>Dense(None,64)—>Dense(None,128)—>Dense (None,3000) outputs a noise reduction signal y 1. The Dense Layer is used for fully connecting the neurons of the previous Layer to realize the linear combination of the characteristics, the Input Layer is used as an Input Layer, and the Dense Layer is used as a fully-connected Layer. In the invention, 6 Dense layers are used successively, the first 3 layers realize the coding effect, and the second 3 layers realize the decoding effect. Coding here means the process of converting the original signal through 3 fully-connected layers into another form, i.e. reducing the original dimension from 1 x 3000 to 1 x 32. Decoding is the inverse process of encoding, namely, the dimension-reduced 1 x 32 dimensional data is increased into 1 x 3000 dimensional data through the last 3 fully-connected layers. Therefore, the whole sense layer is the process for realizing coding and decoding.
Step two: designing a dual-path heterogeneous deep convolutional neural network
The invention designs a double-path heterogeneous deep convolutional neural network which comprises two deep convolutional networks, namely a more complex source domain convolutional network FsAnd a relatively simple target domain convolutional network FtThe two outputs are respectively ysAnd yT。
FsThe network structure is as follows: conv1d —>BatchNorm1d—>ReLU—>MaxPool1d—>Conv1d—>BatchNorm1d—>ReLU—>MaxPool1d—>Conv1d—>BatchNorm1d—>ReLU—>Conv1d—>BatchNorm1d—>ReLU—>Conv1d—>BatchNorm1d—>ReLU—>Linear
FtThe network structure is as follows: conv1d —>ReLU—>Conv1d—>ReLU—>Linear
Here, Conv1d denotes a convolutional layer, BatchNorm1d is a batch normalization layer, MaxPool1d is a one-dimensional pooling layer, ReLU is an activation function, and Linear is a fully-connected layer output layer.
FsAnd FtThe parameters of each layer are as follows:
the design loss function loss is as follows:
loss＝(yS-yT)2
when training, will FsAnd FtOutput ysAnd yTThe loss function loss is connected, and the calculated loss returns to the network FtContinuing training; when the application is classified, F is directly usedtAnd calculating output.
Step three: noise reduction self-encoder training and use
Selecting training sample x without noise from sample base as noise reduction self-encoder FDIntroducing a noise-adding process to pollute the training sample into x1And as FDThe network adopts ADMM algorithm to realize optimization, and output signal y1The loss function is set to (y)1-x)2. After training, network FDThe noise reduction processing may be performed directly on the input signal. The depth noise reduction self-encoder can transform the original signal layer by layer, automatically learn hierarchical feature representation, and avoid the problem of difficult feature selection caused by manual design of a noise reduction algorithm. Meanwhile, the symmetrical network design of the seven fully-connected layers can give consideration to both the calculation efficiency and the denoising effect, and a relatively deeper or shallower network has advantages for the fusion magnet power failure data.
Step four: training of two-way heterogeneous deep convolutional neural network
The training of the two-way heterogeneous deep convolutional neural network comprises the following two steps:
1) training a complex source domain network Fs。
Based on a rolling bearing fault diagnosis data set disclosed by Kaiser university, data with the same length (sampling frequency of 10kHz) as real fault diagnosis data is obtained by sampling, fault types are marked to form a training data set, and the training data set is input into a network FsTo network FsAnd (5) training.
2) Training target domain network FD。
The source domain model and the target domain model are combined into a two-path heterogeneous deep convolutional neural network, parameters of bottom-layer branches of the source domain are kept fixed, and two branches of the network are constrained by a cost function loss to generate the same output for the same input data.
The training of the source domain network and the target domain network needs to reasonably set parameters, and the method is set as follows:
1) training FsThe required data consists of 2000 data samples in class 2, training FDA total of 100 data samples of class 2 are used, which is a typical small sample case.
2)FsBatch size at training 32, FDThe batch processing size is batch _ size 2 during training;
3)Fsand FDThe optimization objective function is binary cross entropy (binary _ cross), the optimization method is a random gradient descent method with momentum, the step size is 0.001, the momentum size is set to 0.9, and the maximum training round number epochs is set to 150.
4) To FDH 5. after the training converged, the trained model is saved as model.
By adopting the two-path heterogeneous deep convolutional neural network, the problem of difference of data characteristics in different fields can be avoided. The two-way heterogeneous deep convolutional network relates to a complex deep network FsAnd a simple shallow network FD. In the case of less fusion superconducting magnet power failure data in practical application, if F is directly usedsTraining may be overfitting, but with a shallow network FDUnder-fitting may occur; and if applicable FsThe transfer learning may be based on different domain dataThe characteristics present a domain drift problem. The patent provides a double-path heterogeneous deep convolutional network, which can use the existing massive data in other fields to train Fs, and then directly train F by using the Fs and fusion superconducting magnet power failure data togetherDNot only can avoid directly using FsThe migration training has the problems of different data types and obvious differences in different fields, and meanwhile, the shallow network FDThe method is not under-fitted and is more applicable.
Step five: data testing
H1, loading a trained model, reading in data to be tested, outputting a test result, calculating test accuracy, and testing two faults of pulse loss and inversion failure in total to obtain 97% accuracy on 400 data.
Through testing, the method can be used for diagnosing two controllable fault types, namely inversion failure and pulse loss, of the fusion superconducting magnet power supply system in real time based on a small amount of training data, and achieves high recognition rate, so that the controllable faults can be eliminated by adopting a control means, and unnecessary shutdown of the fusion superconducting magnet power supply is avoided.
Although illustrative embodiments of the present invention have been described above to facilitate the understanding of the present invention by those skilled in the art, it should be understood that the present invention is not limited to the scope of the embodiments, but various changes may be apparent to those skilled in the art, and it is intended that all inventive concepts utilizing the inventive concepts set forth herein be protected without departing from the spirit and scope of the present invention as defined and limited by the appended claims.
Claims (5)
1. A fusion magnet power supply fault intelligent diagnosis method based on deep learning is characterized in that a fusion magnet power supply is composed of thyristor three-phase full-controlled bridge converter units which are connected in series or in parallel, and each converter unit outputs 6 pulse wave voltage UdAnd output IdCurrent following a given current I in the rated current rangerefThe method is characterized by comprising the following steps:
step 1, using voltage U of converter unitdAnd current IdDeep noise reduction self-encoder F as input signal for fault diagnosis and input to seven-layer full-connection layerDTo UdAnd IdCarrying out noise reduction processing to obtain a signal U subjected to noise reduction processingd' and Id’；
Step 2, signal U after noise reduction processingd' and IdThe method comprises the steps of performing real-time diagnosis on inversion failure and pulse loss fault types of a fusion superconducting magnet high-power supply system through a dual-path heterogeneous deep convolution neural network, and realizing intelligent fault diagnosis of the fusion superconducting magnet high-power supply system under the condition of small samples.
2. A fusion magnet power failure intelligent diagnosis method based on deep learning according to claim 1, characterized in that: the depth noise reduction self-encoder F of the seven fully-connected layers in the step 1DThe structure is Input Layer (None,3000) -)>Dense(None,128)—>Dense(None,64)—>Dense(None,32)—>Dense(None,64)—>Dense(None,128)—>Dense (None,3000) and output the noise reduction signal as y1The system comprises 6 sense layers, wherein the first 3 sense layers realize an encoding effect, the last 3 sense layers realize a decoding effect, the sense layers are in full connection with neurons of the previous Layer to realize linear combination of characteristics, the Input Layer is an Input Layer, and the sense layers are full connection layers.
3. A fusion magnet power failure intelligent diagnosis method based on deep learning according to claim 1, characterized in that: the two-way heterogeneous deep convolutional neural network comprises a source domain convolutional network FsOutput is ysAnd a target domain convolutional network FtOutput is yT；
Source domain convolutional network FsThe network structure of (1) is: conv1d —>BatchNorm1d—>ReLU—>MaxPool1d—>Conv1d—>BatchNorm1d—>ReLU—>MaxPool1d—>Conv1d—>BatchNorm1d—>ReLU—>Conv1d—>BatchNorm1d—>ReLU—>Conv1d—>BatchNorm1d—>ReLU—>Linear；
Target domain convolutional network FtThe network structure of (1) is: conv1d —>ReLU—>Conv1d—>ReLU—>Linear；
Wherein Conv1d is a convolutional layer, BatchNorm1d is a batch normalization layer, MaxPool1d is a one-dimensional pooling layer, ReLU is an activation function, and Linear is a fully-connected layer output layer.
4. A fusion magnet power failure intelligent diagnosis method based on deep learning according to claim 1, characterized in that: the two-way heterogeneous deep convolutional neural network is a source domain convolutional network FsConvolution network F with target domaintThe parameters of each layer are as follows:
5. a fusion magnet power failure intelligent diagnosis method based on deep learning according to claim 1, characterized in that: the two-way heterogeneous deep convolutional neural network has the loss function loss:
loss＝(yS-yT)2
wherein, F is set during trainingsAnd FtOutput ysAnd yTThe loss function loss is connected, and the calculated loss returns to the network FtContinuing training, and directly using F when classifying applicationstAnd calculating output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110998405.9A CN113721162B (en) | 2021-08-27 | 2021-08-27 | Fusion magnet power failure intelligent diagnosis method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110998405.9A CN113721162B (en) | 2021-08-27 | 2021-08-27 | Fusion magnet power failure intelligent diagnosis method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113721162A true CN113721162A (en) | 2021-11-30 |
| CN113721162B CN113721162B (en) | 2023-10-24 |
Family
ID=78678652
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110998405.9A Active CN113721162B (en) | 2021-08-27 | 2021-08-27 | Fusion magnet power failure intelligent diagnosis method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113721162B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5598509A (en) * | 1992-08-28 | 1997-01-28 | Hitachi, Ltd. | Method of configuring a neural network and a diagnosis/recognition system using the same |
| CN107271925A (en) * | 2017-06-26 | 2017-10-20 | 湘潭大学 | The level converter Fault Locating Method of modularization five based on depth convolutional network |
| US20190242936A1 (en) * | 2018-02-05 | 2019-08-08 | Wuhan University | Fault diagnosis method for series hybrid electric vehicle ac/dc converter |
| CN110208721A (en) * | 2019-07-09 | 2019-09-06 | 西南交通大学 | A kind of method for diagnosing faults and device cascading three-level inverter |
| CN210222221U (en) * | 2019-07-09 | 2020-03-31 | 西南交通大学 | Fault diagnosis device of cascaded three-level inverter |
| CN112465030A (en) * | 2020-11-28 | 2021-03-09 | 河南大学 | Multi-source heterogeneous information fusion fault diagnosis method based on two-stage transfer learning |
| CN113125992A (en) * | 2021-04-23 | 2021-07-16 | 合肥工业大学 | NPC three-level inverter fault diagnosis method and system based on DBN |
| CN113283419A (en) * | 2021-04-29 | 2021-08-20 | 国网浙江省电力有限公司湖州供电公司 | Convolutional neural network pointer instrument image reading identification method based on attention |
-
2021
- 2021-08-27 CN CN202110998405.9A patent/CN113721162B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5598509A (en) * | 1992-08-28 | 1997-01-28 | Hitachi, Ltd. | Method of configuring a neural network and a diagnosis/recognition system using the same |
| CN107271925A (en) * | 2017-06-26 | 2017-10-20 | 湘潭大学 | The level converter Fault Locating Method of modularization five based on depth convolutional network |
| US20190242936A1 (en) * | 2018-02-05 | 2019-08-08 | Wuhan University | Fault diagnosis method for series hybrid electric vehicle ac/dc converter |
| CN110208721A (en) * | 2019-07-09 | 2019-09-06 | 西南交通大学 | A kind of method for diagnosing faults and device cascading three-level inverter |
| CN210222221U (en) * | 2019-07-09 | 2020-03-31 | 西南交通大学 | Fault diagnosis device of cascaded three-level inverter |
| CN112465030A (en) * | 2020-11-28 | 2021-03-09 | 河南大学 | Multi-source heterogeneous information fusion fault diagnosis method based on two-stage transfer learning |
| CN113125992A (en) * | 2021-04-23 | 2021-07-16 | 合肥工业大学 | NPC three-level inverter fault diagnosis method and system based on DBN |
| CN113283419A (en) * | 2021-04-29 | 2021-08-20 | 国网浙江省电力有限公司湖州供电公司 | Convolutional neural network pointer instrument image reading identification method based on attention |
Non-Patent Citations (3)
| Title |
|---|
| 张怀峰 等: "面向卫星电源系统的一种新颖异常检测方法", 《宇航学报》, no. 12, pages 90 - 99 * |
| 杭芹: "用于聚变电源的故障诊断算法研究", 《中国博士学位论文全文数据库工程科技Ⅱ辑》, no. 8, pages 12 - 13 * |
| 陈华坤 等: "航空电子设备故障预测特征参数提取方法研究", 《西北工业大学学报》, vol. 35, no. 3, pages 364 - 372 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113721162B (en) | 2023-10-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110866604A (en) | Cleaning method for power transformer state monitoring data | |
| CN110879373B (en) | Oil-immersed transformer fault diagnosis method with neural network and decision fusion | |
| Li et al. | Transformer fault diagnosis based on improved deep coupled dense convolutional neural network | |
| CN116562114A (en) | Power transformer fault diagnosis method based on graph convolution neural network | |
| Yin et al. | Deep learning based feature reduction for power system transient stability assessment | |
| CN116071352A (en) | Method for generating surface defect image of electric power safety tool | |
| Zhao et al. | Multibranch 1-D CNN based on attention mechanism for the DAB converter fault diagnosis | |
| Zhang et al. | Encoding time series as images: A robust and transferable framework for power system DIM identification combining rules and VGGNet | |
| Shen et al. | Online Open-Circuit Fault Diagnosis for Neutral Point Clamped Inverter Based on an Improved Convolutional Neural Network and Sample Amplification Method Under Varying Operating Conditions | |
| Shu et al. | DBF-CNN: A double-branch fusion residual CNN for diagnosis of induction motor broken rotor bar | |
| CN113721162A (en) | Fusion magnet power failure intelligent diagnosis method based on deep learning | |
| CN117592353A (en) | Safety assessment method based on big data characteristic evolution mechanism of battery failure behavior | |
| Wang et al. | Research on double-deck traceability identification method of commutation failure in HVDC system | |
| CN114896883B (en) | Transformer fault diagnosis method based on MEA-SVM classifier | |
| CN116664870A (en) | Power quality disturbance identification method based on transfer learning | |
| He et al. | Partial discharge pattern recognition algorithm based on sparse self-coding and extreme learning machine | |
| Liu et al. | A comparative research on power transformer fault diagnosis based on several artificial neural networks | |
| Liu et al. | Method of extracting gear fault feature based on stacked autoencoder | |
| Hu et al. | Partial discharge pattern recognition of GIS based on CBAM-ResNet | |
| Liu et al. | FSR‐SSL: A fault sample rebalancing framework based on semi‐supervised learning for PV fault diagnosis | |
| CN116304918A (en) | Substation equipment fault identification method and system based on depth forest algorithm | |
| Qu et al. | Study on multi-rbf-svm for transformer fault diagnosis | |
| Wang et al. | Turn-to-turn short circuit of motor stator fault diagnosis using dropout rate improved deep sparse autoencoder | |
| Gu et al. | Power system transient stability assessment based on svm-smote and one-dimensional convolutional neural network | |
| Ding | A Fault Diagnosis Model Based on Attention Convolution 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |