CN116961072B - Multi-DC chain commutation failure recognition method and system based on space-time convolution network - Google Patents
Multi-DC chain commutation failure recognition method and system based on space-time convolution network Download PDFInfo
- Publication number
- CN116961072B CN116961072B CN202310854905.4A CN202310854905A CN116961072B CN 116961072 B CN116961072 B CN 116961072B CN 202310854905 A CN202310854905 A CN 202310854905A CN 116961072 B CN116961072 B CN 116961072B
- Authority
- CN
- China
- Prior art keywords
- time
- space
- commutation failure
- data
- direct current
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 84
- 238000012549 training Methods 0.000 claims abstract description 59
- 239000011159 matrix material Substances 0.000 claims abstract description 51
- 230000008569 process Effects 0.000 claims abstract description 37
- 238000002372 labelling Methods 0.000 claims abstract description 11
- 238000004088 simulation Methods 0.000 claims abstract description 11
- 230000006870 function Effects 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 14
- 238000003860 storage Methods 0.000 claims description 10
- 230000008859 change Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000006399 behavior Effects 0.000 claims description 2
- 238000013528 artificial neural network Methods 0.000 description 22
- 238000013527 convolutional neural network Methods 0.000 description 15
- 238000012360 testing method Methods 0.000 description 14
- 238000010586 diagram Methods 0.000 description 12
- 238000004364 calculation method Methods 0.000 description 7
- 230000008878 coupling Effects 0.000 description 5
- 238000010168 coupling process Methods 0.000 description 5
- 238000005859 coupling reaction Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 101100499229 Mus musculus Dhrsx gene Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000005477 standard model Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 239000011800 void material Substances 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/36—Arrangements for transfer of electric power between ac networks via a high-tension dc link
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- 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
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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
- G06N3/0985—Hyperparameter optimisation; Meta-learning; Learning-to-learn
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Power Engineering (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the field of sequential commutation failure recognition, and provides a multi-direct-current sequential commutation failure recognition method and system based on a space-time convolution network. The method comprises the steps of obtaining historical/simulation generated inverter station converter bus voltage and direct current data; extracting the inversion station converter bus voltage and direct current data according to a certain time interval to obtain a data matrix, labeling data elements in the data matrix with labels, and converting the data matrix into a sample graph; based on the sample graph, adopting a graph convolution network layer to extract spatial features and obtain a low-order feature graph; based on the low-order feature map, extracting time sequence features by adopting a time convolution network layer to obtain a high-order feature map; the high-order feature map passes through a full connection layer and an output layer, and the recognition result of the hoop failure in the training process is predicted; based on the recognition result and the label of the failure of the training process connection loop direction, the super-parameters of the space-time convolution network are optimized in the training process by combining the loss function.
Description
Technical Field
The invention relates to the field of multi-DC-feed receiving-end power grid chained commutation failure recognition, in particular to a multi-DC chained commutation failure recognition method and system based on a space-time convolution network.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Grid commutated high voltage direct current transmission systems (LCC-HVDC, line Commutated Converter High Voltage Direct Current) are widely used in long distance and high capacity transmission due to their cost and technical advantages of trans-regional transmission. But LCC-HVDC has the risk of commutation failure because of its thyristor being a semi-controlled device. With more and more direct current projects densely fed into the receiving-end power grid, the receiving-end area gradually presents a 'strong-straight weak-alternating-current' pattern, and the supporting capability of the alternating-current power grid on the multi-direct-current system is reduced. When the AC system has serious faults, if the voltage drop of the nearby DC commutation bus is caused, the DC commutation failure can be caused; under the action of the direct current control system, the direct current can adjust the trigger angle to avoid commutation failure, but a great amount of reactive power is absorbed from the alternating current system, so that the voltage level of other nearby direct current commutation buses can be further deteriorated, and the commutation failure is propagated among different direct currents, namely, the chain commutation failure occurs.
At present, recognition of chain commutation failure mainly surrounds two aspects: on one hand, the coupling relation strength between the direct current conversion buses is judged, and a corresponding threshold value is set for the coupling relation. When one of the direct currents has commutation failure, and the coupling strength between the direct currents exceeds a threshold value, judging that the chain commutation fails; on the other hand, the information such as the electrical network parameters and the fault positions are used as the input of the neural network, and the commutation failure duration time of different direct currents is predicted to judge whether the cascading commutation failure occurs.
However, in the aspect of identifying the problem of the chain commutation failure, the method for judging the chain commutation failure by setting the coupling threshold has the advantages of intuitiveness and rapidness in judgment, but the setting of the threshold is based on a large amount of practical experience, and the identification accuracy can be reduced to different degrees for different multi-direct current feed-in scenes. The method based on the neural network judges the chained commutation failure by combining the power grid information and the fault information, and overcomes the defect of insufficient adaptability of a threshold value setting method by collecting enough system training samples, but only considers the spatial characteristics among data in the chained commutation failure process, and the accuracy of the chained commutation failure identification is to be improved.
Disclosure of Invention
In order to solve the technical problems in the prior art, the invention provides a multi-DC chain commutation failure identification method and system based on a space-time convolution network, which can identify the multi-DC chain commutation failure in a short time under a multi-DC feed-in scene so as to take countermeasures in time.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
the first aspect of the invention provides a multi-DC chain commutation failure identification method based on a space-time convolution network.
A multi-DC chain commutation failure identification method based on a space-time convolution network comprises the following steps:
Acquiring historical/simulation generated inverter station converter bus voltage and direct current data;
extracting the inversion station converter bus voltage and direct current data according to a certain time interval to obtain a data matrix, labeling data elements in the data matrix with labels, and converting the data matrix into a sample graph;
Based on the sample graph, adopting a graph convolution network layer to extract spatial features and obtain a low-order feature graph; based on the low-order feature map, extracting time sequence features by adopting a time convolution network layer to obtain a high-order feature map; the high-order feature map passes through a full connection layer and an output layer, and the recognition result of the chain commutation failure in the training process is predicted;
Based on the recognition result and the label of the training process chain commutation failure, optimizing the super-parameters of the space-time convolution network in the training process by combining the loss function;
And acquiring real-time inversion station commutation bus voltage and direct current data, and acquiring a recognition result about the chain commutation failure by adopting a trained space-time convolutional neural network.
Further, the behavior time of the data matrix is that the columns of the data matrix are inverter station converter bus voltage and direct current data.
Further, in the process of acquiring the history/simulation generated inverter station converter bus voltage and direct current data, the change condition of each direct current inverter station arc extinguishing angle in the corresponding time is recorded.
Still further, the labeling of the data elements in the data matrix includes: and marking the elements in the data matrix by adopting a conservative marking strategy according to the change condition of the arc extinguishing angle of each direct current inversion station in the corresponding time.
Further, the space-time convolution network comprises a graph convolution network layer, two time convolution network layers, a full connection layer and an output layer.
Further, the process of converting the data matrix into a sample graph includes: and carrying out normalization and graying treatment on the data matrix, and converting the data matrix into a gray scale image, namely a sample image.
Further, the time convolution network layer comprises a plurality of time convolution layers with different hole coefficient combinations and residual connection, and time sequence features of different time scales are extracted through the different time convolution layers.
The second aspect of the invention provides a multi-DC chain commutation failure recognition system based on a space-time convolution network.
A multi-DC chain commutation failure recognition system based on a space-time convolution network comprises:
a data acquisition module configured to: acquiring historical/simulation generated inverter station converter bus voltage and direct current data;
A preprocessing module configured to: extracting the inversion station converter bus voltage and direct current data according to a certain time interval to obtain a data matrix, labeling data elements in the data matrix with labels, and converting the data matrix into a sample graph;
A network training module configured to: based on the sample graph, adopting a graph convolution network layer to extract spatial features and obtain a low-order feature graph; based on the low-order feature map, extracting time sequence features by adopting a time convolution network layer to obtain a high-order feature map; the high-order feature map passes through a full connection layer and an output layer, and the recognition result of the chain commutation failure in the training process is predicted;
an optimization module configured to: based on the recognition result and the label of the training process chain commutation failure, optimizing the super-parameters of the space-time convolution network in the training process by combining the loss function;
An identification module configured to: and acquiring real-time inversion station commutation bus voltage and direct current data, and acquiring a recognition result about the chain commutation failure by adopting a trained space-time convolutional neural network.
A third aspect of the present invention provides a computer-readable storage medium.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the spatio-temporal convolution network based multi-dc-link commutation failure identification method according to the first aspect described above.
A fourth aspect of the invention provides a computer device.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in the spatio-temporal convolution network based multi-dc-link commutation failure identification method according to the first aspect described above when the program is executed.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention recognizes the commutation failure condition in the system in a short time as possible with higher accuracy, failure recognition rate and lower re-judgment rate based on the multi-feed direct current scene, and provides a specific technical support for timely giving an alarm after the occurrence of the sequential commutation failure.
2. The invention adopts artificial intelligence technology to identify the chain commutation failure in the multi-DC feed-in scene, reduces the damage to the data space-time characteristics in the sample generation process, and ensures that the chain commutation failure can be accurately identified in a short time after the occurrence of the chain commutation failure.
3. According to the invention, the spatial features and the time sequence features existing in the input sample are considered, the extracted high-order features are extracted in a targeted manner, and the extracted high-order features are used for identifying the chain commutation failure, so that the identification performance of the algorithm is improved.
4. According to the invention, a plurality of time convolution layers with different hole coefficients are stacked, and the time sequence characteristics with different time scales contained in the input data are extracted through the network structure connected by the residual error, so that the problem of network degradation caused by overlarge network depth is avoided, and meanwhile, the identification accuracy is improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a flowchart of a multi-DC chain commutation failure recognition method based on a space-time convolutional network according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a space-time convolutional neural network according to an embodiment of the present invention;
Fig. 3 is a diagram of an IEEE39 system with two-loop dc feed added according to an embodiment of the present invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
It is noted that the flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of methods and systems according to various embodiments of the present disclosure. It should be noted that each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the logical functions specified in the various embodiments. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by special purpose hardware-based systems which perform the specified functions or operations, or combinations of special purpose hardware and computer instructions.
Example 1
As shown in fig. 1, this embodiment provides a multi-dc-link commutation failure identification method based on a space-time convolutional network, and this embodiment is illustrated by applying the method to a server, and it can be understood that the method may also be applied to a terminal, and may also be applied to a system and a terminal, and implemented through interaction between the terminal and the server. The server can be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and can also be a cloud server for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communication, middleware services, domain name services, security services CDNs, basic cloud computing services such as big data and artificial intelligent platforms and the like. The terminal may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, etc. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the present application is not limited herein. In this embodiment, the method includes the steps of:
Acquiring historical/simulation generated inverter station converter bus voltage and direct current data;
extracting the inversion station converter bus voltage and direct current data according to a certain time interval to obtain a data matrix, labeling data elements in the data matrix with labels, and converting the data matrix into a sample graph;
Based on the sample graph, adopting a graph convolution network layer to extract spatial features and obtain a low-order feature graph; based on the low-order feature map, extracting time sequence features by adopting a time convolution network layer to obtain a high-order feature map; the high-order feature map passes through a full connection layer and an output layer, and the recognition result of the chain commutation failure in the training process is predicted;
Based on the recognition result and the label of the training process chain commutation failure, optimizing the super-parameters of the space-time convolution network in the training process by combining the loss function;
And acquiring real-time inversion station commutation bus voltage and direct current data, and acquiring a recognition result about the chain commutation failure by adopting a trained space-time convolutional neural network.
The embodiment aims to identify the multi-feed DC scene in a short time when the sequential commutation failure occurs, so as to take countermeasures early.
The following describes the present embodiment in detail:
(1) Different types of faults with different severity degrees are applied to different positions of the system so as to simulate time domain signal data of three-phase voltage and direct current of the inversion station converter bus under normal and fault states of the system and change conditions of arc extinguishing angles of the inversion station in corresponding time periods.
(2) In order to preserve the spatial features and the time sequence features in the data, the acquired system data after the faults are arranged into a digital matrix according to a certain sequence to reduce the loss of the time-space features in the data, and the digital matrix is subjected to normalization and graying processing to generate training samples which can be directly used for training the time-space convolutional neural network.
(3) The method comprises the steps of constructing a space-time convolutional neural network, wherein the main structure of the neural network comprises a graph convolutional network layer, a time convolutional network layer, a full connection layer, an output layer and the like, and the three layers are respectively used for extracting spatial features, extracting time features, extracting high-order features for classification and outputting classification results; the time convolution network layers are stacked by the time convolution layers with different hole coefficients so as to give consideration to time sequence characteristics of different time scales, and the time convolution layers are connected through residual errors.
(4) Inputting training samples into a built space-time convolutional neural network to obtain a recognition result of chain commutation; whether the identification performance of the neural network accords with the actual result is tested, if not, the super parameters are required to be regulated according to the test result so that the neural network can realize a more accurate identification result. This process is cycled until the recognition performance of the neural network can meet the requirements in the test.
(5) And acquiring real-time inverter station converter bus voltage and direct current data according to a specified time interval in an actual running power grid, and inputting the real-time inverter station converter bus voltage and direct current data into a trained space-time convolutional neural network to obtain a recognition result about the chain commutation failure.
The whole policy method comprises the following more detailed steps:
S1: acquiring data for training in different scenarios
Through mechanism analysis, the propagation process and principle of the cascading commutation failure are researched, namely, the effect of a direct current self-control system is overlapped due to severe alternating current failure, so that the cascading commutation failure of multi-circuit direct current is caused. Summarizing factors which can influence the recognition result of the chain commutation failure in the process of the chain commutation failure. In order to comprehensively reserve space and time sequence characteristics in the sequential commutation failure process and reduce sample dimensions, three-phase voltage and direct current of a commutation bus are selected as sample characteristic quantities.
Different faults (including single-phase, two-phase and three-phase ground faults) are applied at different positions in the alternating current system, and the severity of the faults is changed by adjusting the magnitude of the ground resistance. And recording the data of the three-phase voltage and the direct current of the inversion station converter bus during the fault period, and recording the change condition of the arc extinguishing angle of each direct current inversion station in the corresponding time.
S2: generation of training samples
The three-phase voltage and the direct current data of the inversion station converter bus generated by simulation are extracted according to a certain time interval (the general time interval of a 50Hz system is 5 ms), and the duration time corresponding to the data is generally 0.2s (the adjustment can be carried out according to the actual requirement). In order to keep the space and time sequence characteristics for identifying the chain commutation failure of the training data, the extracted data are arranged into a matrix form, each row of the matrix represents the data of different direct currents at the same moment, and the larger the number of rows is, the later the moment corresponding to the data is; in the column direction, the three-phase voltages a, b and c and the direct current are sequentially arranged from left to right according to the direct current number in the same direct current.
Carrying out normalization and graying treatment on the data matrix, wherein the normalization adopts a linear normalization method, and the calculation formula is as follows:
（1）
Wherein, Is the original value of the data,/>Respectively, the maximum and minimum values in the data. After normalization and graying, the matrix is saved as a gray scale map so as to be directly used for training the space-time convolutional neural network.
In order to improve the fault recognition rate of the neural network, a conservative label labeling strategy is adopted. In the training group, according to the change condition of the arc extinguishing angle of the converter station in the corresponding time period of the data in the matrix, if the arc extinguishing angle in the corresponding time period has a moment smaller than 7.5 degrees, the corresponding direct current is considered to have commutation failure; and if the arc extinguishing angle in the corresponding time period is always larger than 7.5 degrees, the corresponding direct current phase conversion in the time period is considered to be successful. In the test group, the judgment threshold value of commutation failure is 7 degrees, which is a conservative labeling strategy.
According to different combinations of commutation failures, training samples can be divided into a plurality of categories of successful commutation, single-circuit direct current commutation failure, simultaneous commutation failure, chained commutation failure and the like, in order to ensure the training effect of the neural network, the training samples are balanced among the categories, and the number of samples among different categories is ensured not to be too large so as to reduce the phenomenon of overfitting.
S3: building space-time convolution neural network
The main structure of the space-time convolution neural network comprises a graph convolution network layer, a time convolution network layer, a full connection layer, an output layer and the like, and the whole network structure of the space-time convolution neural network is shown in fig. 2.
S3.1 graph rolled network layer
The graph convolutional network layer is mainly responsible for extracting spatial features in input data, and because the graph belongs to a non-Euclidean space, when convolution is applied to the graph, a convolution kernel is difficult to define, so that a standardized Laplace matrix needs to be calculated first, and the graph is converted into a spectral domain, so that the convolution kernel is easier to obtain. Normalized Laplace matrixThe calculation formula of (2) is as follows:
（2）
Wherein, Is a unit matrix with the order and normalized Laplace matrix/>Consistent,/>Is a neighbor matrix in which elements represent multi-feed interaction factors between different nodes,/>Is a degree matrix in which，/>Is/>/>, In matrixLine/>The elements of the columns have the formula
（3）
Wherein,And/>Respectively represent node/>When the site is disturbed, the first/>And/>The voltage change of the number node, this ratio reflects the degree of coupling tightness between the two nodes. After the normalized Laplace matrix is obtained, the convolution calculation of the graph in the frequency domain can be expressed as:
（4）
Wherein, Is a convolution kernel,/>Is input data,/>Is a matrix/>Feature vector matrix,/>Is a parameter of the convolution kernel extracted from the graph-volume lamination. However, the calculation amount of the method is too large, and the actual requirement is difficult to meet, so that the convolution kernel is approximately converted into a first-order truncated expansion form of the chebyshev polynomial, and the formula (4) can be simplified into:
（5）
S3.2 time convolution network layer
By means of mechanism analysis, time sequence characteristics such as the voltage drop speed of a commutation bus and time sequence characteristics such as the voltage drop amplitude of the commutation bus with a longer time scale can be determined in the process of the sequential commutation failure, and because different time sequence characteristics have larger time scale differences, single time convolution layers are difficult to consider when extracting the time sequence characteristics, the time convolution network layer comprises time convolution layers with different cavity coefficients and are respectively used for extracting the time sequence characteristics with different time scales. The calculation formula of the time convolution layer is as follows:
（6）
Wherein, Is a convolution kernel,/>Is input data,/>Is the size of the convolution kernel,/>Is a hole coefficient that controls the interval of convolution kernel data. The time convolution calculation has strict causal property, and the convolution result at the moment is only related to the information at the previous moment.
In order to preserve the time sequence characteristics of different time scales, the time sequence characteristics are used for identifying the chain commutation failure, meanwhile, the phenomenon that gradient explosion or gradient disappearance is caused due to overlarge depth of a neural network is avoided, and the time convolution layers are connected through residual errors.
S3.3 full connection layer and output layer
The high-order features including the spatial features and the time sequence features extracted by the graph convolution network layer and the time convolution network layer are unfolded and then input into the full-connection layer, and the number of layers of the full-connection layer can be adjusted according to actual requirements, but in general, one to two full-connection layers can meet the identification requirements.
And the final recognition result can be obtained by the result output by the full connection layer through a Softmax activation function.
S4: training of neural networks
The generated training samples are randomly divided into a training group and a testing group according to a certain proportion, wherein the training group is used for training the neural network, and when training is finished, the testing group is used for checking the training effect. And if the test result of the test group cannot reach the practical application standard, the super-parameters of the space-time convolutional neural network need to be adjusted.
Super-parameters can be adjusted in a targeted manner by observing the condition of the training process, and when the loss function (such as cross entropy loss function and the like) and the accuracy rate are fluctuated, the training is finished, and the maximum training period can be considered to be increased. When the accuracy rate always has larger fluctuation and cannot be stabilized in the training process, the learning rate can be considered to be reduced, but the training time can be prolonged by doing so, so that the learning rate can be changed, the initial learning rate is larger, and the optimal solution range can be conveniently and rapidly determined; along with the increase of the iteration times, the learning rate is gradually reduced so as to obtain an accurate optimal solution, and in this way, the learning speed and the accurate finding of the optimal solution can be considered. When the training result is stable, the accuracy rate still cannot meet the requirement, and the structure of the neural network, such as the combination of the size of the convolution kernel and the void coefficient in the time convolution layer, or the number of layers of the full connection layer and the number of neurons contained in each layer, can be properly adjusted. Different solvers can be replaced, the change of the identification performance is observed, and the solver with the best effect is selected.
The above process needs to be repeated, and because of the occasional occurrence of the training process, a group of super parameters are combined to perform multiple training and testing, and the neural network meeting the requirement is not obtained until the average value of the multiple testing results can meet the actual requirement.
S5: system real-time identification
The three-phase voltage and direct current data of the inversion station converter bus in the system running in real time are recorded, the data are collated and data for neural network identification are generated, the specific processing method is the same as the sample generation step in the step S2, and no label is required to be marked on the data.
And inputting the data which is generated in real time and used for recognition into the trained space-time convolutional neural network, so that the recognition result of the chain commutation failure in the corresponding time period of the data can be rapidly obtained.
S6: simulation analysis
Based on CIGRE standard direct current test model, an IEEE 39 node transformation system containing 2-loop direct current is built in PSCAD/EMTDC software, and is used for generating training samples for training a neural network and verifying the validity of a multi-direct current chain commutation failure identification method based on a space-time convolution network. The direct current system adopts a rectification side constant current control and inversion side constant turn-off angle control mode, control parameters are consistent with a standard model, and the structure of the IEEE 39 node transformation system is shown in figure 3.
In order to generate data samples, single-phase grounding, two-phase grounding and three-phase grounding faults are respectively applied to each node, the faults occur in the 2 nd s and last for 0.1s, the grounding resistance value is 1 omega-31 omega, 3594 groups of data are generated, and the data are split into a training group and a testing group at random, wherein the training group comprises 2875 groups of samples, and the testing group comprises 719 groups of samples.
In the aspect of the structure of the neural network, the built space-time convolution neural network consists of a graph roll layer, two time convolution layers, a full connection layer, a Softmax activation function layer and an output layer. Wherein the super parameters of the neural network are shown in table 1.
Table 1: space-time convolutional neural network superparameter
| Super parameter name | Numerical value |
| Initial learning rate | 0.01 |
| Coefficient of learning rate reduction | 0.1 |
| Learning rate reduction cycle | 10 |
| Minimum lot specification | 128 |
| Maximum training period | 100 |
| L2 regularization coefficient | 10-4 |
| Hole coefficient (first layer of first time convolution layer) | 1 |
| Hole coefficient (second layer of first time convolution layer) | 2 |
| Cavity coefficient (first layer of second time convolution layer) | 2 |
| Hole coefficient (second layer of second time convolution layer) | 4 |
| Convolution kernel specification of a temporal convolution layer | 2x1 |
To describe the recognition performance of the space-time convolutional neural network on the chain commutation failure, the recognition accuracy is improvedFailure recognition rate/>Re-judging rate/>Three aspects are evaluated, and a specific calculation formula is as follows:
(6)
(7)
（8）
Wherein, Is the total number of correct identifications, N is the total number of samples used for identification,/>Is the number of times that the fault is correctly identified,/>The single-loop direct current commutation failure is identified as the number of sequential commutation failures or simultaneous commutation failures,/>Total number of commutation failure samples (including single-loop DC commutation failure, simultaneous commutation failure, and sequential commutation failure),/>The commutation success is identified as the number of commutation failures.
Among these indexes, accuracy is a basic evaluation index for identifying various conditions in the test. Since commutation failure can have serious consequences, an important purpose of the proposed method is to distinguish it from normal. The failure recognition rate describes the ability of the space-time convolutional neural network to recognize a failure condition. The judgment in training is more strict than the judgment in testing, that is, the training strategy is relatively conservative, so the re-judgment rate is the degree of conservation describing the proposed method.
In order to embody the feasibility and effectiveness of the proposed method, the proposed method is compared with the recognition performance of a fully connected neural network, an analysis threshold method used in engineering and a critical multi-feed interaction factor method, and the comparison result is shown in table 2.
Table 2: identification performance contrast for chained commutation failure of various methods
As can be seen from the results of Table 2, the performance of the method proposed by the present example in terms of recognition accuracy, failure recognition rate, re-judgment rate, etc. is superior to that of the conventional recognition method, which proves the effectiveness of the proposed method in recognizing the chain commutation failure to a certain extent.
Example two
The embodiment provides a multi-DC chain commutation failure recognition system based on a space-time convolution network.
A multi-DC chain commutation failure recognition system based on a space-time convolution network comprises:
a data acquisition module configured to: acquiring historical/simulation generated inverter station converter bus voltage and direct current data;
A preprocessing module configured to: extracting the inversion station converter bus voltage and direct current data according to a certain time interval to obtain a data matrix, labeling data elements in the data matrix with labels, and converting the data matrix into a sample graph;
A network training module configured to: based on the sample graph, adopting a graph convolution network layer to extract spatial features and obtain a low-order feature graph; based on the low-order feature map, extracting time sequence features by adopting a time convolution network layer to obtain a high-order feature map; the high-order feature map passes through a full connection layer and an output layer, and the recognition result of the chain commutation failure in the training process is predicted;
an optimization module configured to: based on the recognition result and the label of the training process chain commutation failure, optimizing the super-parameters of the space-time convolution network in the training process by combining the loss function;
An identification module configured to: and acquiring real-time inversion station commutation bus voltage and direct current data, and acquiring a recognition result about the chain commutation failure by adopting a trained space-time convolutional neural network.
It should be noted that the data acquisition module, the preprocessing module, the network training module, the optimization module, and the identification module are the same as the examples and application scenarios implemented by the steps in the first embodiment, but are not limited to the disclosure in the first embodiment. It should be noted that the modules described above may be implemented as part of a system in a computer system, such as a set of computer-executable instructions.
Example III
The present embodiment provides a computer readable storage medium having stored thereon a computer program which when executed by a processor implements the steps in the spatio-temporal convolution network based multi-dc-link commutation failure identification method described in the above embodiment.
Example IV
The present embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, where the processor implements the steps in the method for identifying multiple dc-link commutation failures based on a space-time convolutional network according to the above embodiment when executing the program.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, magnetic disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disc, a Read-Only Memory (ROM), a Random access Memory (Random AccessMemory, RAM), or the like.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. The multi-DC chain commutation failure identification method based on the space-time convolution network is characterized by comprising the following steps of:
Acquiring historical/simulation generated inverter station converter bus voltage and direct current data;
extracting the inversion station converter bus voltage and direct current data according to a certain time interval to obtain a data matrix, labeling data elements in the data matrix with labels, and converting the data matrix into a sample graph;
Based on the sample graph, adopting a graph convolution network layer to extract spatial features and obtain a low-order feature graph; based on the low-order feature map, extracting time sequence features by adopting a time convolution network layer to obtain a high-order feature map; the high-order feature map passes through a full connection layer and an output layer, and the recognition result of the chain commutation failure in the training process is predicted;
Based on the recognition result and the label of the training process chain commutation failure, optimizing the super-parameters of the space-time convolution network in the training process by combining the loss function;
acquiring real-time inversion station commutation bus voltage and direct current data, and acquiring a recognition result about chain commutation failure by adopting a trained space-time convolution network;
the space-time convolution network comprises a plurality of time convolution layers with different hole coefficient combinations and residual connection, and time sequence features of different time scales are extracted through the different time convolution layers.
2. The method for identifying multiple direct current chain commutation failures based on space-time convolutional network according to claim 1, wherein the behavior time of the data matrix is the inverter commutation bus voltage and the direct current data.
3. The method for identifying multi-direct-current chain commutation failure based on space-time convolution network according to claim 1, wherein in the process of obtaining the history/simulation generated inversion station commutation bus voltage and direct current data, the change condition of each direct current inversion station arc extinguishing angle in the corresponding time is recorded.
4. The method for identifying multi-dc-link commutation failure based on space-time convolutional network according to claim 3, wherein the labeling of the data elements in the data matrix comprises: and marking the elements in the data matrix by adopting a conservative marking strategy according to the change condition of the arc extinguishing angle of each direct current inversion station in the corresponding time.
5. The method for identifying multi-dc-link commutation failure based on space-time convolutional network according to claim 1, wherein the space-time convolutional network comprises a graph convolutional network layer, two time convolutional network layers, a full connection layer and an output layer.
6. The method for identifying multi-dc-link commutation failure based on space-time convolutional network of claim 1, wherein the process of converting the data matrix into a sample graph comprises: and carrying out normalization and graying treatment on the data matrix, and converting the data matrix into a gray scale image, namely a sample image.
7. The multi-DC chain commutation failure recognition system based on the space-time convolution network is characterized by comprising the following components:
a data acquisition module configured to: acquiring historical/simulation generated inverter station converter bus voltage and direct current data;
A preprocessing module configured to: extracting the inversion station converter bus voltage and direct current data according to a certain time interval to obtain a data matrix, labeling data elements in the data matrix with labels, and converting the data matrix into a sample graph;
A network training module configured to: based on the sample graph, adopting a graph convolution network layer to extract spatial features and obtain a low-order feature graph; based on the low-order feature map, extracting time sequence features by adopting a time convolution network layer to obtain a high-order feature map; the high-order feature map passes through a full connection layer and an output layer, and the recognition result of the chain commutation failure in the training process is predicted;
an optimization module configured to: based on the recognition result and the label of the training process chain commutation failure, optimizing the super-parameters of the space-time convolution network in the training process by combining the loss function;
An identification module configured to: acquiring real-time inversion station commutation bus voltage and direct current data, and acquiring a recognition result about chain commutation failure by adopting a trained space-time convolution network;
the space-time convolution network comprises a plurality of time convolution layers with different hole coefficient combinations and residual connection, and time sequence features of different time scales are extracted through the different time convolution layers.
8. A computer readable storage medium having stored thereon a computer program, which when executed by a processor performs the steps in the spatio-temporal convolution network based multi-dc-link commutation failure identification method according to any one of claims 1 to 6.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the spatio-temporal convolution network based multi-dc-link commutation failure identification method according to any one of claims 1 to 6 when the program is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310854905.4A CN116961072B (en) | 2023-07-12 | 2023-07-12 | Multi-DC chain commutation failure recognition method and system based on space-time convolution network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310854905.4A CN116961072B (en) | 2023-07-12 | 2023-07-12 | Multi-DC chain commutation failure recognition method and system based on space-time convolution network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116961072A CN116961072A (en) | 2023-10-27 |
| CN116961072B true CN116961072B (en) | 2024-06-18 |
Family
ID=88447104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310854905.4A Active CN116961072B (en) | 2023-07-12 | 2023-07-12 | Multi-DC chain commutation failure recognition method and system based on space-time convolution network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116961072B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635732A (en) * | 2018-12-12 | 2019-04-16 | 国网山东省电力公司电力科学研究院 | A kind of continuous commutation failure fault pre-alarming network training method of LCC-HVDC |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20150142763A (en) * | 2014-06-11 | 2015-12-23 | 한국전력공사 | Apparatus for predicting and controlling commutation failure of high voltage direct current system |
| CN112234641B (en) * | 2020-09-21 | 2022-12-23 | 清华大学 | Direct-current commutation failure blocking and preventing control method |
| CN114360254B (en) * | 2022-03-17 | 2022-06-21 | 湖南工商大学 | Space-time-based traffic flow prediction method and related equipment |
-
2023
- 2023-07-12 CN CN202310854905.4A patent/CN116961072B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109635732A (en) * | 2018-12-12 | 2019-04-16 | 国网山东省电力公司电力科学研究院 | A kind of continuous commutation failure fault pre-alarming network training method of LCC-HVDC |
Non-Patent Citations (1)
| Title |
|---|
| 影响多馈入直流***换相失败的电网薄弱区域快速识别方法;肖浩等;中国电机工程学报;20160405;36(07);第1781-1789页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116961072A (en) | 2023-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102663412A (en) | Power equipment current-carrying fault trend prediction method based on least squares support vector machine | |
| Goswami et al. | Predictive model for classification of power system faults using machine learning | |
| CN110726898B (en) | Power distribution network fault type identification method | |
| Xu et al. | Application of artificial neural network to the detection of the transformer winding deformation | |
| CN105425768A (en) | Electric power secondary equipment monitoring device and method | |
| CN105467971A (en) | Electric power secondary equipment monitoring system and method | |
| CN113159345A (en) | Power grid fault identification method and system based on fusion neural network model | |
| CN109061391B (en) | Power grid fault diagnosis method and system based on computer vision tidal current diagram | |
| Jamali et al. | Identification of faulted line section in microgrids using data mining method based on feature discretisation | |
| CN111062569A (en) | Low-current fault discrimination method based on BP neural network | |
| CN116961072B (en) | Multi-DC chain commutation failure recognition method and system based on space-time convolution network | |
| CN117421571A (en) | Topology real-time identification method and system based on power distribution network | |
| WO2018029662A1 (en) | Method and device for wind power curtailment optimization for day-ahead operational planning of a power network | |
| CN116541748A (en) | Power grid fault diagnosis method and system based on artificial intelligence technology | |
| Awasthi et al. | Identification of type of a fault in distribution system using shallow neural network with distributed generation | |
| CN115311509A (en) | Power system transient stability evaluation method and system based on imaging data driving | |
| Karapidakis | Machine learning for frequency estimation of power systems | |
| CN113392589B (en) | High-voltage direct-current converter station fault analysis method and system based on convolutional neural network | |
| CN115494349A (en) | Active power distribution network single-phase earth fault section positioning method | |
| CN113627655B (en) | Method and device for simulating and predicting pre-disaster fault scene of power distribution network | |
| Alhanaf et al. | Fault Detection and Classification in Ring Power System with DG Penetration Using Hybrid CNN-LSTM | |
| CN107679478A (en) | The extracting method and system of transmission line of electricity space load state | |
| CN107508279B (en) | Stability simulation method for power network | |
| CN112287751A (en) | Magnetizing inrush current identification method and device, computer equipment and storage medium | |
| Bhattacharya et al. | Deep fault analysis and subset selection in solar power grids |
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 |