CN118228772B - Distance measurement method, framework, system and medium for actually measured traveling wave of power transmission line - Google Patents
Distance measurement method, framework, system and medium for actually measured traveling wave of power transmission line Download PDFInfo
- Publication number
- CN118228772B CN118228772B CN202410650069.2A CN202410650069A CN118228772B CN 118228772 B CN118228772 B CN 118228772B CN 202410650069 A CN202410650069 A CN 202410650069A CN 118228772 B CN118228772 B CN 118228772B
- Authority
- CN
- China
- Prior art keywords
- attention
- feature map
- module
- traveling wave
- transducer
- 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
- 230000005540 biological transmission Effects 0.000 title claims abstract description 34
- 238000000691 measurement method Methods 0.000 title abstract description 4
- 230000006870 function Effects 0.000 claims abstract description 44
- 230000003044 adaptive effect Effects 0.000 claims abstract description 35
- 230000007246 mechanism Effects 0.000 claims abstract description 31
- 230000004913 activation Effects 0.000 claims abstract description 24
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 22
- 230000002776 aggregation Effects 0.000 claims abstract description 8
- 238000004220 aggregation Methods 0.000 claims abstract description 8
- 230000006835 compression Effects 0.000 claims abstract description 7
- 238000007906 compression Methods 0.000 claims abstract description 7
- 230000005284 excitation Effects 0.000 claims abstract description 7
- 238000005259 measurement Methods 0.000 claims abstract description 6
- 238000000034 method Methods 0.000 claims description 48
- 238000012545 processing Methods 0.000 claims description 42
- 238000011176 pooling Methods 0.000 claims description 41
- 238000010586 diagram Methods 0.000 claims description 21
- 238000012549 training Methods 0.000 claims description 19
- 238000004364 calculation method Methods 0.000 claims description 17
- 238000003860 storage Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000009467 reduction Effects 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 10
- 238000012795 verification Methods 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 9
- 238000001514 detection method Methods 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000009826 distribution Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 description 10
- 230000004048 modification Effects 0.000 description 5
- 238000012986 modification Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 230000004075 alteration Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 241001391944 Commicarpus scandens Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
Classifications
-
- 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
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/085—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution lines, e.g. overhead
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/088—Aspects of digital computing
-
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a distance measurement method, a distance measurement framework, a distance measurement system and a distance measurement medium for actually measured traveling waves of a power transmission line, and belongs to the technical field of relay protection of power systems. The architecture is a novel transducer architecture, and the novel transducer architecture is formed by a convolutional neural network and an adaptive transducer network; based on MBConv convolutional neural network part design, a characteristic aggregation module (CABFAM) of a convolutional attention mechanism is adopted, and a compression and excitation (SE) module in a MBConv structure is replaced by the convolutional attention mechanism (CBAM), so that a channel attention module and a spatial attention module are connected in series to capture spatial correlation between different areas of an image, and meanwhile, the complexity of the image is reduced; siLU (Sigmoid Linear Unit) is introduced as an activation function, so that the nonlinear fitting capacity and learning efficiency of the model are improved.
Description
Technical Field
The invention relates to the technical field of relay protection of power systems, in particular to a distance measuring method, a distance measuring framework, a distance measuring system and a distance measuring medium for actually measured traveling waves of a power transmission line.
Background
The electric power industry is one of the most important basic industries of each country, and has great significance for not only deeply influencing national economic development, but also modern development of the country. Because the direct current transmission line has the defects of long line, complex terrain, bad environment and the like, the direct current transmission line is easy to break down in a direct current power grid.
In the related offline cable fault monitoring method, the fault is checked with high difficulty by using a line inspection mode, permanent faults can be caused, the recovery time is slow, and the fault position on-line monitoring mode based on the traveling wave principle belongs to a high-precision positioning method, and the influence on the operation of the whole system is small, so that the method is widely applied.
In the mode of online monitoring fault positions based on the traveling wave principle, in order to ensure that the traveling wave acquisition device can reliably record fault traveling wave data in weak fault modes such as small fault angles and high-resistance faults, a high-speed recording is often carried out by adopting a mutation starting mode with a lower threshold value, and a large amount of interference clutter can be acquired and stored and recorded, so that the proportion of the interference traveling wave to the fault traveling wave is seriously unbalanced. Therefore, it is important to automatically identify the fault traveling wave from the traveling wave data acquired in a large amount.
With the rapid development of artificial intelligence technology in recent years, the intelligent algorithm has strong learning ability and can perform nonlinear fitting well. However, many existing algorithms still have the problems of low convergence speed, easy local optimum trapping and the like, and have certain limitations on the accuracy and speed of power transmission line fault location.
Therefore, in order to effectively reduce the complexity of extracting the fault traveling wave data feature quantity and improve the accuracy of the fault classification of the power transmission line, a new system based on a transducer architecture needs to be developed.
The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present invention and is not intended to represent an admission that the foregoing is prior art.
Disclosure of Invention
The invention mainly aims to provide a distance measuring method, a distance measuring framework, a distance measuring system and a distance measuring medium for actually measured traveling waves of a power transmission line, and solves the problems in the prior art.
To achieve the above objects, the present invention provides a novel transducer architecture including a convolutional neural network and an adaptive transducer network;
The convolutional neural network adopts a characteristic aggregation module of a convolutional attention mechanism, and replaces a compression and excitation module of a convolutional module MBConv structure with the convolutional attention mechanism;
The adaptive transducer network is based on an adaptive coding hierarchy and consists of a plurality of coding layers, wherein each coding layer consists of a multi-head self-attention layer and a feedforward network layer;
The gaussian error linear unit of the novel fransformer architecture is based on SiLU as an activation function whose output characteristics are taken as inputs to the adaptive fransformer network.
Optionally, the new transducer architecture maps the features of MBConv structure outputs as inputs to self-attention mechanisms, and the calculation formula of each self-attention mechanism includes:
where Q is the query matrix, K is the key matrix, V is the value matrix, and d is the vector dimension of the key.
Optionally, the process of replacing the compression and excitation module of the convolution module MBConv structure with a convolution attention mechanism includes:
processing the input data based on the point convolution check, and increasing the channel number of the input data;
and carrying out feature extraction processing on the data with the increased channel number, and carrying out dimension reduction processing on the data with the feature extraction processing and the point convolution to finish residual connection of input data and output data.
Optionally, the calculation formula when the gaussian error linear unit is used as the activation function includes:
Wherein: mu sum The mean and standard deviation of the normal distribution, respectively.
Optionally, the convolution attention mechanism is composed of a channel attention module and a spatial attention module;
the calculation formula of the characteristic diagram of the channel attention comprises the following steps:
Wherein, The sigmoid function is represented as a function,、Is the weight of the multi-layer perceptron network, r is the reduction ratio, and the hidden activation size of the multi-layer perceptron is，Respectively an associated feature map of the average pooling feature and the maximum pooling feature;
The calculation formula of the spatial attention characteristic diagram comprises the following steps:
Wherein, The sigmoid function is represented as a function,Representing a convolution operation with a convolution kernel size of 7 x 7,AndThe average pooling feature and the maximum pooling feature of the channels of the spatial attention module, respectively.
Optionally, the calculation formula of the adaptive feature map includes:
wherein M S (F) refers to the spatial attention profile, M C (F) refers to the channel attention profile, Representing element-by-element multiplication, F represents the feature map of the input,Is the final refinement output;
the full connection layer of the novel transducer architecture is normalized through softmax, and the formula comprises:
Where ak is the kth input signal in the output layer and e is the natural logarithm.
Optionally, when the coding level of the adaptive transducer network rises, the two transducer models before and after rising evaluate the difference value of the index of evaluating the reliability of the detection result of the original training imageThe defined formula of (2) includes:
wherein, when said Greater than the confidence threshold of the transducer modelWhen the transform coding hierarchy is updated, the updating process includes:
And The maximum coding layer number and the scaling factor for increasing the coding layer number are added each time, respectively, the network layer level is theta,Indicating that the recognition result does not meet the confidence thresholdIs a global sample entropy loss value of (1).
Optionally, the activation function is used for performing nonlinear processing on the neural network parameters, the activation function is a continuously-conductive function, and SiLU is a weighted linear combination of Sigmoid functions;
The calculation formula for the kth SiLU activation ak of the input parameters includes:
Wherein the Sigmoid function is:
。
In order to achieve the above purpose, the present invention provides a ranging method for actually measured traveling waves of a power transmission line, the method comprising the following steps:
After traveling wave fault data are obtained, classifying the traveling wave fault data into a training set, a testing set and a verification set, wherein the proportion of the training set, the testing set and the verification set is 7:3:1;
Preprocessing the training set based on a Gram matrix to obtain A, B and C three-phase time domain image data, wherein A, B and C three phases refer to voltages or currents with three phases offset by 120 degrees;
Inputting the time domain image into a convolutional neural network of the novel transducer architecture, and generating a spatial attention feature map and a channel attention feature map based on a convolutional attention mechanism of the convolutional neural network;
And generating an adaptive feature map according to the spatial attention feature map and the channel attention feature map, and generating a ranging result of traveling wave fault data based on reliability evaluation results of the adaptive feature map and the adaptive transducer network.
Optionally, the step of inputting the time domain image into a convolutional neural network of the novel fransformer architecture, and generating a spatial attention profile and a channel attention profile based on a convolutional attention mechanism of the convolutional neural network includes:
after the time domain image is input into the convolutional neural network, the time domain image is processed through point convolution, the channel number of the time domain image is increased, and normalization processing is performed based on a Gaussian error unit;
Carrying out deep convolution processing on the data with the increased channel number to extract the space characteristic information of the time domain image, and carrying out normalization processing based on a Gaussian error unit;
performing dimension reduction processing on the space characteristic information according to point convolution, and performing normalization processing based on a Gaussian error unit;
And generating the spatial attention characteristic map and the channel attention characteristic map corresponding to the processed time domain image based on the channel attention module and the spatial attention module.
Optionally, the step of generating the spatial attention feature map and the channel attention feature map corresponding to the processed time domain image based on the channel attention module and the spatial attention module includes:
Generating an average pooling feature map and a maximum pooling feature map of the channel attention module according to the channel attention module, inputting the average pooling feature map and the maximum pooling feature map of the channel attention module into a multi-layer perceptron network, and generating the channel attention feature map based on a summation mode of elements;
And generating an average pooling feature map and a maximum pooling feature map of the spatial attention module according to the spatial attention module, and carrying out standard convolution operation on the average pooling feature map and the maximum pooling feature map of the spatial attention module to generate the spatial attention module feature map.
In addition, in order to achieve the above object, the present invention also provides a system, which includes a memory, a processor, and a ranging program for actually measuring traveling wave of a power transmission line stored in the memory and capable of running on the processor, wherein the steps of the ranging method for actually measuring traveling wave of a power transmission line are implemented when the ranging program for actually measuring traveling wave of a power transmission line is executed by the processor.
In addition, in order to achieve the above object, the present invention further provides a computer readable storage medium, where a ranging program for actually measuring a traveling wave of a power transmission line is stored on the computer readable storage medium, and the steps of the ranging method for actually measuring a traveling wave of a power transmission line are implemented when the ranging program for actually measuring a traveling wave of a power transmission line is executed by a processor.
The invention provides a ranging method, a framework, a system and a medium for actually measured traveling waves of a power transmission line, wherein the framework is a novel transducer framework, and a characteristic aggregation module (CABFAM) adopting a convolution attention mechanism is fused with an adaptive transducer coding level to create the novel transducer framework. And then constructing a transducer adjustment mechanism based on the self-adaptive coding level to acquire multi-level differentiated characteristic information. Finally, siLU (Sigmoid Linear Unit) is introduced as an activation function, so that the nonlinear fitting capacity and learning efficiency of the model are improved, and the effect of improving the identification precision and speed of the novel transducer architecture is achieved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description, serve to explain the principles of the invention. In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, and it will be obvious to those skilled in the art that other drawings can be obtained from these drawings without inventive effort.
Fig. 1 is a schematic diagram of an action logic diagram of data processing performed by a novel transducer architecture of a distance measurement method for actually measuring traveling waves of a power transmission line;
FIG. 2 is a schematic diagram of an adaptive transducer coding layer of a novel transducer architecture for a ranging method of an actual measurement traveling wave of a power transmission line;
FIG. 3 is a ROC graph of different fault categories of example results of a ranging method for a measured traveling wave of a transmission line according to the present invention;
fig. 4 is a schematic flow chart of a ranging method for actually measured traveling waves of a power transmission line according to the present invention;
Fig. 5 is a flow chart of a feature aggregation module of a convolution attention mechanism of a novel transducer architecture of the ranging method for actually measuring traveling waves of a power transmission line.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
It should be noted that, the novel transducer architecture of the present invention includes a convolutional neural network and an adaptive transducer network, and the architecture specifically includes:
S1: the convolutional neural network adopts a characteristic aggregation module of a convolutional attention mechanism, and replaces a compression and excitation module of a convolutional module MBConv structure with the convolutional attention mechanism;
s2: the adaptive transducer network is based on an adaptive coding hierarchy and consists of a plurality of coding layers, wherein each coding layer consists of a multi-head self-attention layer and a feedforward network layer;
s3: the Gaussian error linear unit of the novel transducer architecture is based on SiLU (Sigmoid Linear Unit) as an activation function whose output characteristics are used as inputs to the adaptive transducer network
It should be noted that, in the novel transducer architecture described in the present invention, the data set is traveling wave fault data, when traveling wave fault data is acquired, the traveling wave fault data needs to be classified into 16 classes according to fault classes, and is divided into a training set, a test set and a verification set based on a ratio of 7:3:1, that is, the training set scale accounts for 70% of the total data set, the test set scale accounts for 30% of the total data set, and the verification set scale accounts for 10% of the total data set. After classification, the training set is preprocessed, namely, the traveling wave fault data waveform is converted into A, B, C three-phase time domain images in a lossless mode, then the time domain images are used as input data, data processing is carried out as shown in fig. 1, namely, the processing is carried out through a characteristic aggregation module 1 of a convolution attention mechanism, namely, a channel attention module and a space attention module, so that a channel attention characteristic diagram and a space attention characteristic diagram are obtained, then the two characteristic diagrams are synthesized into a self-attention characteristic diagram through a characteristic aggregation module 2, normalization processing is carried out through an activation function SiLU, and then the normalized data are input into a self-adaptive transducer model, so that a final classification result is obtained.
It should be noted that, the new converter architecture maps the features output by MBConv results as the input of the self-attention mechanism, and the calculation formula of each self-attention mechanism includes:
where Q is the query matrix, K is the key matrix, V is the value matrix, and d is the vector dimension of the key.
As an alternative embodiment, based on this, the process of replacing the compression and excitation module of the convolution module MBConv structure with a convolution attention mechanism includes:
And checking input data based on point convolution, processing, increasing the channel number of the input data, then carrying out feature extraction processing on the data with the increased channel number, and carrying out dimension reduction processing based on the data with the feature extraction processing and the point convolution to finish residual connection of the input data and the output data. For example, by adopting the depth separable convolution, firstly, the input data is increased by the number of the point convolution ascending channels with the convolution kernel size of 1×1, then the spatial characteristic information is extracted by the depth convolution with the convolution kernel size of 3×3, finally, the dimension reduction is realized by the point convolution, and the number of the channels in the input process is mapped back to realize residual connection.
In the processing process, after point convolution processing and point convolution dimension reduction processing of a deep convolution processor are performed, the processing is performed through a Gaussian error linear unit, wherein the Gaussian error linear unit introduces a random regularization idea, and a calculation formula of the Gaussian error linear unit is as follows as an activation function:
Wherein: mu sum The mean and standard deviation of the normal distribution, respectively.
Further, after the processed time domain image data is obtained, a self-attention feature map thereof needs to be extracted, and the self-attention feature map is calculated from the channel attention feature map and the spatial attention feature map. The convolution attention mechanism (CBAM) is composed of a channel attention module and a space attention module, so that a channel attention characteristic diagram and a space attention characteristic diagram can be calculated through a calculation formula of the channel attention and the space attention.
Assuming the width of the input feature map is W and the height is H, the channel attention module uses the average pooling and maximum pooling operations to aggregate the feature map spatial information, generating two associated feature maps respectively representing the average pooling feature and the maximum pooling feature、. Inputting the two associated feature maps into a multi-layer perceptron network (MLP), the hidden activation size of the multi-layer perceptron network being set to:
where r is the reduction ratio.
The noted feature map of the output channel is. And finally, outputting the characteristic vector by means of element-by-element summation. In summary, the channel attention is calculated as follows:
Wherein the method comprises the steps of Representing a sigmoid function.、Is the weight of the MLP network, shared by both inputs. The ReLU activation function activates only feature graphs that go through W0.
The spatial attention module generates a spatial attention profile using the spatial relationship of the features and complements the channel attention. Firstly, carrying out average pooling and maximum pooling operation on the input feature images along the channel direction, and splicing the obtained results. Spatial attention profile generated with convolution operationEmbedding channel information of feature graphs through two pooling operations to generate two feature graphsAndRepresenting the average pooling feature and the maximum pooling feature of the channel, respectively. And finally, generating a final spatial attention characteristic diagram through standard convolution operation. The specific calculation mode is as follows:
Wherein the method comprises the steps of The sigmoid function is represented as a function,A convolution operation with a convolution kernel size of 7 x 7 is represented.
After the spatial attention profile and the channel attention profile are obtained, the adaptive profile may be calculated, where the calculation process includes:
Wherein the method comprises the steps of Representing element-by-element multiplication. During element-wise multiplication, the region of interest of a channel is replicated along a spatial dimension and vice versa.Is the final refinement output.
Meanwhile, the full connection layer of the novel transducer architecture is normalized through softmax, and the formula comprises:
Where ak is the kth input signal in the output layer and e is the natural logarithm.
It should be noted that the purpose of the activation function is to nonlinear the neural network parameters. The activation function is continuous and conductive. SiLU functions are weighted linear combinations of Sigmoid functions. The activation ak of the kth SiLU of the input zk is calculated from the sigmoid function multiplied by its input:
Wherein the Sigmoid function is:
As an alternative implementation manner, a schematic diagram of an adaptive transform coding layer is shown in fig. 2, please refer to fig. 2, after the adaptive feature map is processed based on an activation function in a MBConv module, the output feature map set is used as an input of the adaptive transform network, and the adaptive transform network adopts an adaptive coding layer and is composed of a plurality of coding layers, wherein each coding layer is composed of a standard structure, and includes a multi-head attention layer and a feedforward network layer;
The transducer network adopts an adaptive coding level, and when the coding level of the transducer rises from theta 1 to theta 2 (namely the coding level rises), the two transducer models evaluate the difference value of index of reliability of detection results of original training images (total nmax) by using the two transducer models Can be defined as:
If the detection result Exceeding the confidence threshold of the modelIt can be inferred that the target detection results of the original training images need to be further optimized to obtain more reliable results. The update procedure of the transform coding level may be defined as, in the case of network level theta,Θ indicates that the recognition result does not satisfy the reliability thresholdIs a global sample entropy loss value of (1). Network level increment adaptively adjusted during next identificationIt can be calculated as:
wherein: And The maximum coding layer number and the scaling coefficient of the coding layer number are increased each time, and the values of the parameters are set according to the coding layer number which is expected to be increased.
For example, in an optional experimental data, when the maximum value imax=8 of the coding hierarchy and the threshold value max=0.8 of the model reliability are set, verification is performed after training is completed, so as to obtain a classification result, and an ROC curve of the obtained classification result is shown in fig. 3, where specific fault type information is included.
Based on the method, after identification of various traveling wave fault types is completed, the novel transducer model can carry out online monitoring on traveling wave fault data, and can carry out rapid judgment on the fault types when faults occur, so that online traveling wave protection and ranging functions are further realized.
Further, in order to better understand the above technical solutions, exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
Referring to fig. 4, based on a novel transducer architecture, the method for measuring distance of actually measured traveling waves of a power transmission line according to the present invention includes the following steps:
Step S10, after traveling wave fault data are obtained, classifying the traveling wave fault data into a training set, a testing set and a verification set;
It should be noted that, in the embodiment of the present invention, the data set is traveling wave fault data, so when traveling wave fault data is acquired, the traveling wave fault data needs to be classified into 16 classes according to fault types, and is divided into a training set, a test set and a verification set based on a ratio of 7:3:1, that is, the training set scale accounts for 70% of the total data set, the test set scale accounts for 30% of the total data set, and the verification set scale accounts for 10% of the total data set.
Step S20, preprocessing the training set based on a Gram matrix to obtain A, B and C three-phase time domain image data
It should be noted that A, B and C three phases refer to voltages or currents in which three phases are offset from each other by 120 degrees.
Step S30, inputting the time domain image into a convolutional neural network of the novel transducer architecture, and generating a spatial attention feature map and a channel attention feature map based on a convolutional attention mechanism of the convolutional neural network;
In this process, please refer to fig. 5, after inputting the time domain image into the convolutional neural network, processing the time domain image through 1×1 point convolution, increasing the channel number of the time domain image, performing normalization processing based on a Gaussian Error (GELUs) unit, performing 3×3 depth convolution processing on the data with the increased channel number, so as to extract spatial feature information of the time domain image, performing normalization processing based on the gaussian error unit, performing dimension reduction processing based on the convolution attention mechanism according to the 1×1 point convolution, performing normalization processing based on the gaussian error unit, and finally outputting the result.
After the output result is obtained, the spatial attention feature map and the channel attention feature map corresponding to the processed time domain image need to be generated based on the channel attention module and the spatial attention module. In the novel transducer architecture, the convolution attention mechanism of the architecture comprises a channel attention module and a space attention module, so that an average pooling feature map and a maximum pooling feature map of the channel attention module can be generated according to the channel attention module, the average pooling feature map and the maximum pooling feature map of the channel attention module are input into a multi-layer perceptron network, the channel attention feature map is generated based on an element-by-element summation mode, then the average pooling feature map and the maximum pooling feature map of the space attention module are generated according to the space attention module, and standard convolution operation is carried out on the average pooling feature map and the maximum pooling feature map of the space attention module, so that the space attention module feature map is generated.
And S40, generating an adaptive feature map according to the spatial attention feature map and the channel attention feature map, and generating a ranging result of traveling wave fault data based on the adaptive feature map and the reliability evaluation result of the adaptive transducer network.
Based on the detection, the detection processing of the traveling wave fault data is completed through the novel transducer architecture.
Furthermore, it will be appreciated by those of ordinary skill in the art that implementing all or part of the processes in the methods of the above embodiments may be accomplished by computer programs to instruct related hardware. The computer program comprises program instructions, and the computer program may be stored in a storage medium, which is a computer readable storage medium. The program instructions are executed by at least one processor in the control terminal to carry out the flow steps of the embodiments of the method described above.
Therefore, the present invention also provides a computer readable storage medium, where a ranging program for actually measuring a traveling wave of a power transmission line is stored, and when the ranging program for actually measuring the traveling wave of the power transmission line is executed by a processor, the steps of the ranging method for actually measuring the traveling wave of the power transmission line are implemented.
It should be noted that, because the storage medium provided in the embodiments of the present application is a storage medium used for implementing the method in the embodiments of the present application, based on the method described in the embodiments of the present application, a person skilled in the art can understand the specific structure and the modification of the storage medium, and therefore, the description thereof is omitted herein. All storage media adopted by the method of the embodiment of the application belong to the scope of protection of the application.
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 an entirely hardware embodiment, an entirely 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, disk storage, CD-ROM, 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 flowchart and/or block of the flowchart illustrations and/or block diagrams, and combinations of flowcharts 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.
It should be noted that in the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the invention.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (10)
1. The system is characterized by comprising a convolutional neural network and an adaptive transducer network, wherein the system is used for measuring the distance of a real traveling wave of a power transmission line, preprocessing a traveling wave fault data training set to obtain time domain image data, inputting the time domain image data into the convolutional neural network of the transducer architecture, and generating a distance measurement result of the traveling wave fault data;
The convolutional neural network adopts a characteristic aggregation module of a convolutional attention mechanism, and replaces a compression and excitation module of a convolutional module MBConv structure with the convolutional attention mechanism;
The adaptive transducer network is based on an adaptive coding hierarchy and consists of a plurality of coding layers, wherein each coding layer consists of a multi-head self-attention layer and a feedforward network layer;
The Gaussian error linear unit of the transducer architecture is based on SiLU as an activation function, and the output characteristic of the activation function is used as the input of the adaptive transducer network;
The transducer architecture maps the features of MBConv structure outputs as inputs to the self-attention mechanisms, the calculation formula of each self-attention mechanism comprising:
;
Wherein Q is a query matrix, K is a key matrix, V is a value matrix, and d is the vector dimension of the key;
The process of replacing the compression and excitation modules of the convolution module MBConv structure with a convolution attention mechanism includes:
processing the input data based on the point convolution check, and increasing the channel number of the input data;
performing feature extraction processing on the data with the increased channel number, and performing dimension reduction processing based on the data subjected to the feature extraction processing and the point convolution to finish residual connection of input data and output data;
the calculation formula when the Gaussian error linear unit is used as the activation function comprises the following steps:
;
Wherein: mu sum The mean and standard deviation of the normal distribution, respectively.
2. The transducer architecture based system of claim 1, wherein the convolution attention mechanism consists of a channel attention module and a spatial attention module;
the calculation formula of the characteristic diagram of the channel attention comprises the following steps:
;
Wherein, The sigmoid function is represented as a function,、Is the weight of the multi-layer perceptron network, r is the reduction ratio, and the hidden activation size of the multi-layer perceptron is，、Respectively an associated feature map of the average pooling feature and the maximum pooling feature; the calculation formula of the characteristic diagram of the spatial attention comprises the following steps:
;
Wherein, The sigmoid function is represented as a function,Representing a convolution operation with a convolution kernel size of 7 x 7,AndThe average pooling feature and the maximum pooling feature of the channels of the spatial attention module, respectively.
3. The transducer architecture based system of claim 1, wherein the calculation formula of the adaptive feature map comprises:
;
wherein M S (F) refers to the spatial attention profile, M C (F) refers to the channel attention profile, Representing element-by-element multiplication, F represents the feature map of the input,Is the final refinement output;
The full connection layer of the transducer architecture is normalized by softmax, and the formula comprises:
;
Where ak is the kth input signal in the output layer and e is the natural logarithm.
4. The system based on a transducer architecture according to claim 1, wherein when the coding level of the adaptive transducer network is raised, the two transducer models before and after the raising evaluate the difference value of the reliability of the detection result of the original training imageThe defined formula of (2) includes:
;
wherein, when said Greater than the confidence threshold of the transducer modelWhen the transform coding hierarchy is updated, the updating process includes:
；
And The maximum coding layer number and the scaling factor for increasing the coding layer number are added each time, respectively, the network layer level is theta,Indicating that the recognition result does not meet the confidence thresholdIs a global sample entropy loss value of (1).
5. The system based on a Transformer architecture of claim 1, wherein the activation function is used to nonlinear process the neural network parameters, the activation function is a continuously-conductive function, and the SiLU is a weighted linear combination of Sigmoid functions;
The calculation formula for the kth SiLU activation ak of the input parameters includes:
;
Wherein the Sigmoid function is:
。
6. A ranging method for an actual measurement traveling wave of a power transmission line, applied to the system based on a transducer architecture as claimed in any one of claims 1 to 5, the method comprising:
After traveling wave fault data are obtained, classifying the traveling wave fault data into a training set, a testing set and a verification set, wherein the proportion of the training set, the testing set and the verification set is 7:3:1;
Preprocessing the training set based on a Gram matrix to obtain A, B and C three-phase time domain image data, wherein A, B and C three phases refer to voltages or currents with three phases offset by 120 degrees;
Inputting the time domain image into a convolutional neural network of the transducer architecture, and generating a spatial attention feature map and a channel attention feature map based on a convolutional attention mechanism of the convolutional neural network;
And generating an adaptive feature map according to the spatial attention feature map and the channel attention feature map, and generating a ranging result of traveling wave fault data based on reliability evaluation results of the adaptive feature map and the adaptive transducer network.
7. The method of claim 6, wherein the step of inputting the time domain image into a convolutional neural network of the fransformer architecture and generating a spatial attention profile and a channel attention profile based on a convolutional attention mechanism of the convolutional neural network comprises:
after the time domain image is input into the convolutional neural network, the time domain image is processed through point convolution, the channel number of the time domain image is increased, and normalization processing is performed based on a Gaussian error unit;
Carrying out deep convolution processing on the data with the increased channel number to extract the space characteristic information of the time domain image, and carrying out normalization processing based on a Gaussian error unit;
performing dimension reduction processing on the space characteristic information according to point convolution, and performing normalization processing based on a Gaussian error unit;
And generating the spatial attention characteristic map and the channel attention characteristic map corresponding to the processed time domain image based on the channel attention module and the spatial attention module.
8. The method of claim 7, wherein the step of generating the spatial attention profile and the channel attention profile corresponding to the processed time domain image based on the channel attention module and the spatial attention module comprises:
Generating an average pooling feature map and a maximum pooling feature map of the channel attention module according to the channel attention module, inputting the average pooling feature map and the maximum pooling feature map of the channel attention module into a multi-layer perceptron network, and generating the channel attention feature map based on a summation mode of elements;
And generating an average pooling feature map and a maximum pooling feature map of the spatial attention module according to the spatial attention module, and carrying out standard convolution operation on the average pooling feature map and the maximum pooling feature map of the spatial attention module to generate the spatial attention module feature map.
9. A system, the system comprising: the method for measuring the distance of the actual traveling wave of the power transmission line comprises a memory, a processor and a distance measuring program which is stored in the memory and can run on the processor, wherein the distance measuring program of the actual traveling wave of the power transmission line is executed by the processor to realize the steps of the method for measuring the distance of the actual traveling wave of the power transmission line according to any one of claims 6 to 8.
10. A computer readable storage medium, wherein a ranging program for actually measuring traveling wave of a power transmission line is stored on the computer readable storage medium, and when the ranging program for actually measuring traveling wave of the power transmission line is executed by a processor, the steps of the ranging method for actually measuring traveling wave of the power transmission line are implemented.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410650069.2A CN118228772B (en) | 2024-05-24 | 2024-05-24 | Distance measurement method, framework, system and medium for actually measured traveling wave of power transmission line |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410650069.2A CN118228772B (en) | 2024-05-24 | 2024-05-24 | Distance measurement method, framework, system and medium for actually measured traveling wave of power transmission line |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118228772A CN118228772A (en) | 2024-06-21 |
| CN118228772B true CN118228772B (en) | 2024-07-16 |
Family
ID=91509566
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410650069.2A Active CN118228772B (en) | 2024-05-24 | 2024-05-24 | Distance measurement method, framework, system and medium for actually measured traveling wave of power transmission line |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118228772B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114564559A (en) * | 2022-01-12 | 2022-05-31 | 杭州柯林电气股份有限公司 | Method, device, equipment and storage medium for detecting power failure type based on traveling wave |
| CN117972516A (en) * | 2024-01-12 | 2024-05-03 | 昆明理工大学 | Online traveling wave fault identification method based on novel transducer architecture |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN208093769U (en) * | 2016-04-05 | 2018-11-13 | 日本电产株式会社 | radar |
| US20240116516A1 (en) * | 2022-10-10 | 2024-04-11 | Waymo Llc | Methods and Systems for Adjusting Vehicle Behavior based on Estimated Unintentional Lateral Movements |
-
2024
- 2024-05-24 CN CN202410650069.2A patent/CN118228772B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114564559A (en) * | 2022-01-12 | 2022-05-31 | 杭州柯林电气股份有限公司 | Method, device, equipment and storage medium for detecting power failure type based on traveling wave |
| CN117972516A (en) * | 2024-01-12 | 2024-05-03 | 昆明理工大学 | Online traveling wave fault identification method based on novel transducer architecture |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118228772A (en) | 2024-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109614981B (en) | Power system intelligent fault detection method and system based on spearman level-dependent convolutional neural network | |
| CN113569667B (en) | Inland ship target identification method and system based on lightweight neural network model | |
| CN112763214B (en) | Rolling bearing fault diagnosis method based on multi-label zero-sample learning | |
| CN110657984B (en) | Planetary gearbox fault diagnosis method based on reinforced capsule network | |
| CN112184654A (en) | High-voltage line insulator defect detection method based on generation countermeasure network | |
| CN114022586B (en) | Defect image generation method based on countermeasure generation network | |
| CN115953666B (en) | Substation site progress identification method based on improved Mask-RCNN | |
| CN112926728B (en) | Small sample turn-to-turn short circuit fault diagnosis method for permanent magnet synchronous motor | |
| CN110726898A (en) | Power distribution network fault type identification method | |
| CN114565594A (en) | Image anomaly detection method based on soft mask contrast loss | |
| CN110334478A (en) | Machinery equipment abnormality detection model building method, detection method and model | |
| CN117011274A (en) | Automatic glass bottle detection system and method thereof | |
| Yang et al. | Application of improved yolov5 model for strip surface defect detection | |
| CN112419243B (en) | Power distribution room equipment fault identification method based on infrared image analysis | |
| CN112837281B (en) | Pin defect identification method, device and equipment based on cascade convolution neural network | |
| CN117972516A (en) | Online traveling wave fault identification method based on novel transducer architecture | |
| CN117593243A (en) | Compressor appearance self-adaptive detection method guided by reliable pseudo tag | |
| CN117274355A (en) | Drainage pipeline flow intelligent measurement method based on acceleration guidance area convolutional neural network and parallel multi-scale unified network | |
| CN118228772B (en) | Distance measurement method, framework, system and medium for actually measured traveling wave of power transmission line | |
| CN114330450B (en) | Method and system for identifying submarine craft through multi-physical-field fusion detection | |
| CN116630989A (en) | Visual fault detection method and system for intelligent ammeter, electronic equipment and storage medium | |
| CN117011219A (en) | Method, apparatus, device, storage medium and program product for detecting quality of article | |
| CN109782156B (en) | Analog circuit fault diagnosis method based on artificial immune diagnosis network | |
| Lei et al. | Detection of Mechanical Defects of High Voltage Circuit Breaker based on Improved Edge Detection and Deep Learning Algorithms | |
| CN117058443B (en) | Pipeline magnetic flux leakage image identification method based on improved residual error shrinkage 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 |