CN117970224A - CVT error state online evaluation method, system, equipment and medium - Google Patents
CVT error state online evaluation method, system, equipment and medium Download PDFInfo
- Publication number
- CN117970224A CN117970224A CN202410375860.7A CN202410375860A CN117970224A CN 117970224 A CN117970224 A CN 117970224A CN 202410375860 A CN202410375860 A CN 202410375860A CN 117970224 A CN117970224 A CN 117970224A
- Authority
- CN
- China
- Prior art keywords
- value
- data
- feature
- characteristic
- cvt
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000011156 evaluation Methods 0.000 title claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 62
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 23
- 238000000605 extraction Methods 0.000 claims abstract description 20
- 238000013528 artificial neural network Methods 0.000 claims abstract description 15
- 238000013210 evaluation model Methods 0.000 claims abstract description 13
- 230000004913 activation Effects 0.000 claims abstract description 8
- 239000013598 vector Substances 0.000 claims description 70
- 230000007246 mechanism Effects 0.000 claims description 25
- 238000000034 method Methods 0.000 claims description 23
- 238000010276 construction Methods 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 17
- 230000004927 fusion Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 description 4
- 238000012935 Averaging Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 239000002279 physical standard Substances 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a CVT error state online evaluation method, which comprises the following steps: collecting historical characteristic data of a target CVT, adding an error or a normal first label to the historical characteristic data, and constructing a training sample set; constructing a twin neural network, inputting a training sample into the two convolutional neural networks to perform feature extraction, and respectively obtaining a first feature value and a second feature value; calculating a characteristic distance between the first characteristic value and the second characteristic value, inputting the characteristic distance into at least one full-connection layer, and outputting a consistency predicted value by using an activation function; calculating loss based on the difference between the consistency predicted value and the second label, and training the twin neural network based on the loss to obtain a trained CVT error assessment model; and respectively combining the real-time characteristic data of the current target CVT with the error state reference characteristic data and the normal state reference characteristic data to form data pairs, and then inputting the data pairs into a trained CVT error evaluation model to obtain an error evaluation result.
Description
Technical Field
The invention relates to a CVT error state online evaluation method, a CVT error state online evaluation system, CVT error state online evaluation equipment and a CVT error state online evaluation medium, and belongs to the technical field of power equipment state evaluation.
Background
The failure rate of the capacitive voltage transformer (Capacitor Voltage Transformer, CVT) is relatively high and continued operation of the CVT in a failure state will result in loss of trade settlement for the market parties, thus requiring accurate assessment of the CVT's metering error status. Offline verification is a widely used method at present, and the method performs periodic offline verification on the CVT according to the regulations of related regulations, but the offline verification needs to perform non-fault power failure operation on a power system, and frequent power failure is unfavorable for maintaining the reliability of power supply of the system; and the detection equipment such as the booster, the physical standard device and the like has large volume and heavy weight, and is inconvenient to carry to the site for verification. These factors cause a large number of CVTs to be not timely verified, the metering error state of the CVTs is unknown, and hidden danger is brought to the safe operation of the power system. Therefore, on-line evaluation research on metering errors of CVT is conducted, which has important significance for further maintaining fairness of electric energy trade settlement and safe operation of the system.
The prior art, such as the chinese patent with patent number CN114062993a, discloses a CVT error state prediction method based on a time convolution network. The method comprises the following steps: calculating to obtain CVT characteristic parameter values, and generating a time sequence; processing the outlier using Hampel filters; the processed sequence is checked by ADF, and the stability is judged, if the time sequence is non-stable, the time sequence is converted into a stable time sequence through differential exponential smoothing; training the time convolution network by using the training samples; in the test stage, inputting a test sample into a stored network model, outputting an error prediction value of the CVT, comparing the error prediction value with a true value, and checking the validity of the network model; the trained time convolution network model is applied to the state prediction of the CVT, so that whether the future error of the CVT has the possibility of out-of-tolerance can be judged.
The problem with the prior art described above is that the environment in which the CVT operates is variable, the characteristics of the different CVT's required under normal and error conditions are likely to differ significantly, and the prediction of the method described above is likely to be inaccurate in the case of poor sample selection.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a CVT error state online evaluation method, a system, equipment and a medium.
The technical scheme of the invention is as follows:
In one aspect, the present invention provides a CVT error status online evaluation method, including the steps of:
Collecting historical characteristic data of a target CVT, adding an error or normal first label to the historical characteristic data, and then constructing a training sample set, wherein each training sample in the training sample set comprises a group of data pairs, two sample data in the data pairs are randomly extracted from the standard-added historical characteristic data, and meanwhile, a second label is added to the sample data according to whether the data pairs have the same first label or not;
Constructing a twin neural network comprising two convolutional neural networks, respectively inputting two sample data of a training sample into the two convolutional neural networks for feature extraction, and respectively obtaining a first feature value and a second feature value;
Calculating a characteristic distance between the first characteristic value and the second characteristic value, inputting the characteristic distance into at least one full-connection layer, and then outputting a consistency predicted value by using an activation function, wherein the consistency predicted value is used for measuring the similarity between the two inputs;
Calculating loss based on the difference between the consistency predicted value and the second label, and training the twin neural network based on the loss to obtain a trained CVT error assessment model;
And acquiring error state reference characteristic data and normal state reference characteristic data of the target CVT, respectively combining the real-time characteristic data of the current target CVT with the error state reference characteristic data and the normal state reference characteristic data to form data pairs, and inputting the data pairs into a trained CVT error evaluation model to obtain an error evaluation result.
In a preferred embodiment, the step of building the training sample set after adding the error or the normal first label to the historical feature data specifically includes:
acquiring historical characteristic data of target CVT and constructing data set Wherein/>Historical characteristic data representing the ith moment, and adding a first label representing a normal or error to each historical characteristic data;
Let data set From dataset/>Random extraction/>From dataset/>Random extraction/>Wherein T represents the total number of historical moments; construction of data pairs/>If/>And/>Identical to the first tag, then data pair/>Is 1, whereas is 0;
and (5) taking the data pair added with the second label as a training sample and putting the training sample into a training sample set.
As a preferred embodiment, the convolutional neural network comprises a multi-layer one-dimensional convolutional structure, a stitching operation, and a channel attention mechanism, wherein:
the multi-layer one-dimensional convolution structure is used for inputting data, and the convolution kernels of the one-dimensional convolution structures are different in size;
the splicing operation is used for longitudinally splicing the feature vectors output by the multilayer one-dimensional convolution structure to obtain two-dimensional feature vectors;
the channel attention mechanism is used for acquiring the importance degree of each feature vector in the two-dimensional feature vectors, assigning a weight value to each feature vector according to the importance degree, and calculating the weighted two-dimensional feature values through the weight values.
As a preferred embodiment, the step of calculating the feature distance between the first feature value and the second feature value specifically includes:
Taking the absolute value of the difference value between the first characteristic value and the second characteristic value as a third characteristic value;
Taking the first characteristic value as a query vector, taking the second characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the first characteristic value and the third characteristic value by using an attention mechanism to obtain a first correlation characteristic value;
Taking the second characteristic value as a query vector, taking the first characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the second characteristic value and the third characteristic value by using an attention mechanism to obtain a second correlation characteristic value;
And carrying out feature fusion on the first related feature value and the second related feature value to obtain a feature distance.
On the other hand, the invention also provides an online CVT error state assessment system, which is characterized by comprising the following steps:
The system comprises a sample set construction module, a target CVT, a first label and a second label, wherein the sample set construction module is used for acquiring historical characteristic data of the target CVT, and constructing a training sample set after adding an error or a normal first label to the historical characteristic data, each training sample in the training sample set comprises a group of data pairs, two sample data in the data pairs are randomly extracted from the historical characteristic data after adding the standard, and meanwhile, a second label is added to the sample data according to whether the data pairs have the same first label or not;
The network construction module is used for constructing a twin neural network comprising two convolutional neural networks, respectively inputting two sample data of a training sample into the two convolutional neural networks for feature extraction, and respectively obtaining a first feature value and a second feature value;
The feature fusion module is used for calculating the feature distance between the first feature value and the second feature value, inputting the feature distance into at least one full-connection layer, and then outputting a consistency predicted value by using an activation function, wherein the consistency predicted value is used for measuring the similarity between the two inputs;
The network training module is used for calculating loss based on the difference between the consistency predicted value and the second label, training the twin neural network based on the loss and obtaining a trained CVT error evaluation model;
The error evaluation module is used for acquiring error state reference characteristic data and normal state reference characteristic data of the target CVT, respectively combining the real-time characteristic data of the current target CVT with the error state reference characteristic data and the normal state reference characteristic data to form data pairs, and then inputting the data pairs into the trained CVT error evaluation model to obtain an error evaluation result.
In a preferred embodiment, in the sample set construction module, the step of constructing the training sample set after adding the error or the normal first label to the historical feature data specifically includes:
acquiring historical characteristic data of target CVT and constructing data set Wherein/>Historical characteristic data representing the ith moment, and adding a first label representing a normal or error to each historical characteristic data;
Let data set From dataset/>Random extraction/>From dataset/>Random extraction/>Wherein T represents the total number of historical moments; construction of data pairs/>If/>And/>Identical to the first tag, then data pair/>Is 1, whereas is 0;
and (5) taking the data pair added with the second label as a training sample and putting the training sample into a training sample set.
As a preferred embodiment, in the network construction module, the convolutional neural network includes a multi-layer one-dimensional convolutional structure, a splicing operation, and a channel attention mechanism, wherein:
the multi-layer one-dimensional convolution structure is used for inputting data, and the convolution kernels of the one-dimensional convolution structures are different in size;
the splicing operation is used for longitudinally splicing the feature vectors output by the multilayer one-dimensional convolution structure to obtain two-dimensional feature vectors;
the channel attention mechanism is used for acquiring the importance degree of each feature vector in the two-dimensional feature vectors, assigning a weight value to each feature vector according to the importance degree, and calculating the weighted two-dimensional feature values through the weight values.
In a preferred embodiment, in the feature fusion module, the step of calculating the feature distance between the first feature value and the second feature value specifically includes:
Taking the absolute value of the difference value between the first characteristic value and the second characteristic value as a third characteristic value;
Taking the first characteristic value as a query vector, taking the second characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the first characteristic value and the third characteristic value by using an attention mechanism to obtain a first correlation characteristic value;
Taking the second characteristic value as a query vector, taking the first characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the second characteristic value and the third characteristic value by using an attention mechanism to obtain a second correlation characteristic value;
And carrying out feature fusion on the first related feature value and the second related feature value to obtain a feature distance.
In yet another aspect, the present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the CVT error status online assessment method according to any of the embodiments of the present invention when the program is executed by the processor.
In yet another aspect, the present invention further provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements a CVT error status online assessment method according to any of the embodiments of the present invention.
The invention has the following beneficial effects:
1. The method comprises the steps of constructing paired sample data, inputting the data pairs into a twin neural network model, and calculating the similarity between the data pairs; the CVT error evaluation model capable of distinguishing whether the input data pair is consistent is obtained by calculating the loss function according to the similarity and the label value of the data pair, so that the online evaluation of the CVT error state by utilizing the real-time characteristic of the target CVT and the normal and error reference characteristic is realized, the evaluation mode does not have too much requirements on the sample CVT, and the accurate evaluation can be realized by only confirming that the normal and error reference characteristic data is accurate.
2. The invention improves the convolutional neural network, a plurality of one-dimensional convolutional layers (1D-CNN) are arranged in the network, the convolutional kernels of each one-dimensional convolutional structure are different in size, and the operation can extract local features of different sizes of input data, so that the sharing and abstraction of the features are realized, and the change of the input data by the network is more robust and accurate.
3. In the feature fusion, the correlation between 2 two-dimensional feature values and the difference value is calculated by using a 2-time attention mechanism, and finally, the feature distance is obtained by a value-by-value addition averaging method to serve as a fusion feature, so that the correlation between the features can be accurately represented.
Drawings
FIG. 1 is a schematic flow chart of a method according to an embodiment of the invention;
FIG. 2 is a schematic diagram of a convolutional neural network according to an embodiment of the present invention;
Fig. 3 is a schematic flow chart of feature fusion in an embodiment of the invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be understood that the step numbers used herein are for convenience of description only and are not limiting as to the order in which the steps are performed.
It is to be understood that the terminology used in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
The terms "comprises" and "comprising" indicate the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
The term "and/or" refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
Embodiment one:
Referring to fig. 1, the present embodiment provides an online CVT error status assessment method, including the following steps:
S100, acquiring historical characteristic data of a target CVT, adding error or normal first labels to the historical characteristic data, and then constructing a training sample set, wherein each training sample in the training sample set comprises a group of data pairs, two sample data in the data pairs are randomly extracted from the standard-added historical characteristic data, and meanwhile, a second label is added to the sample data according to whether the data pairs have the same first labels;
s200, constructing a twin neural network comprising two convolutional neural networks, respectively inputting two sample data of a training sample into the two convolutional neural networks for feature extraction, and respectively obtaining a first feature value and a second feature value;
S300, calculating a characteristic distance between a first characteristic value and a second characteristic value, inputting the characteristic distance into at least one full-connection layer, and then outputting a consistency predicted value by using an activation function, wherein the consistency predicted value is used for measuring the similarity between the two inputs;
s400, calculating loss based on a difference between the consistency predicted value and the second label, and training the twin neural network based on the loss to obtain a trained CVT error evaluation model;
S500, acquiring error state reference characteristic data and normal state reference characteristic data of a target CVT, and when online evaluation is carried out, only the real-time characteristic data of the current target CVT and the error state reference characteristic data and the normal state reference characteristic data are respectively formed into data pairs and then are input into a trained CVT error evaluation model, the CVT error evaluation model can distribute and output the similarity of the real-time characteristic data and the error state reference characteristic data and the similarity of the real-time characteristic data and the normal state reference characteristic data, and the real-time characteristic data can be judged to be closer to the state by comparing the two similarities, so that an error evaluation result is obtained.
As a preferred implementation manner of this embodiment, the step of constructing the training sample set after adding the error or the normal first label to the historical feature data specifically includes:
acquiring historical characteristic data of target CVT and constructing data set Wherein/>Historical characteristic data representing the ith moment, and adding a first label representing a normal or error to each historical characteristic data;
Let data set From dataset/>Random extraction/>From dataset/>Random extraction/>Wherein T represents the total number of historical moments; construction of data pairs/>If/>And/>Identical to the first tag, then data pair/>If the second tag of (2) is 1And/>Unlike the first tag, then data pair/>Is 0;
and (5) taking the data pair added with the second label as a training sample and putting the training sample into a training sample set.
In one implementation, the present embodiment improves the convolutional neural network, the improved convolutional neural network structure is shown in fig. 2, which includes a multi-layer one-dimensional convolutional structure, a splicing operation, and a channel attention mechanism, wherein:
The multi-layer one-dimensional convolution structure is used for inputting data, and takes the input data as input data For example, will/>Inputting the m feature vectors into m one-dimensional convolution structures to obtain m feature vectors; in the embodiment, the convolution kernels of each one-dimensional convolution structure are different in size, so that local features of input data with different sizes can be extracted through the operation, feature sharing and abstraction are realized, and therefore the change of the network on the input data is more robust and accurate. Before the input data is convolved, it needs to be padded to ensure that the size of the output data of the convolutional layer is consistent with the input data.
Then, the feature vectors extracted from the plurality of one-dimensional convolution structures are longitudinally spliced to obtainTwo-dimensional eigenvectors/>；
Finally, a channel attention mechanism is utilized to obtain a two-dimensional feature vectorThe importance degree of each feature vector is then used to assign a weight value to each feature vector, and the weighted two-dimensional feature value is calculated by the weight value. Thus, the model focuses on some feature vectors, promotes feature vectors which are useful for the current task, and suppresses feature vectors which are less useful for the current task.
In this embodiment, the implementation process of the channel attention mechanism is mainly divided into three steps: ① By global averaging poolingData compression of each channel (row) into a real number; ② Generating a weight value for each channel (row), typically using two fully connected layers to calculate the value, while ensuring that the number of weight values output is the same as the number of inputs; ③ Carrying out sigmoid normalization on the weight obtained in the previous step, and carrying out normalization on the weight and the original binary characteristic/>Multiplying channel by channel (row) to generate weighted two-dimensional features/>。
By the aboveTwo-dimensional feature/>, obtained by feature extractionI.e. the first eigenvalue, for/>Two-dimensional feature/>, obtained by feature extractionThe second characteristic value is obtained.
As a preferred implementation manner of this embodiment, the step of calculating the feature distance between the first feature value and the second feature value specifically includes:
will first characteristic value And a second eigenvalue/>The absolute value of the difference value is taken as a third characteristic value/>; To characterize the first eigenvalue/>Second eigenvalue/>And third eigenvalue/>The correlation between the two is performed by two attentive mechanism operations, referring specifically to fig. 3, the steps are as follows:
First, the first characteristic value As a Query vector Query, the second eigenvalue/>And third eigenvalue/>Respectively serving as a Key vector Key and a Value vector Value, and calculating a first characteristic Value/>, by using an attention mechanismAnd third eigenvalue/>The correlation between the two values obtains a first correlation characteristic value;
Then, the second characteristic value As a Query vector Query, the first eigenvalue/>And third eigenvalue/>Respectively serving as a Key vector Key and a Value vector Value, and calculating a second characteristic Value/>, by using an attention mechanismAnd third eigenvalue/>The correlation between the two values obtains a second correlation characteristic value;
finally, feature fusion is carried out, the first correlation feature value and the second correlation feature value are added element by element, and then average is carried out to obtain the feature distance 。
At the time of obtaining the characteristic distanceAfter that, will/>Input into one or some fully connected layers, and then map the value between 0 and 1 by using Sigmoid activation function to output consistency predicted value/>，/>Similarity between two inputs is measured, if/>And/>Similar (/ >)And/>The first tag is the same), then/>Should be close to 1 if/>And/>Similarity dissimilarity (/ >)And/>Different from the first tag of (c), then/>Should be close to 0.
Based on the above embodiment, in step S400, the second tag of the data pair and the consistency prediction value are utilizedThe loss is calculated by selecting an appropriate loss function, and Contrastive Loss loss functions and Triplte Loss loss functions can be used. Then gradient descent and back propagation, wherein the back propagation firstly updates the parameters of the fully connected layer and then further propagates to the parameters of the convolution layer; through continuous iteration, the twin neural network has the capability of distinguishing whether the input data and the label are consistent.
Embodiment two:
The present embodiment provides an online CVT error status assessment system, which is characterized by comprising:
The system comprises a sample set construction module, a target CVT, a first label and a second label, wherein the sample set construction module is used for acquiring historical characteristic data of the target CVT, and constructing a training sample set after adding an error or a normal first label to the historical characteristic data, each training sample in the training sample set comprises a group of data pairs, two sample data in the data pairs are randomly extracted from the historical characteristic data after adding the standard, and meanwhile, a second label is added to the sample data according to whether the data pairs have the same first label or not; the module is used for implementing the function of step S100 in the first embodiment, and will not be described here again;
the network construction module is used for constructing a twin neural network comprising two convolutional neural networks, respectively inputting two sample data of a training sample into the two convolutional neural networks for feature extraction, and respectively obtaining a first feature value and a second feature value; the module is used for implementing the function of step S200 in the first embodiment, and will not be described in detail herein;
the feature fusion module is used for calculating the feature distance between the first feature value and the second feature value, inputting the feature distance into at least one full-connection layer, and then outputting a consistency predicted value by using an activation function, wherein the consistency predicted value is used for measuring the similarity between the two inputs; the module is used for implementing the function of step S300 in the first embodiment, and will not be described in detail herein;
The network training module is used for calculating loss based on the difference between the consistency predicted value and the second label, training the twin neural network based on the loss and obtaining a trained CVT error evaluation model; the module is used for realizing the function of step S400 in the first embodiment, and will not be described in detail herein;
The error evaluation module is used for acquiring error state reference characteristic data and normal state reference characteristic data of the target CVT, respectively forming data pairs by the real-time characteristic data of the current target CVT and the error state reference characteristic data and the normal state reference characteristic data, and then inputting the data pairs into the trained CVT error evaluation model to obtain an error evaluation result; the module is used to implement the function of step S500 in the first embodiment, and will not be described herein.
As a preferred implementation manner of this embodiment, in the sample set construction module, the step of constructing the training sample set after adding the error or the normal first label to the historical feature data specifically includes:
acquiring historical characteristic data of target CVT and constructing data set Wherein/>Historical characteristic data representing the ith moment, and adding a first label representing a normal or error to each historical characteristic data;
Let data set From dataset/>Random extraction/>From dataset/>Random extraction/>Wherein T represents the total number of historical moments; construction of data pairs/>If/>And/>Identical to the first tag, then data pair/>Is 1, whereas is 0;
and (5) taking the data pair added with the second label as a training sample and putting the training sample into a training sample set.
As a preferred implementation manner of this embodiment, in the network construction module, the convolutional neural network includes a multi-layer one-dimensional convolutional structure, a splicing operation, and a channel attention mechanism, where:
the multi-layer one-dimensional convolution structure is used for inputting data, and the convolution kernels of the one-dimensional convolution structures are different in size;
the splicing operation is used for longitudinally splicing the feature vectors output by the multilayer one-dimensional convolution structure to obtain two-dimensional feature vectors;
the channel attention mechanism is used for acquiring the importance degree of each feature vector in the two-dimensional feature vectors, assigning a weight value to each feature vector according to the importance degree, and calculating the weighted two-dimensional feature values through the weight values.
As a preferred implementation manner of this embodiment, in the feature fusion module, the step of calculating the feature distance between the first feature value and the second feature value specifically includes:
Taking the absolute value of the difference value between the first characteristic value and the second characteristic value as a third characteristic value;
Taking the first characteristic value as a query vector, taking the second characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the first characteristic value and the third characteristic value by using an attention mechanism to obtain a first correlation characteristic value;
Taking the second characteristic value as a query vector, taking the first characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the second characteristic value and the third characteristic value by using an attention mechanism to obtain a second correlation characteristic value;
And carrying out feature fusion on the first related feature value and the second related feature value to obtain a feature distance.
Embodiment III:
The embodiment provides an electronic device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the CVT error state online evaluation method according to any embodiment of the invention when executing the program.
Embodiment four:
The present embodiment proposes a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a CVT error status online evaluation method according to any one of the embodiments of the present invention.
In the embodiments of the present application, "at least one" means one or more, and "a plurality" means two or more. "and/or", describes an association relation of association objects, and indicates that there may be three kinds of relations, for example, a and/or B, and may indicate that a alone exists, a and B together, and B alone exists. Wherein A, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of the following" and the like means any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c may represent: a, b, c, a and b, a and c, b and c or a and b and c, wherein a, b and c can be single or multiple.
Those of ordinary skill in the art will appreciate that the various elements and algorithm steps described in the embodiments disclosed herein can be implemented as a combination of electronic hardware, computer software, and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In several embodiments provided by the present application, any of the functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (hereinafter referred to as ROM), a random access Memory (Random Access Memory hereinafter referred to as RAM), a magnetic disk, or an optical disk, etc., which can store program codes.
The foregoing description is only illustrative of the present invention and is not intended to limit the scope of the invention, and all equivalent structures or equivalent processes or direct or indirect application in other related technical fields are included in the scope of the present invention.
Claims (10)
1. An online CVT error status assessment method, comprising the steps of:
Collecting historical characteristic data of a target CVT, adding an error or normal first label to the historical characteristic data, and then constructing a training sample set, wherein each training sample in the training sample set comprises a group of data pairs, two sample data in the data pairs are randomly extracted from the standard-added historical characteristic data, and meanwhile, a second label is added to the sample data according to whether the data pairs have the same first label or not; constructing a twin neural network comprising two convolutional neural networks, respectively inputting two sample data of a training sample into the two convolutional neural networks for feature extraction, and respectively obtaining a first feature value and a second feature value; calculating a characteristic distance between the first characteristic value and the second characteristic value, inputting the characteristic distance into at least one full-connection layer, and then outputting a consistency predicted value by using an activation function, wherein the consistency predicted value is used for measuring the similarity between the two inputs; calculating loss based on the difference between the consistency predicted value and the second label, and training the twin neural network based on the loss to obtain a trained CVT error assessment model; and acquiring error state reference characteristic data and normal state reference characteristic data of the target CVT, respectively combining the real-time characteristic data of the current target CVT with the error state reference characteristic data and the normal state reference characteristic data to form data pairs, and inputting the data pairs into a trained CVT error evaluation model to obtain an error evaluation result.
2. The CVT error status online assessment method according to claim 1, wherein the step of constructing a training sample set after adding the error or the normal first label to the historical feature data specifically comprises: acquiring historical characteristic data of target CVT and constructing data setWherein/>Historical characteristic data representing the ith moment, and adding a first label representing a normal or error to each historical characteristic data; let data set/>From dataset/>Random extraction/>From dataset/>Random extraction/>Wherein T represents the total number of historical moments; construction of data pairs/>If/>And/>Identical to the first tag, then data pair/>Is 1, whereas is 0; and (5) taking the data pair added with the second label as a training sample and putting the training sample into a training sample set.
3. The CVT error state online assessment method of claim 1, wherein the convolutional neural network comprises a multi-layer one-dimensional convolutional structure, a stitching operation, and a channel attention mechanism, wherein: the multi-layer one-dimensional convolution structure is used for inputting data, and the convolution kernels of the one-dimensional convolution structures are different in size;
the splicing operation is used for longitudinally splicing the feature vectors output by the multilayer one-dimensional convolution structure to obtain two-dimensional feature vectors; the channel attention mechanism is used for acquiring the importance degree of each feature vector in the two-dimensional feature vectors, assigning a weight value to each feature vector according to the importance degree, and calculating the weighted two-dimensional feature values through the weight values.
4. The CVT error state online assessment method according to claim 1, wherein the step of calculating the feature distance between the first feature value and the second feature value specifically comprises:
Taking the absolute value of the difference value between the first characteristic value and the second characteristic value as a third characteristic value;
Taking the first characteristic value as a query vector, taking the second characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the first characteristic value and the third characteristic value by using an attention mechanism to obtain a first correlation characteristic value;
Taking the second characteristic value as a query vector, taking the first characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the second characteristic value and the third characteristic value by using an attention mechanism to obtain a second correlation characteristic value;
And carrying out feature fusion on the first related feature value and the second related feature value to obtain a feature distance.
5. An CVT error status online assessment system, comprising:
The system comprises a sample set construction module, a target CVT, a first label and a second label, wherein the sample set construction module is used for acquiring historical characteristic data of the target CVT, and constructing a training sample set after adding an error or a normal first label to the historical characteristic data, each training sample in the training sample set comprises a group of data pairs, two sample data in the data pairs are randomly extracted from the historical characteristic data after adding the standard, and meanwhile, a second label is added to the sample data according to whether the data pairs have the same first label or not;
The network construction module is used for constructing a twin neural network comprising two convolutional neural networks, respectively inputting two sample data of a training sample into the two convolutional neural networks for feature extraction, and respectively obtaining a first feature value and a second feature value;
The feature fusion module is used for calculating the feature distance between the first feature value and the second feature value, inputting the feature distance into at least one full-connection layer, and then outputting a consistency predicted value by using an activation function, wherein the consistency predicted value is used for measuring the similarity between the two inputs;
The network training module is used for calculating loss based on the difference between the consistency predicted value and the second label, training the twin neural network based on the loss and obtaining a trained CVT error evaluation model;
The error evaluation module is used for acquiring error state reference characteristic data and normal state reference characteristic data of the target CVT, respectively combining the real-time characteristic data of the current target CVT with the error state reference characteristic data and the normal state reference characteristic data to form data pairs, and then inputting the data pairs into the trained CVT error evaluation model to obtain an error evaluation result.
6. The CVT error status online assessment system according to claim 5, wherein the step of constructing the training sample set after adding the error or the normal first label to the historical feature data in the sample set construction module specifically comprises:
acquiring historical characteristic data of target CVT and constructing data set Wherein/>Historical characteristic data representing the ith moment, and adding a first label representing a normal or error to each historical characteristic data;
Let data set From dataset/>Random extraction/>From dataset/>Random decimation inWherein T represents the total number of historical moments; construction of data pairs/>If/>And/>Identical to the first tag, then data pair/>Is 1, whereas is 0;
and (5) taking the data pair added with the second label as a training sample and putting the training sample into a training sample set.
7. The CVT error state online assessment system of claim 5, wherein the convolutional neural network comprises a multi-layer one-dimensional convolutional structure, a splice operation, and a channel attention mechanism in the network building block, wherein:
the multi-layer one-dimensional convolution structure is used for inputting data, and the convolution kernels of the one-dimensional convolution structures are different in size;
the splicing operation is used for longitudinally splicing the feature vectors output by the multilayer one-dimensional convolution structure to obtain two-dimensional feature vectors;
the channel attention mechanism is used for acquiring the importance degree of each feature vector in the two-dimensional feature vectors, assigning a weight value to each feature vector according to the importance degree, and calculating the weighted two-dimensional feature values through the weight values.
8. The CVT error state online assessment system according to claim 5, wherein the feature fusion module calculates the feature distance between the first feature value and the second feature value by:
Taking the absolute value of the difference value between the first characteristic value and the second characteristic value as a third characteristic value;
Taking the first characteristic value as a query vector, taking the second characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the first characteristic value and the third characteristic value by using an attention mechanism to obtain a first correlation characteristic value;
Taking the second characteristic value as a query vector, taking the first characteristic value and the third characteristic value as key vectors and value vectors respectively, and calculating the correlation between the second characteristic value and the third characteristic value by using an attention mechanism to obtain a second correlation characteristic value;
And carrying out feature fusion on the first related feature value and the second related feature value to obtain a feature distance.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the CVT error status online assessment method of any one of claims 1 to 4 when the program is executed by the processor.
10. A computer-readable storage medium having stored thereon a computer program, which when executed by a processor implements the CVT error status online assessment method according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410375860.7A CN117970224B (en) | 2024-03-29 | 2024-03-29 | CVT error state online evaluation method, system, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410375860.7A CN117970224B (en) | 2024-03-29 | 2024-03-29 | CVT error state online evaluation method, system, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117970224A true CN117970224A (en) | 2024-05-03 |
| CN117970224B CN117970224B (en) | 2024-06-21 |
Family
ID=90864887
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410375860.7A Active CN117970224B (en) | 2024-03-29 | 2024-03-29 | CVT error state online evaluation method, system, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117970224B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2020103613A4 (en) * | 2020-11-23 | 2021-02-04 | Agricultural Information and Rural Economic Research Institute of Sichuan Academy of Agricultural Sciences | Cnn and transfer learning based disease intelligent identification method and system |
| CN113612733A (en) * | 2021-07-07 | 2021-11-05 | 浙江工业大学 | Twin network-based few-sample false data injection attack detection method |
| CN114398939A (en) * | 2021-12-06 | 2022-04-26 | 深圳市英维克信息技术有限公司 | Method, device and equipment for detecting system fault and storage medium |
| CN115424331A (en) * | 2022-09-19 | 2022-12-02 | 四川轻化工大学 | Human face relative relationship feature extraction and verification method based on global and local attention mechanism |
| CN115859077A (en) * | 2022-11-10 | 2023-03-28 | 浙江舜云互联技术有限公司 | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions |
| WO2023134402A1 (en) * | 2022-01-14 | 2023-07-20 | 中国科学院深圳先进技术研究院 | Calligraphy character recognition method based on siamese convolutional neural network |
| CN116562114A (en) * | 2023-04-25 | 2023-08-08 | 国网浙江省电力有限公司金华供电公司 | Power transformer fault diagnosis method based on graph convolution neural network |
| CN117115727A (en) * | 2023-07-13 | 2023-11-24 | 国网甘肃省电力公司 | Transformer substation defect judging method and system |
| CN117313525A (en) * | 2023-09-19 | 2023-12-29 | 内蒙古电力(集团)有限责任公司内蒙古超高压供电分公司 | CVT error state qualitative assessment model establishment method, device, equipment and medium |
-
2024
- 2024-03-29 CN CN202410375860.7A patent/CN117970224B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2020103613A4 (en) * | 2020-11-23 | 2021-02-04 | Agricultural Information and Rural Economic Research Institute of Sichuan Academy of Agricultural Sciences | Cnn and transfer learning based disease intelligent identification method and system |
| CN113612733A (en) * | 2021-07-07 | 2021-11-05 | 浙江工业大学 | Twin network-based few-sample false data injection attack detection method |
| CN114398939A (en) * | 2021-12-06 | 2022-04-26 | 深圳市英维克信息技术有限公司 | Method, device and equipment for detecting system fault and storage medium |
| WO2023134402A1 (en) * | 2022-01-14 | 2023-07-20 | 中国科学院深圳先进技术研究院 | Calligraphy character recognition method based on siamese convolutional neural network |
| CN115424331A (en) * | 2022-09-19 | 2022-12-02 | 四川轻化工大学 | Human face relative relationship feature extraction and verification method based on global and local attention mechanism |
| CN115859077A (en) * | 2022-11-10 | 2023-03-28 | 浙江舜云互联技术有限公司 | Multi-feature fusion motor small sample fault diagnosis method under variable working conditions |
| CN116562114A (en) * | 2023-04-25 | 2023-08-08 | 国网浙江省电力有限公司金华供电公司 | Power transformer fault diagnosis method based on graph convolution neural network |
| CN117115727A (en) * | 2023-07-13 | 2023-11-24 | 国网甘肃省电力公司 | Transformer substation defect judging method and system |
| CN117313525A (en) * | 2023-09-19 | 2023-12-29 | 内蒙古电力(集团)有限责任公司内蒙古超高压供电分公司 | CVT error state qualitative assessment model establishment method, device, equipment and medium |
Non-Patent Citations (1)
| Title |
|---|
| 黄桂平 等: "基于深度学习的电容式电压互感器故障诊断", 《四川轻化工大学学报 （自然科学版）》, vol. 36, no. 5, 31 October 2023 (2023-10-31), pages 77 - 83 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117970224B (en) | 2024-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110232240B (en) | Improved transformer top layer oil temperature prediction method | |
| Bernieri et al. | Neural networks and pseudo-measurements for real-time monitoring of distribution systems | |
| CN106663086A (en) | Apparatus and method for ensembles of kernel regression models | |
| CN114490065A (en) | Load prediction method, device and equipment | |
| CN116679211A (en) | Lithium battery health state prediction method | |
| CN115496144A (en) | Power distribution network operation scene determining method and device, computer equipment and storage medium | |
| CN115169809A (en) | Smart city evaluation method and device | |
| CN114385930A (en) | Interest point recommendation method and system | |
| CN113052411A (en) | Data product quality evaluation method and device | |
| CN117970224B (en) | CVT error state online evaluation method, system, equipment and medium | |
| CN109086954A (en) | Prediction technique, device, equipment and medium based on cash flow indicated yield | |
| CN108280485A (en) | A kind of non-rigid method for searching three-dimension model based on spectrogram Wavelet Descriptor | |
| CN112767190A (en) | Phase sequence identification method and device for transformer area based on multilayer stacked neural network | |
| CN111639194A (en) | Knowledge graph query method and system based on sentence vectors | |
| CN116306277A (en) | Landslide displacement prediction method and device and related components | |
| CN108829750A (en) | A kind of quality of data determines system and method | |
| CN115048290A (en) | Software quality evaluation method and device, storage medium and computer equipment | |
| Tizdast et al. | Self-similar but not conformally invariant traces obtained by modified Loewner forces | |
| Modha et al. | Prequential and cross-validated regression estimation | |
| CN111008324A (en) | Travel service pushing method, system and device under big data and readable storage medium | |
| KR102517455B1 (en) | Apparatus and method for calculating risk index based on learning reflecting risk score | |
| Hult et al. | Efficient calculation of risk measures by importance sampling--the heavy tailed case | |
| Zafar et al. | Forecasting inflation using functional time series analysis | |
| CN117612603A (en) | Protein mutation effect prediction method, device, equipment and medium | |
| CN105117328B (en) | DNN code test methods and device |
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 |