CN116933179A - High-voltage direct-current transmission system fault diagnosis method and system based on prototype network - Google Patents
High-voltage direct-current transmission system fault diagnosis method and system based on prototype network Download PDFInfo
- Publication number
- CN116933179A CN116933179A CN202310928195.5A CN202310928195A CN116933179A CN 116933179 A CN116933179 A CN 116933179A CN 202310928195 A CN202310928195 A CN 202310928195A CN 116933179 A CN116933179 A CN 116933179A
- Authority
- CN
- China
- Prior art keywords
- fault
- data
- fault diagnosis
- prototype network
- prototype
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 84
- 230000005540 biological transmission Effects 0.000 title claims abstract description 62
- 238000000034 method Methods 0.000 title claims abstract description 62
- 238000012549 training Methods 0.000 claims abstract description 49
- 238000012795 verification Methods 0.000 claims abstract description 16
- 238000012545 processing Methods 0.000 claims description 28
- 230000008569 process Effects 0.000 claims description 19
- 238000010606 normalization Methods 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000002372 labelling Methods 0.000 claims description 11
- 238000009826 distribution Methods 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000012797 qualification Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 abstract description 10
- 238000013499 data model Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 23
- 238000004590 computer program Methods 0.000 description 8
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 230000009471 action Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000002349 favourable effect Effects 0.000 description 2
- 238000010200 validation analysis Methods 0.000 description 2
- 208000025274 Lightning injury Diseases 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000012258 culturing Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005684 electric field Effects 0.000 description 1
- 238000009422 external insulation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000036962 time dependent Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Human Resources & Organizations (AREA)
- Probability & Statistics with Applications (AREA)
- Water Supply & Treatment (AREA)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
Abstract
The application provides a fault diagnosis method and a fault diagnosis system for a high-voltage direct-current transmission system based on a prototype network, which are characterized in that the prototype network is used for carrying out fault diagnosis, one-dimensional signal data is converted into a two-dimensional characteristic image through treatment of a Gellam angle field (Gramian Angular Field, GAF), a fault diagnosis model of the prototype network HVDC system is established, the model is trained by a training set data model, and feasibility of the prototype network is verified by verification set data. The application solves the problems of fault equipment, fault reasons and fault positioning when the HVDC system breaks down by using a prototype network algorithm, can rapidly and accurately judge the fault type from fault data in the HVDC system, and has important significance for analyzing and solving the fault problems in the HVDC system.
Description
Technical Field
The application relates to the technical field of power transmission safety of power systems, in particular to a fault diagnosis method and system of a high-voltage direct-current power transmission system based on a prototype network.
Background
In the last century, the installed capacity of the world generator is increased by 14.2 times and the generated energy is increased by 7.9 times in the period of 30 years from 50 to 80 years; the annual growth rate of the installed capacity of the generator reaches 9.5%, and the annual growth rate of the generated energy reaches 7.6%, which is equivalent to doubling every ten years. In the new century, people have increasingly depended on electric energy, and the generated energy of each country also presents a state of rapid rise. Meanwhile, the Chinese operators are wide, the territory area is huge, but the main resources for power generation are concentrated in northwest of China, and the power load is concentrated in eastern coastal areas. In this context, the construction of the power transmission system must also be adapted to the construction of the power supply, which means that the transmission capacity, transmission scale, and technical level of the power transmission system all need to be greatly improved. The method of adding a traditional alternating current transmission line and adding new equipment is obviously not suitable for the requirements of new potential states. The high-voltage direct-current transmission technology is widely applied to large-range long-distance point-to-point high-power transmission, such as Zhoushan direct-current transmission engineering, ge Zhouba-Shanghai + -500 kV direct-current transmission engineering, guizhou-Guangdong + -500 kV direct-current transmission engineering and the like, because the high-voltage direct-current transmission technology is favorable for improving the stability of an alternating-current system, can realize asynchronous networking of the alternating-current system, has higher transmission power, has the advantages of rapid and controllable active and reactive power, is favorable for improving the operation capability of the alternating-current system, limits the short-circuit capacity of the alternating-current system, has low line cost, has small operation loss and the like. The high-voltage direct-current transmission system consists of a transmitting end alternating-current system, a rectifying station, an inversion station, a direct-current transmission line and a first-stage alternating-current system. Many fault problems may occur in the operation process, and the fault problems are generally classified into converter faults, direct current line faults and alternating current part faults.
The traditional high-voltage alternating current line fault diagnosis mainly uses the information such as fault wave recording data and the like to manually analyze the waveform diagram of the analog quantity and the switching value by redrawing the waveforms of the voltage and current (analog quantity) and the protection action (switching value) during fault, and has the advantages of single means, low efficiency and long time consumption. The correctness of the manual evaluation result is directly related to the experience of an expert, and the objectivity and the comprehensiveness of the analysis result are difficult to ensure. And a large amount of actual case accumulation experience is needed for culturing an expert, but the complex alternating fault cases are fewer, when the system has complex faults or relay protection, and a plurality of uncertain factors such as misoperation or refusal operation occur, the conventional fault diagnosis cannot meet the actual requirements, and the normal operation of the system is influenced and even a large amount of economic losses are caused.
Disclosure of Invention
The application provides a fault diagnosis method and a fault diagnosis system for a high-voltage direct-current transmission system based on a prototype network aiming at solving the problem of how to quickly and accurately identify the type of faults occurring in the high-voltage direct-current transmission system.
In a first aspect, the present application provides a fault diagnosis method for a hvdc transmission system based on a prototype network, the method comprising:
collecting fault data of a high-voltage direct-current transmission system, and labeling the fault data;
dividing fault data with labels into a training set and a verification set; the training set comprises a support set and a query set;
converting one-dimensional signal data into a two-dimensional characteristic image through the treatment of a gram angle field, and establishing a prototype network HVDC system fault diagnosis model;
inputting the training set to the prototype network HVDC system fault diagnosis model for model training;
and the prototype network HVDC system fault diagnosis model after model training is verified to be qualified through the verification set is used for fault diagnosis of the high-voltage direct-current transmission system.
The application learns a function F through a prototype network, namely a meta learning algorithm, and the function F is learned on a training set, wherein F can automatically learn a function F, the input of F is a picture set, the output is a function F, the input of F is a picture, and the output is a label. The collected fault wave-recording one-dimensional signal data are converted into two-dimensional image data by adopting a gram angle field, and then the data are trained by adopting a prototype network model, so that a fault diagnosis algorithm of the high-voltage direct-current transmission system based on the prototype network is obtained, and the fault type of the high-voltage direct-current transmission system is rapidly and accurately identified.
In some implementations, the labeling the fault data is specifically:
classifying the fault data according to fault types; the fault types at least comprise fault types, fault levels, fault reasons, fault influence ranges or fault solutions.
In some implementations, the converting the one-dimensional signal data into the two-dimensional feature image by a gladhand angle field process includes:
s101: data normalization, scaling the data range to either [ -1,1] or [0,1];
s102: converting the normalized sequence data into a polar coordinate system;
s103: calculating the angle sum between the different points to identify the time correlation of the different points in time;
the angle and formula are:
in the above-mentioned method, the step of,for the ith and jth fault data x i And x j Angle value of (2); /> Representing time-series data before normalization processing; />Is time series data after standardized processing; i is an identity matrix, and the diagonal is a matrix with all 1's and the other elements being 0's.
In some implementations, the converting the one-dimensional signal data into the two-dimensional feature image by a gladhand angle field process further includes:
s201: data normalization, scaling the data range to either [ -1,1] or [0,1];
s202: converting the normalized sequence data into a polar coordinate system;
s203: calculating the angle difference between different points to identify the time correlation of the different points in time;
the angle difference formula is as follows:
in the above-mentioned method, the step of,for the ith and jth fault data x i And x j Angle value of (2); /> Representing time-series data before normalization processing; />Is time series data after standardized processing; i is an identity matrix, and the diagonal is a matrix with all 1's and the other elements being 0's.
In some implementations, the building a prototype network HVDC system fault diagnosis model includes:
s301: using encoders for each picture in the support set and the query set, respectivelyProcessing is carried out, and an encoding representation of each picture is learned;
s302: average value processing is carried out on the Embeddings coding representation under each category of the support set, and prototype representation of each category is obtained;
s303: similarity calculation is carried out on the Embedding code representation under each category of the support set and the prototype representation of each category of the support set;
s304: activating the similarity into probability distribution by adopting a softmax function to obtain a classification label of the query set picture;
s305: and performing cross entropy loss calculation on the classification labels and the true value labels of the query set pictures, and completing the training of an epoode through gradient back propagation.
In some implementations, the similarity calculation is formulated as:
in the method, in the process of the application,encoding representations for the eimbeddings under each category of the support set; c is a prototype representation of each class of the support set.
The similarity is activated into probability distribution by adopting a softmax function, and the formula is as follows:
in a second aspect, the present application further provides a fault diagnosis system for a hvdc transmission system based on a prototype network, the system comprising:
the acquisition unit is used for acquiring fault data of the high-voltage direct-current transmission system;
the first preprocessing unit is used for labeling the fault data;
the second preprocessing unit divides the fault data with the labels into a training set and a verification set;
the first processing unit is used for converting the one-dimensional signal data into a two-dimensional characteristic image through the treatment of a gram angle field and establishing a prototype network HVDC system fault diagnosis model;
the optimization processing unit inputs the training set to the fault diagnosis model of the prototype network HVDC system to perform model training; inputting the verification set into a trained prototype network HVDC system fault diagnosis model to perform qualification verification;
and the second processing unit is used for carrying out fault diagnosis on the high-voltage direct-current transmission system according to the optimized fault diagnosis model of the prototype network HVDC system.
In a third aspect, the application also provides a HVDC system fault diagnosis model employing prototype network based high voltage dc as described above.
In a fourth aspect, the present application also provides a computer storage medium, where the computer program, when executed by a processor, implements the method for diagnosing a fault of a hvdc transmission system based on a prototype network as described above.
Compared with the prior art, the technical scheme of the application has the beneficial effects that:
according to the application, the collected fault wave-recording one-dimensional signal data is converted into two-dimensional image data by adopting a gram angle field, and then the data is trained by adopting a prototype network model, so that a fault diagnosis algorithm of the high-voltage direct-current transmission system based on the prototype network is obtained, and the fault type of the high-voltage direct-current transmission system is rapidly and accurately identified.
Drawings
Fig. 1 is a flow chart of a fault diagnosis method of a high-voltage direct-current transmission system based on a prototype network.
Fig. 2 is a flowchart of the fault diagnosis algorithm described in fig. 1.
Fig. 3 is a diagram of a process of converting a one-dimensional signal into a two-dimensional image using a gladhand angle field according to the present application.
Fig. 4 is a graph illustrating a Convolutional Neural Network (CNN) according to the present application.
Fig. 5 is a prototype representation of the present application.
Fig. 6 is a schematic diagram of a fault diagnosis system of a hvdc transmission system based on a prototype network according to the present application.
Detailed Description
The following description of the embodiments of the present application 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 application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Embodiment one:
referring to fig. 1-2, a fault diagnosis method for a hvdc transmission system based on a prototype network according to a first embodiment of the present application is provided, where the method includes:
step 1: collecting fault data of a high-voltage direct-current transmission system, and labeling the fault data;
the step of labeling the fault data is specifically as follows:
classifying the fault data according to fault types; the fault types at least comprise fault types, fault levels, fault reasons, fault influence ranges or fault solutions.
Step 2: dividing fault data with labels into a training set and a verification set; the training set comprises a support set and a query set;
preferably, the labeling process is performed according to sample data, and the labeling process is specifically classified as:
(1) A converter failure;
(2) Direct current line fault;
(3) An ac system failure;
in this embodiment, 80% of the data is selected as the training set and 20% of the data is selected as the validation set.
Step 3: converting one-dimensional signal data into a two-dimensional characteristic image through the treatment of a gram angle field (Gramian Angular Field, GAF), and establishing a prototype network HVDC system fault diagnosis model;
the specific examples of the glamer angle field GAF are: converting the scaled 1D sequence data from a rectangular coordinate system to a polar coordinate system, and step 3: converting one-dimensional signal data into a two-dimensional characteristic image through the treatment of a gram angle field (Gramian Angular Field, GAF), and establishing a prototype network HVDC system fault diagnosis model; for each point in time value, it can be mapped to a point on the polar coordinate system. Polar coordinates convert a numerical value into coordinates in two dimensions, distance and angle.
The time correlation of the different points in time is then identified by taking into account the angle and/or difference between the different points. Depending on whether the angle and or the angle difference are made, in the present application, two implementation methods are preferably employed: corresponding to the angle sum GASF (Gramian Angular Summation Field) and corresponding to the angle difference GADF (Gramian Angular Difference Field). Normalization is performed on the GAF image, the pixel value range of the image is adjusted to be between 0 and 1, and a min-max normalization or other normalization methods can be adopted. The resulting normalized GAF image can be regarded as a two-dimensional representation of features reflecting the features of the original one-dimensional signal data in terms of time correlation.
By converting one-dimensional signal data into two-dimensional GAF images, HVDC system fault diagnosis models can be built in the prototype network using these images as input data. The prototype network can realize diagnosis and classification of faults by learning and analyzing patterns and features in the GAF image.
Step 4: inputting the training set to the prototype network HVDC system fault diagnosis model for model training;
training is performed using a training set input model. Forward propagation and backward propagation are repeated in an iterative optimization manner. The specific training process comprises the following steps:
input samples: a batch of samples in the training set is input into the prototype network model.
Forward propagation: and carrying out forward propagation on the input sample to obtain the prediction output of the model.
Calculating loss: and calculating the value of the loss function according to the prediction output and the sample label, and evaluating the performance of the model on the current batch.
Back propagation: and carrying out back propagation according to gradient information of the loss function, and updating parameters of the model to minimize the loss function.
Repeating the iteration: repeating the steps, dividing the training set into a plurality of batches, and performing repeated iterative training on the whole training set until a preset training round number is reached or convergence condition is reached.
The training set is input into a prototype network HVDC system fault diagnosis model to perform model training, so that the model can learn and understand the mode and the characteristic of the HVDC system fault in the GAF image, and accurate fault diagnosis is realized.
Step 5: and the prototype network HVDC system fault diagnosis model after model training is verified to be qualified through the verification set is used for fault diagnosis of the high-voltage direct-current transmission system.
In some implementations, HVDC systems consist essentially of: the system comprises a transmitting end alternating current system, a rectifying station, an inversion station, a direct current transmission line and a first-stage alternating current system.
Preferably, the types of faults common in HVDC systems are:
a. inverter failure:
valve short-circuit faults generally refer to the fact that the converter valve is suddenly and greatly climbed due to the born reverse voltage or the external insulation of the converter valve is damaged, so that the capacity of blocking the converter valve in the forward and reverse directions is lost, namely the converter valve can be conducted under the forward and reverse voltages.
b. Direct current line fault:
HVDC engineering is often applied to cross-region power transmission, and among various types of faults occurring on a dc line, the probability of occurrence of a short-circuit fault to ground is the largest, and it can account for approximately 80% or more of the faults of the dc line. Many factors can cause the short circuit fault of the direct current circuit to the ground, and lightning stroke, pollution, branch contact line and the like are not required to occur frequently. When a short circuit to ground fault occurs, the current rapidly rises due to the instant release of the electric field energy originally stored on the line, and the magnitude of the current is related to the distance between the fault occurrence point and the rectifying station, and the closer the fault occurrence point is to the rectifying station, the smaller the ground resistance of the outlet of the rectifying station is, and the larger the short circuit current is.
c. Ac part failure:
an alternating current system phase-to-ground short circuit fault or a single-phase to-ground short circuit fault at the rectifying side and an alternating current system two-phase short circuit fault or a single-phase to-ground short circuit fault at the inverting side. When any two phases of the alternating current system are short-circuited, the two phases form short circuits, and two-phase short-circuit current occurs in the alternating current system. If a fault occurs on the rectifying side, the rectifier cannot complete normal commutation work due to loss of two-phase voltage, so that current and voltage on the direct current line are rapidly reduced, and the transmission power is linearly reduced. If the fault occurs on the inversion side, the inverter commutation failure is caused, and the current of the direct current transmission line is increased and the current of the alternating current system is reduced in the morning.
When the AC system has a single-phase to-ground short circuit fault, the short circuit current can be directly led to pass through the converter valve conducted in the converter and then reaches the neutral end of the DC side through the grounding grid and the grounding system, so that a short circuit loop is formed, and the fault characteristic of the short circuit loop is similar to that of the converter valve. If a fault occurs on the rectifying side, it is also necessary to prevent dc loop resonance that may be caused by the second harmonic entering the dc line.
And collecting fault conditions from the existing HVDC system, and analyzing fault wave recording to obtain fault data.
In some implementations, as shown in fig. 3, the converting the one-dimensional signal data into a two-dimensional feature image by a gladhand field process includes:
s101: data normalization, scaling the data range to either [ -1,1] or [0,1];
or alternatively, the first and second heat exchangers may be,
wherein x is i Representing fault data; x is x min Representing a fault data minimum; x is x max Representing a fault data maximum;andrepresenting the normalized fault data.
S102: converting the normalized sequence data into a polar coordinate system;
the normalized sequence data is converted into a coordinate system, namely, a numerical value is regarded as an included angle cosine value, a timestamp is regarded as a radius, and the formula is as follows:
s103: calculating the angle sum between the different points to identify the time correlation of the different points in time;
the angle and formula are:
in the above, phi i ，φ j For the ith and jth fault data x i And x i Angle value of (2); representing time-series data before normalization processing; />Is time series data after standardized processing; i is an identity matrix, and the diagonal is a matrix with all 1's and the other elements being 0's.
In some implementations, the converting the one-dimensional signal data into the two-dimensional feature image by a gladhand angle field process further includes:
s201: data normalization, scaling the data range to either [ -1,1] or [0,1];
s202: converting the normalized sequence data into a polar coordinate system;
s203: calculating the angle difference between different points to identify the time correlation of the different points in time;
the angle difference formula is as follows:
in the above-mentioned method, the step of,for the ith and jth fault data x i And x j Angle value of (2); /> Representing time-series data before normalization processing; />Is time series data after standardized processing; i is an identity matrix, and the diagonal is a matrix with all 1's and the other elements being 0's.
In some implementations, the building a prototype network HVDC system fault diagnosis model includes:
s301: using encoders for each picture in the support set and the query set, respectivelyProcessing is carried out, and an encoding representation of each picture is learned;
s302: average processing is carried out on the Embedding code representation under each category of the support set to obtain a prototype representation of each category, as shown in FIG. 5;
s303: similarity calculation is carried out on the Embedding code representation under each category of the support set and the prototype representation of each category of the support set;
s304: activating the similarity into probability distribution by adopting a softmax function to obtain a classification label of the query set picture;
s305: and performing cross entropy loss calculation on the classification labels and the true value labels of the query set pictures, and completing the training of an epoode through gradient back propagation.
In some implementations, the establishing a prototype network HVDC system fault diagnosis model includes the following calculation processes:
(1) Assuming the original dataset is D, the internal samples are shown in the form of { (x) 1 ,y 1 ),(x 2 ,y 2 ),…,(x n ,y n ) Where x represents a vector representation and y represents a class label.
(2) For each class, n sample points are randomly generated for it from the total sample set, and for each class, a support set is generated as S.
(3) Similarly, n sample points are randomly selected from the total sample set for each class to generate a query set Q, containing one support set and one query set for each epoode.
(4) For sample points within the support set, the coding formula is usedTo generate a prototype representation for each class, here coding formula +.>There may be any way of information extraction. Such as CNN, LSTM, etc., the present application employs a CNN model, as shown in fig. 4.
(5) For each class, we generate its prototype representation as:
(6) A code for the query set is also generated for the query set.
(7) Further, calculating Euclidean distance conditions of prototype representations of the query set and the support set;
(8) Finally, the probability P of the current sample belonging to each class is calculated ω (y=k|x) calculation method using softmax here:
(9) Finally, the loss function is calculated as
The original assembled network is optimized by minimizing the loss function.
Prototype network model training process: prototype network (Prototypical Network) is a prototype-based metric learning method for classification tasks without or with weak supervision. The main idea is to build class boundaries by learning prototypes of samples and assign test samples to classes to which prototypes closest thereto belong. The following is a training process of the prototype network model:
1. data preparation: a training dataset is first prepared, comprising input samples and their corresponding class labels. The class labels for each sample need to be discrete, limited classes.
2. Prototype calculation: and calculating the average value vector of all samples of each category for each category to obtain a prototype of the category. The prototype vector is the average of the sample vectors for each class and is used to represent the center position of that class.
3. Distance measurement: a distance measurement method (such as euclidean distance, cosine similarity, etc.) is selected for measuring the distance between the test sample and each class prototype.
4. Loss function definition: a loss function is defined, typically a cross entropy loss function. The loss function is used to measure the classification performance of the model for the training samples.
5. Model training: the model is trained by means of iterative optimization. In each iteration, a batch of samples (mini-batch) is randomly selected from the training data set, the distance between the samples and the class prototype is calculated, and model parameters are updated by using a loss function.
6. Model evaluation: the performance of the model is evaluated using a validation set or test set. The classification accuracy or other indicator of the model for the new sample is calculated.
7. Super parameter tuning: different model hyper-parameters (e.g., learning rate, batch size, etc.) are tried, and the best hyper-parameters are selected by verifying the performance of the set.
8. Model application: the new samples are classified using the trained prototype network model.
By learning prototypes of the samples, the prototype network can capture the feature distribution between categories, thus achieving efficient classification. In particular implementations, adjustments and modifications may be made as appropriate to the particular task and characteristics of the data set.
The application learns a function F through a prototype network, namely a meta learning algorithm, and the function F is learned on a training set, wherein F can automatically learn a function F, the input of F is a picture set, the output is a function F, the input of F is a picture, and the output is a label. The collected fault wave-recording one-dimensional signal data are converted into two-dimensional image data by adopting a gram angle field, and then the data are trained by adopting a prototype network model, so that a fault diagnosis algorithm of the high-voltage direct-current transmission system based on the prototype network is obtained, and the fault type of the high-voltage direct-current transmission system is rapidly and accurately identified.
Embodiment two:
the application also provides a fault diagnosis system of the HVDC transmission system based on the prototype network, as shown in fig. 6, the system comprises:
the acquisition unit is used for acquiring fault data of the high-voltage direct-current transmission system;
the first preprocessing unit is used for labeling the fault data;
the second preprocessing unit divides the fault data with the labels into a training set and a verification set;
the first processing unit is used for converting the one-dimensional signal data into a two-dimensional characteristic image through the treatment of a gram angle field and establishing a prototype network HVDC system fault diagnosis model; the corresponding GAF method, such as angle and GASF or angle difference GADF, is used to convert the one-dimensional data into a two-dimensional time-dependent feature image. These images will be used as inputs for building a prototype network HVDC system fault diagnosis model.
The optimization processing unit inputs the training set to the fault diagnosis model of the prototype network HVDC system to perform model training; inputting the verification set into a trained prototype network HVDC system fault diagnosis model to perform qualification verification; and through an iterative optimization mode, the model learns the mode and the characteristic of the HVDC system fault according to the input GAF image.
And the second processing unit is used for carrying out fault diagnosis on the high-voltage direct-current transmission system according to the optimized fault diagnosis model of the prototype network HVDC system.
Preferably, a fault diagnosis model of a prototype network HVDC system is established, which comprises the following procedures:
(1) Assuming that the original data set is D, for each epi-code, one support set and one query set are contained;
(2) Extracting the characteristics of each picture by using an encoder;
(3) Calculating a prototype of each category in the support set;
(4) Calculating the similarity between each query set picture Embedding and each category;
(5) Optimizing the loss function.
The output result is the fault type in the current power transmission system.
From fault data acquisition to final fault diagnosis, the prototype network HVDC system fault diagnosis model can be effectively utilized to automatically identify and predict faults of the high-voltage direct-current power transmission system, and diagnosis efficiency and accuracy are improved.
Embodiment III:
the application also provides a fault diagnosis model of the HVDC system, which adopts the fault diagnosis model of the prototype network HVDC system in the fault diagnosis method of the high-voltage direct-current power transmission system based on the prototype network. The model models the behavior of the HVDC system in normal and various fault states by building a prototype network. The prototype network is then trained using the fault data to enable the network to identify and classify various fault types.
The basic steps of the model include:
HVDC system data is collected in normal and fault conditions.
A prototype network may be constructed using neural networks or the like.
The data is input into the prototype network for training, so that the prototype network learns the characteristics of different fault modes.
The accuracy and effectiveness of the model is verified and can be assessed using the test dataset.
The actual fault diagnosis and classification is performed using the trained models.
By using the prototype network-based HVDC system fault diagnosis model, operation and maintenance personnel can detect faults in real time by monitoring the behaviors and states of the system and take measures in time to repair, so that the reliability and stability of the HVDC system are improved.
Embodiment four:
the application also provides a computer storage medium, and the computer program is executed by a processor to realize the fault diagnosis method of the high-voltage direct-current transmission system based on the prototype network.
The computer storage medium may be any medium suitable for storing and executing a computer program, such as a hard disk drive, a solid state disk, a memory, and the like.
In the computer storage medium, one or more computer programs may be stored, which are loaded and executed by a processor. The programs realize a fault diagnosis method of the HVDC transmission system based on the prototype network. In particular, these programs may include algorithms and logic for data processing, model construction, training, and fault diagnosis.
When the processors execute these computer programs, they will read the corresponding input data (HVDC system data), build a prototype network according to predefined algorithms and logic, and train and tune the prototype network. The processor may then use the trained model to receive real-time HVDC system data and perform fault diagnosis and classification based on the output of the prototype network.
In summary, the present application can implement a prototype network-based fault diagnosis method for a HVDC system on a processor by using a computer program in such a computer storage medium, thereby improving reliability and stability of the HVDC system.
Although the illustrative embodiments have been described herein with reference to the accompanying drawings, it is to be understood that the above illustrative embodiments are merely illustrative and are not intended to limit the scope of the present application thereto. Various changes and modifications may be made therein by one of ordinary skill in the art without departing from the scope and spirit of the application. All such changes and modifications are intended to be included within the scope of the present application as set forth in the appended claims.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of 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.
In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described device embodiments are merely illustrative, e.g., the division of the elements is merely a logical functional division, and there may be additional divisions when actually implemented, e.g., multiple elements or components may be combined or integrated into another device, or some features may be omitted or not performed.
Various component embodiments of the application may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that some or all of the functions of some of the modules according to embodiments of the present application may be implemented in practice using a microprocessor or Digital Signal Processor (DSP). The present application can also be implemented as an apparatus program (e.g., a computer program and a computer program product) for performing a portion or all of the methods described herein. Such a program embodying the present application may be stored on a computer readable medium, or may have the form of one or more signals. Such signals may be downloaded from an internet website, provided on a carrier signal, or provided in any other form.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
While the application has been described in conjunction with the specific embodiments above, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art in light of the foregoing description. Accordingly, all such alternatives, modifications, and variations are included within the spirit and scope of the following claims.
Claims (10)
1. A prototype network-based fault diagnosis method for a high-voltage direct-current transmission system, the method comprising:
collecting fault data of a high-voltage direct-current transmission system, and labeling the fault data;
dividing fault data with labels into a training set and a verification set; the training set comprises a support set and a query set;
converting one-dimensional signal data into a two-dimensional characteristic image through the treatment of a gram angle field, and establishing a prototype network HVDC system fault diagnosis model;
inputting the training set to the prototype network HVDC system fault diagnosis model for model training;
and the prototype network HVDC system fault diagnosis model after model training is verified to be qualified through the verification set is used for fault diagnosis of the high-voltage direct-current transmission system.
2. The fault diagnosis method for a prototype network-based hvdc transmission system according to claim 1, wherein said labeling said fault data comprises:
classifying the fault data according to fault types; the fault types at least comprise fault types, fault levels, fault reasons, fault influence ranges or fault solutions.
3. The fault diagnosis method for a prototype network-based hvdc transmission system according to claim 2, wherein said converting one-dimensional signal data into a two-dimensional characteristic image by means of a gladhand field process comprises:
s101: data normalization, scaling the data range to either [ -1,1] or [0,1];
s102: converting the normalized sequence data into a polar coordinate system;
s103: calculating the angle sum between the different points to identify the time correlation of the different points in time;
the angle and formula are:
in the above-mentioned method, the step of,for the ith and jth fault data x i And x j Angle value of (2); /> Representing time-series data before normalization processing; />Is time series data after standardized processing; i identity matrix, and a matrix with all 1's on the diagonal and other elements 0.
4. The prototype network-based hvth system fault diagnosis method according to claim 2, wherein the converting one-dimensional signal data into a two-dimensional feature image by a gladhand field process further comprises:
s201: data normalization, scaling the data range to either [ -1,1] or [0,1];
s202: converting the normalized sequence data into a polar coordinate system;
s203: calculating the angle difference between different points to identify the time correlation of the different points in time;
the angle difference formula is as follows:
in the above-mentioned method, the step of,for the ith and jth fault data x i And x j Angle value of (2); /> Representing time-series data before normalization processing; />Is time series data after standardized processing; i identity matrix, and a matrix with all 1's on the diagonal and other elements 0.
5. The prototype network-based fault diagnosis method for a HVDC system according to any one of claims 3 and 4, wherein said modeling the prototype network HVDC system fault diagnosis comprises:
s301: using encoders for each picture in the support set and the query set, respectivelyProcessing is carried out, and an encoding representation of each picture is learned;
s302: average value processing is carried out on the Embeddings coding representation under each category of the support set, and prototype representation of each category is obtained;
s303: similarity calculation is carried out on the Embedding code representation under each category of the support set and the prototype representation of each category of the support set;
s304: activating the similarity into probability distribution by adopting a softmax function to obtain a classification label of the query set picture;
s305: and performing cross entropy loss calculation on the classification labels and the true value labels of the query set pictures, and completing the training of an epoode through gradient back propagation.
6. The fault diagnosis method for a prototype network-based hvdc transmission system in accordance with claim 5, wherein said similarity calculation is formulated as:
in the method, in the process of the application,encoding representations for the eimbeddings under each category of the support set; c is a prototype representation of each class of the support set.
7. The method for diagnosing faults in a hvdc transmission system based on a prototype network according to claim 6, wherein the similarity is activated as a probability distribution using a softmax function, and the formula is:
8. a system of a prototype network-based fault diagnosis method for a hvdc transmission system according to any one of claims 1-7, said system comprising:
the acquisition unit is used for acquiring fault data of the high-voltage direct-current transmission system;
the first preprocessing unit is used for labeling the fault data;
the second preprocessing unit divides the fault data with the labels into a training set and a verification set;
the first processing unit is used for converting the one-dimensional signal data into a two-dimensional characteristic image through the treatment of a gram angle field and establishing a prototype network HVDC system fault diagnosis model;
the optimization processing unit inputs the training set to the fault diagnosis model of the prototype network HVDC system to perform model training; inputting the verification set into a trained prototype network HVDC system fault diagnosis model to perform qualification verification;
and the second processing unit is used for carrying out fault diagnosis on the high-voltage direct-current transmission system according to the optimized fault diagnosis model of the prototype network HVDC system.
9. A model for HVDC system fault diagnosis, characterized in that it employs a prototype network HVDC system fault diagnosis model in a prototype network-based method for HVDC system fault diagnosis according to any one of claims 1-7.
10. A computer storage medium, characterized in that the program, when executed by a processor, implements a prototype network-based fault diagnosis method for a hvdc transmission system according to any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310928195.5A CN116933179A (en) | 2023-07-26 | 2023-07-26 | High-voltage direct-current transmission system fault diagnosis method and system based on prototype network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310928195.5A CN116933179A (en) | 2023-07-26 | 2023-07-26 | High-voltage direct-current transmission system fault diagnosis method and system based on prototype network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116933179A true CN116933179A (en) | 2023-10-24 |
Family
ID=88387585
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310928195.5A Pending CN116933179A (en) | 2023-07-26 | 2023-07-26 | High-voltage direct-current transmission system fault diagnosis method and system based on prototype network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116933179A (en) |
-
2023
- 2023-07-26 CN CN202310928195.5A patent/CN116933179A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Trakas et al. | Spatial risk analysis of power systems resilience during extreme events | |
| Kang et al. | Big data analytics in China's electric power industry: modern information, communication technologies, and millions of smart meters | |
| CN108376982B (en) | Load phase sequence identification method and device | |
| CN110726898B (en) | Power distribution network fault type identification method | |
| CN113992153B (en) | Visual real-time monitoring distributed management system of photovoltaic power station | |
| CN113011481A (en) | Electric energy meter function abnormity evaluation method and system based on decision tree algorithm | |
| CN115905959B (en) | Defect factor-based power circuit breaker relevance fault analysis method and device | |
| CN113987294A (en) | CVT (continuously variable transmission) online fault diagnosis method based on genetic optimization GRU (generalized regression Unit) neural network | |
| CN115453267A (en) | Fault diagnosis system for electric power information system | |
| CN110968703B (en) | Method and system for constructing abnormal metering point knowledge base based on LSTM end-to-end extraction algorithm | |
| CN113937764A (en) | Low-voltage distribution network high-frequency measurement data processing and topology identification method | |
| Zhang et al. | Encoding time series as images: A robust and transferable framework for power system DIM identification combining rules and VGGNet | |
| Li et al. | Incipient Fault Detection in Power Distribution System: A Time–Frequency Embedded Deep-Learning-Based Approach | |
| CN112083275A (en) | Distribution network fault type identification method and system | |
| CN116933179A (en) | High-voltage direct-current transmission system fault diagnosis method and system based on prototype network | |
| CN111060755A (en) | Electromagnetic interference diagnosis method and device | |
| CN115908082A (en) | Enterprise pollution discharge monitoring method and device based on electricity utilization characteristic indexes | |
| Domínguez et al. | Analysis of electric power consumption using Self-Organizing Maps. | |
| CN115542070A (en) | Distribution network line fault positioning method and storage medium | |
| CN110245617B (en) | Artificial intelligence analysis method based on transient recording waveform | |
| Zhang et al. | Cyber-physical-social systems for wind power operation and maintenance | |
| CN117639107B (en) | Power quality assessment method and system for power distribution network | |
| Wang et al. | Fault Diagnosis of Wind Turbine Generator with Stacked Noise Reduction Autoencoder Based on Group Normalization | |
| Guo et al. | Graph Mining for Classifying and Localizing Solar Panels in Distribution Grids | |
| Wu et al. | A deep recognition network of capacitor voltage transformer based on dilated convolution and Bi-LSTM |
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 |