CN111881781B - Transformer winding deformation classification method based on scanning impedance method and support vector machine - Google Patents

Transformer winding deformation classification method based on scanning impedance method and support vector machine Download PDF

Info

Publication number
CN111881781B
CN111881781B CN202010657569.0A CN202010657569A CN111881781B CN 111881781 B CN111881781 B CN 111881781B CN 202010657569 A CN202010657569 A CN 202010657569A CN 111881781 B CN111881781 B CN 111881781B
Authority
CN
China
Prior art keywords
deformation
frequency
impedance
winding
amplitude
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202010657569.0A
Other languages
Chinese (zh)
Other versions
CN111881781A (en
Inventor
李振华
张阳坡
蒋伟辉
张宇杰
黄悦华
李振兴
徐艳春
邾玢鑫
杨楠
刘颂凯
张磊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Three Gorges University CTGU
Original Assignee
China Three Gorges University CTGU
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China Three Gorges University CTGU filed Critical China Three Gorges University CTGU
Priority to CN202010657569.0A priority Critical patent/CN111881781B/en
Publication of CN111881781A publication Critical patent/CN111881781A/en
Application granted granted Critical
Publication of CN111881781B publication Critical patent/CN111881781B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
    • G01B7/00—Measuring arrangements characterised by the use of electric or magnetic techniques
    • G01B7/16—Measuring arrangements characterised by the use of electric or magnetic techniques for measuring the deformation in a solid, e.g. by resistance strain gauge
    • G01B7/18—Measuring arrangements characterised by the use of electric or magnetic techniques for measuring the deformation in a solid, e.g. by resistance strain gauge using change in resistance
    • 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
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Signal Processing (AREA)
  • Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)

Abstract

Based on a scanning impedance method and a transformer winding deformation classification method of a support vector machine, short-circuit impedance is obtained through testing and calculation in a wide frequency domain range under a normal operation condition. Constructing a scanning impedance amplitude-frequency curve and a phase-frequency curve of a normal transformer as fingerprint tracks according to the obtained amplitude and phase data of the short-circuit impedance; when the transformer has a winding deformation fault, constructing a scanning impedance amplitude-frequency curve and a phase-frequency curve in the state by adopting the same method under the same test condition; determining a characteristic vector based on statistical indexes and scanning impedance waveform characteristics under the condition of a frequency sweeping impedance complex number, and constructing a characteristic vector library according to obtained characteristic curves of different winding deformations; according to the feature vectors extracted from the winding deformation features of different types, the winding deformation is classified by using a support vector machine optimized by adopting a particle swarm algorithm, the deformation type of the winding is finally determined according to the classification result, and a new way is provided for detecting the micro deformation of the transformer winding.

Description

Transformer winding deformation classification method based on scanning impedance method and support vector machine
Technical Field
The invention relates to the field of detection of winding states of power transformers, in particular to a transformer winding deformation classification method based on a scanning impedance method and a support vector machine.
Background
The transformer is used as key equipment of a power transmission line, plays a role in converting voltage to ensure better transmission of electric energy, and can cause great impact on a local or even an integral power system when the transformer fails in operation. The transformer is deformed due to human misoperation, external force, other natural events and other nonresistible factors during transportation, installation and operation, so that the state detection of the transformer is very important in the power department, and especially the prevention and detection of gradual faults such as winding deformation and the like are particularly important. According to statistics, about 60% of transformer faults are winding deformation, the normal operation of the transformer cannot be influenced by slight deformation of the winding, but the gradual accumulation of the winding deformation and the slight deformation can cause more serious deformation and even damage of the transformer, so that not only can great economic loss be caused, but also the stability of a power system can be influenced. It is necessary to accurately and effectively detect the micro-deformation of the transformer winding.
The Sweep Frequency Impedance (SFI) method is a detection method with good effect by combining two practical application methods according to the defects of various current transformer state detection methods in practical application by Liuyong, Yanfan and the like of the Sian university of traffic. A Short Circuit Impedance (SCI) method and a Frequency Response Analysis (FRA) method are a novel nondestructive testing method for detecting deformation of transformer windings, which has been proposed in recent years. However, currently, research based on the scanning impedance method still remains on the detection method itself, and there is less research on extracting differentiated feature vectors from data obtained by the scanning impedance method to classify faults. Therefore, the characteristic vectors which are comprehensive, reliable, new and high in information and can better explain the scanning impedance data are extracted, a proper intelligent classification method is provided according to the characteristic vectors, and the completion of the typical fault classification of the transformer provides strong support for the offline state line detection at the present stage and even the online monitoring of the transformer in the future.
Disclosure of Invention
The invention provides a transformer winding deformation classification method based on a scanning impedance method and a support vector machine, which is simple in test procedure and good in repeatability; and the signal-to-noise ratio is higher, the diagnostic information is rich, and the sensitivity is better. Particularly, the transformer can respond to the tiny deformation of the winding in time, is convenient for the maintainers to rapidly process, and provides a new method for the electric power practitioners to detect the state of the transformer.
The technical scheme adopted by the invention is as follows:
the transformer winding deformation classification method based on the scanning impedance method and the support vector machine comprises the following steps:
step 1: under the normal condition of the transformer operation, the head end and the tail end of one side of the transformer winding are in short circuit, and a sine voltage frequency sweeping signal is injected into the other side of the transformer winding
Figure BDA0002577299290000021
After the signals are amplified, the head end and the tail end of the transformer winding are connected with a sampling resistor R to obtain excitation signals of the head end of the transformer winding
Figure BDA0002577299290000022
End response signal
Figure BDA0002577299290000023
And short-circuit current
Figure BDA0002577299290000024
The short circuit impedance Z is obtained by calculationk(jω)。
Step 2: short-circuit impedance Z obtained according to step 1kAmplitude | Z of (j ω)kAnd (j omega) I and the phase omega test data, and constructing a scanning impedance amplitude-frequency curve and a phase-frequency curve of the normal transformer as fingerprint tracks.
And step 3: when the transformer has slight deformation fault of the winding, under the same test condition, the same method is adopted to measure excitation signals of the head ends of the transformer winding under different deformation conditions
Figure BDA0002577299290000025
End response signal
Figure BDA0002577299290000026
And short-circuit current
Figure BDA0002577299290000027
The short circuit impedance Z is obtained by calculationkf(jω)。
And 4, step 4: short circuit impedance Z obtained according to step 3kfAmplitude | Z of (j ω)kf(j ω) | and phase ωfThe method comprises the steps of constructing scanning impedance amplitude-frequency curves and phase-frequency curve tracks of the transformer in different deformation states and constructing a winding micro-deformation characteristic curve library.
And 5: comparing and analyzing the scanning impedance amplitude-frequency curve and the phase-frequency curve track constructed in the step (4) with the fingerprint track constructed in the step (2); and under the condition of complex frequency sweep impedance, determining a frequency sweep impedance real part correlation coefficient CC1, a frequency sweep impedance imaginary part correlation coefficient CC2, a resonance point vector Euclidean distance VD and an amplitude-frequency resonance point weighting frequency function value W based on statistical indexes and scanning impedance waveform characteristicsf1Phase frequency zero weighted frequency function value Wf2Forming a characteristic vector by the amplitude deviation of the short-circuit impedance at the power frequency and the phase deviation of the short-circuit impedance at the power frequency; and constructing a winding micro-deformation characteristic vector library according to different winding deformation curves.
Step 6: and (5) classifying the winding deformation by using a support vector machine optimized by a particle swarm algorithm according to a characteristic vector library constructed by utilizing the trace characteristic difference of the scanning impedance amplitude-frequency curve and the phase-frequency curve of the slight deformation of different types of windings in the step 5, and finally determining the deformation type of the transformer winding according to the classification result.
The invention discloses a transformer winding deformation classification method based on a scanning impedance method and a support vector machine, which has the following technical effects:
1) the invention provides a method for classifying winding micro-deformation based on scanning impedance amplitude-frequency curve and phase-frequency curve data, which is characterized in that statistical indexes and waveform characteristics are fused with each other under the condition of scanning impedance complex numbers to establish a characteristic vector with obvious discrimination, then a support vector machine for solving the problems of small samples, nonlinearity and high-dimensional mode identification is utilized to classify the winding micro-deformation, and a particle swarm algorithm is utilized to optimize the vector machine, so that the accuracy of fault classification is further improved.
2) The method is based on the scanning impedance ZkAmplitude | Z of (j ω)kAnd (j omega) I and the phase omega test data to obtain a scanning impedance amplitude-frequency curve and simultaneously obtain a scanning impedance phase-frequency curve. By using the method, the amplitude-frequency and phase-frequency curves of the swept-frequency impedance of the transformer and the short-circuit impedance value can be obtained through one-time test, richer and more sensitive fault characteristic information is possessed, and errors caused by multiple wiring are effectively reduced. And finally determining the health state of the winding by carrying out winding deformation classification on the scanning impedance curve track.
3) According to the method, under the condition of complex frequency sweeping impedance, non-redundant derivative values with strong informativeness and obvious discrimination are determined as feature vectors on the basis of statistical indexes with high sensitivity for identifying the winding fault type and the waveform characteristics of the frequency sweeping impedance, and compared with feature parameters extracted under the condition of single amplitude frequency or single phase frequency, the method is more sensitive and more accurate in feature information, and is more beneficial to classification of winding micro-deformation.
4) The method adopts the support vector machine as a classification algorithm, and simultaneously adopts the particle swarm optimization to optimize the support vector machine, so that the success rate of classification of the micro deformation of the winding is higher.
5) The method is based on the scanning impedance complex eigenvalue and the support vector machine optimized by the particle swarm optimization to classify the winding micro-deformation, and a new way is provided for the detection of the transformer winding micro-deformation.
Drawings
Fig. 1 is a flow chart of classification of winding micro-deformation according to the present invention.
Fig. 2(a) is a trace diagram of the amplitude-frequency curve of the scanning impedance when the winding is normal according to the present invention;
fig. 2(b) is a trace diagram of the phase-frequency curve of the scanning impedance when the winding is normal.
Fig. 3 is a comparison graph of the amplitude-frequency curve of the scan impedance proposed by the present invention and the amplitude-frequency curve of the conventional frequency response.
FIG. 4(a) is a comparison graph of the trace of the amplitude-frequency curve of the scanned impedance when 3 typical fault conditions occur in the winding and the normal condition;
fig. 4(b) is a comparison graph of the trace of the amplitude-frequency curve of the scanned impedance when 3 typical fault conditions occur in the winding and the normal condition.
Fig. 5 is a graph comparing the amplitude-frequency correlation coefficient CC of the winding bump scanning impedance with the real part correlation coefficient CC1 and the imaginary part correlation coefficient CC2 of the scanning impedance under complex conditions.
FIG. 6 is a classification diagram of support vector machine sample results after optimization by particle swarm optimization.
Detailed Description
The transformer winding deformation classification method based on the scanning impedance method and the support vector machine comprises the following steps:
step 1: under the normal operation condition of the transformer, the head end and the tail end of one side of a transformer winding are in short circuit, and a sine voltage frequency sweeping signal is injected into the other side of the transformer winding
Figure BDA0002577299290000031
After the signals are amplified, the head end and the tail end of the transformer winding are connected with a sampling resistor R to obtain excitation signals of the head end of the transformer winding
Figure BDA0002577299290000032
End response signal
Figure BDA0002577299290000033
And short-circuit current
Figure BDA0002577299290000034
The short circuit impedance Z is obtained by calculationk(jω)。
In the step 1, the excitation signal is sent to the head end of the transformer winding
Figure BDA0002577299290000035
End response signal
Figure BDA0002577299290000036
Short circuit current
Figure BDA0002577299290000037
Short circuit impedance Zk(j ω) is a transformer windingRelated test data in the normal state of the group are acquired through a data acquisition card to obtain excitation signals of the head end of the transformer winding
Figure BDA0002577299290000038
End response signal
Figure BDA0002577299290000039
Short circuit current
Figure BDA00025772992900000310
From which the short-circuit impedance Z is obtained by mathematical calculationk(j ω). The calculation formula is as follows:
Figure BDA0002577299290000041
step 2: short-circuit impedance Z obtained according to step 1kAmplitude | Z of (j ω)kAnd (j omega) I and the phase omega test data, and constructing a scanning impedance amplitude-frequency curve and a phase-frequency curve of the normal transformer as fingerprint tracks.
In the step 2, an impedance amplitude-frequency curve and a phase-frequency curve are scanned, wherein the amplitude-frequency curve is a short-circuit impedance amplitude curve under the frequency of 10 Hz-1 MHz, and the phase-frequency curve is a short-circuit impedance phase value curve under the frequency of 10 Hz-1 MHz.
And step 3: when the transformer has slight deformation fault of the winding, under the same test condition, the same method is adopted to measure excitation signals of the head ends of the transformer winding under different deformation conditions
Figure BDA0002577299290000042
End response signal
Figure BDA0002577299290000043
And short-circuit current
Figure BDA0002577299290000044
The short circuit impedance Z is obtained by calculationkf(jω)。
In the step 3, under the same test conditions, the following steps are carried out:
step 3, the position and the size of the sinusoidal voltage frequency sweeping signal applied by the transformer are the same as those of the voltage signal applied in the step 1;
the size of the sampling resistor connected to the head end and the tail end of the transformer in the step 3 is completely the same as that of the sampling resistor connected to the head end and the tail end in the step 1;
the other test conditions affecting the test result in step 3 and step 1 are also identical. On-site test is carried out each time, attention is paid to avoid interference of electromagnetic signals, good contact of joints is guaranteed, and in addition, test conditions such as whether an iron core is grounded, residual magnetism of the iron core, a sleeve pipe state and a winding wiring mode are kept the same.
And 4, step 4: short circuit impedance Z obtained according to step 3kfAmplitude | Z of (j ω)kf(j ω) | and phase ωfThe method comprises the steps of testing data, constructing scanning impedance amplitude-frequency curves and phase-frequency curve tracks of the transformer in different deformation states, and constructing a winding micro-deformation characteristic curve library. And 5: comparing and analyzing the scanning impedance amplitude-frequency curve and the phase-frequency curve track constructed in the step (4) with the fingerprint track constructed in the step (2); and under the condition of complex frequency sweep impedance, determining a frequency sweep impedance real part correlation coefficient CC1, a frequency sweep impedance imaginary part correlation coefficient CC2, a resonance point vector Euclidean distance VD and an amplitude-frequency resonance point weighting frequency function value W based on statistical indexes and scanning impedance waveform characteristicsf1Phase frequency zero weighted frequency function value Wf2Forming a characteristic vector by the amplitude deviation of the short-circuit impedance at the power frequency and the phase deviation of the short-circuit impedance at the power frequency; and constructing a winding micro-deformation characteristic vector library according to different winding deformation curves.
In the step 5, the scanning impedance amplitude-frequency curve and the phase-frequency curve track constructed in the step 4 are compared and analyzed with the fingerprint track constructed in the step 2, and after comparison and analysis, the following results can be found: the overall curve trend, amplitude-frequency curve resonance point size and frequency position, phase-frequency curve zero frequency position and short circuit impedance amplitude and phase at power frequency of different winding deformation type curve tracks and fingerprint tracks are different, and the characteristic parameters of the characteristic vectors in the step 5 are characteristic representation of winding deformation conditions.
Step 6: and (5) classifying the winding deformation by using a support vector machine optimized by a particle swarm algorithm according to a characteristic vector library constructed by utilizing the trace characteristic difference of the scanning impedance amplitude-frequency curve and the phase-frequency curve of the slight deformation of different types of windings in the step 5, and finally determining the deformation type of the transformer winding according to the classification result.
In step 6, classifying the winding deformation types by using the feature vectors through a support vector machine, including:
winding deformation is the most common fault type of the transformer, and the common winding deformation types include winding short circuit and winding bulge. Axial bending, overall displacement, wire deflection, wire cake collapse, etc. The method comprises the steps of firstly adopting a support vector machine to classify, setting different types of winding deformation feature vector class labels as 1, 2, 3, … … and n in sequence before classification based on feature vector libraries of different types of winding deformation, setting the same type of winding deformation feature vectors as the same class labels, selecting partial data of different classes as a training set, and using the rest part as a test set. And after the intelligent algorithm finishes classification, checking a classification result graph of the test set to obtain a classification result. Because the discriminant function is required to be obtained by training a sample during first classification, partial data of different classes are required to be set as a training set, and the discriminant function is obtained by training during subsequent classification, so that classification can be directly carried out based on the discriminant function obtained by the first training without setting a new training set.
The specific first classification process is as follows: firstly, setting category labels for a plurality of groups of winding deformation feature vectors of different types extracted based on statistical indexes and wave characteristics, then loading a plurality of groups of feature vector data extracted based on the statistical indexes and the wave characteristics into matlab, selecting half of the data as a training set to carry out classification model training through an SVM, and selecting the remaining half as a test set to carry out classification label prediction. After sample data is loaded, in order to facilitate processing and accelerate convergence speed, the [ 01 ] interval normalization processing is carried out on the sample characteristic data, the radial basis function is used as a kernel function to solve classification decision function parameters of training data, a linear discriminant function is further obtained, then the prediction sample is substituted into the discriminant function to carry out classification effect validation, a class label of the sample is obtained through calculation, and winding deformation classification is completed. When training an SVM with a radial basis kernel, there are two parameters that must be considered: a penalty parameter c and a kernel function parameter g of svmtrain of the support vector machine. The parameter c corresponds the misclassification of the training samples to the simplicity of the decision surface, which determines the smoothness of the decision plane and also determines how much freedom is given to the model to select samples as support vectors to correctly classify all training samples. The parameter g defines the radius of influence of a single training sample and can be seen as the inverse of the radius of influence of the sample selected by the model support vector. The selection of the parameters c and g will directly affect the classification result of the support vector machine.
In order to improve the fitting precision and generalization capability of the support vector machine, solve the optimal classification surface and optimize the punishment parameter c and the kernel function parameter g of svmtrain, the invention adopts the particle swarm optimization algorithm with high convergence speed to optimize the parameters c and g conveniently. The algorithm is based on a random solution, an optimal solution is found through iteration, and in each iteration, the particles update themselves by tracking two extreme values. The first extreme value is the optimal solution found by the particles, the other extreme value is the optimal solution found by the whole population at present, the global optimum is found by following the optimal value found at present, and the parameters c and g are used for finding a balanced optimal point between the structural risk of the support vector machine and the sample error. And finally determining the deformation type of the winding by classifying the winding deformation result.
After the SVM classifies the winding deformation, different types of winding deformation can be classified into different categories, the same type of winding deformation can be classified into the same category, each category has a corresponding category label, and the deformation type of a specific winding can be determined by looking at the category label of the category.
Different types of winding deformations can be classified into different categories, in particular: because the winding deformation types are numerous, the windings of different types can be classified according to requirements, for example, the deformation winding bulges, the axial bending and the integral displacement of the windings of different types are classified, micro deformation points are respectively arranged at the head end and the tail end of the winding and the middle position of the winding during simulation, the deformation degree of each deformation point is from 3 percent to 25 percent, and 20 groups of deformation data of different deformation degrees are obtained at 3 deformation positions; micro deformation points are respectively arranged at the head end, the tail end and the middle part of the winding for the axial bending of the winding, the deformation degree of each deformation point is from 3 percent to 25 percent, and 20 groups of deformation data with different deformation degrees are obtained at 3 deformation positions; micro deformation points are respectively arranged at the head end, the tail end and the middle part of the winding for the integral displacement of the winding, the deformation degree of each deformation point is from 3% to 25%, and 20 groups of deformation data with different deformation degrees are obtained at 3 deformation positions; the method comprises the steps that 60 groups of deformation samples are obtained in total through 3 typical winding deformation, 60 groups of characteristic vectors are finally obtained, category labels are set for the characteristic vectors when the 60 groups of data are classified, three groups of 3 categories are required to be set due to the fact that three different types of winding deformation conditions are classified, 20 groups of winding bulge characteristic vector categories with different deformation degrees, which are obtained at 3 deformation positions, are all 1, and each characteristic vector of the 20 groups of winding bulge deformation has one category label which is 1; 20 sets of axial bending characteristic vector category labels with different deformation degrees, which are obtained from 3 deformation positions, are all 2, namely each characteristic vector of the 20 sets of winding bulge deformation has a category label of 2; 20 groups of integral displacement characteristic vector category labels with different deformation degrees, which are acquired at 3 deformation positions, are all 3, namely each characteristic vector of the 20 groups of winding bulge deformations has a category label of 3.
The same type of winding deformation can be assigned to the same category, specifically: the winding deformation of the same type in different degrees which occurs in different positions is the same type, for example, micro deformation points are respectively arranged at the head end and the middle position of the winding on the winding bulge in simulation, the deformation degree of each deformation point is from 3% to 25%, 20 groups of deformation data with different deformation degrees are obtained at 3 deformation positions, the 20 groups of data are winding deformation of the same type, namely the winding bulge, the characteristic vector of the winding deformation of the same type is set to be the same type label, namely, each characteristic vector of the deformation of the 20 groups of winding bulge has a type label of 1.
Each category has a corresponding category label, specifically: before classification, winding deformation feature vector class labels of different types are sequentially set to be 1, 2, 3, … … and n, and winding deformation feature vectors of the same type are set to be the same class labels. The category labels are the symbol codes of the same category, and the logic of classification is to classify the windings of the same type into one category, namely to classify the feature vectors of the same category labels into one category.
The type of deformation specifically refers to: winding deformation is the most common fault type of the transformer, and the common winding deformation types include winding short circuit and winding bulge. Axial bending, overall displacement, lead deflection, coil collapse, and the like.
In order to further carry out qualitative analysis on the classification accuracy of winding micro-deformation based on a scanning impedance method and a support vector machine, a simulation model of a transformer with a certain model is established in circuit simulation software Pspice. The method simulates the slight deformation conditions of different degrees and different types of the windings at different positions of the normal windings and the windings, and compares the scanning impedance curve with the frequency response amplitude-frequency curve, thereby proving that the characteristics of the scanning impedance curve are richer and more sensitive.
Fig. 2(a) is a trace diagram of the amplitude-frequency curve of the scanning impedance when the winding is normal. Fig. 2(b) is a trace diagram of the phase-frequency curve of the scanning impedance when the winding is normal. Compared with the traditional method of taking an amplitude-frequency curve as the basis of judging the winding fault, the method simultaneously obtains the amplitude-frequency curve and the phase-frequency curve of the scanning impedance through the graph 2(a) and the graph 2(b), and the characteristic information is richer.
Fig. 3 is a comparison graph of the amplitude-frequency curve of the scan impedance proposed by the present invention and the amplitude-frequency curve of the conventional frequency response. As can be seen from fig. 3, compared with SCI, SFI can obtain not only the winding short-circuit impedance when the sweep frequency is 50Hz, but also richer status feature information of amplitude frequency and phase frequency in the wide frequency domain; in contrast to FRA, it can be theoretically demonstrated that the swept impedance magnitude-frequency curve is approximately symmetric about 10lg50 with the frequency response magnitude-frequency curve. After the correlation coefficients of the two are compared, for the same micro winding deformation, the sweep frequency impedance amplitude-frequency curve has smaller correlation coefficient, more obvious change and stronger sensitivity than the frequency response amplitude-frequency curve. Therefore, the method is more beneficial to judging the state and deformation type of the winding, and is particularly beneficial to detecting micro deformation.
Fig. 4(a) and 4(b) are a comparison graph of the amplitude-frequency curve trace of the scanned impedance and a comparison graph of the phase-frequency curve trace of the scanned impedance in the case of 3 typical fault conditions and normal conditions of the winding. Comparing each amplitude-frequency curve track of fig. 4(a), it can be seen that the overall trend of each fault curve and the normal curve, the magnitude and frequency position of the resonance point of the amplitude-frequency curve, and the short-circuit impedance amplitude at power frequency are changed, and the index change directions and amplitudes of each fault curve track are different; comparing the amplitude-frequency curve traces of fig. 4(b), it can be seen that the overall trend of the various fault curves and the normal curve, the phase frequency zero position, and the short-circuit impedance phase at power frequency are changed, and the direction and amplitude of the index change of the various fault curve traces are also different.
Fig. 5 is a graph comparing the amplitude-frequency correlation coefficient CC of the winding bump scanning impedance with the real part correlation coefficient CC1 and the imaginary part correlation coefficient CC2 of the scanning impedance under complex conditions. The traditional method for extracting the characteristics by only utilizing amplitude-frequency or phase-frequency image data is not comprehensive enough, and the winding state can be more truly described only by simultaneously considering the amplitude-frequency and phase-frequency characteristics. Fig. 5 compares CC, CC1, and CC2 under the condition of winding bulge, and the differences among the three indexes are shown in the form of reciprocal, and it can be seen from fig. 5 that the improved statistical indexes under complex numbers not only can accurately express the characteristic information of winding deformation, but also have higher sensitivity than the original data indexes.
FIG. 6 is a classification diagram of the support vector machine test set results after optimization by particle swarm optimization. The classification results of fig. 6 are calculated to obtain that the classification success rates of the winding micro-deformation classification method provided by the invention for the set 3 winding deformation test samples are respectively 100%, 100% and 90.00%, and the overall classification rate of 3 types of slight deformation reaches 96.667%. Therefore, each fault classification effect and the whole classification effect are ideal, and the advantages of the method in the field of monitoring of the micro-deformation state of the transformer winding are verified.
Through the realization of the steps one by one, the classification of the micro deformation of the transformer winding can be realized, and finally the deformation type of the winding is determined according to the classification result.
Meanwhile, the fault track characteristics of the scanning impedance amplitude-frequency curve and the scanning impedance phase-frequency curve become larger along with the increase of the deformation severity of the winding, and the classification success rate is higher. Therefore, if winding deformation fault test is carried out on the same batch and model of transformer windings in advance before delivery, a characteristic database of the deformation scanning impedance amplitude-frequency curve and phase-frequency curve track of the model of transformer windings is constructed, and when the winding deformation fault occurs to the running transformer windings, the winding deformation can be classified according to the method and the deformation type of the winding is finally determined.

Claims (6)

1. The transformer winding deformation classification method based on the scanning impedance method and the support vector machine is characterized by comprising the following steps of:
step 1: under the normal operation condition of the transformer, the head end and the tail end of one side of a transformer winding are in short circuit, and a sine voltage frequency sweeping signal is injected into the other side of the transformer winding
Figure FDA0002577299280000011
After the signals are amplified, the head end and the tail end of the transformer winding are connected with a sampling resistor R to obtain excitation signals of the head end of the transformer winding
Figure FDA0002577299280000012
End response signal
Figure FDA0002577299280000013
And short-circuit current
Figure FDA0002577299280000014
The short circuit impedance Z is obtained by calculationk(jω)；
Step 2: short-circuit impedance Z obtained according to step 1kAmplitude | Z of (j ω)kTest data of (j omega) I and phase omega to construct normal transformerScanning an impedance amplitude-frequency curve and a phase-frequency curve to be used as fingerprint tracks;
and step 3: when the transformer has slight deformation fault of the winding, under the same test condition, the same method is adopted to measure excitation signals of the head ends of the transformer winding under different deformation conditions
Figure FDA0002577299280000015
End response signal
Figure FDA0002577299280000016
And short-circuit current
Figure FDA0002577299280000017
The short circuit impedance Z is obtained by calculationkf(jω)；
And 4, step 4: short circuit impedance Z obtained according to step 3kfAmplitude | Z of (j ω)kf(j ω) | and phase ωfConstructing a scanning impedance amplitude-frequency curve and a phase-frequency curve track under different deformation states and constructing a winding micro-deformation characteristic curve library;
and 5: comparing and analyzing the scanning impedance amplitude-frequency curve and the phase-frequency curve track constructed in the step (4) with the fingerprint track constructed in the step (2); and under the condition of complex frequency sweep impedance, determining a frequency sweep impedance real part correlation coefficient CC1, a frequency sweep impedance imaginary part correlation coefficient CC2, a resonance point vector Euclidean distance VD and an amplitude-frequency resonance point weighting frequency function value W based on statistical indexes and scanning impedance waveform characteristicsf1Phase frequency zero weighted frequency function value Wf2Forming a characteristic vector by the amplitude deviation of the short-circuit impedance at the power frequency and the phase deviation of the short-circuit impedance at the power frequency; constructing a winding micro-deformation characteristic vector library according to different winding deformation curves;
step 6: and (5) classifying the winding deformation by using a support vector machine optimized by a particle swarm algorithm according to a characteristic vector library constructed by utilizing the trace characteristic difference of the scanning impedance amplitude-frequency curve and the phase-frequency curve of the slight deformation of different types of windings in the step 5, and finally determining the deformation type of the transformer winding according to the classification result.
2. The transformer winding deformation classification method based on the scanning impedance method and the support vector machine according to claim 1, is characterized in that: in the step 1, the excitation signal is sent to the head end of the transformer winding
Figure FDA0002577299280000018
End response signal
Figure FDA0002577299280000019
Short circuit current
Figure FDA00025772992800000110
Short circuit impedance Zk(j omega) for related test data of the transformer winding in the normal state, acquiring excitation signals of the head end of the transformer winding through a data acquisition card
Figure FDA00025772992800000111
End response signal
Figure FDA00025772992800000112
Short circuit current
Figure FDA00025772992800000113
From which the short-circuit impedance Z is obtained by mathematical calculationk(jω)。
3. The transformer winding deformation classification method based on the scanning impedance method and the support vector machine according to claim 1, is characterized in that: in the step 2, an impedance amplitude-frequency curve and a phase-frequency curve are scanned, wherein the amplitude-frequency curve is a short-circuit impedance amplitude curve under the frequency of 10 Hz-1 MHz, and the phase-frequency curve is a short-circuit impedance phase value curve under the frequency of 10 Hz-1 MHz.
4. The transformer winding deformation classification method based on the scanning impedance method and the support vector machine according to claim 1, is characterized in that: in the step 3, under the same test conditions, the following means:
step 3, the position and the size of the sinusoidal voltage frequency sweeping signal applied by the transformer are the same as those of the voltage signal applied in the step 1;
the size of the sampling resistor connected to the head end and the tail end of the transformer in the step 3 is completely the same as that of the sampling resistor connected to the head end and the tail end in the step 1;
the other test conditions affecting the test result in step 3 and step 1 are also identical.
5. The transformer winding deformation classification method based on the scanning impedance method and the support vector machine according to claim 1, is characterized in that: in the step 5, the scanning impedance amplitude-frequency curve and the phase-frequency curve track constructed in the step 4 are compared and analyzed with the fingerprint track constructed in the step 2, and after comparison and analysis, the following results can be found: the overall curve trend, amplitude-frequency curve resonance point size and frequency position, phase-frequency curve zero frequency position and short circuit impedance amplitude and phase at power frequency of different winding deformation type curve tracks and fingerprint tracks are different, and the characteristic parameters of the characteristic vectors in the step 5 are characteristic representation of winding deformation conditions.
6. The transformer winding deformation classification method based on the scanning impedance method and the support vector machine according to claim 1, characterized in that: in the step 6, the winding deformation types are classified by using the characteristic vectors through the support vector machine, and the punishment parameter c and the kernel function parameter g of svmtrain of the support vector machine influence the classification result of the support vector machine; the particle swarm algorithm is used for optimizing the classification, so that the classification accuracy is improved; and finally determining the deformation type of the winding by classifying the winding deformation result.
CN202010657569.0A 2020-07-09 2020-07-09 Transformer winding deformation classification method based on scanning impedance method and support vector machine Active CN111881781B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010657569.0A CN111881781B (en) 2020-07-09 2020-07-09 Transformer winding deformation classification method based on scanning impedance method and support vector machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010657569.0A CN111881781B (en) 2020-07-09 2020-07-09 Transformer winding deformation classification method based on scanning impedance method and support vector machine

Publications (2)

Publication Number Publication Date
CN111881781A CN111881781A (en) 2020-11-03
CN111881781B true CN111881781B (en) 2022-05-06

Family

ID=73151484

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010657569.0A Active CN111881781B (en) 2020-07-09 2020-07-09 Transformer winding deformation classification method based on scanning impedance method and support vector machine

Country Status (1)

Country Link
CN (1) CN111881781B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113238165A (en) * 2021-02-09 2021-08-10 重庆大学 Reactor turn-to-turn short circuit fault detection method based on IFRA and SVM

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010060253A1 (en) * 2008-11-25 2010-06-03 上海市电力公司 The system and method for detecting the state of the transformer winding by utilizing excitation of the constant current sweep frequency power source
CN111273199A (en) * 2018-12-05 2020-06-12 国网四川省电力公司眉山供电公司 Intelligent detection method for transformer winding deformation based on sweep frequency impedance curve identification

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2010060253A1 (en) * 2008-11-25 2010-06-03 上海市电力公司 The system and method for detecting the state of the transformer winding by utilizing excitation of the constant current sweep frequency power source
CN111273199A (en) * 2018-12-05 2020-06-12 国网四川省电力公司眉山供电公司 Intelligent detection method for transformer winding deformation based on sweep frequency impedance curve identification

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
变压器绕组变形的扫频短路阻抗法研究;薛亮红等;《国外电子测量技术》;20170715(第07期);全文 *
基于粒子群算法优化支持向量机的变压器绕组变形分类方法;刘伟等;《高压电器》;20200316(第03期);全文 *

Also Published As

Publication number Publication date
CN111881781A (en) 2020-11-03

Similar Documents

Publication Publication Date Title
Liu et al. Classifying transformer winding deformation fault types and degrees using FRA based on support vector machine
Tarimoradi et al. Novel calculation method of indices to improve classification of transformer winding fault type, location, and extent
Chen et al. Intelligent early structural health prognosis with nonlinear system identification for RFID signal analysis
CN106646096B (en) Transformer fault classification and recognition methods based on vibration analysis method
CN110006645B (en) Multi-source fusion high-voltage circuit breaker mechanical fault diagnosis method
CN110376522B (en) Motor fault diagnosis method of data fusion deep learning network
CN111914883B (en) Spindle bearing state evaluation method and device based on deep fusion network
CN110792563A (en) Wind turbine generator blade fault audio monitoring method based on convolution generation countermeasure network
CN111398798B (en) Circuit breaker energy storage state identification method based on vibration signal interval feature extraction
CN111881781B (en) Transformer winding deformation classification method based on scanning impedance method and support vector machine
CN115758200A (en) Vibration signal fault identification method and system based on similarity measurement
Mao et al. Transformer winding type recognition based on FRA data and a support vector machine model
CN112149569B (en) Voiceprint fault diagnosis method of transformer based on fuzzy C-means clustering algorithm
CN114814501B (en) On-line diagnosis method for capacitor breakdown fault of capacitor voltage transformer
CN109991508A (en) A kind of transformer winding state diagnostic method based on kinematic nonlinearity characteristic sequence
CN115905832A (en) Transformer fault diagnosis method based on vibration signal clustering identification
CN113805018A (en) Intelligent identification method for partial discharge fault type of 10kV cable of power distribution network
CN117630797A (en) Ammeter health state detection method, system and storage medium based on working current
CN113052249B (en) Transformer winding fault type identification method based on support vector machine and current deviation coefficient
Ridi et al. Unseen appliances identification
CN115577249A (en) Transformer acoustic signal identification method, system and medium with multi-view feature fusion
CN113255771B (en) Fault diagnosis method and system based on multi-dimensional heterogeneous difference analysis
US11867763B2 (en) Method for monitoring circuit breaker and apparatus and internet of things using the same
CN115291156A (en) Online detection system and detection method for error characteristics of voltage transformer
Li et al. Classification and location of transformer winding deformations using genetic algorithm and support vector machine

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
EE01 Entry into force of recordation of patent licensing contract
EE01 Entry into force of recordation of patent licensing contract

Application publication date: 20201103

Assignee: Yichang Yizhixing Technology Co.,Ltd.

Assignor: CHINA THREE GORGES University

Contract record no.: X2023980034895

Denomination of invention: Transformer winding deformation classification method based on scanning impedance method and support vector machine

Granted publication date: 20220506

License type: Common License

Record date: 20230426