A current transformer error correction method and device based on a classification SVM

Abstract

The application discloses a current transformer error correction method and device based on classification SVM, obtains the secondary side current output by the current transformer, generates the secondary side current waveform corresponding to the secondary side current, calculates the waveform asymmetry coefficient based on the secondary side current waveform, judges the specific fault type of the current transformer according to the waveform asymmetry coefficient, calculates the harmonic content of the secondary side current when the fault type is determined as the primary side current existing direct current bias, inputs the harmonic content and the waveform asymmetry coefficient into the pre-trained SVM classification model, so that the SVM classification model outputs the first current transformer saturation value, selects the first RBF model corresponding to the first current transformer saturation value, so that the first RBF model corrects the secondary side current waveform and outputs the secondary side current correction waveform, compared with the prior art, the technical scheme of the application can improve the accuracy of the secondary side current waveform correction.

Description

Technical Field

[0001] This invention relates to the technical field of power systems, and in particular to a method and apparatus for error correction of current transformers based on classification SVM. Background Technology

[0002] Current transformers are the basic measurement units in power systems, and their measurement accuracy is crucial for ensuring the fairness of electricity price settlement between the power grid and users. Due to the nonlinear excitation characteristics of current transformers, when the primary current in a line is too large, the secondary current will exhibit varying degrees of distortion and waveform loss. Furthermore, with the large-scale integration of new energy sources into the grid, the DC component content in the line will significantly increase. When the bias flux generated by the DC component is superimposed on the primary sinusoidal fundamental component, it is easier for the current transformer to saturate. At this time, the transformer cannot linearly transmit the primary current, and the secondary current cannot linearly reflect the change in the primary current, resulting in a large error in the measurement results.

[0003] To address this issue, there are currently two main solutions. The first solution is to propose a new design concept for current transformers after analyzing the performance of various current transformers and ferromagnetic materials. The second solution is to use a nonlinear correction method to perform real-time correction on the secondary output of the current transformer to improve the measurement accuracy of the transformer.

[0004] Scholars both domestically and internationally have proposed many methods for saturation compensation of current transformers, which can be divided into three categories according to different principles: distortion current compensation method that takes into account excitation current, distortion current saturation compensation method based on regression, and nonlinear distortion current compensation method based on artificial intelligence algorithm. However, the above three mainstream technical solutions all have certain defects and cannot be widely applied in practical systems. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a current transformer error correction method and device based on classification SVM, which can improve the accuracy of secondary side current waveform correction.

[0006] To address the aforementioned technical problems, this invention provides a current transformer error correction method based on classification SVM, comprising:

[0007] Obtain the secondary current output by the current transformer, generate the secondary current waveform corresponding to the secondary current, and calculate the waveform asymmetry coefficient based on the secondary current waveform.

[0008] Based on the waveform asymmetry coefficient, the fault type of the current transformer is determined, wherein the fault type includes DC bias in the primary side current and excessive amplitude of the primary side current.

[0009] When the fault type is determined to be DC bias in the primary current, the harmonic content of the secondary current is calculated, and the harmonic content and the waveform asymmetry coefficient are input into the pre-trained SVM classification model so that the SVM classification model outputs the first current transformer saturation value.

[0010] A first RBF model corresponding to the saturation value of the first current transformer is selected so that the first RBF model can correct the secondary current waveform and output the secondary current correction waveform.

[0011] In one possible implementation, the waveform asymmetry coefficient is calculated based on the secondary-side current waveform, specifically including:

[0012] Multiple periodic waveforms in the secondary current waveform are sampled to obtain multiple positive half-wave waveform sampling points and multiple negative half-wave waveform sampling points corresponding to each periodic waveform.

[0013] Based on the preset area calculation formula, the first area enclosed by every three positive half-wave waveform sampling points and the time axis in the plurality of positive half-wave waveform sampling points is calculated to obtain a plurality of first areas;

[0014] And calculate the second area enclosed by every three negative half-wave waveform sampling points and the time axis in the plurality of negative half-wave waveform sampling points to obtain a plurality of second areas;

[0015] The maximum value among the plurality of first areas is selected to obtain the first maximum area; the maximum value among the plurality of second areas is selected to obtain the second maximum area.

[0016] Input the first maximum area value and the second maximum area value into the waveform asymmetry coefficient calculation formula to obtain the waveform asymmetry coefficient corresponding to each period waveform;

[0017] The preset area calculation formula is as follows:

[0018]

[0019] In the formula, S is the area, n is the waveform sampling point, i2 is the secondary current, and Δt is the sampling interval time;

[0020] The formula for calculating the waveform asymmetry coefficient is as follows:

[0021]

[0022] In the formula, σ is the waveform asymmetry coefficient, and S + S is the first maximum area. ― This is the second largest area.

[0023] In one possible implementation, the fault type of the current transformer is determined based on the waveform asymmetry coefficient, specifically including:

[0024] Obtain the waveform asymmetry coefficient corresponding to a preset number of cycles in the secondary current waveform. When it is determined that the waveform asymmetry coefficient remains constant, the fault type of the current transformer is determined to be that the primary current amplitude is too large.

[0025] When it is determined that the waveform asymmetry coefficient gradually increases, the fault type of the current transformer is judged to be that there is DC bias in the primary side current.

[0026] In one possible implementation, calculating the harmonic content of the secondary current specifically includes:

[0027] The secondary side current is subjected to Fourier transform processing to obtain the spectrum function of the secondary side current, and the harmonic amplitude of the secondary side current is obtained based on the spectrum function.

[0028] Substitute the harmonic amplitude into the preset harmonic content calculation formula to calculate the harmonic content of the secondary current and the harmonic content of the preset order.

[0029] The preset formula for calculating harmonic content is as follows:

[0030]

[0031]

[0032] In the formula, THD represents the harmonic content. i For the harmonic content of a preset number, A n Let A be the amplitude of the nth harmonic. i Let be the amplitude of the i-th harmonic.

[0033] In one possible implementation, the training process of the SVM classification model specifically includes:

[0034] Obtain the distorted secondary current waveforms corresponding to different distorted secondary currents, and calculate the distortion waveform asymmetry coefficient and harmonic components corresponding to the distorted secondary current waveforms.

[0035] Using the asymmetry coefficient of the distorted waveform and the harmonic components as inputs, and the current transformer saturation value of the current transformer as output, the initial SVM classification model is trained until the misclassification rate of the initial SVM classification model is less than the misclassification rate limit, thus generating an SVM classification model. The current transformer saturation value includes 0.2-level saturation value and 0.5-level saturation value.

[0036] In one possible implementation, before selecting the first RBF model corresponding to the saturation value of the first current transformer, the method further includes:

[0037] Establish an RBF model based on the saturation value of the current transformer;

[0038] When the saturation value of the current transformer is 0.2, a 0.2-level RBF model is established; when the saturation value of the current transformer is 0.5, a 0.5-level RBF model is established.

[0039] The present invention also provides a current transformer error correction device based on classification SVM, comprising: a waveform asymmetry coefficient calculation module, a fault type judgment module, a current transformer saturation value determination module, and a secondary current waveform correction module.

[0040] The waveform asymmetry coefficient calculation module is used to obtain the secondary current output by the current transformer, generate the secondary current waveform corresponding to the secondary current, and calculate the waveform asymmetry coefficient based on the secondary current waveform.

[0041] The fault type determination module is used to determine the fault type of the current transformer based on the waveform asymmetry coefficient, wherein the fault type includes DC bias in the primary side current and excessive amplitude of the primary side current.

[0042] The current transformer saturation value determination module is used to calculate the harmonic content of the secondary current when the fault type is determined to be DC bias in the primary current, and input the harmonic content and the waveform asymmetry coefficient into the pre-trained SVM classification model so that the SVM classification model outputs the first current transformer saturation value.

[0043] The secondary current waveform correction module is used to select the first RBF model corresponding to the saturation value of the first current transformer, so that the first RBF model corrects the secondary current waveform and outputs the secondary current correction waveform.

[0044] In one possible implementation, the fault type determination module is used to calculate a waveform asymmetry coefficient based on the secondary side current waveform, specifically including:

[0045] Multiple periodic waveforms in the secondary current waveform are sampled to obtain multiple positive half-wave waveform sampling points and multiple negative half-wave waveform sampling points corresponding to each periodic waveform.

[0046] Based on the preset area calculation formula, the first area enclosed by every three positive half-wave waveform sampling points and the time axis in the plurality of positive half-wave waveform sampling points is calculated to obtain a plurality of first areas;

[0047] And calculate the second area enclosed by every three negative half-wave waveform sampling points and the time axis in the plurality of negative half-wave waveform sampling points to obtain a plurality of second areas;

[0048] The maximum value among the plurality of first areas is selected to obtain the first maximum area; the maximum value among the plurality of second areas is selected to obtain the second maximum area.

[0049] Input the first maximum area value and the second maximum area value into the waveform asymmetry coefficient calculation formula to obtain the waveform asymmetry coefficient corresponding to each period waveform;

[0050] The preset area calculation formula is as follows:

[0051]

[0052] In the formula, S is the area, n is the waveform sampling point, i2 is the secondary current, and Δt is the sampling interval time;

[0053] The formula for calculating the waveform asymmetry coefficient is as follows:

[0054]

[0055] In the formula, σ is the waveform asymmetry coefficient, and S + S is the first maximum area. ― This is the second largest area.

[0056] In one possible implementation, the fault type determination module is used to determine the fault type of the current transformer based on the waveform asymmetry coefficient, specifically including:

[0057] Obtain the waveform asymmetry coefficient corresponding to a preset number of cycles in the secondary current waveform. When it is determined that the waveform asymmetry coefficient remains constant, the fault type of the current transformer is determined to be that the primary current amplitude is too large.

[0058] When it is determined that the waveform asymmetry coefficient gradually increases, the fault type of the current transformer is judged to be that there is DC bias in the primary side current.

[0059] In one possible implementation, the current transformer saturation value determination module is used to calculate the harmonic content of the secondary current, specifically including:

[0060] The secondary side current is subjected to Fourier transform processing to obtain the spectrum function of the secondary side current, and the harmonic amplitude of the secondary side current is obtained based on the spectrum function.

[0061] Substitute the harmonic amplitude into the preset harmonic content calculation formula to calculate the harmonic content of the secondary current and the harmonic content of the preset order.

[0062] The preset formula for calculating harmonic content is as follows:

[0063]

[0064]

[0065] In the formula, THD represents the harmonic content. i For the harmonic content of a preset number, A n Let A be the amplitude of the nth harmonic. i Let be the amplitude of the i-th harmonic.

[0066] In one possible implementation, the training process of the SVM classification model specifically includes:

[0067] Obtain the distorted secondary current waveforms corresponding to different distorted secondary currents, and calculate the distortion waveform asymmetry coefficient and harmonic components corresponding to the distorted secondary current waveforms.

[0068] Using the asymmetry coefficient of the distorted waveform and the harmonic components as inputs, and the current transformer saturation value of the current transformer as output, the initial SVM classification model is trained until the misclassification rate of the initial SVM classification model is less than the misclassification rate limit, thus generating an SVM classification model. The current transformer saturation value includes 0.2-level saturation value and 0.5-level saturation value.

[0069] In one possible implementation, before selecting the first RBF model corresponding to the saturation value of the first current transformer, the secondary current waveform correction module further includes:

[0070] Establish an RBF model based on the saturation value of the current transformer;

[0071] When the saturation value of the current transformer is 0.2, a 0.2-level RBF model is established; when the saturation value of the current transformer is 0.5, a 0.5-level RBF model is established.

[0072] The present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the current transformer error correction method based on classification SVM as described in any of the preceding claims.

[0073] The present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the current transformer error correction method based on classification SVM as described in any of the preceding claims.

[0074] This invention provides a current transformer error correction device based on classification SVM, which has the following advantages compared with the prior art:

[0075] By acquiring the secondary current output from the current transformer, a secondary current waveform corresponding to the secondary current is generated. Based on the secondary current waveform, a waveform asymmetry coefficient is calculated. The fault type of the current transformer is determined according to the waveform asymmetry coefficient. When the fault type is determined to be DC bias in the primary current, the harmonic content of the secondary current is calculated. The harmonic content and the waveform asymmetry coefficient are input into a pre-trained SVM classification model so that the SVM classification model outputs a first current transformer saturation value. A first RBF model corresponding to the first current transformer saturation value is selected so that the first RBF model corrects the secondary current waveform and outputs a corrected secondary current waveform. Compared with the prior art, the technical solution of this invention can improve the accuracy of secondary current waveform correction, and the selection of the corresponding first RBF model based on the first current transformer saturation value can improve the data processing speed. Attached Figure Description

[0076] Figure 1 This is a flowchart illustrating an embodiment of a current transformer error correction method based on classification SVM provided by the present invention.

[0077] Figure 2 This is a schematic diagram of an embodiment of a current transformer error correction device based on classification SVM provided by the present invention;

[0078] Figure 3 This is a schematic diagram of the actual connection of a current transformer in one embodiment of the present invention;

[0079] Figure 4 This is a schematic diagram of the secondary current waveform in one embodiment of the present invention under two conditions: the primary side rated current has a 40% DC bias component and the primary side current is 150% of the rated current.

[0080] Figure 5 This is a schematic diagram of the secondary current waveform under different primary side amplitude values ​​in one embodiment of the present invention;

[0081] Figure 6This is a schematic diagram illustrating the mechanism of excessive primary current amplitude in one embodiment of the present invention;

[0082] Figure 7 This is a schematic diagram of the mechanism of a current transformer under DC bias in one embodiment of the present invention;

[0083] Figure 8 This is a secondary side current waveform under different DC biases in one embodiment of the present invention;

[0084] Figure 9 This is a schematic diagram of the waveform asymmetry coefficient under two conditions: the primary side DC bias degree continuously increases and the primary side current amplitude continuously increases, in one embodiment of the present invention.

[0085] Figure 10 This is a schematic diagram illustrating the classification prediction effect of a fault classification model in one embodiment of the present invention;

[0086] Figure 11 This is a schematic diagram showing the ratio difference and phase difference results of a current transformer for measurement under different saturation levels in one embodiment of the present invention;

[0087] Figure 12 This is a schematic diagram of the simulation results of the ratio difference and phase difference of the current transformer under different DC component values ​​in one embodiment of the present invention;

[0088] Figure 13 This is a schematic diagram of the secondary current waveform under three conditions: distortion, normal, and correction, in one embodiment of the present invention.

[0089] Figure 14 This refers to the amplitude and phase of the secondary current fundamental wave under three conditions: distortion, normal, and correction, in one embodiment of the present invention. Detailed Implementation

[0090] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0091] Example 1, see Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of a current transformer error correction method based on classification SVM provided by the present invention, as shown below. Figure 1 As shown, the method includes steps 101-104, as detailed below:

[0092] Step 101: Obtain the secondary current output by the current transformer, generate the secondary current waveform corresponding to the secondary current, and calculate the waveform asymmetry coefficient based on the secondary current waveform.

[0093] In one embodiment, such as Figure 3 As shown, Figure 3 This is a schematic diagram of the actual connection of a current transformer; in the diagram, the number of turns of the primary winding is N1, and the current flowing through it is i1; the number of turns of the secondary winding is N2, and the current is i2; the secondary load resistance and inductance are R2 and L2 respectively; the core area is S.

[0094] Based on the working principle of current transformers and according to Ampere's circuital law, we have:

[0095] Hl = N1i1―N2i2;

[0096] In the formula, H represents the circumferential magnetic field strength in the iron core, and l is the average length of the magnetic circuit.

[0097] If the permeability of the iron core is μ, then the magnetic induction intensity B and the magnetic field intensity H satisfy the following relationship:

[0098] B = μH.

[0099] According to the laws of electromagnetic induction and Ohm's law, the voltage on the secondary circuit of a current transformer is:

[0100]

[0101] In the formula, the magnetic flux φ = BS.

[0102] In one embodiment, when the primary current contains a DC component I DC If the amplitude of the primary current is too large, the current transformer core will change from operating in the linear region to operating in the saturation region.

[0103] The secondary current waveforms were analyzed under two conditions: one with a 40% DC bias component in the primary side rated current and the other with a primary side current of 150% of the rated current. Figure 4 As shown, Figure 4 The diagram shows the secondary current waveforms under two conditions: the primary side rated current has a 40% DC bias component and the primary side current is 150% of the rated current. As can be seen from the diagram, although the secondary current waveform is significantly distorted in both cases, when there is a DC bias, the secondary current waveform exhibits a clear asymmetry between the positive and negative half-waves.

[0104] In one embodiment, in order to quantitatively analyze the degree of asymmetry of the positive and negative half-waves of the secondary current within one cycle, a waveform asymmetry coefficient σ is introduced for measurement. The area S enclosed by every three sampling points in the positive and negative half-waves of the secondary current waveform and the time t axis is calculated, and the maximum value of the enclosed area in the positive and negative half-waves is divided to obtain the waveform asymmetry coefficient σ within one cycle.

[0105] Specifically, by sampling multiple periodic waveforms in the secondary current waveform, multiple positive half-wave waveform sampling points and multiple negative half-wave waveform sampling points corresponding to each periodic waveform are obtained. Based on a preset area calculation formula, a first area enclosed by every three positive half-wave waveform sampling points and the time axis is calculated to obtain multiple first areas. A second area enclosed by every three negative half-wave waveform sampling points and the time axis is calculated to obtain multiple second areas. The maximum value among the multiple first areas is selected to obtain the first maximum area value, and the maximum value among the multiple second areas is selected to obtain the second maximum area value. The first maximum area value and the second maximum area value are input into the waveform asymmetry coefficient calculation formula to obtain the waveform asymmetry coefficient corresponding to each periodic waveform.

[0106] Preferably, the plurality of periodic waveforms are three-period waveforms.

[0107] In one embodiment, the preset area calculation formula is as follows:

[0108]

[0109] In the formula, S is the area, n is the waveform sampling point, i2 is the secondary current, and Δt is the sampling interval time.

[0110] In one embodiment, the formula for calculating the waveform asymmetry coefficient is as follows:

[0111]

[0112] In the formula, σ is the waveform asymmetry coefficient, and S + S is the first maximum area. ― This is the second largest area.

[0113] Step 102: Determine the fault type of the current transformer based on the waveform asymmetry coefficient, wherein the fault type includes DC bias in the primary side current and excessive amplitude of the primary side current.

[0114] In one embodiment, in PSCAD software, the rated current of the current transformer is set to 2kA. Simulation analysis for primary side currents ranging from 70% to 150% of the rated current reveals that as the primary side rated current increases, the degree of waveform defects and distortion in the secondary side current increases continuously. However, the secondary side current waveform remains symmetrical in both positive and negative half-waves. Figure 5 As shown, Figure 5 This is a schematic diagram of the secondary current waveform under different primary side amplitude values.

[0115] The transmission characteristics of current transformers are significantly affected by the nonlinearity of the excitation current i(t). For example... Figure 6 As shown, Figure 6 This is a schematic diagram of the mechanism of excessive primary current amplitude. When the primary current is the rated current, the amplitude of the excitation current i(t) is small, and the excitation current is sinusoidal. Therefore, the secondary current can linearly represent the primary current with an error of approximately zero. When the primary current amplitude is too large, the magnetic flux generated in the iron core is large, which reduces the excitation impedance, causes the excitation current to increase rapidly, and causes distortion, resulting in obvious waveform defects in the secondary current. However, since the increase in magnetic flux in the iron core is independent of the polarity of the primary current, the waveform asymmetry coefficient σ of the secondary current waveform is 1.

[0116] In one embodiment, the analysis focuses on the case where the current transformer contains a DC component on the primary side. For metering CTs, the nonlinear effect of the CT excitation branch needs to be considered, and this effect becomes more significant in the presence of DC bias. Under DC bias, the DC component will generate DC bias flux in different directions depending on its polarity, with the positive polarity DC bias flux Φ dc For example, let's analyze this; the bias magnetic flux and the main magnetic flux Φ generated by the AC component... dc The superposition of these phases forms the total magnetic flux density of the iron core under biased magnetization.

[0117] The excitation characteristic curve and output excitation current waveform of the current transformer with a DC component on the primary side are shown below. Figure 7 As shown, Figure 7 This is a schematic diagram of the mechanism of a current transformer under DC bias; for example... Figure 7 As shown, the solid line represents the current waveform of the core excitation branch without a DC component on the primary side; the dashed line represents the excitation current waveform corresponding to the presence of DC bias on the primary side. When a positive DC bias flux Φ exists... dc The magnetic flux density generated by the positive half-cycle of the primary current will be related to Φ dcThe superposition of these components increases the actual magnetic flux of the iron core; the bias flux has a demagnetizing effect on the magnetic flux density generated by the negative half-wave on the primary side. As the DC bias component increases, the positive half-cycle is more likely to enter saturation. When the iron core operating point enters saturation, the excitation current will increase significantly, causing a spike in the positive half-cycle and severe waveform distortion. However, due to the demagnetizing effect, the negative half-axis waveform still exhibits a sine wave. This waveform asymmetry will continuously increase with the increase of the DC bias flux, and the change process is as follows: Figure 8 As shown, Figure 8 These are the secondary current waveforms under different DC biases.

[0118] Based on the above analysis, it can be seen that as the DC bias on the primary side increases, the waveform asymmetry coefficient σ also increases; however, when the amplitude of the primary side current increases, the waveform asymmetry coefficient σ remains constant; the relationship is as follows: Figure 9 As shown, Figure 9 This is a schematic diagram of the waveform asymmetry coefficient under two conditions: the DC bias degree of the primary side continuously increases and the amplitude of the primary side current continuously increases.

[0119] In one embodiment, digital compensation of current transformers is often performed on the current transformer output over a period of time rather than on each instantaneous value in real time. This is determined by two factors: firstly, if compensation is performed on each current sampling point, it will consume a lot of CPU time, which is difficult to implement in working systems with low CPU processing speed and many sampling points; secondly, it is difficult to determine whether the transformer core is in a saturated state based on the independent current transformer output value. Therefore, in this embodiment, the sampling start point is the point where the secondary output current of the current transformer crosses zero and the current change direction is positive. The current transformer output value is sampled within three cycles to determine the type of fault of the power transformer.

[0120] Specifically, the waveform asymmetry coefficients corresponding to a preset number of periodic waveforms in the secondary current waveform are obtained. When it is determined that the waveform asymmetry coefficients remain constant, the fault type of the current transformer is determined to be DC bias in the primary current. When it is determined that the waveform asymmetry coefficients gradually increase, the fault type of the current transformer is determined to be DC bias in the primary current.

[0121] Preferably, the preset number of periodic waveforms are three periodic waveforms obtained from sampling.

[0122] In one embodiment, the Support Vector Machine (SVM) is a typical two-class classifier that is often used to solve binary classification problems.

[0123] The specific form of SVM is as follows: Training sample set T = {(x i ,yi )∈(X×Y) n}where x i ∈X=R in y i ∈Y={―1,1}i={1,2…n},x i If these are feature vectors, then there must exist a classification hyperplane that satisfies the distribution of positive and negative samples in the training sample set T on both sides of the hyperplane, as shown below:

[0124]

[0125] In the formula, w T Let be the coefficient vector of the hyperplane, and b be the intercept of the hyperplane normal vector.

[0126] SVM classification commonly employs two methods: hard margin and soft margin. Hard margin seeks a hyperplane to binary classify all sample points. While the majority of data samples are linearly separable, only a few outliers may prevent the finding of an optimal hyperplane. Therefore, soft margin classification often introduces a slack variable ζ. i By selecting an appropriate parameter C, the SVM classification method can achieve an acceptable misclassification rate, as shown below:

[0127]

[0128] In the formula, C is the penalty coefficient, which satisfies C>0. The larger C is, the heavier the penalty.

[0129] For the classification of nonlinear data, it is often necessary to use a kernel function that satisfies the Mercer condition to map the sample data to a high-dimensional feature space H, and then find the corresponding hyperplane in the high-dimensional space for classification.

[0130] Commonly used kernel functions include: polynomial kernel function, linear kernel function, radial basis function (RBF) kernel function, and sigmoid kernel function. In SVM classification, the RBF is typically used as the kernel function, mainly due to its excellent versatility and ability to approximate any function with small errors.

[0131] Preferably, when determining the fault type of a current transformer, the fault type of the current transformer can also be determined based on the fault differentiation algorithm of SVM.

[0132] In one embodiment, an initial SVM fault classification model is constructed, with waveform asymmetry coefficients as input and fault type as output. The initial SVM fault classification model is trained until it converges, thus obtaining the SVM fault classification model.

[0133] Specifically, in PSCAD, set the current transformer parameters as follows: N1 = 10, N2 = 100, l = 471 mm, S = 600 mm 2 The primary rated current is I. N =2000A, with a frequency f=50Hz; secondary load R2=0.8Ω, L=1.91mH. Using the waveform asymmetry coefficient as a classification index, the training results are as follows: Figure 10 As shown, Figure 10 The diagram illustrates the classification and prediction performance of the fault classification model. In the diagram, fault 1 indicates that the primary current has DC bias, and fault 0 indicates that the primary current has DC bias. It can be seen that the fault classification model can effectively distinguish between the two faults.

[0134] Step 103: When the fault type is determined to be DC bias in the primary current, calculate the harmonic content of the secondary current, and input the harmonic content and the waveform asymmetry coefficient into the pre-trained SVM classification model so that the SVM classification model outputs the first current transformer saturation value.

[0135] In one embodiment, as the saturation of the current transformer increases, the harmonic components in the secondary current output waveform not only have a continuously increasing amplitude of individual components, but also a continuously increasing harmonic order. Therefore, the saturation of the current transformer can be classified by using the asymmetry coefficient and the content of the 2nd to 5th harmonics in the secondary output waveform of the current transformer as classification indicators.

[0136] In one embodiment, according to Fourier series theory, any periodic function that satisfies the Dirichlet condition can be decomposed into an algebraic sum of DC, infinitely many sine and cosine functions; therefore, the secondary side current is subjected to Fourier transform processing to obtain the spectrum function of the secondary side current, and based on the spectrum function, the harmonic amplitude of the secondary side current is obtained; wherein, the spectrum function is as follows:

[0137]

[0138]

[0139] In the formula, w = 2π / T is called the fundamental angular frequency, a0 is its DC component, and a n and b n These are the amplitudes of its sine and cosine components, respectively.

[0140] Simplifying the above equation, we get:

[0141]

[0142]

[0143] In the formula, A0 is the DC component contained in the periodic signal, A n It is the amplitude of the nth harmonic. It is its initial phase.

[0144] In one embodiment, the harmonic amplitude is substituted into a preset harmonic content calculation formula to calculate the harmonic content of the secondary current and the harmonic content of a preset order.

[0145] The preset formula for calculating harmonic content is as follows:

[0146]

[0147]

[0148] In the formula, THD represents the harmonic content. i For the harmonic content of a preset number, A n Let A be the amplitude of the nth harmonic. i Let be the amplitude of the i-th harmonic.

[0149] Preferably, for the harmonic content of the preset number of harmonics, the harmonic content of the 2nd to 5th harmonics is selected.

[0150] In one embodiment, the training process of the SVM classification model specifically includes: acquiring the distorted secondary current waveforms corresponding to different distorted secondary currents, calculating the distortion waveform asymmetry coefficients and harmonic components corresponding to the distorted secondary current waveforms, so as to generate model training sample data.

[0151] Specifically, the 2nd to 5th harmonics are selected as the classification criteria for the multi-class SVM index. In PSCAD, the DC bias current is set under the conditions of 5%, 15%, 25%, and 40% of the rated current on the primary side. The corresponding simulation is performed and the harmonic content of the secondary current output by the current transformer under each condition is calculated. This is so that the DC bias current of the primary side current can be determined based on the obtained harmonic content.

[0152] In one embodiment, the initial SVM classification model is trained using the asymmetry coefficient of the distorted waveform and the harmonic components as inputs and the current transformer saturation value of the current transformer as output, until the misclassification rate of the initial SVM classification model is less than the misclassification rate limit, thereby generating an SVM classification model.

[0153] In one embodiment, the saturation value of the current transformer is defined to include a 0.2-level saturation value and a 0.5-level saturation value; and the ratio difference and phase difference of the current transformer for measurement are set at different saturation value levels, such as... Figure 11 As shown, Figure 11This diagram illustrates the ratio difference and phase difference results of current transformers used for measurement under different saturation levels. It also defines that when there is a DC component on the primary side, the ratio difference of the current transformer meets the accuracy requirements of a 0.5 class transformer, and the current transformer is in a slightly saturated state. When the ratio difference of the current transformer does not meet the accuracy requirements of a 0.5 class transformer, the transformer is in a heavily saturated state.

[0154] In one embodiment, when training the SVM classification model, the ratio difference and phase difference of the current transformer are also calculated, wherein the calculation formulas for the ratio difference and the phase difference are as follows:

[0155]

[0156]

[0157] In the formula, K n Given the rated turns ratio, I1 and I2 are the fundamental effective values ​​of the primary and secondary currents, respectively. These are the fundamental phases of the primary and secondary currents, respectively.

[0158] In one embodiment, simulations were also performed on the ratio difference and phase difference of current transformers with different DC components on the primary side. The results of the ratio difference and phase difference of the current transformers were obtained when the DC component was 0%, 1%, 2%, 5%, 8%, and 10%, as shown below. Figure 12 As shown, Figure 12 This is a schematic diagram of the simulation results of the ratio difference and phase difference of the current transformer under different DC component values.

[0159] In one embodiment, the current transformer is set to be in a slightly saturated state when there is a DC bias of 0-5% on the primary side; and in a heavily saturated state when there is a DC bias of more than 5% on the primary side.

[0160] In one embodiment, twenty sets of sample data are selected for model training in two cases: when the current transformer is in a slightly saturated state and when the current transformer is in a heavily saturated state. After the harmonic content and the waveform asymmetry coefficient are input into the pre-trained SVM classification model, the SVM classification model first determines the DC component of the primary current in the current transformer based on the harmonic content and the waveform asymmetry coefficient, and then determines the first current transformer saturation value based on the DC component and outputs it.

[0161] Step 104: Select the first RBF model corresponding to the saturation value of the first current transformer, so that the first RBF model can correct the secondary side current waveform and output the secondary side current correction waveform.

[0162] In one embodiment, before selecting the first RBF model corresponding to the saturation value of the first current transformer, the method further includes: establishing an RBF model based on the saturation value of the current transformer.

[0163] Specifically, when the saturation value of the current transformer is 0.2, a 0.2-level RBF model is established; when the saturation value of the current transformer is 0.5, a 0.5-level RBF model is established.

[0164] Preferably, corresponding RBF models are established for the output of different current transformer saturation values. This not only avoids the need for a large number of data samples when only one model is established, which has a better correction effect when the data is concentrated, but also results in a poor correction effect when the data amplitude varies greatly; in addition, by setting multiple RBF models, the number of training samples for each neural network can be reduced, which can improve the training speed of the model.

[0165] In one embodiment, a first RBF model corresponding to the saturation value of the first current transformer is selected; specifically, when the saturation value of the first current transformer is a 0.2 level saturation value, the first RBF model is a 0.2 level RBF model, and when the saturation value of the first current transformer is a 0.5 level saturation value, the first RBF model is a 0.5 level RBF model.

[0166] In one embodiment, when the first RBF model corrects the secondary current waveform, it mainly corrects the fundamental amplitude and fundamental phase of the secondary current waveform so that the secondary current waveform under the condition of distortion is corrected to the secondary current waveform under the condition of normal.

[0167] As an example in this embodiment: when the primary side is superimposed with a 40% DC bias component of the rated current, the secondary current waveform output by the current transformer will inevitably be distorted. After obtaining the current transformer value, the secondary current waveform is corrected by selecting the corresponding RBF model, and the correction effect is as follows. Figure 13 As shown, Figure 13 These are schematic diagrams of the secondary current waveforms under three conditions: distortion, normal, and correction; and the fundamental amplitude and phase of the secondary current, distortion current, and correction current corresponding to the primary rated current are shown below. Figure 14 As shown, Figure 14 It refers to the fundamental amplitude and phase of the secondary current under three conditions: distortion, normal, and correction.

[0168] In one embodiment, the neural network consists of many interconnected neurons. Each neuron represents a specific propagation function, commonly referred to as the activation function; the links between any two neurons are represented using specific weights, which essentially constitute the memory storage part of the neural network. The RBF neural network (Radial Basis Function, RBF), unlike feedback neural networks, propagates information layer by layer forward, that is, a unidirectional propagation process from the input layer to the hidden layers and finally to the output layer. Its topological structure is as follows: Figure 14 As shown, Figure 14 This is a schematic diagram of the RBF neural network topology; in the diagram, {X_((1)),X_((2))…X_((n))} are the input values ​​of the RBF neural network, and {Y_((1)),Y_((2))…Y_((n))} are the predicted values ​​of the RBF neural network. ij and w jk These are the weights of the RBF neural network. An RBF neural network expresses a functional mapping from n independent variables to m dependent variables. Unlike traditional BP neural networks, which are sensitive to initial parameters and prone to getting trapped in local optima, RBF neural networks have only one hidden layer, resulting in fast convergence and good adaptability to nonlinear problems. When training with sample data, the input signal is processed layer by layer from the input layer through the hidden layers, finally obtaining the training results at the output layer. The trained RBF neural network is then tested using test set data. If the output layer does not produce the expected results, the number of hidden layer nodes and the model are modified, continuously adjusting the network weights and thresholds. Finally, the model with the best test performance is selected as the final corrected model.

[0169] For RBF neural networks, the performance evaluation metrics are mainly determined by two aspects: firstly, their correction effect on secondary current waveforms with varying degrees of distortion; and secondly, the speed of the network output. Training an RBF neural network model primarily involves establishing the activation functions for the hidden and output layers. The activation functions for neurons in the hidden layer are generally radial basis functions (RBFs). RBFs are radially symmetric around their center point and are non-negative and non-linear, thus enabling them to achieve good non-linear mapping and providing good fitting for non-linear signals. The activation functions between the output and hidden layers are often linear. This is mainly to ensure rapid network convergence, avoid getting trapped in local minima, and enable the network to better interpret the information contained in the input layer signals.

[0170] The performance of an RBF neural network largely depends on the nonlinear mapping capability of its hidden layers. This mapping capability primarily depends on the number of nodes, their center positions, and their width. Too few nodes result in poor nonlinear mapping and insufficient model fit; conversely, too many nodes lead to overfitting and poor generalization ability. Common methods for selecting the center positions include self-organizing center selection, supervised center selection, and random center selection. For determining the weights, methods such as orthogonal least squares and pseudo-inverse methods are commonly used.

[0171] Preferably, the self-organizing center selection algorithm is an unsupervised learning method. This method clusters all input samples to obtain the center c of the radial basis of each hidden node. i The algorithm steps are as follows:

[0172] Step 1: Given an initial number of categories K, randomly generate initial classifications and calculate their centroids c. i (k);

[0173] Step 2: Calculate the distance between each sample and the initial cluster center, and assign the sample to the cluster center closest to it, that is, the dataset is divided into K classes;

[0174] Step 3: Recalculate the cluster centers for the dataset that has been divided into K classes;

[0175] Step 4: If the cluster centers change or the number of iterations has not been reached, return to Step 2;

[0176] Step 5: When the radial basis function is selected as a Gaussian function, its width σ i This can be derived from the following formula:

[0177]

[0178] In the formula, c max This represents the maximum distance between the K cluster centers.

[0179] Preferably, for the pseudo-inverse method to calculate the output weights,

[0180] When the hidden layer node center c is calculated i and the corresponding width σ i Then, output the weight vector w. i This can be calculated using the pseudo-inverse method. Assume the input vector is X. (i) Then the hidden layer output of the j-th node can be calculated using equation (10):

[0181]

[0182] In the formula, φj () is the activation function of the j-th node in the hidden layer.

[0183] The output matrix of the hidden layer can be represented by equation (11):

[0184]

[0185] The output weight vector W = {w1, w2, ... w} of the RBF network j The relationship between} and the network's output vector Y is shown in equation (12):

[0186] Y = ΦW;

[0187] Then W can be calculated as:

[0188] W=(Φ T Φ) ―1 Φ T Y;

[0189] In the formula, (Φ T Φ) ―1 Φ T Often written as Φ + This is called the pseudo-inverse matrix.

[0190] In one embodiment, to better address the complex operating conditions of new power systems, the fault types of current transformers are first classified. Different RBF models are then set for the saturation degree of the current transformers based on the fault type to perform error correction. Thus, when the data input to the input layer changes significantly, the number of nodes, center, width, and weights can be adjusted by matching an appropriate algorithm model. In other words, key parameters will autonomously change with the input data to determine new centers and weights, further improving the RBF neural network's ability to process new data. Simultaneously, it eliminates the need to retrain the RBF neural network, improving data processing efficiency.

[0191] Example 2, see Figure 2 , Figure 2 This is a schematic diagram of an embodiment of a current transformer error correction device based on classification SVM provided by the present invention, as shown below. Figure 2 As shown, the device includes a waveform asymmetry coefficient calculation module 201, a fault type judgment module 202, a current transformer saturation value determination module 203, and a secondary current waveform correction module 204, as detailed below:

[0192] The waveform asymmetry coefficient calculation module 201 is used to obtain the secondary current output by the current transformer, generate the secondary current waveform corresponding to the secondary current, and calculate the waveform asymmetry coefficient based on the secondary current waveform.

[0193] The fault type determination module 202 is used to determine the fault type of the current transformer based on the waveform asymmetry coefficient, wherein the fault type includes DC bias in the primary side current and excessive amplitude of the primary side current.

[0194] The current transformer saturation value determination module 203 is used to calculate the harmonic content of the secondary current when the fault type is determined to be DC bias in the primary current, and input the harmonic content and the waveform asymmetry coefficient into the pre-trained SVM classification model so that the SVM classification model outputs the first current transformer saturation value.

[0195] The secondary current waveform correction module 204 is used to select the first RBF model corresponding to the saturation value of the first current transformer, so that the first RBF model corrects the secondary current waveform and outputs the secondary current correction waveform.

[0196] In one embodiment, the fault type judgment module 202 is used to calculate the waveform asymmetry coefficient based on the secondary side current waveform, specifically including: sampling multiple periodic waveforms in the secondary side current waveform to obtain multiple positive half-wave waveform sampling points and multiple negative half-wave waveform sampling points corresponding to each periodic waveform; calculating a first area enclosed by every three positive half-wave waveform sampling points and the time axis based on a preset area calculation formula, to obtain multiple first areas; and calculating a second area enclosed by every three negative half-wave waveform sampling points and the time axis, to obtain multiple second areas; selecting the maximum value among the multiple first areas to obtain a first area maximum value, and selecting the maximum value among the multiple second areas to obtain a second area maximum value; inputting the first area maximum value and the second area maximum value into the waveform asymmetry coefficient calculation formula to obtain the waveform asymmetry coefficient corresponding to each periodic waveform; wherein, the preset area calculation formula is as follows:

[0197]

[0198] In the formula, S is the area, n is the waveform sampling point, i2 is the secondary current, and Δt is the sampling interval time;

[0199] The formula for calculating the waveform asymmetry coefficient is as follows:

[0200]

[0201] In the formula, σ is the waveform asymmetry coefficient, and S + S is the first maximum area. ― This is the second largest area.

[0202] In one embodiment, the fault type determination module 202 is used to determine the fault type of the current transformer based on the waveform asymmetry coefficient. Specifically, it includes: obtaining the waveform asymmetry coefficient corresponding to a preset number of periodic waveforms in the secondary current waveform; when it is determined that the waveform asymmetry coefficient remains constant, determining that the fault type of the current transformer is that the primary current amplitude is too large; when it is determined that the waveform asymmetry coefficient gradually increases, determining that the fault type of the current transformer is that the primary current has DC bias.

[0203] In one embodiment, the current transformer saturation value determination module 203 is used to calculate the harmonic content of the secondary side current, specifically including: performing Fourier transform processing on the secondary side current to obtain the spectrum function of the secondary side current; obtaining the harmonic amplitude of the secondary side current based on the spectrum function; substituting the harmonic amplitude into a preset harmonic content calculation formula to calculate the harmonic content of the secondary side current and the harmonic content of a preset order; wherein, the preset harmonic content calculation formula is as follows:

[0204]

[0205]

[0206] In the formula, THD represents the harmonic content. i For the harmonic content of a preset number, A n Let A be the amplitude of the nth harmonic. i Let be the amplitude of the i-th harmonic.

[0207] In one embodiment, the training process of the SVM classification model specifically includes: acquiring the distorted secondary current waveforms corresponding to different distorted secondary currents, calculating the distorted waveform asymmetry coefficient and harmonic components corresponding to the distorted secondary current waveforms; training the initial SVM classification model with the distorted waveform asymmetry coefficient and the harmonic components as inputs and the current transformer saturation value of the current transformer as output, until the misclassification rate of the initial SVM classification model is less than the misclassification rate limit, and generating an SVM classification model, wherein the current transformer saturation value includes a 0.2-level saturation value and a 0.5-level saturation value.

[0208] In one embodiment, before selecting the first RBF model corresponding to the saturation value of the first current transformer, the secondary current waveform correction module 204 further includes: establishing an RBF model based on the saturation value of the current transformer; establishing a 0.2-level RBF model when the saturation value of the current transformer is 0.2-level; and establishing a 0.5-level RBF model when the saturation value of the current transformer is 0.5-level.

[0209] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0210] It should be noted that the above-described embodiment of the current transformer error correction device based on classification SVM is merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0211] Based on the above-described embodiments of the current transformer error correction method based on classification SVM, another embodiment of the present invention provides a current transformer error correction terminal device based on classification SVM. The current transformer error correction terminal device based on classification SVM includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the current transformer error correction method based on classification SVM of any embodiment of the present invention.

[0212] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the classification-based SVM-based current transformer error correction terminal device.

[0213] The current transformer error correction terminal device based on classification SVM can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The current transformer error correction terminal device based on classification SVM may include, but is not limited to, a processor and a memory.

[0214] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the current transformer error correction terminal equipment based on classification SVM, connecting all parts of the equipment via various interfaces and lines.

[0215] The memory can be used to store the computer program and / or modules. The processor implements various functions of the current transformer error correction terminal device based on classification SVM by running or executing the computer program and / or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0216] Based on the above embodiments of the current transformer error correction method based on classification SVM, another embodiment of the present invention provides a storage medium, the storage medium including a stored computer program, wherein, when the computer program is running, the device where the storage medium is located executes the current transformer error correction method based on classification SVM of any embodiment of the present invention.

[0217] In this embodiment, the storage medium is a computer-readable storage medium, and the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0218] In summary, the present invention provides a current transformer error correction method and apparatus based on classification SVM. This method acquires the secondary current output from the current transformer, generates a secondary current waveform corresponding to the secondary current, calculates a waveform asymmetry coefficient based on the waveform waveform, determines the fault type of the current transformer based on the waveform asymmetry coefficient, calculates the harmonic content of the secondary current when the fault type is determined to be DC bias in the primary current, and inputs the harmonic content and the waveform asymmetry coefficient into a pre-trained SVM classification model so that the SVM classification model outputs a first current transformer saturation value, selects a first RBF model corresponding to the first current transformer saturation value, and uses the first RBF model to correct the secondary current waveform, outputting a corrected secondary current waveform. Compared with the prior art, the technical solution of the present invention can improve the accuracy of secondary current waveform correction, and the selection of the corresponding first RBF model based on the first current transformer saturation value can improve the data processing speed.

[0219] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A current transformer error correction method based on classification SVM, characterized in that, include: Obtain the secondary current output by the current transformer, generate the secondary current waveform corresponding to the secondary current, and calculate the waveform asymmetry coefficient based on the secondary current waveform. Based on the waveform asymmetry coefficient, the fault type of the current transformer is determined, wherein the fault type includes DC bias in the primary side current and excessive amplitude of the primary side current. When the fault type is determined to be DC bias in the primary current, the harmonic content of the secondary current is calculated, and the harmonic content and the waveform asymmetry coefficient are input into the pre-trained SVM classification model so that the SVM classification model outputs the first current transformer saturation value. A first RBF model corresponding to the saturation value of the first current transformer is selected so that the first RBF model can correct the secondary current waveform and output the secondary current correction waveform.

2. The current transformer error correction method based on classification SVM as described in claim 1, characterized in that, Based on the secondary current waveform, the waveform asymmetry coefficient is calculated, specifically including: Multiple periodic waveforms in the secondary current waveform are sampled to obtain multiple positive half-wave waveform sampling points and multiple negative half-wave waveform sampling points corresponding to each periodic waveform. Based on the preset area calculation formula, the first area enclosed by every three positive half-wave waveform sampling points and the time axis in the plurality of positive half-wave waveform sampling points is calculated to obtain a plurality of first areas; And calculate the second area enclosed by every three negative half-wave waveform sampling points and the time axis in the plurality of negative half-wave waveform sampling points to obtain a plurality of second areas; The maximum value among the plurality of first areas is selected to obtain the first maximum area; the maximum value among the plurality of second areas is selected to obtain the second maximum area. Input the first maximum area value and the second maximum area value into the waveform asymmetry coefficient calculation formula to obtain the waveform asymmetry coefficient corresponding to each period waveform; The preset area calculation formula is as follows: ； In the formula, For area, For waveform sampling points, For the time corresponding to the nth sampling point, This is the secondary side current. This is the sampling interval time; The formula for calculating the waveform asymmetry coefficient is as follows: ； In the formula, The waveform asymmetry coefficient, The first maximum area, This is the second largest area.

3. The current transformer error correction method based on classification SVM as described in claim 2, characterized in that, Based on the waveform asymmetry coefficient, the fault type of the current transformer can be determined, specifically including: Obtain the waveform asymmetry coefficient corresponding to a preset number of cycles in the secondary current waveform. When it is determined that the waveform asymmetry coefficient remains constant, the fault type of the current transformer is determined to be that the primary current amplitude is too large. When it is determined that the waveform asymmetry coefficient gradually increases, the fault type of the current transformer is judged to be that there is DC bias in the primary side current.

4. The current transformer error correction method based on classification SVM as described in claim 1, characterized in that, The calculation of the harmonic content of the secondary current specifically includes: The secondary side current is subjected to Fourier transform processing to obtain the spectrum function of the secondary side current, and the harmonic amplitude of the secondary side current is obtained based on the spectrum function. Substitute the harmonic amplitude into the preset harmonic content calculation formula to calculate the harmonic content of the secondary current and the harmonic content of the preset order. The preset formula for calculating harmonic content is as follows: ； ； In the formula, Harmonic content, The harmonic content is set to the preset number. Let be the amplitude of the i-th harmonic, and n be the highest harmonic number considered.

5. The current transformer error correction method based on classification SVM as described in claim 1, characterized in that, The training process of the SVM classification model specifically includes: Obtain the distorted secondary current waveforms corresponding to different distorted secondary currents, and calculate the distortion waveform asymmetry coefficient and harmonic components corresponding to the distorted secondary current waveforms. Using the asymmetry coefficient of the distorted waveform and the harmonic components as inputs, and the current transformer saturation value of the current transformer as output, the initial SVM classification model is trained until the misclassification rate of the initial SVM classification model is less than the misclassification rate limit, thus generating an SVM classification model. The current transformer saturation value includes 0.2-level saturation value and 0.5-level saturation value.

6. The current transformer error correction method based on classification SVM as described in claim 5, characterized in that, Before selecting the first RBF model corresponding to the saturation value of the first current transformer, the method further includes: Establish an RBF model based on the saturation value of the current transformer; When the saturation value of the current transformer is 0.2, a 0.2-level RBF model is established; when the saturation value of the current transformer is 0.5, a 0.5-level RBF model is established.

7. A current transformer error correction device based on classification SVM, characterized in that, include: The module includes a waveform asymmetry coefficient calculation module, a fault type judgment module, a current transformer saturation value determination module, and a secondary current waveform correction module. The waveform asymmetry coefficient calculation module is used to obtain the secondary current output by the current transformer, generate the secondary current waveform corresponding to the secondary current, and calculate the waveform asymmetry coefficient based on the secondary current waveform. The fault type determination module is used to determine the fault type of the current transformer based on the waveform asymmetry coefficient, wherein the fault type includes DC bias in the primary side current and excessive amplitude of the primary side current. The current transformer saturation value determination module is used to calculate the harmonic content of the secondary current when the fault type is determined to be DC bias in the primary current, and input the harmonic content and the waveform asymmetry coefficient into the pre-trained SVM classification model so that the SVM classification model outputs the first current transformer saturation value. The secondary current waveform correction module is used to select the first RBF model corresponding to the saturation value of the first current transformer, so that the first RBF model corrects the secondary current waveform and outputs the secondary current correction waveform.

8. The current transformer error correction device based on classification SVM as described in claim 7, characterized in that, The fault type determination module is used to calculate the waveform asymmetry coefficient based on the secondary side current waveform, specifically including: Multiple periodic waveforms in the secondary current waveform are sampled to obtain multiple positive half-wave waveform sampling points and multiple negative half-wave waveform sampling points corresponding to each periodic waveform. Based on the preset area calculation formula, the first area enclosed by every three positive half-wave waveform sampling points and the time axis in the plurality of positive half-wave waveform sampling points is calculated to obtain a plurality of first areas; And calculate the second area enclosed by every three negative half-wave waveform sampling points and the time axis in the plurality of negative half-wave waveform sampling points to obtain a plurality of second areas; The maximum value among the plurality of first areas is selected to obtain the first maximum area; the maximum value among the plurality of second areas is selected to obtain the second maximum area. Input the first maximum area value and the second maximum area value into the waveform asymmetry coefficient calculation formula to obtain the waveform asymmetry coefficient corresponding to each period waveform; The preset area calculation formula is as follows: ； In the formula, For area, For waveform sampling points, For the time corresponding to the nth sampling point, This is the secondary side current. This is the sampling interval time; The formula for calculating the waveform asymmetry coefficient is as follows: ； In the formula, The waveform asymmetry coefficient, The first maximum area, This is the second largest area.

9. A terminal device, characterized in that, The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the current transformer error correction method based on classification SVM as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the current transformer error correction method based on classification SVM as described in any one of claims 1 to 6.

Images