CN115600061B - Inverter zero voltage drop data processing method based on machine learning - Google Patents
Inverter zero voltage drop data processing method based on machine learning Download PDFInfo
- Publication number
- CN115600061B CN115600061B CN202211598077.4A CN202211598077A CN115600061B CN 115600061 B CN115600061 B CN 115600061B CN 202211598077 A CN202211598077 A CN 202211598077A CN 115600061 B CN115600061 B CN 115600061B
- Authority
- CN
- China
- Prior art keywords
- sequence
- value
- noise
- sub
- degree
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
- G01R19/0092—Arrangements for measuring currents or voltages or for indicating presence or sign thereof measuring current only
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The invention relates to the technical field of inverters, in particular to a zero voltage drop data processing method of an inverter based on machine learning, which comprises the following steps: obtaining current in an alternating current side voltage transformer circuit of the inverter to form a current data sequence, calculating a grouping length according to a maximum value and a minimum value in the current data sequence, and grouping the current data sequence to obtain a current data matrix; performing singular value decomposition on the current data matrix to obtain a plurality of sub-matrices, calculating the association degree of the two sub-matrices, grouping the sub-matrices according to the association degree, and processing the sub-matrices in the grouping to obtain a diagonal element sequence; calculating a noise evaluation value, and deleting the diagonal element sequence with the value smaller than an evaluation threshold value; and then calculating the noise degree, further determining the weight, weighting and summing all diagonal element sequences according to the weight to obtain a noise-removing current sequence, determining whether the abnormality exists, and processing abnormal data. The invention can remove noise, so that the current detection precision is higher.
Description
Technical Field
The invention relates to the technical field of inverters, in particular to a zero voltage drop data processing method of an inverter based on machine learning.
Background
With the reform and the deepening development of the power system, more and more people pay more attention to the enhancement of customer demand side management, so that the improvement of the electric energy calculation loop with large user quantity is more important. However, in order to solve the problem, in the prior art, whether the voltage drop is too large is often judged according to the current magnitude, but the voltage drop is interfered by noise when the current is collected, and at the moment, the judgment of the voltage drop according to the current is inaccurate.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide a machine learning-based inverter zero voltage drop data processing method, which adopts the following technical scheme:
obtaining current in an alternating current side voltage transformer circuit of the inverter within a set time period to form a current data sequence; forming an extreme value position sequence by the position serial numbers corresponding to all the maximum values and the minimum values in the current data sequence; calculating the grouping length according to the position interval between the adjacent maximum values and the position interval between the adjacent minimum values in the extreme value position sequence; grouping the current data sequence by utilizing the grouping length to obtain a current data matrix;
performing singular value decomposition on the current data matrix to obtain a plurality of sub-matrices, wherein each sub-matrix corresponds to a singular value; obtaining the correlation degree of the two sub-matrixes according to the difference value of any two sub-matrixes; dividing the submatrices with the correlation degree smaller than the degree threshold into a group, and further obtaining a plurality of groups of all the submatrices;
processing the groups by using an inverse diagonal averaging algorithm to obtain a plurality of diagonal element sequences, wherein each group corresponds to one diagonal element sequence, and each element in the diagonal element sequences corresponds to a singular value of a sub-matrix; obtaining a noise evaluation value according to singular values corresponding to elements in each diagonal element sequence and the sequence length, and deleting the sequence of which the noise evaluation value is smaller than an evaluation threshold value;
for the diagonal element sequence left after the deletion operation, obtaining the noise degree according to the difference value of the elements in the sequence; determining weight according to the noise degree, and performing weighted summation on all diagonal element sequences by using the weight to obtain a noise-removing current sequence; and judging whether abnormal current exists according to the noise-removing current sequence, and processing the abnormal current by using a zero voltage drop device.
Preferably, the method for acquiring the packet length specifically includes:
when the first extreme value and the last extreme value in the extreme value position sequence are both maximum values or both minimum values, calculating the grouping length, and expressing the grouping length as follows by using a formula:
wherein the content of the first and second substances,which indicates the length of the packet or packets,the value of the 2c-1 th element in the extreme position sequence is shown,represents the value of the 2c +1 th element in the extreme position sequence,representing the value of the 2 c-th element in the extremum position sequence,representing the value of the 2c +2 th element in the extreme position sequence, wherein r represents the total number of the elements in the extreme position sequence;
when the first extreme value in the extreme value position sequence is a maximum value and the last extreme value is a minimum value, or the first extreme value is a minimum value and the last extreme value is a maximum value, calculating the grouping length, and expressing the grouping length as follows by using a formula:
wherein the content of the first and second substances,which indicates the length of the packet or packets,the value of the 2c-1 th element in the extreme position sequence is shown,represents the value of the 2c +1 th element in the extreme position sequence,representing the value of the 2 c-th element in the extremum position sequence,representing the value of the 2c +2 th element in the extremum position sequence, and r representing the total number of elements in the extremum position sequence.
Preferably, the method for obtaining the degree of association between the two sub-matrices specifically includes:
wherein the content of the first and second substances,indicating the degree of association of the sub-matrix h and the sub-matrix l,the value representing the i-th element of the sub-matrix h,the value of the i-th element of the sub-matrix l is represented, and H represents the size of the sub-matrix.
Preferably, the method for acquiring a plurality of groups of all the submatrices specifically includes:
dividing the two sub-matrixes with the correlation degree smaller than the degree threshold into a group, and calculating the average matrix of the two sub-matrixes; acquiring the association degree of the average matrix and other sub-matrices, and recording the sub-matrices with the association degree smaller than a degree threshold value into the group; calculating an average matrix and a new average matrix of the submatrices recorded in the group, continuously acquiring the association degree of the new average matrix and other submatrices, and calculating the average matrix once every time one submatrix is recorded in the group until all the submatrices are traversed to obtain a group of matrixes; a plurality of packets are obtained in the same way.
Preferably, the method for acquiring the noise evaluation value specifically includes:
wherein the content of the first and second substances,represents the noise evaluation value of the diagonal element sequence m,representing the mean of all singular values corresponding to the diagonal element sequence m,represents the maximum value of the singular value means corresponding to all diagonal element sequences,indicates the length of the diagonal element sequence m,representing the maximum of the lengths of all diagonal element sequences.
Preferably, the method for acquiring the noise level specifically includes:
and respectively subtracting the numerical values of two adjacent elements in the diagonal element sequence to obtain a difference sequence, calculating the noise degree according to the elements in the difference sequence, and expressing the noise degree by a formula as follows:
wherein the content of the first and second substances,is the noise level of the diagonal element sequence n,representing the mean of all elements in the difference sequence corresponding to the diagonal element sequence n,representing the largest element in the difference sequence corresponding to the diagonal element sequence n.
Preferably, the determining the weight according to the noise level specifically includes:
and when the noise degree of the diagonal element sequence is smaller than the noise threshold, taking the noise degree as the weight corresponding to the diagonal element sequence.
The embodiment of the invention at least has the following beneficial effects:
the invention adopts the zero voltage drop device to solve the problem of overlarge voltage drop of the secondary circuit, and the overlarge voltage drop can cause abnormal current, so that the acquired current is subjected to abnormal detection. However, in the acquisition process, a noise influence detection result appears, so that the invention uses a singular spectrum analysis method to remove noise, when the singular spectrum analysis method is used, the grouping length is calculated according to the position interval between adjacent maximum values and adjacent minimum values in the acquired current data sequence, the sequence is grouped to obtain a current data matrix, and then the singular value decomposition method is used for processing; the decomposed sub-matrices are grouped by the calculated degree of association, and weights are set adaptively at the time of reconstruction. The method has higher adaptability and precision of denoising, so that the current detection precision is higher, the condition of false detection is avoided, and the voltage drop can be accurately judged according to the current.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a method for processing zero-voltage drop data of an inverter based on machine learning according to the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description, the structure, the features and the effects of the inverter zero-voltage-drop data processing method based on machine learning according to the present invention are provided with reference to the accompanying drawings and the preferred embodiments. In the following description, the different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following describes a specific scheme of the inverter zero-voltage-drop data processing method based on machine learning in detail with reference to the accompanying drawings.
The specific scenes aimed by the invention are as follows: the direct current side of the inverter is connected with a direct current energy storage power supply and used for converting direct current of the energy storage power supply into alternating current, and the alternating current side of the inverter is connected with alternating current equipment for supplying power. And a voltage transformer is arranged on the alternating current side and used for detecting alternating current voltage.
In the inverter circuit, the inverter is used to convert the dc power into ac power, and the dc voltage utilization rate is low. The definition of the direct current voltage utilization rate is as follows: the direct-current voltage utilization rate refers to the ratio of the fundamental wave amplitude of the voltage of an output line in an inverter circuit to the input voltage. The output capacity of the motor control system can be improved due to the fact that the direct-current voltage utilization rate is high, in order to obtain the good direct-current voltage utilization rate, firstly, SPWM is adopted for modulation, and in addition, the modulation ratio is 1. The maximum voltage utilization rate of the bus is only 86.6 percent at this time. In order to achieve higher utilization rate, firstly, 3-order harmonic injection is used, then SPWM modulation is carried out, the modulation ratio is still 1, and at the moment, the voltage utilization rate of the bus can be improved by 15.47 percent on the basis. Thus, the problem of the utilization rate of the direct-current voltage is solved, and the current is collected after the problem.
Referring to fig. 1, a flowchart of a method for processing zero-voltage drop data of an inverter based on machine learning according to an embodiment of the present invention is shown, where the method includes the following steps:
acquiring current in an alternating current side voltage transformer circuit of an inverter within a set time period to form a current data sequence; forming an extreme value position sequence by the position serial numbers corresponding to all the maximum values and the minimum values in the current data sequence; calculating the grouping length according to the position interval between the adjacent maximum values and the position interval between the adjacent minimum values in the extreme value position sequence; and grouping the current data sequence by using the grouping length to obtain a current data matrix.
Firstly, it should be noted that the zero voltage drop management device is a zero voltage drop intelligent electric quantity device designed according to actual requirements of vast power consumers, and because of the situations of large voltage drop of a secondary circuit of a voltage transformer, inaccurate electric energy calculation and the like, the device aims to eliminate voltage drop and reduce electric energy calculation errors. If the voltage drop in the secondary circuit is large, the current in the circuit can be measured at the moment, and whether the problem of large voltage drop of the secondary circuit occurs or not is judged according to the current condition. The current transformer is used for measuring the current in the secondary loop, and the noise data exists in the acquired current data due to the fact that the acquisition of the current transformer is interfered by a narrow-band signal and a white noise signal, so the noise reduction method can be used for denoising the current data.
Then, the current in the circuit within the set time period is obtained, and in the embodiment, the current data within 5min is collected to form a current data sequence. According to the idea of the singular spectrum analysis algorithm, the current data sequence needs to be converted into a matrix form, so that the current data sequence needs to be subjected to proper grouping operation.
Specifically, the current data sequence is fitted to a curve, wherein the horizontal axis in the coordinate system of the curve is the position serial number of each element in the sequence, and the vertical axis is the value of each element in the sequence, that is, the magnitude of the current. Obtaining a function of the curve asThe derivative function of this function is derived so that the derivative function is 0, whereby the extreme point of the curve can be obtained. For a fluctuating curve, a maximum value and a minimum value inevitably exist, if a plurality of extreme points exist, a maximum value point and a minimum value point inevitably exist, and the obtained abscissa of the extreme points forms an extreme value position sequence and is recorded as an extreme value position sequenceWherein, in the step (A),representing the first extreme value and r representing the total number of extreme values.
Since there is necessarily a minimum value between two adjacent maximum values, similarly, there is also a maximum value between two adjacent minimum values. And acquiring the current magnitude corresponding to each element in the extreme value position sequence corresponding to the current data sequence. If the first extreme valueThe corresponding current data is less than the second extreme valueAnd corresponding current data, wherein the first extreme point is a minimum point, and the second extreme point is a maximum point. If the first extreme valueThe corresponding current data is greater than the second extreme valueAnd corresponding current data, wherein the first extreme point is a maximum point, and the second extreme point is a minimum point. According to the method, whether each element in the extreme value position sequence corresponds to the maximum value point or the minimum value point can be obtained.
In this embodiment, the eigenvalues of the matrix are obtained by using the singular value decomposition method, and in order to remove noise data in the current better, the grouping needs to include both the maximum value point and the minimum value point as much as possible. Meanwhile, both abnormal current data and noise data have extreme values, and a proper grouping length needs to be acquired, so that the noise data is not grouped in one group as much as possible, and the disadvantage of subsequent processing is avoided.
And finally, calculating the grouping length according to the position interval between the adjacent maximum values and the position interval between the adjacent minimum values in the extreme value position sequence, and when the first extreme value and the last extreme value in the extreme value position sequence are both maximum values or both minimum values, calculating the grouping length, wherein the grouping length is expressed by a formula:
wherein the content of the first and second substances,which indicates the length of the packet or packets,the value of the 2c-1 element in the extreme value position sequence, namely the position serial number corresponding to the 2c-1 extreme value,represents the value of the 2c +1 th element in the extreme position sequence,representing the value of the 2 c-th element in the extremum position sequence,and representing the value of the 2c +2 th element in the extremum position sequence, wherein r represents the total number of elements in the extremum position sequence, namely the total number of extremums contained in the extremum position sequence.
When the first extreme value in the extreme value position sequence is a maximum value and the last extreme value is a minimum value, or the first extreme value is a minimum value and the last extreme value is a maximum value, calculating the grouping length, and expressing the grouping length as follows by using a formula:
wherein the content of the first and second substances,which indicates the length of the packet or packets,the value of the 2c-1 th element in the extreme position sequence is shown,represents the value of the 2c +1 th element in the extreme position sequence,representing the value of the 2 c-th element in the extremum position sequence,representing the value of the 2c +2 th element in the extremum position sequence, and r representing the total number of elements in the extremum position sequence.
It should be noted that, because the current data sequence needs to be grouped to obtain a corresponding matrix form, the grouping needs to ensure that the data length of each group is the same, that is, there cannot be data that is not grouped in isolation. While the current data sequence is grouped with the resulting grouping length, the current data sequence cannot necessarily be divided evenly, i.e. the sequence length cannot necessarily be divided exactly by the grouping length. Therefore, the packet length needs to be adjusted in size so that the current data sequence can be divided uniformly. In this embodiment, the current data matrix is formed by gradually increasing the value of the packet length until the length of the current data sequence is divisible by the packet length to obtain a corresponding packet.
Performing singular value decomposition on the current data matrix to obtain a plurality of sub-matrices, wherein each sub-matrix corresponds to a singular value; obtaining the correlation degree of the two sub-matrixes according to the difference value of any two sub-matrixes; and dividing the submatrices with the correlation degree smaller than the degree threshold into a group, and further obtaining a plurality of groups of all the submatrices.
First, singular value decomposition is performed on the current data matrix to obtain a plurality of sub-matrices. The singular value decomposition algorithm is a known algorithm, and is only described briefly in this embodiment. Has a decomposition formula ofU and V are unit orthogonal matrices,is a transposed matrix of the matrix V,the method comprises the steps of forming a diagonal matrix for current data matrix eigenvalues, wherein each element in the diagonal matrix is a singular value of a current data matrix.
Specifically, the singular value decomposition of the current data matrix is expressed by the formula:
wherein, X is a current data matrix,for the k-th element in the matrix U,for the k-th element in the matrix V,is a matrixThe k-th element in (a), o, is the number of ranks of the current data matrix X, i.e. the number of non-zero singular values of the current data matrix X,andrespectively, the 1 st sub-matrix and the o-th sub-matrix.
Then, the degree of correlation between the two sub-matrices is calculated, and is expressed by the formula:
wherein the content of the first and second substances,the degree of association between the sub-matrix h and the sub-matrix l is expressed, and the degree of association can be expressed as the degree of correlation between the two matrices.The value representing the i-th element of the sub-matrix h,the value of the ith element of the sub-matrix l is represented, and H represents the size of the sub-matrix.
Finally, dividing the two sub-matrixes with the correlation degree smaller than the degree threshold into a group, and calculating the average matrix of the two sub-matrixes; acquiring the association degree of the average matrix and other sub-matrices, and recording the sub-matrices with the association degree smaller than a degree threshold value into the group; calculating an average matrix and a new average matrix of the submatrices recorded in the group, continuously acquiring the association degree of the new average matrix and other submatrices, and calculating the average matrix once every time one submatrix is recorded in the group until all the submatrices are traversed to obtain a group of matrixes; a plurality of packets are obtained in the same manner.
Specifically, in this embodiment, the value of the degree threshold is 2, and an implementer can set the value of the degree threshold according to an actual situation. When the degree of association is less than the degree thresholdI.e. byThen, the submatrix h and the submatrix l are divided into a group, and the average matrix of the two matrixes is calculated according to the numerical values of the elements of the submatrix h and the submatrix l. Obtaining any other submatrix p except the submatrix h and the submatrix l, and calculating an average matrixAnd the degree of association of the submatrix p, when the degree of association is less than a degree threshold, the submatrix p is recorded into a group containing the submatrix h and the submatrix l, and an average matrix is calculatedAverage matrix of sum sub-matrix p. Continuously calculating any other sub-matrix and average matrixAnd recording the submatrices with the correlation degree smaller than the degree threshold value into the group, calculating an average matrix of the submatrices in the group every time one submatrix is recorded into the group until all the submatrices are traversed to obtain a group of matrixes, and finishing the grouping of all the submatrices according to the same method to obtain a plurality of groups. Wherein, one group comprises a plurality of submatrices, and each submatrix corresponds to a singular value.
It should be noted that when the calculated association degree is greater than the degree threshold, other sub-matrices need to be selected to continue calculating the association degree, and the operation of recording the sub-matrices in the groups is not executed until the obtained association degree is less than the degree threshold.
Processing the plurality of groups by using an inverse diagonal averaging algorithm to obtain a plurality of diagonal element sequences, wherein each group corresponds to one diagonal element sequence, and each element in the diagonal element sequences corresponds to a singular value of a sub-matrix; and obtaining a noise evaluation value according to the singular value and the sequence length corresponding to the elements in each diagonal element sequence, and deleting the sequence of which the noise evaluation value is smaller than the evaluation threshold value.
Specifically, after obtaining a plurality of groups, processing the sub-matrices in each group by using an inverse diagonal averaging method to obtain a sequence, where the inverse diagonal averaging method is a known technique, and only a simple description is made here, that is, all the sub-matrices in a group form a comprehensive matrix, and the diagonals of the comprehensive matrix are solved according to a formula to obtain the sequence of the group, and then obtain the sequence of each group as a diagonal element sequence.
It should be noted that some sequences in the diagonal element sequences of each group are acquired to belong to noise sequences, and in the singular spectrum analysis algorithm, a sequence formed by numerical values with very small singular values is used as a noise sequence.
And obtaining a noise evaluation value according to the singular value corresponding to the element in each diagonal element sequence and the sequence length, wherein the noise evaluation value is expressed by a formula as follows:
wherein the content of the first and second substances,represents the noise evaluation value of the diagonal element sequence m,represents the mean of all singular values corresponding to the diagonal element sequence m,representing all diagonal elementsThe maximum value of the singular value means corresponding to the element sequence,indicates the length of the diagonal element sequence m,representing the maximum of the lengths of all diagonal element sequences.
Since the singular value corresponding to the sequence that may be a noise point is small, the sequence with a shorter length is more likely to be a sequence including a noise point. Therefore, when the noise evaluation value is smaller than the evaluation threshold, the evaluation threshold is 0.1 in this embodiment, that is, when the noise evaluation value is smaller than the evaluation thresholdThen, the diagonal element sequence is considered to possibly contain noise points, and the sequence is deleted. The subsequent deletion operation on the sequence does not affect the subsequent operation because the subsequent accumulation reconstruction needs to be carried out on each diagonal element sequence to obtain a new sequence. Meanwhile, the evaluation threshold value implementer can set according to actual conditions.
Step four, obtaining the noise degree of the diagonal element sequence left after the deletion operation according to the difference value of the elements in the sequence; determining weight according to the noise degree, and performing weighted summation on all diagonal element sequences by using the weight to obtain a noise-removing current sequence; and judging whether abnormal current exists according to the noise-removing current sequence, and processing the abnormal current by using a zero voltage drop device.
Specifically, for the diagonal element sequence remaining after the deletion operation, there may be noise points, and since the noise points are isolated points and have a large difference from other data, the screening is performed according to the difference of the elements in the sequence. Specifically, for a diagonal element sequence, the difference is made between the first element and the second element of the sequence, the difference is made between the second element and the third element, the difference is made between the third element and the fourth element, and so on, and all the obtained differences form the difference sequence corresponding to the diagonal element sequence. Calculating the noise degree according to elements in the difference value sequence, and expressing the noise degree by a formula as follows:
wherein the content of the first and second substances,is the noise level of the diagonal element sequence n,representing the mean of all elements in the difference sequence corresponding to the diagonal element sequence n,representing the largest element in the difference sequence corresponding to the diagonal element sequence n. Degree of noiseThe larger the value of (a), the smaller the probability that the diagonal element sequence has noise points.
And when the noise degree of the diagonal element sequence is smaller than the noise threshold, taking the noise degree as the weight corresponding to the diagonal element sequence.
In this embodiment, the noise threshold value is 0.8, the first value is 1, and an implementer can determine the noise threshold value and the first value according to an actual situation. In particular, whenWhen the diagonal element sequence n does not contain any noise point at all, the weight corresponding to the diagonal element sequence n is set to 1 and recorded as(ii) a When in useThe noise level of the diagonal element sequenceAs the corresponding weight, it is recorded as. Since the presence of abnormal data in the current may cause a noise level less than the noise threshold, and the abnormal current is monitored, the sequence of such conditions is given a weight equal to the noise level.
According to the corresponding weight of each diagonal element sequence, carrying out weighted summation on each sequence, and expressing the sum as follows by a formula:
wherein W is a noise removing current sequence, namely W is a current sequence after noise removal,the weight corresponding to the diagonal element sequence n,for the diagonal element sequence N, N represents the number of diagonal element sequences.
If the voltage drop is large, a large abnormal current exists at this time, and since the voltage drop and the current are large, the voltage drop and the current are only generated for a circuit with a long circuit, under the condition of the same rated voltage and power, the embodiment obtains normal current sequence data by measuring the current in the short circuit, compares the normal current sequence data with the noise-removing current sequence, and judges whether the abnormal current exists in the current. Wherein, the implementer can select a proper method for comparison and judgment according to the actual situation.
When the current is detected to be abnormal, because the current data in the set time period is collected in the embodiment, namely when the current in the set time period has the abnormal data, the zero-voltage-drop metering function is automatically cut off, and the secondary analog voltage signal meter provided by the original voltage transformer continues to operate, the metering precision is ensured, and the safe operation of the device is also ensured. A constant-value voltage source is connected into a secondary loop by using the constant-value compensation type compensator, and the voltage source is adjusted according to the thickness of the non-open circuit so as to counteract the influence of secondary voltage drop.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.
Claims (1)
1. A machine learning-based inverter zero voltage drop data processing method is characterized by comprising the following steps:
obtaining current in an alternating current side voltage transformer circuit of the inverter within a set time period to form a current data sequence; forming an extreme value position sequence by the position serial numbers corresponding to all the maximum values and the minimum values in the current data sequence; calculating the grouping length according to the position interval between the adjacent maximum values and the position interval between the adjacent minimum values in the extreme value position sequence; grouping the current data sequence by utilizing the grouping length to obtain a current data matrix;
performing singular value decomposition on the current data matrix to obtain a plurality of sub-matrices, wherein each sub-matrix corresponds to a singular value; obtaining the correlation degree of the two sub-matrixes according to the difference value of any two sub-matrixes; dividing the submatrices with the correlation degree smaller than the degree threshold into a group, and further obtaining a plurality of groups of all the submatrices;
processing the groups by using an inverse diagonal averaging algorithm to obtain a plurality of diagonal element sequences, wherein each group corresponds to one diagonal element sequence, and each element in the diagonal element sequences corresponds to a singular value of a sub-matrix; obtaining a noise evaluation value according to singular values corresponding to elements in each diagonal element sequence and the sequence length, and deleting the sequence of which the noise evaluation value is smaller than an evaluation threshold value;
for the diagonal element sequence left after the deletion operation, obtaining the noise degree according to the difference value of the elements in the sequence; determining weight according to the noise degree, and performing weighted summation on all diagonal element sequences by using the weight to obtain a noise-removing current sequence; judging whether abnormal current exists according to the noise-removing current sequence, and processing the abnormal current by using a zero voltage drop device;
the method for acquiring the packet length specifically comprises the following steps:
when the first extreme value and the last extreme value in the extreme value position sequence are both maximum values or both minimum values, calculating the grouping length, and expressing the grouping length as follows by using a formula:
wherein the content of the first and second substances,indicates the packet length, and>represents the value of the 2c-1 th element in the extremum position sequence, is selected>Represents the value of the 2c +1 element in the extremum position sequence and is then selected>Represents the value of the 2 c-th element in the extremum position series, takes place in a manner which is characteristic of the number of extrema in the extremum position series>Represents the 2c + in the extreme position sequenceValues of 2 elements, r representing the total number of elements in the extremum position sequence;
when the first extreme value in the extreme value position sequence is a maximum value and the last extreme value in the extreme value position sequence is a minimum value, or the first extreme value is a minimum value and the last extreme value in the extreme value position sequence is a maximum value, calculating the grouping length, and expressing the grouping length by a formula as follows:
wherein the content of the first and second substances,indicates the packet length, and>represents the value of the 2c-1 th element in the extremum position sequence, and/or the value of a member in the extremum position sequence>Represents the value of the 2c +1 element in the extremum position sequence and is then selected>Represents the value of the 2 c-th element in the extremum position series, takes place in a manner which is characteristic of the number of extrema in the extremum position series>Representing the value of the 2c +2 th element in the extreme position sequence, wherein r represents the total number of the elements in the extreme position sequence;
the method for acquiring the association degree of the two sub-matrixes comprises the following specific steps:
wherein the content of the first and second substances,represents the degree of association of the sub-matrix h and the sub-matrix l>Denotes the value of the i-th element of the sub-matrix h, < > H>The value of the ith element of the sub-matrix l is represented, and H represents the size of the sub-matrix; />
The method for acquiring the plurality of groups of all the submatrices specifically comprises the following steps:
dividing the two sub-matrixes with the correlation degree smaller than the degree threshold into a group, and calculating the average matrix of the two sub-matrixes; acquiring the association degree of the average matrix and other sub-matrices, and recording the sub-matrices with the association degree smaller than a degree threshold value into the group; calculating an average matrix and a new average matrix of the submatrices recorded in the group, continuously acquiring the association degree of the new average matrix and other submatrices, and calculating the average matrix once every time one submatrix is recorded in the group until all the submatrices are traversed to obtain a group of matrixes; obtaining a plurality of groups according to the same method;
the method for acquiring the noise evaluation value specifically comprises the following steps:
wherein the content of the first and second substances,represents the noise evaluation value of the diagonal element sequence m, is greater than>Represents the mean of all singular values corresponding to the diagonal element sequence m, <' > or>Represents the maximum value of the singular value mean corresponding to all diagonal element sequences, and/or represents the value of the diagonal element sequence>Represents the length of the diagonal element sequence m, < > or >>Represents the maximum of the lengths of all diagonal element sequences;
the method for acquiring the noise degree specifically comprises the following steps:
respectively subtracting the numerical values of two adjacent elements in the diagonal element sequence to obtain a difference value sequence, calculating the noise degree according to the elements in the difference value sequence, and expressing the noise degree by a formula as follows:
wherein the content of the first and second substances,for the noise level of the diagonal element sequence n, <' >>Represents the mean value of all elements in the difference sequence corresponding to the diagonal element sequence n, < > or >>Representing the largest element in the difference sequence corresponding to the diagonal element sequence n;
the determining the weight according to the noise degree specifically includes:
and when the noise degree of the diagonal element sequence is smaller than the noise threshold, taking the noise degree as the weight corresponding to the diagonal element sequence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211598077.4A CN115600061B (en) | 2022-12-14 | 2022-12-14 | Inverter zero voltage drop data processing method based on machine learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211598077.4A CN115600061B (en) | 2022-12-14 | 2022-12-14 | Inverter zero voltage drop data processing method based on machine learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115600061A CN115600061A (en) | 2023-01-13 |
CN115600061B true CN115600061B (en) | 2023-03-28 |
Family
ID=84854136
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211598077.4A Active CN115600061B (en) | 2022-12-14 | 2022-12-14 | Inverter zero voltage drop data processing method based on machine learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115600061B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116908524B (en) * | 2023-09-13 | 2023-12-01 | 中国建筑科学研究院有限公司 | Abnormal sensing monitoring system of building electrical system based on artificial intelligence |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105938578A (en) * | 2016-04-13 | 2016-09-14 | 合肥工业大学 | Large-scale photovoltaic power station equivalent modeling method based on clustering analysis |
CN108828405A (en) * | 2018-06-06 | 2018-11-16 | 西南交通大学 | A kind of electric transmission line fault detection method based on random matrix |
CN113836704A (en) * | 2021-09-06 | 2021-12-24 | 广州大学 | Time series noise reduction method and device based on improved singular spectrum analysis |
Family Cites Families (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2008148075A1 (en) * | 2007-05-24 | 2008-12-04 | Alexander George Parlos | Machine condition assessment through power distribution networks |
CN102769336B (en) * | 2012-08-08 | 2015-04-01 | 江苏银佳企业集团有限公司 | Single-phase inverter for emergency power supply |
CN105323036A (en) * | 2014-08-01 | 2016-02-10 | 中国移动通信集团公司 | Method and device for performing singular value decomposition on complex matrix and computing equipment |
US11770197B2 (en) * | 2015-07-17 | 2023-09-26 | Origin Wireless, Inc. | Method, apparatus, and system for accurate wireless monitoring |
US11639981B2 (en) * | 2015-07-17 | 2023-05-02 | Origin Wireless, Inc. | Method, apparatus, and system for movement tracking |
CN105116208B (en) * | 2015-08-18 | 2018-03-06 | 昆明理工大学 | A kind of extra-high voltage DC transmission system commutation failure method for diagnosing faults |
CN105353305B (en) * | 2015-11-17 | 2018-01-12 | 浙江大学 | Rotor fault of eccentricity diagnostic method based on complete adaptive matrix beam |
CN108595783A (en) * | 2018-04-03 | 2018-09-28 | 南京航空航天大学 | A kind of design method of small fault estimating system for CRH2 type high ferro inverters |
US11108795B2 (en) * | 2018-05-25 | 2021-08-31 | At&T Intellectual Property I, L.P. | Intrusion detection using robust singular value decomposition |
CN111585544A (en) * | 2019-02-15 | 2020-08-25 | 广东石油化工学院 | Method and filter for filtering power signal impulse noise |
CN111585542A (en) * | 2019-02-15 | 2020-08-25 | 广东石油化工学院 | Method and filter for filtering power signal impulse noise |
CN110161370A (en) * | 2019-04-25 | 2019-08-23 | 国网辽宁省电力有限公司 | A kind of electric network fault detection method based on deep learning |
CN110321921A (en) * | 2019-05-09 | 2019-10-11 | 济南浪潮高新科技投资发展有限公司 | A method of it is abnormal that detection being decomposed based on time series segmentation and tensor CP |
CN110661237B (en) * | 2019-07-03 | 2021-10-29 | 天津大学 | Current differential protection method suitable for power distribution network under self-synchronization condition |
US11303579B1 (en) * | 2019-09-10 | 2022-04-12 | University Of South Florida | System and method for admittance model identification for inverter-based resources |
CN111740575B (en) * | 2020-07-01 | 2023-06-09 | 电子科技大学 | Inverter model parameter self-adaptive identification method based on steepest descent method |
CN112381351B (en) * | 2020-10-15 | 2022-10-11 | 广西电网有限责任公司 | Power utilization behavior change detection method and system based on singular spectrum analysis |
CN112421586A (en) * | 2020-10-30 | 2021-02-26 | 国网甘肃省电力公司电力科学研究院 | Protection starting judgment method and device based on Hankel matrix singular value decomposition |
CN113156269A (en) * | 2021-05-12 | 2021-07-23 | 国网山东省电力公司电力科学研究院 | Fault location method for direct-current hybrid line |
CN114518488A (en) * | 2022-01-19 | 2022-05-20 | 浙江正泰仪器仪表有限责任公司 | Distributed light storage charging and discharging system for high-precision internet-of-things intelligent electric meter based on electric energy parameter search measurement |
CN114897109B (en) * | 2022-07-14 | 2022-09-16 | 山东鑫亚装备制造有限公司 | Mower abnormity monitoring and early warning method |
-
2022
- 2022-12-14 CN CN202211598077.4A patent/CN115600061B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105938578A (en) * | 2016-04-13 | 2016-09-14 | 合肥工业大学 | Large-scale photovoltaic power station equivalent modeling method based on clustering analysis |
CN108828405A (en) * | 2018-06-06 | 2018-11-16 | 西南交通大学 | A kind of electric transmission line fault detection method based on random matrix |
CN113836704A (en) * | 2021-09-06 | 2021-12-24 | 广州大学 | Time series noise reduction method and device based on improved singular spectrum analysis |
Also Published As
Publication number | Publication date |
---|---|
CN115600061A (en) | 2023-01-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116702081B (en) | Intelligent inspection method for power distribution equipment based on artificial intelligence | |
CN115600061B (en) | Inverter zero voltage drop data processing method based on machine learning | |
KR101352204B1 (en) | Apparatus and method for classification of power quality disturbances at power grids | |
CN116703183B (en) | Novel power distribution network system power load assessment method | |
CN117390573B (en) | Wind turbine generator operation abnormality early warning method based on time sequence prediction | |
CN112582978B (en) | Leakage protection method and device and charging pile | |
CN113759206A (en) | Method and system for judging fault type of power distribution network | |
CN113721070A (en) | Residual current detection device and residual current detection method | |
Yılmaz et al. | A new signal processing-based islanding detection method using pyramidal algorithm with undecimated wavelet transform for distributed generators of hydrogen energy | |
Imani et al. | A novel MODWT-based fault detection and classification scheme in VSC-HVDC transmission line | |
CN117491938B (en) | Current transformer detection control method of photovoltaic energy storage system | |
CN112072609B (en) | Method for adjusting action current of leakage protector | |
JP2019158674A (en) | Arc detection device, control method therefor, control program, and DC power supply system | |
CN116780557A (en) | Power distribution network voltage control method and system based on clustering partition algorithm | |
Barreto et al. | Wavelet-based fault detection in grid-connected photovoltaic systems | |
Jose et al. | Smart Grid Data Compression of Power Quality Events using Wavelet Transform | |
CN115021676A (en) | Photovoltaic system direct current arc fault detection method, device, equipment and medium | |
CN110794210B (en) | Method and device for judging voltage harmonic isolation effect, power supply system, computer equipment and storage medium | |
Shang et al. | Efficiency analysis of data compression of power system transients using wavelet transform | |
KR100333305B1 (en) | Method for detecting voltage sag by using wavelet transform | |
CN114123164B (en) | Method and device for calculating short-circuit capacity of power system and terminal equipment | |
CN115808600B (en) | DC arc detection method based on mathematical morphology and pattern recognition | |
Peilin et al. | Power quality detection and discrimination in distributed power system based on wavelet transform | |
CN110048393B (en) | Transient voltage disturbance suppression method and device based on direct current commutation influence degree | |
CN117421610B (en) | Data anomaly analysis method for electric energy meter running state early warning |
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 |