Fault detection method for power distribution system based on large number theorem and data similarity

Abstract

The application discloses a power distribution system fault detection method based on a large number theorem and data similarity, and comprises the following steps: S1, collecting current data; S2, setting a time step and a sampling interval, and obtaining a current data segment according to the time step and the sampling interval; S3, calculating a distribution model of each current data segment according to the large number theorem; S4, evaluating the similarity of different current data segments through KL divergence calculation, and obtaining a relation matrix; and S5, obtaining an evaluation measure of a power distribution system state according to the relation matrix, and detecting a fault. The application adopts the large number theorem and the data similarity to detect the power distribution system fault, has the characteristics of high accuracy, and is suitable for detecting the power distribution system fault which is difficult to be identified by a traditional method in a smart grid.

Description

Technical Field

[0001] This invention relates to the field of intelligent technology for power grid fault diagnosis, and specifically to a fault detection method for power distribution systems based on the law of large numbers and data similarity. Background Technology

[0002] Power distribution line faults can be classified into three types based on their nature: single-phase grounding, phase-to-phase short circuit, and grounding-phase-to-phase short circuit. Based on their time frame, they can be categorized into transient faults and permanent faults. Transient faults occur when the conditions for a phase-to-phase short circuit are disrupted by an electric arc after discharge, thus preventing the formation of a short circuit. Permanent faults, on the other hand, cannot have their short-circuit conditions disrupted by an electric arc and require human intervention.

[0003] High resistance faults are one of the common faults in power distribution lines. The causes of high resistance faults can be mainly divided into two types: one is that the power distribution line breaks during operation and comes into contact with high impedance; the other is that the normally operating power distribution line breaks and comes into contact with objects around the line. Both of these situations lead to the occurrence of high resistance faults in power distribution lines.

[0004] When a high-resistance fault occurs, the current level is significantly lower than that caused by a short circuit. This has a certain impact on the online fault identification of power distribution lines. Traditional current protection has a low detection rate for this type of fault, so it cannot be adjusted in time, which can lead to more serious faults in the power distribution system, such as line short circuits and fires.

[0005] In summary, traditional detection methods are insufficient to detect high-resistance faults in power distribution systems. Therefore, new methods and techniques are needed to analyze data when the current level does not change significantly in order to detect power distribution system faults and prevent the growth and spread of faults. Summary of the Invention

[0006] The technical problem this invention aims to solve is that traditional current protection is only suitable for short-circuit faults in power distribution systems where the current increases significantly, and has a low detection rate for high-resistance faults generated by the system. New methods and means are urgently needed. This invention provides a power distribution system fault detection method based on the law of large numbers and data similarity. According to the law of large numbers, the change in current value follows a Gaussian distribution. KL divergence is used to calculate the similarity of current values ​​over different time periods. The power distribution system state is obtained by analyzing the relationship between current values ​​over different time periods, thereby detecting faults. This method is characterized by high accuracy and is suitable for detecting power distribution system faults in smart grids that are difficult to identify using traditional methods.

[0007] The technical solution adopted by this invention to solve its technical problem is: a method for detecting faults in a power distribution system based on the law of large numbers and data similarity, comprising the following steps: S1, collecting current data. S2, Set time step and sampling interval According to the time step and sampling interval Obtain current data segment ,in S3. Calculate the distribution model of each current data segment according to the law of large numbers, given the number of current data segments to be calculated. S4. Evaluate different current data segments using KL divergence calculation. Similarity to obtain the relationship matrix S5. Based on the relation matrix The evaluation measure of the power distribution system status is obtained, and one or more of methods 1, 2 and 3 are used to detect faults.

[0008] The current data is discrete data, obtained by measuring the current value of a specific line in the power distribution system over a certain period of time.

[0009] The faults include, but are not limited to, high-resistance faults, single-phase grounding faults, and phase-to-phase open circuits, etc.

[0010] Step S2 includes: S21, setting the time step. and sampling interval The time step refers to the current data segment. Length, sampling interval Refers to the segment of current data. The time interval; S22, based on the time step and sampling interval Current data The data was extracted, and a total of [data points] were extracted. There are several current data segments, and the current data segments are recorded as follows: .

[0011] Step S3 includes: S31, solving the current data segment using the maximum likelihood estimation method. mean vector Covariance Matrix ,Right now S32, Obtain the current data segment The distribution is ,in Represents current data segment The first in Individual values; S33, using KL divergence to calculate different current data segments. and Distribution similarity : S34. Calculate the similarity of all current data segments and construct a relation matrix. .

[0012] The method described in step S5 is: calculating the relation matrix. eigenvalues , and Then calculate the evaluation function. The evaluation measurement results are obtained.

[0013] The eigenvalues , and This includes, but is not limited to, specific elements in a matrix, or the minimum, maximum, or median, or eigenvalues ​​or elements in the eigenvectors of a matrix, or specific constants.

[0014] The evaluation measurement results are obtained through an evaluation function. The value is obtained when If the value falls within a specific range, the corresponding fault is detected.

[0015] The second method described in step S5 is: taking Current data segment during normal operation of the power distribution system and the current data segment to be measured The similarity, if If the value is higher than the threshold, then For a normally functioning power distribution system, if the value is below a threshold... The power distribution system corresponding to the fault.

[0016] The third method described in step S5 is: calculating the actual distribution of current data segments when the power distribution system is operating normally. and distribution function ,calculate and European distance Calculate the actual distribution of the current data segment to be measured. and distribution function European distance Compare the differences between the two When the difference value is below the threshold, the current data segment to be measured corresponds to a normally functioning power distribution system; when the difference value is above the threshold, the current data segment to be measured corresponds to a faulty power distribution system.

[0017] The beneficial effects of this invention are as follows: According to the law of large numbers and the central limit theorem, the distribution of current data follows a Gaussian distribution. The similarity of different current data segments can be calculated using KL divergence, thereby analyzing whether the current change is normal and effectively detecting high-resistance faults that are difficult to identify by traditional current protection methods. Attached Figure Description

[0018] Figure 1This is a flowchart of the power distribution system fault detection method of the present invention.

[0019] Figure 2 This is a schematic diagram of the current data segment obtained by the power distribution system fault detection method of the present invention.

[0020] Figure 3 This is the current data segment of an embodiment of the present invention. , and .

[0021] Figure 4 This is the current data segment of an embodiment of the present invention. Distribution model .

[0022] Figure 5 This is the current data segment of an embodiment of the present invention. Distribution model .

[0023] Figure 6 This is the current data segment of an embodiment of the present invention. Distribution model . Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0025] like Figure 1 As shown, the power distribution system fault detection method based on the law of large numbers and data similarity includes the following steps: S1, collecting current data; S2, setting the time step and sampling interval, and obtaining current data segments according to the time step and sampling interval; S3, calculating the distribution model of each current data segment according to the law of large numbers; S4, evaluating the similarity of different current data segments through KL divergence calculation, and obtaining the relationship matrix; S5, obtaining the evaluation measure of the power distribution system state based on the relationship matrix, and detecting faults.

[0026] like Figure 2 As shown, the collected current data The data is discrete, lasting 4 seconds, and sampled at a frequency of 1 kHz. The time step is set. 0.5s and sampling interval With a sampling interval of 1 second, multiple current data segments can be obtained. It is worth noting that if the sampling interval is less than the time step, the data in these segments will partially overlap, and this method will still be effective in this case.

[0027] As can be seen, a significant feature of this invention is that it addresses the problem of low detection rate of high-resistance faults in traditional current protection. Based on the law of large numbers, it calculates that the change in current value follows a Gaussian distribution, and uses KL divergence to calculate the similarity of current values ​​in different time periods. The power distribution system status is obtained by analyzing the relationship between current values ​​in different time periods, and faults are detected.

[0028] Example

[0029] This embodiment employs a fault detection method based on the law of large numbers and data similarity to... Figure 2 The current data shown An evaluation and measurement were performed to detect high-resistance faults.

[0030] like Figure 3 As shown, set the time step. 0.5s and sampling interval The current data segment is obtained in 1 second. The following is based on Figure 1 The steps S3 to S5 shown are for fault detection.

[0031] First, calculate the current data segment. Distribution model. Current data segment. , and Distribution model , and Each as Figure 4 , Figure 5 and Figure 6 As shown.

[0032] Next, the similarity of different current data segments was evaluated using KL divergence calculation to obtain the relationship matrix. .

[0033] Finally, an assessment measure of the power distribution system's state is obtained based on the relationship matrix, and one or more of methods one, two, and three are used to detect faults. The following sections describe how methods one and two are used to detect faults in the power distribution system.

[0034] Method 1: Calculate the relation matrix eigenvalues , and Then calculate the evaluation function to obtain Since the value of the evaluation function is less than 0.7 (where 0.7 is the threshold of the evaluation function, determined experimentally), the current data is judged... Abnormality; there is a fault in the power distribution system.

[0035] Method 2: Given This is a normal current data segment; the threshold was determined to be 0.5 based on experiments. [Examination / Investigation] If the current is greater than the threshold of 0.5, then To ensure the power distribution system is operating normally, examine... If the current is less than the threshold of 0.5, then the corresponding power distribution system has a fault.

[0036] At this point, the current data is complete. The evaluation measures are used to detect high-resistance faults.

[0037] Based on the preferred embodiments of the present invention described above, those skilled in the art can make various changes and modifications without departing from the technical concept of the invention. The technical scope of this invention is not limited to the contents of the specification, but must be determined by the scope of the claims.

Claims

1. A method for fault detection in power distribution systems based on the law of large numbers and data similarity, characterized in that, This includes the following steps: S1, Acquiring current data ; S2, Set time step and sampling interval According to the time step and sampling interval Obtain current data segment ,in S3. Calculate the distribution model of each current data segment according to the law of large numbers, given the number of current data segments to be calculated. ; S4. Evaluate different current data segments using KL divergence calculation. Similarity to obtain the relationship matrix ; S5. Based on the relation matrix The evaluation measure of the power distribution system status is obtained, and one or more of methods 1, 2 and 3 are used to detect faults.

2. The power distribution system fault detection method as described in claim 1, characterized in that, The current data is discrete data, obtained by measuring the current value of a specific line in the power distribution system over a certain period of time.

3. The power distribution system fault detection method as described in claim 1, characterized in that, The faults include, but are not limited to, high-resistance faults, single-phase grounding faults, and phase-to-phase open circuits, etc.

4. The power distribution system fault detection method as described in claim 1, characterized in that, Step S2 includes: S21, setting the time step. and sampling interval The time step refers to the current data segment. Length, sampling interval Refers to the segment of current data. The time interval; S22, based on the time step and sampling interval Current data The data was extracted, and a total of [data points] were extracted. There are several current data segments, and the current data segments are recorded as follows: .

5. The power distribution system fault detection method as described in claim 1, characterized in that, Step S3 includes: S31, solving the current data segment using the maximum likelihood estimation method. mean vector Covariance Matrix ,Right now S32, Obtain the current data segment The distribution is ,in Represents current data segment The first in Individual values; S33, using KL divergence to calculate different current data segments. and Distribution similarity : S34. Calculate the similarity of all current data segments and construct a relation matrix. .

6. The power distribution system fault detection method as described in claim 1, characterized in that, The method described in step S5 is: calculating the relation matrix. eigenvalues , and Then calculate the evaluation function. The evaluation measurement results are obtained.

7. The power distribution system fault detection method as described in claims 1 and 6, characterized in that, The eigenvalues , and This includes, but is not limited to, specific elements in a matrix, or the minimum, maximum, or median, or eigenvalues ​​or elements in the eigenvectors of a matrix, or specific constants.

8. The power distribution system fault detection method as described in claims 1 and 6, characterized in that, The evaluation measurement results are obtained through an evaluation function. The value is obtained when If the value falls within a specific range, the corresponding fault is detected.

9. The power distribution system fault detection method as described in claim 1, characterized in that, The second method described in step S5 is: taking Current data segment during normal operation of the power distribution system and the current data segment to be measured The similarity, if If the value is higher than the threshold, then For a normally functioning power distribution system, if the value is below a threshold... The power distribution system corresponding to the fault.

10. The power distribution system fault detection method as described in claim 1, characterized in that, The third method described in step S5 is: calculating the actual distribution of current data segments when the power distribution system is operating normally. and distribution function ,calculate and European distance Calculate the actual distribution of the current data segment to be measured. and distribution function European distance Compare the differences between the two When the difference value is below the threshold, the current data segment to be measured corresponds to a normally functioning power distribution system; when the difference value is above the threshold, the current data segment to be measured corresponds to a faulty power distribution system.

Images