A distribution transformer fault detection method and system

Abstract

The present invention relates to the field of data processing technology, and in particular to a distribution transformer fault detection method and system. The method includes: acquiring data sequences of various position areas on the distribution transformer in real time; determining the noise performance degree of each data point in each data sequence; correcting the noise performance degree of the data point according to the correlation between the data sequence and the preset local range sequence of the corresponding data point in other data sequences in its neighborhood; determining the noise influence degree of each data sequence; determining the number of adaptive segmentation trees when the corresponding position area is detected using the RRCF model according to the noise influence degree of each data sequence; and performing detection through the RRCF model to perform distribution transformer fault detection. By analyzing the noise influence degree of each data sequence, the number of segmentation trees in the adaptive RRCF model is adjusted, thereby saving distribution transformer fault detection computing resources, improving computing efficiency and detection accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a distribution transformer fault detection method and system. Background Art

[0002] As an important equipment in the power system, the technological innovation of distribution transformers has always attracted much attention. In recent years, with the development of smart grids, new energy and other fields, the technological innovation of distribution transformers has also shown a diversified trend. Local area discharge detection of transformers is a common fault detection method. By detecting local area discharge, potential problems can be discovered early, so that measures can be taken to avoid more serious faults. By detecting local area discharge, its operating costs can be reduced, providing a strong guarantee for the stable operation of the power system.

[0003] The patent document with the announcement number CN113469257B discloses a distribution transformer fault detection method and system. The distribution transformer fault detection method adopts a distribution transformer fault detection model based on a robust multi-core extreme learning machine-autoencoder, and obtains a detection result based on the monitoring data acquired in real time during the operation of the distribution transformer, so as to solve the data imbalance problem and the noise interference problem, thereby improving the fault detection rate and reducing the fault false alarm rate.

[0004] However, compared with the detection model in the above patent document, the Robust Random Cut Forest (RRCF) algorithm is more suitable for real-time data streams and has higher computational efficiency. It can quickly detect anomalies in data streams. At the same time, the implementation of the RRCF algorithm is relatively simple and easy to deploy and maintain. For real-time systems, simple implementation means lower maintenance costs.

[0005] The RRCF algorithm is an anomaly detection algorithm for dynamic data volume. The algorithm dynamically maintains the distribution of random segmentation trees on the data stream and judges the degree of anomaly based on the impact of the insertion or deletion of new sample points on the complexity of the segmentation tree. When using RRCF to detect anomalies in the regional discharge data of distribution transformers, since the regional discharge data is time series data, the RRCF algorithm needs to regularly update the model when processing time series data to adapt to data changes. When the model is updated each time, the traditional RRCF model often uses a fixed number of segmentation trees for anomaly detection. However, due to environmental problems around distribution transformers, the noise impact at different times is different. If a fixed number of segmentation trees is used, the RRCF algorithm will not be able to adapt to the changes in transformer local discharge data, resulting in poor anomaly detection results and affecting the accuracy of distribution transformer fault detection. Summary of the invention

[0006] In order to solve the problem that a fixed number of segmentation trees will cause the RRCF algorithm to be unable to adapt to changes in transformer partial discharge data, thereby resulting in poor anomaly detection effect and affecting the accuracy of distribution transformer fault detection, the present invention provides a distribution transformer fault detection method and system.

[0007] In a first aspect, the present invention provides a distribution transformer fault detection method, which adopts the following technical solution:

[0008] A distribution transformer fault detection method comprises: acquiring data sequences of various position areas on the distribution transformer in real time; determining the noise performance degree of each data point in each data sequence by curve fitting for all data points and the next data point within a preset local range of each data point, wherein the noise performance degree is positively correlated with the error during curve fitting; correcting the noise performance degree of the data point according to the correlation between the data sequence and the sequences of the preset local range of the corresponding data points in other data sequences in its neighborhood, wherein the corrected noise performance degree is positively correlated with the noise performance degree of the data point and negatively correlated with the correlation; determining the noise influence degree of each data sequence, wherein the noise influence The degree of noise is positively correlated with the corrected noise performance of each data point in the data sequence; the number of adaptive segmentation trees when the corresponding position area is detected using the RRCF model is determined according to the noise influence degree of each data sequence, the first number of adaptive segmentation trees is positively correlated with the preset number of segmentation trees and the first noise influence degree of the data sequence, and the subsequent number of adaptive segmentation trees is positively correlated with the previous number of adaptive segmentation trees and the ratio of the noise influence degree of the data sequence this time and the previous time; the data sequence of the distribution transformer in each position area is detected by the number of adaptive segmentation trees of the RRCF model to obtain the abnormal value of each position area, so as to detect the fault of the distribution transformer.

[0009] The beneficial effects are: by acquiring data in real time, evaluating and correcting the degree of noise manifestation, and utilizing the number of adaptive segmentation trees of the RRCF model, the abnormalities of the transformer at different locations can be effectively identified and quantified, thereby providing more accurate and reliable fault detection. Therefore, this method improves the accuracy of fault detection by comprehensively considering the degree of noise manifestation and its changing trend in the data sequence; at the same time, this method helps to timely discover potential problems through refined processing and dynamic adjustment, reduce maintenance costs, and improve the operating safety of the transformer.

[0010] Furthermore, the curve fitting is performed using the least square method. During the curve fitting, the abscissa is the time sequence of the data points in each data sequence, and the ordinate is the value of the data point in each data sequence.

[0011] Furthermore, the noise performance degree of each data point in each data sequence satisfies the following relationship:

[0012] ; In the formula, Indicates In the data sequence The noise level of each data point is Indicates In the data sequence The number of data points in the preset local range of data points, Indicates In the data sequence The function of the fitting curve of all data points in the preset local range of data points when the least squares method is used for curve fitting, Indicates In the data sequence The preset local range of data points data points and The function of the line connecting the data points when fitting the curve using the least squares method is: represents the integral symbol, Represents the differential symbol.

[0013] The beneficial effect is that by calculating the difference between the fitting curve and the line connecting the data points, the degree of noise performance of each data point in the data sequence is quantified, thereby helping to identify and process data points with large noise, improving the accuracy of data analysis and modeling, and thus improving the reliability of prediction or analysis results.

[0014] Furthermore, the corrected noise performance level satisfies the following relationship:

[0015] ; In the formula, Indicates In the data sequence The corrected noise performance of the data points is Indicates In the data sequence The noise level of each data point is Indicates The number of data sequences in the neighborhood of a data sequence, Indicates In the data sequence A sequence of preset local ranges of data points, Indicates The first Within the data series and The time sequence of the data points corresponds to the sequence of the preset local range of data points. represents the natural exponential function, Represents the Pearson correlation coefficient function.

[0016] The beneficial effect is that by combining the Pearson correlation coefficient and exponential decay, the corrected noise representation degree can more accurately reflect the noise impact of the data point in its neighborhood, thereby reducing the interference of noise on the detection results, and by considering the correlation of neighborhood data sequences, it helps to more accurately identify abnormal data points, thereby improving the reliability and accuracy of fault detection.

[0017] Furthermore, the noise influence degree of each data sequence satisfies the following relationship:

[0018] ; In the formula, Indicates The degree of noise influence on a data sequence, Indicates The number of data points in a data sequence, Indicates In the data sequence The corrected noise performance of each data point.

[0019] The beneficial effect is that by averaging the degree of corrected noise performance in the entire data sequence, a global noise impact assessment is provided, which can more comprehensively reflect the overall noise level in the data sequence and avoid the deviation of local noise from the overall assessment.

[0020] Furthermore, the number of the first adaptive segmentation trees satisfies the following relationship:

[0021] ; In the formula, Indicates The number of first adaptive segmentation trees corresponding to the position segment of the data sequence, Indicates the number of preset partition trees. Indicates the first The degree of noise influence on a data sequence, Represents the ceiling function.

[0022] The beneficial effects are: the number of segmentation trees is adaptively adjusted according to the degree of noise influence, making the RRCF model more flexible, able to optimize the segmentation strategy according to data characteristics, and adapt to different noise environments; by dynamically adjusting the number of segmentation trees, the model can more accurately capture data features in data sequences with greater noise, and maintain efficient processing in data sequences with less noise, thereby improving overall performance and stability.

[0023] Furthermore, the number of subsequent adaptive segmentation trees each time satisfies the following relationship:

[0024] ; In the formula, Indicates The number of adaptive segmentation trees corresponding to the position segment of the data sequence, Indicates The number of adaptive segmentation trees for the corresponding position segment of the data sequence, Indicates this The degree of noise influence on a data sequence, Indicates the last The degree of noise influence on a data sequence, Represents the ceiling function.

[0025] The beneficial effects are: by dynamically adjusting the number of segmentation trees according to the degree of noise influence, computing resources can be allocated more effectively. Increasing the number of segmentation trees when the noise is high can process data more carefully, while reducing the number of segmentation trees when the noise is low can save computing resources and improve computing efficiency. By adjusting the number of segmentation trees each time, it is possible to avoid using a large number of tree structures in each step, reducing computing time and storage requirements, and improving overall processing efficiency.

[0026] Furthermore, the distribution transformer fault detection includes: in response to an abnormal value being greater than a preset abnormal threshold, determining that the location area is abnormal, and issuing an early warning prompt to complete the distribution transformer fault detection.

[0027] In a second aspect, the present invention provides a distribution transformer fault detection system, which adopts the following technical solution:

[0028] A distribution transformer fault detection system comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned distribution transformer fault detection method is implemented.

[0029] By adopting the above technical solution, the above-mentioned distribution transformer fault detection method is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0030] The present invention has the following technical effects:

[0031] Traditional RRCF often uses a fixed number of segmentation trees. In the partial discharge data of distribution transformers, the surrounding environment and the operating status of the distribution transformer will cause changes in the noise impact. Therefore, if a fixed number of segmentation trees is used, the RRCF model will detect abnormal points as normal or normal points as abnormal. By analyzing the degree of noise impact of each data sequence, the number of segmentation trees in the adaptive RRCF model is reduced when the noise impact is low, saving computing resources and improving computing efficiency; when the noise impact is high, the number of segmentation trees is increased to improve the sensitivity of the model and better capture abnormal signals, significantly improving the efficiency of RRCF algorithm anomaly detection, more accurately identifying anomalies, thereby reducing false alarms and missed alarms and improving detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.

[0033] Figure 1 The present invention is a flow chart of a method for detecting a distribution transformer fault in an embodiment of the present invention.

[0034] Figure 2 It is a coordinate diagram of curve fitting using the least squares method in a distribution transformer fault detection method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0036] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0037] The embodiment of the present invention discloses a distribution transformer fault detection method, referring to Figure 1 , comprising steps S1 to S6:

[0038] S1: Acquire the data sequence of each location area on the distribution transformer in real time.

[0039] It should be noted that, because the discharge of the distribution transformer is accompanied by the release of sound waves, an ultrasonic sensor can be used to capture the sound waves to reflect the discharge condition of the distribution transformer.

[0040] Ultrasonic sensors are installed at equal intervals on the casing of the distribution transformer in order to quickly locate the fault location when an abnormality occurs and to make the calculation more accurate through the relationship between the discharge data at multiple locations.

[0041] Implementers can set the interval and acquisition frequency of ultrasonic sensors according to actual implementation conditions, for example, the interval is 0.3m and the acquisition frequency is 20Hz.

[0042] S2: Determine the degree of noise performance of each data point in each data sequence.

[0043] The noise performance degree of each data point in each data sequence is determined by performing curve fitting on all data points and the next data point within a preset local range of each data point.

[0044] Specifically, the curve fitting is performed using the least square method. During the curve fitting, the abscissa is the time sequence of the data points in each data sequence, and the ordinate is the value of the data point in each data sequence.

[0045] Specifically, the noise performance degree of each data point in each data sequence satisfies the following relationship:

[0046] ;

[0047] In the formula, Indicates In the data sequence The noise level of each data point is Indicates In the data sequence The number of data points in the preset local range of data points, Indicates In the data sequence The function of the fitting curve of all data points in the preset local range of data points when the least squares method is used for curve fitting, Indicates In the data sequence The preset local range of data points data points and The function of the line connecting the data points when fitting the curve using the least squares method is: represents the integral symbol, Represents the differential symbol.

[0048] The implementer can set the length of the local range according to the actual implementation situation, for example, the length of 21 data points centered on the current data point; if the current data point is an edge data point, it will be discarded accordingly. For example, if the current data point is the first data point in the data sequence, the missing part is the left side of it. data points, represents the floor function, then the local range of the data point is the non-missing part, that is, the right side data point and the data point itself.

[0049] Considering that the preset local range lengths of different data points are inconsistent, the mean value of the area difference between the fitting curve and the data point connection line within the preset local range is used to represent the error in fitting. Figure 2 , the error is a straight line , , and curve The enclosed area.

[0050] S3: Correcting the noise performance of the data point according to the correlation between the data sequence and a preset local range sequence of corresponding data points in other data sequences in its neighborhood.

[0051] It should be noted that if an abnormality occurs in a data point, the data sequence will also change dramatically, and its fitting error may also be large, so the noise performance level needs to be corrected. Since the discharge data is detected by an ultrasonic sensor, if an abnormality occurs in a data point, the data points of other adjacent data sequences will also be affected. Therefore, the correlation of the data changes around a point can be used to correct the noise performance level. If the correlation between the data change of a data point and the data change of other data sequences around it is higher, it means that it is more likely to be a real abnormality, so its noise performance level is lower. Among them, the correlation can be expressed by the Pearson correlation coefficient of all data points in a preset local range between the current data point and the data points in other surrounding data sequences that correspond to the time sequence of the data point. The larger the Pearson correlation coefficient, the higher the correlation.

[0052] Specifically, the corrected noise performance level satisfies the following relationship:

[0053] ;

[0054] In the formula, Indicates In the data sequence The corrected noise performance of the data points is Indicates In the data sequence The noise level of each data point is Indicates The number of data sequences in the neighborhood of a data sequence, Indicates In the data sequence A sequence of preset local ranges of data points, Indicates The first Within the data series and The time sequence of the data points corresponds to the sequence of the preset local range of data points. represents the natural exponential function, Represents the Pearson correlation coefficient function.

[0055] S4: Determine the degree of noise influence on each data sequence.

[0056] It should be noted that after obtaining the corrected noise performance degree of each data point in the data sequence, the degree of noise influence on the data sequence can be obtained.

[0057] Specifically, the noise influence degree of each data sequence satisfies the following relationship:

[0058] ;

[0059] In the formula, Indicates The degree of noise influence on a data sequence, Indicates The number of data points in a data sequence, Indicates In the data sequence The corrected noise performance of each data point.

[0060] S5: Determine the number of adaptive segmentation trees when using the RRCF model to detect the corresponding location area according to the degree of noise influence of each data sequence.

[0061] It should be noted that traditional RRCF often uses a fixed number of segmentation trees. In the partial discharge data of distribution transformers, the surrounding environment and the operating status of the distribution transformer will cause changes in the noise impact. Therefore, if a fixed number of segmentation trees is used, the RRCF model will detect abnormal points as normal or normal points as abnormal. By analyzing the degree of noise impact of each data sequence, the number of segmentation trees in the adaptive RRCF model is adjusted. The greater the noise impact, the greater the number of segmentation trees can be increased to improve the sensitivity of the model to better step abnormal data. The lower the noise impact, the smaller the number of segmentation trees can be reduced to save computing resources.

[0062] Specifically, the number of the first adaptive segmentation trees satisfies the following relationship:

[0063] ;

[0064] In the formula, Indicates The number of first adaptive segmentation trees corresponding to the position segment of the data sequence, Indicates the number of preset partition trees. Indicates the first The degree of noise influence on a data sequence, Represents the ceiling function.

[0065] The implementer can set the number of segmentation trees and the number of data points in the segmentation trees according to the actual implementation situation. For example, the number of segmentation trees is 200 and the number of data points in the segmentation trees is 80.

[0066] Specifically, the number of subsequent adaptive segmentation trees each time satisfies the following relationship:

[0067] ;

[0068] In the formula, Indicates The number of adaptive segmentation trees corresponding to the position segment of the data sequence, Indicates The number of adaptive segmentation trees for the corresponding position segment of the data sequence, Indicates this The degree of noise influence on a data sequence, Indicates the last The degree of noise influence on a data sequence, Represents the ceiling function.

[0069] Among them, when When >1, it means that the noise influence becomes greater, so the number of segmentation trees is increased. When <1, it means that the influence of noise becomes smaller, so the number of segmentation trees is reduced. =1, it means that the degree of noise influence has not changed, and the number of segmentation trees remains unchanged.

[0070] Implementers can set the update interval of the adaptive partition tree in the RRCF model according to actual implementation conditions, for example, 3 minutes.

[0071] S6: Detection is performed through the RRCF model to detect distribution transformer faults.

[0072] The data sequence of the distribution transformer in each location area is detected by the number of adaptive segmentation trees of the RRCF model to obtain the abnormal value of each location area in order to perform distribution transformer fault detection.

[0073] Specifically, the distribution transformer fault detection includes:

[0074] In response to the abnormal value being greater than the preset abnormal threshold, the location area is determined to be abnormal, and an early warning prompt is issued to complete the distribution transformer fault detection.

[0075] An embodiment of the present invention further discloses a distribution transformer fault detection system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a distribution transformer fault detection method according to the present invention is implemented.

[0076] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0077] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, or any other medium that can be used to store the required information and can be accessed by an application, module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device.

[0078] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0079] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A distribution transformer fault detection method, characterized in that: include: Obtain data sequences of each location area on the distribution transformer in real time; By performing curve fitting on all data points and the next data point within the preset local range of each data point, the noise performance degree of each data point in each data sequence is determined. The noise performance degree is positively correlated with the error during curve fitting. The noise performance degree of each data point in each data sequence satisfies the following relationship: , where Indicates In the data sequence The noise level of each data point is Indicates In the data sequence The number of data points in the preset local range of data points, Indicates In the data sequence The function of the fitting curve of all data points in the preset local range of data points when the least squares method is used for curve fitting, Indicates In the data sequence The preset local range of data points data points and The function of the line connecting the data points when fitting the curve using the least squares method is: represents the integral symbol, represents a differential sign; the noise performance degree of the data point is corrected according to the correlation between the data sequence and the corresponding data point in other data sequences in its neighborhood within a preset local range, and the corrected noise performance degree is positively correlated with the noise performance degree of the data point and negatively correlated with the correlation; Determine the noise influence degree of each data sequence, and the noise influence degree is positively correlated with the corrected noise performance degree of each data point in the data sequence; determine the number of adaptive segmentation trees when the corresponding position area is detected using the RRCF model according to the noise influence degree of each data sequence, the first adaptive segmentation tree number is positively correlated with the preset segmentation tree number and the first noise influence degree of the data sequence, and the subsequent adaptive segmentation tree number is positively correlated with the previous adaptive segmentation tree number and the ratio of the noise influence degree of the data sequence this time and the previous time; The data sequence of the distribution transformer in each location area is detected by the number of adaptive segmentation trees of the RRCF model to obtain the abnormal value of each location area in order to perform distribution transformer fault detection.

2. A distribution transformer fault detection method according to claim 1, characterized in that: The curve fitting is performed using the least square method. During the curve fitting, the horizontal axis is the time sequence of the data points in each data sequence, and the vertical axis is the value of the data points in each data sequence.

3. A distribution transformer fault detection method according to claim 1, characterized in that: The corrected noise performance level satisfies the following relationship: ； In the formula, Indicates In the data sequence The corrected noise performance of the data points is Indicates In the data sequence The noise level of each data point is Indicates The number of data sequences in the neighborhood of a data sequence, Indicates In the data sequence A sequence of preset local ranges of data points, Indicates The first Within the data series and The time sequence of the data points corresponds to the sequence of the preset local range of data points. represents the natural exponential function, Represents the Pearson correlation coefficient function.

4. A distribution transformer fault detection method according to claim 1, characterized in that: The noise influence degree of each data sequence satisfies the following relationship: ； In the formula, Indicates The degree of noise influence on a data sequence, Indicates The number of data points in a data sequence, Indicates In the data sequence The corrected noise performance of each data point.

5. A distribution transformer fault detection method according to claim 1, characterized in that: The number of the first adaptive segmentation trees satisfies the following relationship: ； In the formula, Indicates The number of first adaptive segmentation trees corresponding to the position segment of the data sequence, Indicates the number of preset partition trees. Indicates the first The degree of noise influence on a data sequence, Represents the ceiling function.

6. A distribution transformer fault detection method according to claim 1, characterized in that: The number of subsequent adaptive segmentation trees each time satisfies the following relationship: ； In the formula, Indicates The number of adaptive segmentation trees corresponding to the position segment of the data sequence, Indicates The number of adaptive segmentation trees for the corresponding position segment of the data sequence, Indicates this The degree of noise influence on a data sequence, Indicates the last The degree of noise influence on a data sequence, Represents the ceiling function.

7. A distribution transformer fault detection method according to claim 1, characterized in that: The distribution transformer fault detection comprises: In response to the abnormal value being greater than the preset abnormal threshold, the location area is determined to be abnormal, and an early warning prompt is issued to complete the distribution transformer fault detection.

8. A distribution transformer fault detection system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a distribution transformer fault detection method according to any one of claims 1 to 7 is implemented.

Images