Substation secondary circuit fault monitoring method

Abstract

The invention relates to the field of secondary circuit monitoring, in particular to a transformer substation secondary circuit fault monitoring method. The method comprises the following steps: acquiring preprocessed operation data of secondary circuit equipment, constructing an initial isolated tree according to the operation data in a preset time period, inputting the operation data at a future moment beyond the preset time period into the initial isolated tree as a change sample, and calculating the change degree of any node after the change sample is input into the initial isolated tree; the nodes with the change degree larger than a preset change threshold value serve as research nodes, the node change uniformity of the initial isolated tree is calculated according to the change degree of the research nodes, and in response to the situation that the node change uniformity is smaller than a reconstruction threshold value, the initial isolated tree is reconstructed to obtain an updated isolated tree; and calculating and updating an abnormal score of the operation data in the isolated tree based on the isolated forest, and completing fault monitoring. According to the technical scheme, the accuracy of a fault monitoring result can be improved.

Description

Technical Field

[0001] The present invention relates to the field of secondary circuit monitoring, and more specifically, to a method for monitoring secondary circuit faults in a transformer substation. Background Art

[0002] Secondary circuits usually include relay protection circuits, automation control circuits, measurement circuits, communication circuits, etc., which ensure real-time monitoring of substations, fault detection, and the stability of power systems. Although secondary circuits seem to be auxiliary, their healthy operation directly affects the safety of substations and the reliability of power grids. Especially when power equipment fails, secondary circuits can respond quickly through rapid detection and automatic protection, thereby preventing the fault from expanding.

[0003] Isolation forest is a machine learning algorithm for anomaly detection. Its core idea is to "isolate" data points by randomly splitting the data space. Isolation forest builds multiple trees, each of which gradually divides the data into different regions by randomly selecting features and cutting points until all data points are "isolated". Anomalies usually require fewer splitting steps to be isolated, so their path length in the tree is shorter. Compared with traditional distance- or density-based anomaly detection methods, isolation forest is more efficient in processing large-scale data, does not depend on the distribution of data, and is suitable for high-dimensional and sparse data sets.

[0004] The existing Chinese patent application document with publication number CN118554450A discloses a method and device for monitoring the safe operation status of a power grid based on big data, wherein the above application document constructs an isolation forest model for all data slice groups; and calculates the abnormal score of each data slice group in the isolation forest model, and in response to the abnormal score being greater than the set value, the corresponding data slice group is abnormal data, and at this time, the power grid operation status of the data slice group corresponding to the time sequence is abnormal. However, the aging of secondary circuit equipment will increase the electrical contact resistance, resulting in a collective offset of the collected data, which in turn leads to inaccurate abnormal detection results when using the isolation forest for abnormal detection. Summary of the invention

[0005] In order to solve the problem of inaccurate abnormal monitoring results, the present invention proposes a substation secondary circuit fault monitoring method.

[0006] The present invention discloses a method for monitoring faults in secondary circuits of transformer substations, comprising: obtaining pre-processed operating data of secondary circuit equipment, constructing an initial isolated tree according to the operating data in a preset time period, inputting operating data at future moments outside the preset time period as change samples into the initial isolated tree, and calculating the degree of change of any node after the change samples are input into the initial isolated tree; taking nodes whose degree of change is greater than a preset change threshold as research nodes, calculating the node change uniformity of the initial isolated tree according to the degree of change of the research nodes, and in response to the node change uniformity being less than a reconstruction threshold, reconstructing the initial isolated tree to obtain an updated isolated tree; and calculating the abnormal score of the operating data in the updated isolated tree based on the isolation forest to complete fault monitoring.

[0007] The initial isolation tree is constructed using the operating data within the preset period, and the degree of node change is evaluated by inputting the change samples at future moments, so as to monitor the changes of the system in real time. When the uniformity of node change is lower than the set reconstruction threshold, the reconstruction of the isolation tree is automatically triggered to ensure that the tree structure can reflect the latest system dynamics and avoid the decrease of monitoring accuracy due to the change of data distribution.

[0008] Preferably, the preprocessing is to fill in the missing running data using a mean filling method.

[0009] Preferably, the degree of change satisfies the relationship: , Represents the initial isolated tree Middle Layer Node The degree of change, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The number of running data, Represents the initial isolated tree Middle The total number of nodes in the layer, Represents the normalization function.

[0010] It can not only measure the change in the operating data of a single node before and after the change, but also compare the changes of different nodes to ensure that the monitoring of the equipment status has high sensitivity and accuracy. By quantifying the degree of change, it helps to discover potential operating anomalies, thereby providing protection for subsequent fault monitoring and improving the accuracy of monitoring results.

[0011] Preferably, the degree of change also satisfies the relationship: , Represents the initial isolated tree Middle Layer Node The degree of change, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The number of running data, Represents the initial isolated tree Middle The total number of nodes in the layer, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The variance of the running data, Represents the normalization function.

[0012] By further introducing the consideration of data variance, the degree of change of each node in the initial isolated tree is comprehensively evaluated, which not only measures the relative change in the running data of the node before and after the change, but also combines the change in data volatility.

[0013] Preferably, the node change uniformity includes: for any initial isolated tree, calculating the mean change degree of all research nodes in the initial isolated tree, respectively calculating the absolute difference between the change degree of each research node in the initial isolated tree and the mean change degree, and using the cumulative value of all absolute differences through negative correlation mapping as the node change uniformity. This method can effectively reflect the uniformity of changes in each node in the tree structure.

[0014] By calculating the mean of the degree of change of all research nodes, and then evaluating the degree of deviation of each node change, if the degree of change of a node is far from the mean, the corresponding absolute difference is large. Finally, these differences are converted into the impact on the uniformity of node change through negative correlation mapping.

[0015] Preferably, the node change uniformity also satisfies the relationship: , Represents the initial isolated tree The uniformity of node changes, Represents the initial isolated tree Research Node The depth of the layer, Represents the initial isolated tree The average layer depth of all research nodes in , Represents the initial isolated tree The total number of research nodes, Represents the initial isolated tree Research Node The degree of change, Represents the initial isolated tree The mean change degree of all research nodes in represents the normalization function, Represents an exponential function.

[0016] The mean deviation of node depth is calculated by normalization to reflect the distribution characteristics of nodes in the tree structure. Then, the mean deviation of node change degree is weighted by exponential function to enhance the influence of nodes with uneven change degree on uniformity.

[0017] Preferably, the reconstructing the initial isolated tree to obtain the updated isolated tree includes: using the change samples input into the initial isolated tree at the rightmost side in the time sequence as the running data in the updated preset time period to construct the updated isolated tree.

[0018] Preferably, completing fault monitoring includes: setting an abnormality threshold, and generating and sending an alarm signal in response to an average abnormality score of the operating data in a continuous preset time period being greater than the abnormality threshold.

[0019] Beneficial effects of the present invention: The present invention can flexibly adapt to data changes and detect possible anomalies or faults in the loop by constructing an initial isolated tree and introducing change samples in future periods. By calculating the degree of change and uniformity of change of each node and reconstructing in response to nodes with excessive changes, the system can effectively update the isolated tree model, thereby improving the accuracy and sensitivity of fault detection. In addition, the method also sets an abnormality score threshold to ensure that when a fault occurs, an alarm signal can be automatically triggered, helping operation and maintenance personnel to respond quickly, reducing the risk of outages and fault expansion, and improving the reliability and safety of substations. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, 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-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 The present invention is a flowchart of a method for monitoring secondary circuit faults in a substation according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] 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.

[0022] 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.

[0023] The present invention provides a method for monitoring secondary circuit faults in a substation. Figure 1 As shown, a substation secondary circuit fault monitoring method includes steps S1 to S3, which are described in detail below.

[0024] S1, obtain the pre-processed operating data of the secondary circuit equipment, build an initial isolated tree based on the operating data in the preset time period, input the operating data at future times outside the preset time period as change samples into the initial isolated tree, and calculate the degree of change of any node after the change sample is input into the initial isolated tree.

[0025] It should be noted that in a substation, the secondary circuit refers to the auxiliary circuit related to the control, protection, monitoring, dispatching and other functions of the power equipment, which usually cannot directly carry power transmission. The secondary circuit data refers to the voltage, current, temperature and other parameters in the circuit. By monitoring the secondary circuit data in the substation, potential fault risks can be discovered in time, and measures can be taken in advance to prevent major accidents.

[0026] In one embodiment, a remote terminal unit (RTU) system is used to collect the initial operation data of the secondary circuit, and the preset collection frequency is 1 Hz. Preferably, the initial operation data includes voltage, current and temperature.

[0027] The initial running data is preprocessed to obtain preprocessed running data, and the preprocessing is to fill the missing initial running data with a mean filling method.

[0028] It should be noted that the secondary circuit equipment of the substation, such as relays, protection devices, control equipment and communication equipment, usually needs to operate around the clock. Long-term high-load operation will lead to heat accumulation in electrical components and lines, thereby accelerating the aging process of the equipment.

[0029] However, the aging of equipment may cause the collected data to shift. For example, the aging of equipment such as cables, terminal blocks, and relays will increase the electrical contact resistance, which in turn causes the temperature to rise. This will cause the collected operating data to shift collectively, resulting in inaccurate anomaly detection results when using the isolation forest algorithm for anomaly detection.

[0030] Exemplarily, the preset time period is set to 10 minutes. An initial isolation tree is constructed according to the operation data in the preset time period.

[0031] It should be noted that, since the initial isolated tree has been determined, the segmentation threshold of each layer in the initial isolated tree has been determined, so the input of the changed sample will not generate new nodes in the isolated tree.

[0032] The operation data of future moments outside the preset time period are input as change samples into the initial isolated tree. For example, if the preset time period is 10 minutes, the operation data after the 10th minute is used as the change sample. The change degree of any node after the change sample is input into the initial isolated tree is calculated, and the change degree satisfies the relationship: , Represents the initial isolated tree Middle Layer Node The degree of change, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The number of running data, Represents the initial isolated tree Middle The total number of nodes in the layer, Represents the normalization function.

[0033] The above calculation method can not only measure the change in the operating data of a single node before and after the change, but also compare the changes of different nodes to ensure that the monitoring of the equipment status has high sensitivity and accuracy. By quantifying the degree of change, it is helpful to discover potential operating anomalies, thereby providing protection for subsequent fault monitoring and improving the accuracy of monitoring results.

[0034] In one embodiment, the degree of change also satisfies the relationship: , Represents the initial isolated tree Middle Layer Node The degree of change, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The number of running data, Represents the initial isolated tree Middle The total number of nodes in the layer, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The variance of the running data, Represents the normalization function.

[0035] By further introducing the consideration of data variance, the degree of change of each node in the initial isolated tree is comprehensively evaluated, which not only measures the relative change in the running data of the node before and after the change, but also combines the change in data volatility.

[0036] One measures the change in the number of node data, and the other measures the change in the variance of node operation data, which can more comprehensively reflect the dynamic changes of nodes. This not only captures the changes in the number of data, but also takes into account the changes in the amplitude of data fluctuations, enhances the sensitivity and accuracy of node changes, and helps to more accurately identify abnormal fluctuations and potential risks in the operation status of equipment.

[0037] S2, taking nodes whose degree of change is greater than a preset change threshold as research nodes, calculating the node change uniformity of the initial isolated tree according to the degree of change of the research nodes, and in response to the node change uniformity being less than a reconstruction threshold, reconstructing the initial isolated tree to obtain an updated isolated tree.

[0038] In one embodiment, a research node is obtained. Exemplarily, the preset change threshold is set to 0.7.

[0039] The node change uniformity of the initial isolated tree is calculated according to the change degree of the research node. The node change uniformity includes: for any initial isolated tree, calculating the mean change degree of all research nodes in the initial isolated tree, calculating the absolute difference between the change degree of each research node in the initial isolated tree and the mean change degree, and using the cumulative value of all absolute differences through negative correlation mapping as the node change uniformity.

[0040] It should be noted that with the continuous input of change samples, the node characteristics in the initial isolated tree will change. If the degree of change of each node in the initial isolated tree is large, it means that there is a fault in the secondary circuit equipment at this time, resulting in large fluctuations in the data points; if the nodes with large changes in the isolated tree are concentrated in a certain part of the isolated tree, it means that the collected data drifts due to equipment aging. At this time, it means that the initial isolated tree no longer meets the data distribution characteristics and the isolated tree needs to be reconstructed.

[0041] Exemplarily, in response to the node change uniformity being less than a reconstruction threshold, the reconstruction threshold is set to 0.4, and the initial isolated tree is reconstructed to obtain an updated isolated tree, including: using the change sample input into the initial isolated tree on the rightmost side in chronological order as the operating data in the updated preset time period to construct an updated isolated tree.

[0042] In one embodiment, the node change uniformity also satisfies the relationship: , Represents the initial isolated tree The uniformity of node changes, Represents the initial isolated tree Research Node The depth of the layer, Represents the initial isolated tree The average layer depth of all research nodes in , Represents the initial isolated tree The total number of research nodes, Represents the initial isolated tree Research Node The degree of change, Represents the initial isolated tree The mean change degree of all research nodes in represents the normalization function, Represents an exponential function.

[0043] By normalizing and calculating the mean deviation of the node layer depth, we can reflect the distribution characteristics of the nodes in the tree structure, and then use the exponential function to weight the mean deviation of the node change degree, which enhances the impact of nodes with uneven change degrees on uniformity. This can not only measure the uniformity of the node layer depth, but also comprehensively consider the amplitude and volatility of node changes, help identify those nodes that have changed significantly or unevenly after the change, and effectively reflect the overall change uniformity of the initial isolated tree.

[0044] S3, based on the isolation forest calculation, updates the anomaly score of the running data in the isolation tree to complete fault monitoring.

[0045] In one embodiment, the anomaly score of the running data in the isolation tree is updated based on the isolation forest calculation. The calculation of the anomaly score is a well-known technology for those skilled in the art and will not be described in detail here.

[0046] An abnormality threshold is set, and in response to the average abnormality score of the running data in a continuous preset period being greater than the abnormality threshold, an alarm signal is generated and sent.

[0047] Exemplarily, the anomaly threshold is set to 0.8.

[0048] 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.

[0049] 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 method for monitoring secondary circuit faults in a substation, characterized in that: include: Obtain the pre-processed operating data of the secondary circuit equipment, build an initial isolated tree based on the operating data in a preset period, input the operating data at future times outside the preset period as a change sample into the initial isolated tree, and calculate the degree of change of any node after the change sample is input into the initial isolated tree; Nodes whose degree of change is greater than a preset change threshold are taken as research nodes, and the node change uniformity of the initial isolated tree is calculated according to the degree of change of the research nodes. In response to the node change uniformity being less than a reconstruction threshold, the initial isolated tree is reconstructed to obtain an updated isolated tree; Based on the isolation forest calculation, the anomaly score of the running data in the isolation tree is updated to complete the fault monitoring.

2. A substation secondary circuit fault monitoring method according to claim 1, characterized in that: The preprocessing is to fill in the missing running data using the mean filling method.

3. A substation secondary circuit fault monitoring method according to claim 1, characterized in that: The degree of change satisfies the relationship: , Represents the initial isolated tree Middle Layer Node The degree of change, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The number of running data, Represents the initial isolated tree Middle The total number of nodes in the layer, Represents the normalization function.

4. A substation secondary circuit fault monitoring method according to claim 1, characterized in that: The degree of change also satisfies the relationship: , Represents the initial isolated tree Middle Layer Node The degree of change, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The number of running data, Represents the initial isolated tree Middle The total number of nodes in the layer, and They represent the initial isolated tree before and after the sample is changed and input into the initial isolated tree. Middle Layer Node The variance of the running data, Represents the normalization function.

5. A substation secondary circuit fault monitoring method according to claim 1, characterized in that: The node change uniformity includes: For any initial isolated tree, the mean change degree of all research nodes in the initial isolated tree is calculated, and the absolute difference between the change degree of each research node in the initial isolated tree and the mean change degree is calculated respectively. The cumulative value of all absolute differences is used as the result of negative correlation mapping as the node change uniformity.

6. A substation secondary circuit fault monitoring method according to claim 1, characterized in that: The node change uniformity also satisfies the relationship: , Represents the initial isolated tree The uniformity of node changes, Represents the initial isolated tree Research Node The depth of the layer, Represents the initial isolated tree The average layer depth of all research nodes in , Represents the initial isolated tree The total number of research nodes, Represents the initial isolated tree Research Node The degree of change, Represents the initial isolated tree The mean change degree of all research nodes in represents the normalization function, Represents an exponential function.

7. A substation secondary circuit fault monitoring method according to claim 1, characterized in that: The reconstruction of the initial isolated tree to obtain an updated isolated tree comprises: The change samples that are input to the rightmost of the initial isolation tree in chronological order are used as the running data in the preset period after the update to construct the updated isolation tree.

8. A substation secondary circuit fault monitoring method according to claim 1, characterized in that: The completion of fault monitoring comprises: An abnormality threshold is set, and in response to the average abnormality score of the running data in a continuous preset period being greater than the abnormality threshold, an alarm signal is generated and sent.

Images