Unbalance degree characteristic analysis-based line breakage fault positioning method

Abstract

The invention provides a line breakage fault positioning method based on unbalance degree characteristic analysis, and the method specifically comprises the steps: constructing a target voltage data sample based on historical open-phase alarm data and a historical fault breakage event; based on an unbalance degree index, extracting an unbalance degree feature and correspondingly correcting a state label of the target voltage data sample, and constructing a sample data set; constructing a fault diagnosis model, and training the fault diagnosis model based on the sample data set; and when the open-phase alarm signal is collected, obtaining a fault positioning result according to the corresponding real-time line data in combination with the fault diagnosis model. According to the method, after the target voltage data sample is obtained, the unbalance degree feature of the target voltage data sample is further extracted, the corresponding state label is corrected through the extracted unbalance degree feature, and then the corresponding sample data set is constructed, so that the deviation between the training sample label and the reality is reduced, and the accuracy of subsequent fault positioning is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid fault location, and in particular to a line disconnection fault location method based on imbalance characteristic analysis. Background Art

[0002] Line disconnections not only cause regional power outages but can also trigger secondary disasters such as equipment damage. Therefore, rapid and accurate fault location is crucial for ensuring stable power system operation. Fault location technology based on electrical parameters, with its advantages of non-contact detection and strong real-time performance, is widely used in power system operation and maintenance. Therefore, existing technologies often use historical data to extract characteristics of electrical parameters such as voltage and current. These characteristics are then used to construct fault diagnosis models using machine learning algorithms to achieve rapid and automated location of line disconnections.

[0003] To ensure the model can efficiently learn the mapping between electrical parameter characteristics and line faults, it is often necessary to add specific labels to the model's training samples. The accuracy of the label content directly affects the recognition accuracy of the trained model. However, existing electrical data samples are mostly labeled based on existing fault records or fixed thresholds. Electrical parameters are generally subject to dynamic changes. Labels based on simple fault conclusions or thresholds often deviate from reality and fail to truly reflect the actual state of power grid operation. Deviations in sample labels directly affect the accuracy of subsequent fault location, making it difficult to ensure the stability and safety of power grid operation. Summary of the Invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art in which the training sample labels used for fault location using electrical parameters and machine learning algorithms deviate from the actual ones, resulting in low fault location accuracy. A line break fault location method based on imbalance feature analysis is provided. After obtaining the target voltage data sample, its imbalance feature is further extracted, and the corresponding state label is corrected by the extracted imbalance feature. Then, the corresponding sample data set is constructed to reduce the deviation between the training sample label and the actual one, thereby ensuring the accuracy of subsequent fault location.

[0005] The purpose of the present invention is achieved through the following technical solutions: The line disconnection fault location method based on imbalance characteristic analysis includes: Construct target voltage data samples based on historical phase loss alarm data and historical fault disconnection events; Based on the imbalance index, the imbalance features are extracted and the state labels of the target voltage data samples are corrected accordingly to construct a sample data set; Build a fault diagnosis model and train it based on the sample data set; When a phase loss alarm signal is collected, the fault location result is obtained based on the corresponding real-time line data combined with the fault diagnosis model.

[0006] Furthermore, the target voltage data sample is constructed based on the historical phase loss alarm data and historical fault disconnection events, including: Filter data extraction lines based on historical phase loss alarm data; Obtain the distribution transformer terminal equipment information of the data extraction line, and correspondingly retrieve the three-phase voltage data of each distribution transformer terminal; The association relationship between the distribution transformer terminal and each historical fault disconnection event is obtained. Based on the corresponding timestamp, the corresponding status label and associated event information are added to the three-phase voltage data of each distribution transformer terminal to construct the target voltage data sample.

[0007] Furthermore, the target voltage data sample is constructed based on the historical phase loss alarm data and historical fault disconnection events, further comprising: According to the topology information of each distribution transformer terminal, the distribution transformer terminals with associated relationships are combined to obtain a distribution transformer terminal group; Constructing a three-phase voltage data group based on the three-phase voltage data of each distribution transformer terminal in the distribution transformer terminal group; Constructing a corresponding three-phase voltage data group sequence according to the corresponding three-phase voltage data group in chronological order; Based on the corresponding timestamps and the association relationship between each distribution transformer terminal in the distribution transformer terminal group and each historical fault disconnection event, the status label and associated event information are added to the corresponding three-phase voltage data group to construct the target voltage data sample.

[0008] Furthermore, the method of extracting the imbalance feature based on the imbalance index and correspondingly correcting the state label of the target voltage data sample to construct a sample data set includes: Based on the imbalance index, the imbalance index value of each target voltage data sample is calculated according to the corresponding three-phase voltage data; Based on the imbalance index value, extract the three-phase imbalance characteristics of each target voltage data sample; Based on the corresponding associated event information and combined with the three-phase imbalance characteristics, the state label of the corresponding target voltage data sample is modified; A sample data set is constructed based on the three-phase imbalance features extracted from each target voltage data sample and the corrected state labels.

[0009] Furthermore, the state label of the corresponding target voltage data sample is corrected based on the corresponding associated event information and in combination with the three-phase imbalance feature, including: Construct a time window based on the corresponding associated event information; Based on the time window, identify the change trend of the three-phase imbalance characteristics of each target voltage data sample; Correct the corresponding status label according to the changing trend of the three-phase imbalance characteristic.

[0010] Furthermore, the construction of the fault diagnosis model and training of the fault diagnosis model based on the sample data set include: Based on the sample data set, the correlation between the three-phase imbalance characteristics and the fault disconnection event is calculated; Combined with the corresponding correlation degree, the three-phase imbalance characteristics and the corresponding corrected state labels are selected from the sample data set to generate a training data set; Based on the decision tree algorithm, a fault diagnosis model is constructed and trained using a training data set.

[0011] Furthermore, the method of combining the corresponding correlation levels and screening the three-phase imbalance characteristics and the corresponding corrected state labels from the sample data set to generate a training data set includes: Based on a preset threshold, the three-phase unbalance characteristics are screened as input variables of the fault diagnosis model; The eigenvalues ​​of the screened three-phase unbalanced features and the corresponding corrected state labels are used as training samples to construct a training data set.

[0012] Furthermore, the three-phase imbalance index includes at least three-phase imbalance, positive sequence voltage component and negative sequence voltage component.

[0013] Furthermore, the imbalance characteristics at least include a time domain characteristic of each imbalance indicator.

[0014] Furthermore, the fault location result includes at least the location of the faulty device and the corresponding three-phase unbalance characteristic value.

[0015] The beneficial effects of the present invention are: (1) Target voltage data samples are constructed based on historical phase loss alarm data and historical fault disconnection events. This systematically integrates multi-source historical data, fully explores normal operation information and fault information under different working conditions, and effectively expands the diversity and quantity of data samples. At the same time, features are extracted and state labels are corrected based on the imbalance index. Then, a sample data set is constructed based on the extracted features and the corrected state labels, making the constructed sample data set more consistent with the actual operating conditions of the corresponding line. The optimized sample data set provides more comprehensive and accurate training data for the fault diagnosis model, improves the model's generalization ability in complex power grid environments, reduces the probability of misjudgment and missed judgment caused by sample defects, ensures the accuracy of fault location, and ensures the stability and safety of power grid operation.

[0016] (2) During the correction process, the dynamic changes of the imbalance characteristics before and after the fault occur are taken into consideration, and the status label content is adjusted so that the corrected status label can more accurately describe the actual operating status of the distribution transformer. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below with reference to the accompanying drawings and examples.

[0019] Example: When a line breaks, the three-phase voltage balance on both sides of the fault point is immediately disrupted, and the voltage difference between the healthy phase and the faulty phase increases significantly. The three-phase voltage imbalance, as a direct indicator of the degree of asymmetry in the three-phase system, can sensitively reflect this voltage difference in real time. In the actual power grid operating environment, factors such as equipment switching may cause fluctuations in electrical parameters, but these fluctuations typically do not cause significant changes in the three-phase voltage imbalance. Therefore, this example uses the three-phase voltage imbalance characteristic as a parameter basis, combined with a machine learning algorithm to achieve fault location.

[0020] However, when a fault occurs, the change in the three-phase voltage imbalance has obvious timing characteristics. At the moment of the fault, the three-phase voltage imbalance will suddenly change, and as the fault persists, this imbalance will remain at a high level or show a specific change trend. If the status label of the corresponding voltage data sample is set based on a simple fault conclusion or threshold, it is easy for the label to deviate from the actual situation, affecting the subsequent fault location accuracy.

[0021] Based on this, this embodiment proposes a line disconnection fault location method based on imbalance characteristic analysis, such as Figure 1 As shown, including: Construct target voltage data samples based on historical phase loss alarm data and historical fault disconnection events; Based on the imbalance index, the imbalance features are extracted and the state labels of the target voltage data samples are corrected accordingly to construct a sample data set; Build a fault diagnosis model and train it based on the sample data set; When a phase loss alarm signal is collected, the fault location result is obtained based on the corresponding real-time line data combined with the fault diagnosis model.

[0022] In order to improve the pertinence of the constructed fault diagnosis model, historical phase loss alarm data is obtained to determine the line where the line disconnection fault has occurred. Voltage data of the line is extracted, and the extracted data is further processed in combination with historical fault disconnection events to construct the initial target voltage data sample for fault diagnosis model training.

[0023] Specifically, the target voltage data sample is constructed based on historical phase loss alarm data and historical fault disconnection events, including: Filter data extraction lines based on historical phase loss alarm data; Obtain the distribution transformer terminal equipment information of the data extraction line, and correspondingly retrieve the three-phase voltage data of each distribution transformer terminal; The association relationship between the distribution transformer terminal and each historical fault disconnection event is obtained. Based on the corresponding timestamp, the corresponding status label and associated event information are added to the three-phase voltage data of each distribution transformer terminal to construct the target voltage data sample.

[0024] In the field of power system fault diagnosis, the accuracy and relevance of fault diagnosis models depend on high-quality training data that closely matches actual fault scenarios. Therefore, by mining fault information from historical data, accurately locating the faulty line, and extracting voltage data closely related to the fault, we provide representative training samples for the fault diagnosis model.

[0025] Historical phase loss alarm data is an important clue for screening faulty lines. During the operation of the power system, when a line is disconnected, a phase loss alarm is usually triggered. These alarms record key factors such as the time and location of the fault. By screening and analyzing them, the line where the line disconnection fault occurred can be determined.

[0026] In this embodiment, three-phase imbalance is selected as the primary electrical parameter for line break fault diagnosis. The three-phase voltage data from the distribution transformer terminal equipment precisely reflects changes in the line's three-phase imbalance. Therefore, after identifying the faulty line, the three-phase voltage data from the distribution transformer terminal equipment is extracted to construct the voltage data samples required for fault diagnosis model training.

[0027] In order to ensure the accuracy of the fault diagnosis model in identifying line break faults, the distribution transformer terminal is further associated with each historical fault line break event, and corresponding status labels and associated event information are added to the corresponding three-phase voltage data based on the timestamp, so that each set of voltage data is corresponded to a specific fault scenario, so that the constructed target voltage data sample can truly reflect the power system status when the fault occurs. In the subsequent fault diagnosis model training process, the true mapping relationship between the line break fault and the three-phase voltage data can be learned, thereby improving the accuracy and pertinence of the diagnosis of the line break fault.

[0028] By comparing the location information of distribution transformer terminals with the locations of historical fault disconnection events and combining the topological relationships of the corresponding devices, an association between the distribution transformer terminals and each historical fault disconnection event can be established. Based on the timestamp, the three-phase voltage data of each distribution transformer terminal is matched with the corresponding historical fault disconnection event. For voltage data collected within the fault occurrence time range, corresponding status labels (such as "normal operation before disconnection," "abnormal state during disconnection," and "normal operation after disconnection") are added based on the fault type and status, along with associated event information such as event number and fault cause. The three-phase voltage data with these status labels and associated event information is then organized and stored to form a complete target voltage data sample for subsequent fault diagnosis model training.

[0029] Considering that distribution transformer terminals in power systems do not exist in isolation, but are connected through lines to form a complex topological structure, this topological relationship determines the close electrical connection and mutual influence between the terminals. The three-phase voltage data of a single distribution transformer terminal can only reflect the local line status, while the data of multiple related distribution transformer terminals can show the overall operation of the regional power line from a more macro perspective. Therefore, based on historical phase loss alarm data and historical fault disconnection events, the target voltage data sample is constructed, which also includes: According to the topology information of each distribution transformer terminal, the distribution transformer terminals with associated relationships are combined to obtain a distribution transformer terminal group; Constructing a three-phase voltage data group based on the three-phase voltage data of each distribution transformer terminal in the distribution transformer terminal group; Constructing a corresponding three-phase voltage data group sequence according to the corresponding three-phase voltage data group in chronological order; Based on the corresponding timestamps and the association relationship between each distribution transformer terminal in the distribution transformer terminal group and each historical fault disconnection event, the status label and associated event information are added to the corresponding three-phase voltage data group to construct the target voltage data sample.

[0030] The topological information of distribution transformer terminals reflects their location and connectivity within the power network. For example, distribution transformer terminals on the same feeder or power branch are electrically interconnected. A fault near one terminal will affect the voltage data of other connected terminals. By analyzing topological information, these connected terminals are grouped into distribution transformer terminal groups, ensuring that the data within the group is inherently physically connected. For example, using a breadth-limited search or depth-first search algorithm, starting with one distribution terminal, the topological information is used to search for distribution transformer terminals that are electrically connected to it, thereby constructing a distribution transformer terminal group. Based on this data, three-phase voltage data sets and sequences are constructed to more accurately reflect the overall operational status changes of the regional power lines before and after the fault.

[0031] Furthermore, by constructing a three-phase voltage data set, the dispersed three-phase voltage data within the distribution transformer terminal group is integrated to form a data set that reflects the overall voltage status of the region. Different distribution transformer terminals are interconnected in terms of geographic location and connectivity, and their voltage data changes are inherently linked. This integrated data set can more comprehensively reflect the operating status of the regional lines.

[0032] Moreover, when a fault occurs, the line voltage change is not instantaneous, but rather an evolutionary process. Therefore, a three-phase voltage data set sequence is constructed in chronological order to simulate the operation process of the power system over time, so that the data has dynamic characteristics in the time dimension.

[0033] Furthermore, based on the timestamps and the association between distribution transformer terminals and historical fault disconnection events, status labels and associated event information are added to the three-phase voltage data sets. This aligns the data set sequences with actual fault scenarios, enabling the fault diagnosis model to learn not only the voltage characteristics at the time of the fault, but also the dynamic patterns of fault occurrence and development, improving the model's diagnostic capabilities for complex fault scenarios.

[0034] Taking the fault time range as an example, the three-phase voltage data group can be added with corresponding status labels, such as "close to the fault point - fault occurring - abnormal state", "affected area - post-fault recovery stage - normal operation", etc., based on the association between each distribution transformer terminal in the distribution transformer terminal group and the fault event, such as the distance from the fault point and the degree of impact of the fault. The corresponding associated event information, such as the fault event number, fault type, and fault cause, can be added.

[0035] By constructing the distribution transformer terminal group as described above and aggregating the three-phase voltage data of each distribution transformer terminal within the distribution transformer terminal group to build a three-phase voltage data group and sequence, the overall operating characteristics and fault evolution patterns of the line in the time dimension can be effectively captured. Furthermore, historical fault disconnection events are combined to give the data contextual meaning, thereby providing more systematic and dynamic training samples for the fault diagnosis model.

[0036] The target voltage data samples are constructed by respectively combining the three-phase voltage data corresponding to a single distribution transformer terminal and a distribution transformer terminal group consisting of multiple associated distribution transformer terminals to improve the integrity of the data samples and thus improve the generalization ability of the subsequent fault diagnosis model.

[0037] And considering that the status labels constructed above are directly constructed through historical fault disconnection events, there may be situations where they cannot accurately reflect the actual operating conditions. Therefore, while extracting features from the three-phase voltage data, the status labels are further corrected according to the extracted features to construct a sample data set that is more in line with reality and ensure the accuracy of the subsequent fault diagnosis model.

[0038] Specifically, the method of extracting the imbalance feature based on the imbalance index and correspondingly correcting the state label of the target voltage data sample to construct a sample data set includes: Based on the imbalance index, the imbalance index value of each target voltage data sample is calculated according to the corresponding three-phase voltage data; Based on the imbalance index value, extract the three-phase imbalance characteristics of each target voltage data sample; Based on the corresponding associated event information and combined with the three-phase imbalance characteristics, the state label of the corresponding target voltage data sample is modified; A sample data set is constructed based on the three-phase imbalance features extracted from each target voltage data sample and the corrected state labels.

[0039] The imbalance indexes in this embodiment include at least three-phase imbalance, positive-sequence voltage component, and negative-sequence voltage component.

[0040] For the set imbalance index, the corresponding index value is calculated based on the corresponding three-phase voltage data.

[0041] Specifically, according to the symmetrical component theory, the three-phase voltage is decomposed into zero sequence , positive sequence and negative sequence Three components: ; ; ; Among them, a=1∠120° is the rotation factor of the symmetric coordinate transformation, V + is the positive sequence voltage component; V a is the voltage of phase A; aV b To set the B phase voltage V b The quantity after the rotation phase transformation according to the rotation factor a; a 2 =1∠240°, which is another rotation factor obtained based on the rotation factor a. 2 V c To set the C phase voltage V c According to another rotation factor a 2 The quantity after rotating phase transformation; V - is the negative sequence voltage component; a 2 V b To set the B phase voltage V b According to another rotation factor a 2 The quantity after phase transformation; aV c To set the C phase voltage V c The quantity after phase transformation according to the rotation factor a.

[0042] Based on the positive sequence voltage amplitude and negative sequence voltage amplitude, the three-phase imbalance can be further calculated. , and its calculation formula is: .

[0043] Based on the above imbalance index, the imbalance index value of each target voltage data sample is calculated according to the corresponding three-phase voltage data, and then the corresponding imbalance feature is extracted.

[0044] The imbalance characteristics in this embodiment include at least the time domain characteristics of each imbalance indicator, such as peak value, variance and slope, so as to reflect the variation characteristics of the three-phase voltage imbalance in the time dimension.

[0045] Among them, the positive sequence voltage amplitude and the negative sequence voltage amplitude are calculated respectively according to the positive sequence voltage component and the negative sequence voltage component: ; ; in, represents the amplitude of the positive sequence voltage component; represents the square of the real part of the positive sequence voltage; represents the square of the imaginary part of the positive sequence voltage, It represents the negative sequence voltage amplitude; represents the square of the real part of the negative sequence voltage; It represents the square of the imaginary part of the negative sequence voltage.

[0046] Determine the corresponding positive sequence voltage peak and negative sequence voltage peak according to the amplitude calculation results: ; ; in, is the peak value of the positive sequence voltage component; max is the maximum value function, It is the highest voltage value that the positive sequence voltage can reach during the entire observation period; is the instantaneous amplitude of the positive sequence voltage component at any time t, It represents the peak value of the negative sequence voltage component; It is the highest voltage value that the negative sequence voltage can reach during the entire change cycle; is the instantaneous amplitude of the negative sequence voltage component at any time t.

[0047] Then, the maximum value of the voltage unbalance, that is, the peak value of the three-phase unbalance, is determined through the positive-sequence voltage peak value and the negative-sequence voltage peak value.

[0048] Calculate the variance of the positive sequence voltage component and the negative sequence voltage component according to the amplitude calculation results: ; ; in, is the variance of the positive sequence voltage; is the amplitude of the positive sequence voltage measured at different times t; for The average value of is the number of data points, For one of the data points, is the variance of the negative sequence voltage; is the amplitude of the negative sequence voltage measured at different times t; for The average value of .

[0049] The fluctuation degree characteristics of the voltage imbalance are further determined by the variance of the positive sequence voltage component and the variance of the negative sequence voltage component. The variance value and the fluctuation degree characteristics of the voltage imbalance show a linear relationship. The larger the variance, the greater the fluctuation degree of the voltage imbalance.

[0050] Calculate the slope of the positive sequence voltage component and the negative sequence voltage component over time based on the amplitude calculation results: ; ; in, and are the amplitude changes of the positive sequence voltage component and the negative sequence voltage component at adjacent time points, is the time interval between two adjacent measurement time points.

[0051] The changing trend characteristics of the voltage imbalance are then determined by the slopes of the positive-sequence voltage component and the negative-sequence voltage component. The value represents the magnitude of the changing trend, and the positive and negative signs reflect the increasing or decreasing trend of the voltage imbalance.

[0052] After the three-phase imbalance feature is extracted, the status tag is verified and corrected based on the corresponding associated event information and the change of the three-phase imbalance feature.

[0053] Specifically, based on the corresponding associated event information and combined with the three-phase imbalance characteristics, the state label of the corresponding target voltage data sample is corrected, including: Construct a time window based on the corresponding associated event information; Based on the time window, identify the change trend of the three-phase imbalance characteristics of each target voltage data sample; Correct the corresponding status label according to the changing trend of the three-phase imbalance characteristic.

[0054] Power system faults typically progress from inception to development. For example, a line break fault may progress through stages such as insulation aging, partial discharge, and gradual fracture. The three-phase imbalance characteristics exhibit distinct trends in each stage. Capturing these trends through time windows allows for more accurate fault location. Furthermore, time windows are constructed using the timestamps of historical line break events to align target voltage data samples with known fault scenarios. This allows for more accurate identification of characteristic changes before and after the fault occurs, and corresponding state labels are modified to better align with the actual fault evolution logic.

[0055] When identifying the changing trend of the three-phase imbalance characteristics, the changing trend can be identified through trend analysis methods such as the sliding average method and the linear regression method.

[0056] If, in the time window before a line break fault occurs, the overall trend of change in the three-phase imbalance characteristics is a gradual upward trend, the corresponding status label should be "normal operation" because it was originally before the line break fault event. In response to the current trend of change in the three-phase imbalance characteristics, it is further modified to "normal operation - fault warning". If, in the time window after a line break fault occurs, the trend of change in the three-phase imbalance characteristics has not returned to the normal range and is fluctuating, the corresponding status label should be "normal operation" because it was originally after the line break fault event. In response to the trend of change in the corresponding three-phase imbalance characteristics, it is further modified to "abnormal state - residual fault".

[0057] After completing the corresponding state label correction, in order to optimize the training efficiency and accuracy of the subsequent fault diagnosis model, the correlation between each three-phase imbalance feature and the fault disconnection event is further calculated, and then the data for fault diagnosis model training is screened to establish a training data set, and then the corresponding fault diagnosis model is constructed for corresponding training and learning.

[0058] The method of constructing a fault diagnosis model and training the fault diagnosis model based on a sample data set includes: Based on the sample data set, the correlation between the three-phase imbalance characteristics and the fault disconnection event is calculated; Combined with the corresponding correlation degree, the three-phase imbalance characteristics and the corresponding corrected state labels are selected from the sample data set to generate a training data set; Based on the decision tree algorithm, a fault diagnosis model is constructed and trained using a training data set.

[0059] The original feature set constructed from the extracted three-phase imbalance features may contain a large amount of redundant or irrelevant information, such as repeated features at different time scales and noise features caused by environmental interference. These will increase the complexity of model training and reduce the generalization ability. Therefore, the correlation between each three-phase imbalance feature and the fault disconnection event is first calculated. Then, based on the corresponding correlation, the three-phase imbalance features for model training and learning are screened from the sample data set.

[0060] Specifically, the Pearson correlation coefficient is used to quantify the degree of linear correlation between the three-phase unbalance characteristics and the fault disconnection event.

[0061] The calculation formula is: ; in, is the Pearson correlation coefficient, used to quantify the eigenvalue Line disconnection events the strength and direction of the linear relationship between These are voltage-related characteristic values ​​such as voltage unbalance, positive sequence voltage component, negative sequence voltage component, peak value, variance, slope, etc. For the The voltage unbalance value obtained by the measurement is For line disconnection events, Indicates that a line disconnection has occurred. Indicates that there is no line disconnection. is the eigenvalue The average value of For variables The average value of .

[0062] The correlation coefficient finally calculated can measure the degree of correlation between each feature and line disconnection, and <0.3 indicates a weak correlation, 0.3≤ <0.5 indicates a moderate correlation. ≥0.5 indicates a strong correlation.

[0063] After the correlation degree is calculated, the three-phase imbalance characteristics and the corresponding corrected state labels are filtered from the sample data set based on the corresponding correlation degree to generate a training data set, including: Based on a preset threshold, the three-phase unbalance characteristics are screened as input variables of the fault diagnosis model; The eigenvalues ​​of the screened three-phase unbalanced features and the corresponding corrected state labels are used as training samples to construct a training data set.

[0064] In this embodiment, the preset threshold is set to 0.5, and only three-phase unbalance features that are strongly correlated with fault disconnection events are selected as training samples to ensure the training and learning efficiency and accuracy of the fault diagnosis model.

[0065] The decision tree algorithm is then used to construct and train the fault diagnosis model. Specifically, the decision tree can be constructed using the CART algorithm, and then the data in the constructed training data set is standardized and divided into a training set and a test set. Then, the tree depth, the minimum number of leaf node samples and other parameters are adjusted through cross-validation to achieve training and learning of the fault diagnosis model.

[0066] The trained fault diagnosis model can be put into use. Specifically, the corresponding fault diagnosis model can be triggered by collecting a phase loss alarm signal. When a phase loss alarm signal is collected, the corresponding real-time line data is input into the fault diagnosis model to obtain the fault location result.

[0067] Specifically, the fault location result includes at least the location of the faulty device and the corresponding three-phase unbalance characteristic value.

[0068] By identifying the location of the faulty equipment and the corresponding three-phase imbalance characteristic values, the fault location and fault type can be effectively located, the efficiency of subsequent fault handling can be optimized, and the stability and safety of distribution network operation can be ensured.

[0069] The embodiment described above is only a preferred solution of the present invention and does not limit the present invention in any form. Other variations and modifications are possible without exceeding the technical solution described in the claims.

Claims

1. A line disconnection fault location method based on imbalance characteristic analysis is characterized by: include: Construct target voltage data samples based on historical phase loss alarm data and historical fault disconnection events; Based on the imbalance index, the imbalance features are extracted and the state labels of the target voltage data samples are corrected accordingly to construct a sample data set; Build a fault diagnosis model and train it based on the sample data set; When a phase loss alarm signal is collected, the fault location result is obtained based on the corresponding real-time line data combined with the fault diagnosis model.

2. The line disconnection fault location method based on imbalance characteristic analysis according to claim 1, characterized in that: The target voltage data sample is constructed based on the historical phase loss alarm data and historical fault disconnection events, including: Filter data extraction lines based on historical phase loss alarm data; Obtain the distribution transformer terminal equipment information of the data extraction line, and correspondingly retrieve the three-phase voltage data of each distribution transformer terminal; The association relationship between the distribution transformer terminal and each historical fault disconnection event is obtained. Based on the corresponding timestamp, the corresponding status label and associated event information are added to the three-phase voltage data of each distribution transformer terminal to construct the target voltage data sample.

3. The line disconnection fault location method based on imbalance characteristic analysis according to claim 2, characterized in that: The method of constructing a target voltage data sample based on historical phase loss alarm data and historical fault disconnection events further includes: According to the topology information of each distribution transformer terminal, the distribution transformer terminals with associated relationships are combined to obtain a distribution transformer terminal group; Constructing a three-phase voltage data group based on the three-phase voltage data of each distribution transformer terminal in the distribution transformer terminal group; Constructing a corresponding three-phase voltage data group sequence according to the corresponding three-phase voltage data group in chronological order; Based on the corresponding timestamps and the association relationship between each distribution transformer terminal in the distribution transformer terminal group and each historical fault disconnection event, the status label and associated event information are added to the corresponding three-phase voltage data group to construct the target voltage data sample.

4. The line disconnection fault location method based on imbalance characteristic analysis according to claim 2 or 3, characterized in that: The method of extracting the imbalance feature based on the imbalance index and correspondingly correcting the state label of the target voltage data sample to construct a sample data set includes: Based on the imbalance index, the imbalance index value of each target voltage data sample is calculated according to the corresponding three-phase voltage data; Based on the imbalance index value, extract the three-phase imbalance characteristics of each target voltage data sample; Based on the corresponding associated event information and combined with the three-phase imbalance characteristics, the state label of the corresponding target voltage data sample is modified; A sample data set is constructed based on the three-phase imbalance features extracted from each target voltage data sample and the corrected state labels.

5. The line disconnection fault location method based on imbalance characteristic analysis according to claim 4 is characterized in that: The state label of the corresponding target voltage data sample is corrected based on the corresponding associated event information and in combination with the three-phase imbalance characteristic, including: Construct a time window based on the corresponding associated event information; Based on the time window, identify the change trend of the three-phase imbalance characteristics of each target voltage data sample; Correct the corresponding status label according to the changing trend of the three-phase imbalance characteristic.

6. The line disconnection fault location method based on imbalance characteristic analysis according to claim 1, characterized in that: The method of constructing a fault diagnosis model and training the fault diagnosis model based on a sample data set includes: Based on the sample data set, the correlation between the three-phase imbalance characteristics and the fault disconnection event is calculated; Combined with the corresponding correlation degree, the three-phase imbalance characteristics and the corresponding corrected state labels are selected from the sample data set to generate a training data set; Based on the decision tree algorithm, a fault diagnosis model is constructed and trained using a training data set.

7. The line disconnection fault location method based on imbalance characteristic analysis according to claim 1, characterized in that: The method of combining the corresponding correlation levels and screening the three-phase imbalance characteristics and the corresponding corrected state labels from the sample data set to generate a training data set includes: Based on a preset threshold, the three-phase unbalance characteristics are screened as input variables of the fault diagnosis model; The eigenvalues ​​of the screened three-phase unbalanced features and the corresponding corrected state labels are used as training samples to construct a training data set.

8. The line disconnection fault location method based on imbalance characteristic analysis according to claim 1, characterized in that: The imbalance index includes at least three-phase imbalance, positive sequence voltage component and negative sequence voltage component.

9. The line disconnection fault location method based on imbalance characteristic analysis according to claim 1, characterized in that: The imbalance characteristics at least include a time domain characteristic of each imbalance indicator.

10. The line disconnection fault location method based on imbalance characteristic analysis according to claim 1, characterized in that: The fault location result includes at least the location of the faulty device and the corresponding three-phase unbalance characteristic value.

Images