Urban rail power supply system fault early warning method based on multi-source data fusion

Abstract

The invention relates to the technical field of power monitoring, in particular to an urban rail power supply system fault early warning method based on multi-source data fusion, which comprises the following steps: acquiring a thermogram, extracting a temperature drift trajectory, marking a thermal abnormal point, identifying abnormal data by combining voltage sampling, analyzing the overlapping property of abnormal time and direction, and positioning a fault overlapping point. And identifying the linkage early warning area through spatial statistics. According to the method, the continuous drift trajectory of the temperature gravity center in the traction transformer thermogram is recognized, the thermal abnormal points are extracted, the voltage sudden change features are analyzed in combination with the voltage sampling sequences in the same time period, the composite recognition mode of time synchronization and trend direction comparison is achieved, the fault symptom positioning accuracy is effectively improved, and the fault symptom positioning accuracy is improved. By aggregating the fault coincidence points in the plurality of power supply sections and executing the spatio-temporal clustering operation, the capability of sensing common anomalies in the region can be significantly improved, and a potential linkage fault region can be clearly identified based on joint calculation of a spatial adjacent structure and a statistical aggregation degree.

Description

Technical Field

[0001] The present invention relates to the technical field of power monitoring, and in particular to a fault early warning method for an urban rail power supply system based on multi-source data fusion. Background Art

[0002] The field of power monitoring technology involves real-time collection, analysis and status assessment of various electrical parameters in the power system, such as current, voltage, frequency and power, to achieve effective control of the operating status of power equipment and risk warning.

[0003] Among them, the urban rail power supply system fault warning method refers to identifying potential abnormal conditions by analyzing the monitoring data of the contact network and traction power supply circuit in response to possible fault signs such as voltage anomalies, current mutations, and contact network disconnections in the rail transit power supply network.

[0004] Existing technologies only monitor and analyze single parameters such as voltage and current of the contact network and traction power supply circuit, making it difficult to capture the multi-dimensional performance characteristics of different types of faults in a timely manner before they occur. In the early stage when the surface of the traction transformer abnormally heats up and has not yet caused obvious voltage fluctuations, it is often difficult to make an early warning response, resulting in the fault being discovered only when it develops to a critical state, which can easily cause power outages or system chain failures. In the case of synchronous anomalies at multiple points in the region, existing technologies lack the ability to identify the temporal and spatial distribution patterns between anomalies, and cannot form a basis for multi-section linkage judgment, resulting in delayed early warning of the overall system and limited response range. A typical scenario is that multiple transformers show similar abnormal trends in a short period of time but are not identified by the system, eventually forming a cross-section hidden fault chain. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a fault early warning method for an urban rail power supply system based on multi-source data fusion.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solution: a method for early warning of urban rail power supply system faults based on multi-source data fusion, comprising the following steps: S1: Collect continuous thermal images of the track traction transformer surface within a specified period, identify the continuous drift trajectory of the center of gravity of the traction transformer surface temperature, and mark the thermal anomaly point set; S2: Obtaining a concentrated recording time period of the thermal anomaly point, extracting a voltage sampling sequence on the high-voltage side of the traction transformer within the same time period, and identifying a voltage anomaly data group; S3: Analyze whether the time points at which the thermal anomaly point set and the voltage anomaly data set generate anomalies within the same time period overlap, determine whether the temperature center of gravity drift direction is consistent with the voltage mutation direction, and mark the traction transformer fault coincidence point; S4: Summarize the coincidence points of the traction transformer faults in multiple power supply sections, analyze whether multiple transformers have similar sudden changes within a specified time, and construct a regional anomaly aggregation set; S5: Taking the coordinates of each abnormal traction transformer and the fault occurrence time of the regional abnormal aggregation set as input, a spatial proximity structure is constructed, the distribution of fault points in the structure is analyzed by Ripley's K function, and the linkage warning area is marked.

[0007] As a further solution of the present invention, the step of obtaining the thermal anomaly point set is specifically as follows: S111: Collect continuous thermal images of the surface of the track traction transformer within a specified period, select a pixel group in the central area of ​​each frame whose temperature exceeds the ambient temperature, determine the center of gravity of the temperature distribution of each frame based on the relationship between the temperature of the pixel point and the image position, and generate a temperature center of gravity trajectory sequence; S112: Based on the temperature center of gravity trajectory sequence, tracking the change path of the center of gravity position between consecutive frames, identifying the position where the direction turns in the path, locating the time sequence position of the change according to the image frame number, extracting the coordinate points and time information representing the change behavior, and generating a drift change position set; S113: Based on the coordinates and time data of each change point in the drift change position set and in combination with the boundary of the traction transformer surface monitoring area, invalid points outside the area range are screened out to generate a thermal anomaly point set.

[0008] As a further solution of the present invention, the steps for obtaining the voltage anomaly data set are specifically as follows: S211: calling the time data corresponding to each abnormal point in the thermal abnormal point set, determining the start and end time range, and extracting the voltage side sampling data sequence of the traction transformer voltage side within the same time period; S212: Identify the starting position where the voltage exceeds the voltage mutation threshold based on the time distribution of continuous data points and voltage changes in the voltage-side sampling data sequence, extract the fluctuation direction trend, duration, and magnitude change indicators corresponding to each change, and generate a voltage fluctuation feature set; S213: Screening the change segments of each change indicator in the voltage fluctuation feature set, and classifying them in chronological order to generate a voltage anomaly data group.

[0009] As a further solution of the present invention, the step of obtaining the traction transformer fault coincidence point is specifically as follows: S311: Calling all time information recorded in the thermal anomaly point set and the voltage anomaly data set, respectively establishing corresponding time axis sequences, and filtering anomaly records that occur simultaneously in the same time period by comparing the interval relationship between each time point in the two sequences, to generate an anomaly time overlap data table; S312: Extracting the continuous coordinate change direction and the numerical change direction of the corresponding voltage data according to the thermal anomaly point and abnormal voltage corresponding to each overlapping time point in the abnormal time overlapping data table, determining whether the change directions are the same, and screening a direction consistency mark list; S313: Based on the matching point information with consistent direction and overlapping time in the direction consistency mark list, the information is integrated according to the traction transformer number and time sequence to generate the traction transformer fault coincidence point.

[0010] As a further solution of the present invention, the steps of obtaining the regional anomaly aggregation set are specifically as follows: S411: Retrieving the time information and physical coordinates recorded in the traction transformer fault coincidence point, extracting the power supply section identifier of each fault point and dividing it into groups, counting the time distribution range and spatial location set of the fault occurrence in each group, and generating a section time-space distribution data table; S412: Clustering the coordinates and time of each fault point in the segment spatiotemporal distribution data table using the DBSCAN clustering algorithm to generate a spatiotemporal clustering boundary indicator group; S413: Based on the cluster information divided from the spatiotemporal clustering boundary indicator group, extract the corresponding power supply section number and the number of transformers in the cluster, merge the cluster sets that meet both section centralization and fault type consistency, and generate a regional anomaly aggregation set.

[0011] As a further solution of the present invention, the steps for obtaining the linkage warning area are specifically as follows: S511: calling the coordinate information and corresponding fault occurrence time of each abnormal traction transformer in the regional abnormality aggregation set, recording the straight-line distance and time tag sequence between all transformers, and generating a spatiotemporal relationship structure data group; S512: According to the spatiotemporal relationship structure data set, a clustering radius interval based on spatial distance and a time weight factor are set, distribution characteristics between fault points are statistically analyzed, and Ripley's K function is input to calculate the clustering degree of traction transformer fault coincidence points in geographic space, thereby generating a set of spatial clustering statistical curves; S513: Determine the degree of deviation between the spatial aggregation statistical curve group and the theoretical uniform distribution curve, screen out spatial areas where the deviation exceeds the set aggregation intensity judgment benchmark value, and output them as linkage warning areas.

[0012] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by identifying the continuous drift trajectory of the temperature center of gravity in the traction transformer thermal image and extracting thermal anomaly points, and combining the voltage sampling sequence in the same time period to analyze the voltage mutation characteristics, a composite identification method of time synchronization and trend direction comparison is realized, which effectively improves the accuracy of fault sign positioning. By aggregating the fault coincidence points in multiple power supply sections and performing spatiotemporal clustering operations, the ability to perceive common anomalies in the region can be significantly improved. Based on the joint calculation of spatial proximity structure and statistical aggregation, potential linkage fault areas can be clearly identified, and the linkage early warning capability of spatial correlation and time evolution trend between multiple transformers is enhanced, avoiding the problems of misjudgment and missed judgment caused by single-point anomaly judgment, and improving the efficiency and response accuracy of urban rail power supply systems to potential fault diffusion trends in large-scale and multi-segment contexts. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 Schematic diagram of the process of the urban rail power supply system fault early warning method based on multi-source data fusion of the present invention; Figure 2 This is a schematic diagram of the process of marking a thermal anomaly point set according to the present invention; Figure 3 This is a schematic diagram of the process of identifying voltage anomaly data groups according to the present invention; Figure 4 A schematic diagram of the process of marking the traction transformer fault coincidence point according to the present invention; Figure 5 A schematic diagram of the process of constructing a regional anomaly aggregation set according to the present invention; Figure 6 This is a flow chart of marking linkage warning areas according to the present invention. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0015] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0016] See also Figure 1The present invention provides a technical solution: a method for early warning of urban rail power supply system faults based on multi-source data fusion, comprising the following steps: S1: Collect continuous thermal images of the track traction transformer surface within a specified period, identify the continuous drift trajectory of the center of gravity of the traction transformer surface temperature, and mark the thermal anomaly point set; S2: Obtain the concentrated recording time period of the thermal anomaly points, extract the voltage sampling sequence on the high-voltage side of the traction transformer within the same time period, and identify the voltage anomaly data group; S3: Analyze whether the time points at which the thermal anomaly point set and the voltage anomaly data set generate anomalies within the same time period overlap, determine whether the temperature center of gravity drift direction is consistent with the voltage mutation direction, and mark the traction transformer fault coincidence point; S4: Summarize the coincidence points of traction transformer faults in multiple power supply sections, analyze whether multiple transformers have similar sudden changes within a specified time, and construct a regional anomaly aggregation set; S5: Take the coordinates and fault occurrence time of each abnormal traction transformer in the regional anomaly aggregation set as input, build a spatial proximity structure, analyze the distribution of fault points in the structure through Ripley's K function, and mark the linkage warning area.

[0017] See also Figure 2 , the specific steps for obtaining the thermal anomaly point set are: S111: Collect continuous thermal images of the surface of the track traction transformer within a specified period, select a pixel group in the central area of ​​each frame whose temperature exceeds the ambient temperature, determine the center of gravity of the temperature distribution of each frame based on the relationship between the temperature of the pixel point and the image position, and generate a temperature center of gravity trajectory sequence; When collecting continuous thermal images of the surface of the track traction transformer, a thermal imager is installed at the monitoring site and set to continuously shoot the transformer at a frequency of 5 frames per second. The acquisition time is set to 10 minutes, and a total of 3000 frames of thermal images will be obtained. For each frame of the image, its central area is selected as the analysis area. For example, the overall image size is 800×600 pixels, and the central area is set to a square area with a width and height of 200 pixels. The infrared temperature value of each pixel is obtained, and then the temperature value of each pixel is compared with the currently measured ambient temperature value. For example, if the ambient temperature is measured to be 32 degrees Celsius, the pixels with a temperature greater than 32 degrees Celsius are selected as the high For warm pixels, the coordinate positions and temperature values ​​of all high-temperature pixels are counted. The concentration of these pixels in the image is calculated using a temperature-weighted method to determine the temperature center of gravity. For example, if one pixel has a temperature of 36 degrees Celsius and another has a temperature of 38 degrees Celsius, their corresponding positions are located to the right and below the center of the image, respectively. By multiplying the temperature value by the pixel position for all high-temperature pixels, adding up all the pixels and dividing by the total temperature value, the temperature center of gravity coordinate position of the current frame is obtained. The above operation is repeated for all 3000 frames of images to form a continuous center of gravity trajectory sequence, which is used to represent the central direction path of the transformer heat source changing over time.

[0018] S112: Based on the temperature center of gravity trajectory sequence, the path of the center of gravity position change between consecutive frames is tracked, the location where the direction turns in the path is identified, and the temporal position of the change is located according to the image frame number. The coordinate points and time information representing the change behavior are extracted to generate a set of drift change positions; According to the above center of gravity trajectory sequence, the center of gravity coordinate position at each moment is read frame by frame and the corresponding time point or frame number is recorded. For example, the center of gravity of the first frame is at the left position of the image, the second frame is at the center-right position, and the third frame becomes the upper right position. The center trajectory path is recorded. In the continuous change of the path, the direction trend between the positions of adjacent center of gravity points is observed segment by segment. For example, the center of gravity moves to the right from frame 1 to frame 2, and moves to the upper right from frame 2 to frame 3. The overall direction is upper right. If the position suddenly changes to the upper left starting from frame 4, it is determined that a turning point has occurred. When identifying the turning point Based on the significance of the change in the horizontal or vertical direction, if the angle of the change direction changes by more than 30 degrees on the basis of the original direction, it is considered a significant turning point. Based on this judgment standard, all image frame positions where the turning occurs are gradually found and the frame number and the center of gravity coordinate information at that time are recorded. For example, the 50th frame has a significant transformation from the lower left direction to the upper right direction, and its center of gravity coordinate position is slightly above the center of the image. Then this coordinate and the frame number together form a drift change point. Finally, after integrating all the turning points, a drift change position set is constructed.

[0019] S113: Based on the coordinates and time data of each change point in the drift change position set and the boundary of the traction transformer surface monitoring area, invalid points outside the area are screened out to generate a thermal anomaly point set; After obtaining the drift change position set, each data point in the set needs to be compared with the monitoring area boundary. The boundary is usually mapped based on the actual physical size of the traction transformer. For example, the monitoring area is defined as the area from horizontal coordinates 100 to 300 and vertical coordinates 50 to 250 on the image. The coordinate value of each drift point is checked one by one to see if it falls within this area. For example, if the coordinates of a drift point are 120 horizontally and 130 vertically, it is clearly within the specified area and is retained. If the coordinates of a point are 320 horizontally, it is outside the boundary and needs to be deleted. In addition, it is necessary to combine the time information to screen whether the abnormal points have temporal continuity. For example, if a drift point occurs at the 1500th frame, that is, at the 5th minute, and another drift point occurs at the 1510th frame, that is, 2 seconds after the 5th minute, the time interval between the two drift points is 2 seconds. If the change time interval is less than 2 seconds, it may be a normal heat flow fluctuation rather than an abnormal drift, and the point can be selected to be eliminated. Through the dual verification method of regional boundaries and time, the valid drift points that are both within the area and meet the change timing requirements are finally retained, and these points are integrated to form the thermal anomaly point set.

[0020] See also Figure 3 , the specific steps for obtaining the voltage abnormality data group are: S211: Calling the time data corresponding to each abnormal point in the thermal abnormal point set, determining the start and end time range, and extracting the voltage side sampling data sequence of the traction transformer voltage side within the same time period; Call the time data of the hot anomaly point concentration, first extract the time of all anomaly points, and record the timestamp corresponding to each anomaly point in the time series. For example, if the anomaly points appear at 300 seconds, 304 seconds, and 309 seconds, the minimum time point is determined to be 300 seconds and the maximum is 309 seconds. Set the analysis start and end time range from 300 seconds to 309 seconds as the extraction interval for subsequent voltage sampling data. Then read the voltage side data in the corresponding time period in the voltage measurement system. The sampling frequency is set to 100 times per second. Then 900 voltage sampling points will be extracted in this 9-second time period. These points are arranged in chronological order to form a voltage side sampling data sequence. This sequence is the basic information source for subsequent judgment of voltage anomalies. In this process, it is necessary to ensure that the sampling data is completely aligned with the hot anomaly point time, that is, the time corresponding to the first point in the sampling data is 300 seconds, and the last point is 309 seconds. If there is microsecond-level time information in the hot anomaly point concentration.

[0021] S212: Based on the time distribution of continuous data points and voltage changes in the voltage-side sampling data sequence, identify the starting position where the voltage exceeds the voltage mutation threshold, extract the fluctuation direction trend, duration, and numerical amplitude change indicators corresponding to each change, and generate a voltage fluctuation feature set; In the voltage-side sampling data sequence, the continuous change of the voltage value at each moment is analyzed point by point. By calculating the voltage difference between two adjacent sampling points, it is identified when the voltage change exceeds the mutation threshold. For example, if the mutation threshold is set to 2 volts, if the voltage at one sampling point is 220 volts and the next point is 217.8 volts, the change is 2.2 volts, which has exceeded the set threshold. In this case, this sampling point is recorded as the mutation starting point. The subsequent sampling points are tracked, and it is recorded whether the voltage is continuously decreasing, continuously increasing, or fluctuating. The duration is calculated based on the time interval. For example, if the mutation lasts for 0.6 seconds, the direction is decreasing, and the total voltage drop is 5 volts, then this segment is recorded as a complete voltage fluctuation event. Subsequently, all mutation starting points in the subsequent data are analyzed in sequence and their directions, durations, and amplitudes are extracted. Finally, all fluctuation events are sorted into a voltage fluctuation feature set. Each set of feature data contains a direction label (increasing or decreasing), start time, end time, and value change.

[0022] S213: Screening the change segments of each change indicator in the voltage fluctuation feature set, and classifying them in chronological order to generate a voltage anomaly data group; For each record in the voltage fluctuation feature set, each record is screened item by item according to the change index. The screening rules set dual standards based on the change amplitude and duration. For example, a segment with a voltage change amplitude greater than 3 volts and a duration of more than 0.3 seconds is identified as a significant voltage fluctuation event. For example, a segment with a voltage drop of 4.1 volts and a duration of 0.5 seconds is retained if it meets the conditions, while another segment with a voltage rise of only 1.5 volts is eliminated. After the screening is completed, all retained items are arranged and classified in order of the starting time. If multiple fluctuation events occur with a time interval of less than 0.5 seconds, they are merged into a continuous event for processing. Finally, they are sorted according to the classification standard and time to generate a voltage anomaly data group.

[0023] See also Figure 4 ,The specific steps for obtaining the traction transformer fault coincidence point are as follows: S311: Call all the time information recorded in the thermal anomaly point set and the voltage anomaly data set, establish corresponding time axis sequences respectively, compare the interval relationship between each time point in the two sequences, filter out anomaly records that appear simultaneously in the same time period, and generate an anomaly time overlap data table; The time information contained in both the thermal anomaly point set and the voltage anomaly data set is retrieved. First, all recorded times in the two data sets are extracted into two independent timeline sequences. Each point in the thermal anomaly point set contains timestamp information. For example, if the first time point in the thermal anomaly point set is 302 seconds, and the second is 305 seconds, and an abnormal interval in the voltage anomaly data starts at 303 seconds and lasts until 306 seconds, then the thermal anomaly point at 305 seconds falls within that abnormal interval, thus forming an overlap. However, 302 seconds falls before the start of that interval and does not meet the overlap criteria. The criteria for determining overlap are: if a thermal anomaly time point falls between the start and end times of any voltage anomaly interval (including the start and end times themselves), it is considered a time overlap; if it is before the start time or after the end time, it does not constitute an overlap. Using this logic, the temporal relationships between all thermal anomaly points and voltage anomaly intervals are compared, filtering out overlapping records and integrating them into an abnormality time overlap data table. Each record indicates the overlapping time point, the corresponding thermal image anomaly information, and the voltage anomaly content for that period.

[0024] S312: Extract the continuous coordinate change direction and the value change direction of the corresponding voltage data based on the thermal anomaly point and abnormal voltage corresponding to each overlapping time point in the abnormal time overlap data table, determine whether the change directions are the same, and filter the direction consistency mark list; Based on the overlapping records in the abnormal time overlapping data table, the thermal anomaly point coordinate information and voltage value change information of each record are extracted. For the thermal anomaly part, the two-dimensional direction vector of the heat source movement is calculated by the coordinate difference between the current frame and the previous frame, and the polar angle corresponding to the vector direction (calculated using the inverse tangent function, in degrees) is used as a reference for the heat source movement direction. In order to standardize the direction determination, the 360° circumference is divided into 8 direction areas, which are specifically divided as follows: right direction (→): The angle range is , upper right direction (↗): the angle range is , Upward direction (↑): Angle range is , upper left direction (↖): angle range is , left direction (←): angle range is and , lower left direction (↙): the angle range is , down direction (↓): angle range is , lower right direction (↘): the angle range is . To ensure that each main direction (except the left) covers a range of 45°, based on the periodicity of the angle from -180° to +180°, the left direction (←) is defined as the union of two intervals near +180° and -180°. This area extends 22.5° "upward" (counterclockwise) from the center line, and the range is +157.5°. Therefore, from +157.5° to +180° belongs to the left direction. Corresponding to [+157.5°, +180°]. Similarly, this area extends 22.5° "downward" (clockwise) from the center line, and the range is -157.5°. Therefore, from -180° to -157.5° also belongs to the left direction. Corresponding to (-180°, -157.5°).

[0025] For voltage data, if the voltage value of the current frame decreases relative to the previous frame, the voltage direction is marked as "decreasing"; if the voltage value increases, the direction is marked as "increasing." The heat source direction is matched with the voltage change direction: if the heat source movement direction is "down," "lower right," or "lower left" and the voltage direction is "decreasing," or if the heat source movement direction is "up," "upper right," or "upper left" and the voltage direction is "increasing," the direction is considered consistent; otherwise, the direction is inconsistent. By determining the direction consistency of all overlapping records item by item, the items with consistent directions are selected and a direction consistency marker list is constructed. This list includes the overlapping time point, the direction of the thermal anomaly movement, the voltage change direction, and the consistency judgment result.

[0026] S313: Based on the matching point information in the direction consistency mark list that has consistent directions and overlaps in time, the information is integrated in order of traction transformer numbers and time to generate the traction transformer fault coincidence point; Points with consistent directions and time overlap are extracted from the direction consistency mark list, grouped according to the traction transformer number in the record, and each group of data is arranged in chronological order. All points that meet the conditions are sequentially integrated to form a set of fault coincidence points of the traction transformer. Each fault coincidence point has the traction transformer identification, thermal anomaly coordinates, voltage change amplitude and occurrence time. After integration, it can be formed that, for example, a coincidence point occurs at 305 seconds when a heat source moves to the upper right and the voltage drops at the same time for a certain numbered transformer, and another coincidence point occurs at 312 seconds when a left shift occurs for another numbered transformer and the voltage rises at the same time. Finally, all coincidence points are systematically generated and summarized in this structured recording method.

[0027] See also Figure 5 ,The specific steps for obtaining the regional anomaly aggregation set are: S411: Retrieving the time information and physical coordinates recorded in the traction transformer fault coincidence point, extracting the power supply section identifier of each fault point and dividing it into groups, counting the time distribution range and spatial location set of the fault occurrence in each group, and generating a section time-space distribution data table; The time information and physical coordinates recorded in the traction transformer fault coincidence points are retrieved. First, based on the spatial coordinates of each fault point, its physical location in the track power supply system is mapped to the corresponding power supply section identifier. For example, if the coordinates of a fault point fall within the coordinate range of the third power supply section, its power supply section identifier is set to 3. All fault points are classified and grouped according to the power supply section. Each group represents a set of faults in a power supply section. Then, time statistics are performed on the fault points within each group, and the earliest and latest fault occurrence times within the group are extracted to form the time distribution range of the power supply section. For example, if the start time of a group fault is 600 seconds and the end time is 660 seconds, the time range of the group is 60 seconds. The spatial coordinate positions of all fault points in the group are further extracted and unified as the spatial distribution set of the group to form the spatiotemporal record of faults within each power supply section. Finally, all section numbers, corresponding time ranges, and spatial coordinate sets are integrated and output to construct a section spatiotemporal distribution data table.

[0028] S412: Clustering the coordinates and time of each fault point in the section spatiotemporal distribution data table using the DBSCAN clustering algorithm to generate a spatiotemporal cluster boundary indicator group; In order to perform spatiotemporal clustering analysis on the fault coincidence points of traction transformers, the density-based spatial clustering algorithm DBSCAN is adopted. The fault points contain three types of data dimensions: horizontal coordinates (Unit: meter), vertical coordinate (Unit: meter) and time information (Unit: seconds). Since the three have different physical dimensions, the three-dimensional data must first be normalized to a dimensionless standardized space before distance calculation and density determination can be performed.

[0029] For each fault point , whose three-dimensional attributes are the original vectors , converted to a dimensionless standard vector using the following normalization formula : , , ; in, :Indicates the The original horizontal coordinates, vertical coordinates and timestamps of each fault point; : The minimum and maximum values ​​of the horizontal coordinate in the data set; : The minimum and maximum values ​​of the vertical axis; : The minimum and maximum values ​​of time; : are the normalized horizontal coordinate, vertical coordinate and time respectively.

[0030] After normalization, any two fault points and The joint spatiotemporal distance between is defined as follows: ; in, : No. The normalized three-dimensional vector of the fault point; : No. The normalized three-dimensional vector of the fault point; : indicates a point with dot Joint distance in normalized space; all difference units are unified as "dimensionless", so the distance value is also a unitless scalar with a value range of between.

[0031] There are three traction transformer fault points, and their original data are as follows: Point A: , , , point B: , , , point C: , , .

[0032] Compute the global minimum and maximum: , →Horizontal range 90 meters, , →Vertical range 100 meters, , →Time range: 40 seconds.

[0033] Normalize the three points separately: Point A: , , ; Point B: , , ; Point C: , , ; Compute the joint distance between points: Points A and B: ; Points A and C: ; Points B and C: ; Set DBSCAN parameters: Neighborhood radius (unitless, applicable to normalized space), minimum number of neighbors ; Clustering judgment result: The distance between A and B is , the two are neighbors; the distances between A and C, and between B and C are greater than , do not constitute a neighborhood; the neighborhoods of A and B both contain at least two points (including themselves), satisfying , so they are all core points; C has no neighbors, does not meet the core point conditions, and is not in any neighborhood, so it is determined to be an outlier.

[0034] Based on the above analysis, the DBSCAN algorithm outputs the following cluster boundary information: Cluster 1: contains points A and B, indicating that the density is close in the normalized space-time space; Noise point: Point C is considered an abnormal isolated point because its distance from other points is too large.

[0035] First, a normalization formula is used to normalize the horizontal, vertical, and temporal position data of each fault point to a unified dimensionless interval. This resolves the physical dimension inconsistencies in the original data and ensures comparability in the spatial and temporal dimensions in subsequent calculations. Next, based on these normalized values, the standardized three-dimensional Euclidean distance formula is used to calculate the relative proximity between any two points in space and time. This result indicates whether the two fault points are close enough in space and time to potentially belong to the same density cluster. Each distance value is used to determine whether it falls within a specified radius, thereby determining the number of neighbors of a point. This is then combined with a threshold for the minimum number of neighbors to identify core points. Finally, a clustering algorithm is used to determine which points are connected to form clusters and which points are identified as isolated outliers due to insufficient neighborhoods. The entire process, which integrates step-by-step normalization, distance measurement, neighborhood determination, core point identification, and cluster assignment, achieves efficient cluster identification of traction transformer fault points in both spatial and temporal dimensions.

[0036] S413: Based on the cluster information divided from the spatiotemporal clustering boundary indicator group, extract the corresponding power supply section number and transformer quantity in the cluster, merge the cluster sets that meet both section concentration and fault type consistency, and generate a regional anomaly aggregation set; First, read the list of all fault points contained in each cluster, extract the power supply section number corresponding to each fault point, and match it with the preset section mapping table through the coordinate range. For example, if the coordinates of all fault points in a cluster fall between 100 and 160 meters horizontally, and the coordinate segment is mapped to the 3rd power supply section in the system, then the corresponding section number of the cluster is 3. Then count the number of transformer numbers associated with the fault points in the cluster. If a cluster contains multiple fault point records, but some transformer numbers are repeated, the repeated numbers need to be eliminated and only the total number of unique numbers is counted. For example, if there are six fault points in the cluster, corresponding to A01, A02, A01, A03, A03, and A04 respectively, then the count There are four transformers. After completing these data extraction, the section concentration judgment is performed on each cluster, that is, whether all fault points in the cluster belong to the same power supply section. This can be achieved by judging whether there are multiple different section numbers. If only one number appears, it is concentrated, and if multiple numbers appear, it is non-concentrated. Then, the clusters that meet the section concentration are judged for fault type consistency. The method is to extract the fault type field marked by each fault point in the cluster and perform classification statistics. If the proportion of a certain type reaches a preset consistency threshold, such as 90%, it is judged to be consistent. If it does not reach the threshold, the cluster is excluded. Finally, the clusters that meet both conditions after the above screening are merged and unified into a regional anomaly aggregation set.

[0037] See also Figure 6 The specific steps for obtaining the linkage warning area are as follows: S511: Call the coordinate information and corresponding fault occurrence time of each abnormal traction transformer in the regional abnormality aggregation set, record the straight-line distance and time label sequence between all transformers, and generate a spatiotemporal relationship structure data group; The coordinate information and fault occurrence time of each abnormal traction transformer in the regional abnormal aggregation set are called. First, the two-dimensional spatial coordinate value of each traction transformer in the aggregation set and its corresponding fault occurrence time point are extracted one by one. For example, a transformer is numbered T001, the coordinates are (128, 92), and the fault time is 620 seconds. This information is input as a data point. After traversing all abnormal transformers, an original data set containing the coordinates and time of all transformers is constructed. Then, the spatial straight-line distance between any two transformers is calculated in turn. The two-dimensional Euclidean distance formula is used to combine and calculate all point pairs to generate a complete distance matrix. Each row represents the spatial distance between a transformer and all other transformers. For example, if there are five abnormal transformers, a symmetrical distance matrix with 5 rows and 5 columns will be formed. The main diagonal is zero, which means the distance itself is zero. At the same time, the fault time of each transformer is recorded in chronological order, and the corresponding time values ​​are used to form a set of time label sequences to ensure that the spatial distance and time point are matched one by one. Finally, the distance matrix and the time series are jointly constructed into a structured data group to characterize the relative position relationship between abnormal transformers in space and the order of events in time, forming a spatiotemporal relationship structure data group.

[0038] S512: Based on the spatiotemporal relationship structure data set, set the clustering radius interval and time weight factor based on spatial distance, calculate the distribution characteristics of the fault points, input Ripley's K function to calculate the clustering degree of the traction transformer fault coincidence points in the geographic space, and generate a set of spatial clustering statistical curves; Based on the spatiotemporal relationship structure data set, spatial statistical methods are used to quantitatively analyze the clustering characteristics of traction transformer fault points within geographic space. Ripley's K function is used as an analytical tool. First, the analysis radius interval and time weighting factor must be set. The analysis radius represents the spatial threshold between observation points for being considered "neighbors." The time weighting factor is used to adjust the influence of fault timing in spatial distance statistics to truly reflect spatiotemporal density. The closer the faulty transformers are to each other and the closer the fault times are, the greater their weight and the greater their contribution to clustering in the spatial statistics. Conversely, if the time interval is large, the effect of this point pair is weakened.

[0039] The expression of Ripley's K function with time weighting term is as follows: ; in, : At the aggregation scale of Under the condition of m, the statistical value of spatial aggregation between abnormal traction transformers, the larger the value, the stronger the aggregation; : The total area of ​​the observation area, for example, for a 200m × 200m track segment, it is 40,000 square meters; : The number of traction transformers included in the abnormal aggregation set in the current area. For example, a total of 3 abnormal transformers were collected; :Indicates the The transformer and The straight-line Euclidean distance between transformers, in meters, is calculated based on their horizontal and vertical coordinates; : Distance threshold judgment function, when the distance between the two transformers is less than or equal to the set radius When , the value is 1, otherwise it is 0; : No. With the The time weight factor between transformers is expressed as: , :Respectively With the The time when a transformer fails, in seconds; : Time decay coefficient, which controls the influence of time difference on the aggregation intensity, and the unit is the inverse of one second; the value of the weight factor is between 0 and 1, which means that the closer the time between the fault points is, the greater the The closer it is to 1, the higher the contribution; the larger the time difference, the closer the weight value is to 0. Set the analysis radius 10 meters, 20 meters, 30 meters, etc., as the multi-scale aggregation observation range. Time attenuation coefficient It can be set to 0.1, which means that the weight drops to about 0.367 when the time difference is 10 seconds. This value can be determined based on historical accident characteristics. For example, if linkage within 10 seconds between traction transformers is more common, it can be set to 0.1. If the response is faster, the value can be increased.

[0040] By using multiple different radii The values ​​are calculated to form a spatial clustering statistical curve. Each point on this curve represents the spatial clustering intensity of transformer fault points within the current clustering radius. This statistical curve is compared with the theoretical uniform distribution curve and used in subsequent steps to determine the degree of clustering deviation and identify spatial areas with abnormal clustering risks.

[0041] First, a complete spatiotemporal data structure is constructed based on the spatial coordinates and fault time of each abnormal traction transformer. Then, within multiple defined clustering radii, the straight-line distances and fault time differences between all transformers are analyzed pairwise. For each transformer pair whose distance does not exceed the current clustering radius, a weight corresponding to the time difference is calculated. The closer the time difference, the closer the weight is to 1; the larger the time difference, the exponentially decreasing the weight, approaching 0. All pairs of points that meet the spatial proximity condition are weighted by the time difference, forming a clustering strength evaluation weighted by spatiotemporal closeness. Next, all weighted counts are normalized, and the clustering statistics for the current clustering radius are calculated based on the area of ​​the entire observation area and the total number of transformers. This process is repeated at multiple radius scales, forming a complete spatial clustering statistics curve that characterizes the overall spatial distribution characteristics of the fault points.

[0042] The theoretical uniform distribution curve, used as a comparison, assumes a completely random and uniform distribution of the same number of transformers within the same area. This is calculated by multiplying the transformer density by the neighborhood area corresponding to the current clustering radius. Under this uniformity assumption, the probability of neighboring transformers is continuously and smoothly distributed, with the statistical result increasing exponentially with the square of the radius. In other words, at each radius, the theoretically calculated number of neighboring points increases as the area (i.e., the square of the radius), constructing a theoretical distribution curve that exhibits regularity and lacks clustering. This theoretical curve serves as a baseline for actual clustering behavior and provides a reference for measuring the degree of deviation from actual clustering.

[0043] Finally, the actual calculated aggregation statistics at each aggregation radius are compared one by one with the theoretical values ​​of the corresponding radius to determine whether the spatial aggregation intensity at the current scale exceeds the normal random fluctuation range, which is used to identify whether there are significant abnormal aggregation areas in the space.

[0044] S513: Determine the degree of deviation between the spatial aggregation statistical curve group and the theoretical uniform distribution curve, select the spatial areas where the deviation exceeds the set aggregation intensity judgment benchmark value, and output them as linkage warning areas; First, read the actual aggregation statistical curve and the theoretical uniform distribution curve to ensure that the two are constructed based on the same aggregation radius sequence. On this basis, the corresponding value difference is calculated for each aggregation radius point. The difference calculation method is to subtract the theoretical curve value from the actual curve value to obtain the deviation amplitude at each aggregation scale. For example, when the aggregation radius is 20 meters, the actual statistical value is 8500 and the theoretical value is 7000, then the deviation amplitude is 1500. Then set the aggregation intensity judgment benchmark value to determine whether the deviation is significant. The benchmark value can be set to a percentage, such as 20%, which means that when the actual value is greater than the theoretical value, the deviation amplitude is 1500. If the theoretical value is 20% or higher, the deviation is considered significant. The deviation amplitude at each aggregation scale is converted into a percentage of the theoretical value. For example, the above-mentioned 1500 deviation corresponds to a theoretical value of 7000, and the percentage is 21.4%. If it is higher than the set threshold of 20%, it is considered a significant deviation at this scale. Then, the spatial radius corresponding to all significant deviation aggregation scales is mapped back to the spatial area range, and the spatial coordinate coverage area is marked in combination with the fault point location. Finally, the spatial fragments formed at all significant deviation aggregation scales are merged and these spatial areas are summarized and output as linkage warning areas.

[0045] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A fault warning method for urban rail power supply system based on multi-source data fusion, characterized in that: The following steps are involved: S1: Collect continuous thermal images of the track traction transformer surface within a specified period, identify the continuous drift trajectory of the center of gravity of the traction transformer surface temperature, and mark the thermal anomaly point set; S2: Obtaining a concentrated recording time period of the thermal anomaly point, extracting a voltage sampling sequence on the high-voltage side of the traction transformer within the same time period, and identifying a voltage anomaly data group; S3: Analyze whether the time points at which the thermal anomaly point set and the voltage anomaly data set generate anomalies within the same time period overlap, determine whether the temperature center of gravity drift direction is consistent with the voltage mutation direction, and mark the traction transformer fault coincidence point; S4: Summarize the coincidence points of the traction transformer faults in multiple power supply sections, analyze whether multiple transformers have similar sudden changes within a specified time, and construct a regional anomaly aggregation set; S5: Taking the coordinates of each abnormal traction transformer and the fault occurrence time of the regional abnormal aggregation set as input, a spatial proximity structure is constructed, the distribution of fault points in the structure is analyzed by Ripley's K function, and the linkage warning area is marked.

2. The urban rail power supply system fault early warning method based on multi-source data fusion according to claim 1 is characterized in that: The steps for obtaining the thermal anomaly point set are specifically as follows: S111: Collect continuous thermal images of the surface of the track traction transformer within a specified period, select a pixel group in the central area of ​​each frame whose temperature exceeds the ambient temperature, determine the center of gravity of the temperature distribution of each frame based on the relationship between the temperature of the pixel point and the image position, and generate a temperature center of gravity trajectory sequence; S112: Based on the temperature center of gravity trajectory sequence, tracking the change path of the center of gravity position between consecutive frames, identifying the position where the direction turns in the path, locating the time sequence position of the change according to the image frame number, extracting the coordinate points and time information representing the change behavior, and generating a drift change position set; S113: Based on the coordinates and time data of each change point in the drift change position set and in combination with the boundary of the traction transformer surface monitoring area, invalid points outside the area range are screened out to generate a thermal anomaly point set.

3. The urban rail power supply system fault early warning method based on multi-source data fusion according to claim 2 is characterized in that: The steps for obtaining the voltage abnormality data group are specifically as follows: S211: calling the time data corresponding to each abnormal point in the thermal abnormal point set, determining the start and end time range, and extracting the voltage side sampling data sequence of the traction transformer voltage side within the same time period; S212: Identify the starting position where the voltage exceeds the voltage mutation threshold based on the time distribution of continuous data points and voltage changes in the voltage-side sampling data sequence, extract the fluctuation direction trend, duration, and magnitude change indicators corresponding to each change, and generate a voltage fluctuation feature set; S213: Screening the change segments of each change indicator in the voltage fluctuation feature set, and classifying them in chronological order to generate a voltage anomaly data group.

4. The urban rail power supply system fault early warning method based on multi-source data fusion according to claim 3 is characterized in that: The steps for obtaining the traction transformer fault coincidence point are specifically as follows: S311: Calling all time information recorded in the thermal anomaly point set and the voltage anomaly data set, respectively establishing corresponding time axis sequences, and filtering anomaly records that occur simultaneously in the same time period by comparing the interval relationship between each time point in the two sequences, to generate an anomaly time overlap data table; S312: Extracting the continuous coordinate change direction and the numerical change direction of the corresponding voltage data according to the thermal anomaly point and abnormal voltage corresponding to each overlapping time point in the abnormal time overlapping data table, determining whether the change directions are the same, and screening a direction consistency mark list; S313: Based on the matching point information with consistent direction and overlapping time in the direction consistency mark list, the information is integrated according to the traction transformer number and time sequence to generate the traction transformer fault coincidence point.

5. The urban rail power supply system fault early warning method based on multi-source data fusion according to claim 4 is characterized in that: The steps for obtaining the regional anomaly aggregation set are specifically as follows: S411: Retrieving the time information and physical coordinates recorded in the traction transformer fault coincidence point, extracting the power supply section identifier of each fault point and dividing it into groups, counting the time distribution range and spatial location set of the fault occurrence in each group, and generating a section time-space distribution data table; S412: Clustering the coordinates and time of each fault point in the segment spatiotemporal distribution data table using the DBSCAN clustering algorithm to generate a spatiotemporal clustering boundary indicator group; S413: Based on the cluster information divided from the spatiotemporal clustering boundary indicator group, extract the corresponding power supply section number and the number of transformers in the cluster, merge the cluster sets that meet both section centralization and fault type consistency, and generate a regional anomaly aggregation set.

6. The urban rail power supply system fault early warning method based on multi-source data fusion according to claim 5 is characterized in that: The steps for obtaining the linkage warning area are as follows: S511: calling the coordinate information and corresponding fault occurrence time of each abnormal traction transformer in the regional abnormality aggregation set, recording the straight-line distance and time tag sequence between all transformers, and generating a spatiotemporal relationship structure data group; S512: According to the spatiotemporal relationship structure data set, a clustering radius interval based on spatial distance and a time weight factor are set, distribution characteristics between fault points are statistically analyzed, and Ripley's K function is input to calculate the clustering degree of traction transformer fault coincidence points in geographic space, thereby generating a set of spatial clustering statistical curves; S513: Determine the degree of deviation between the spatial aggregation statistical curve group and the theoretical uniform distribution curve, screen out spatial areas where the deviation exceeds the set aggregation intensity judgment benchmark value, and output them as linkage warning areas.

Images