Method for diagnosing transformer component faults based on machine learning

Abstract

The present application relates to the technical field of transformer, specifically to a transformer component fault diagnosis method based on machine learning, comprising the following steps: analyzing medium data collected by a transformer component, comparing dielectric loss factor sequence and offset characteristics, training to generate offset time sequence characteristic set, screening abnormal period and aggregating key discharge characteristics, obtaining abnormal clustering signal group, inducing trend evolution sequence, layering optimization path structure, and outputting fault link positioning data.In the present application, through comprehensive analysis of multiple parameters such as dielectric loss, discharge, temperature rise, gas and vibration, abnormal distribution is actively screened, key node trend change is captured in real time, and historical and real-time characteristics under different operation scenarios are fused, so that complex link evolution and implicit abnormalities are efficiently analyzed, the fault evolution relationship between nodes is structured and output, the direct application of the diagnosis result in full-link tracing and multi-node trend judgment is supported, and the fault diagnosis pertinence and hidden danger identification ability are effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transformers, in particular to a transformer component fault diagnosis method based on machine learning. BACKGROUND

[0002] The technical field of transformers relates to the design, manufacture, operation and maintenance of transformer equipment in the process of electric energy conversion, transmission and distribution, the core matters of this field include performance optimization and reliability improvement of transformer core, coil, insulation structure, cooling system and other components, and also cover monitoring of transformer operating state, development of maintenance strategy, and research and application of fault diagnosis and preventive measures, among which, traditional transformer fault diagnosis refers to identifying abnormal state of transformer components based on physical quantity change detection means such as thermal imaging detection, oil chromatographic analysis and acoustic emission monitoring, aiming at identifying component-level faults such as local overheating of transformer core, coil insulation aging and cooling system failure, means such as oil gas component analysis, partial discharge detection, core vibration signal extraction and temperature rise test are usually used to realize diagnosis and preliminary identification of different component faults of transformer.

[0003] In the process of daily equipment detection and operation management, the prior art relies on a single index as the basis for fault, which is difficult to form global trend monitoring and link traceability judgment, and is prone to miss local linkage hidden dangers in the face of multiple types of abnormalities or complex evolution phenomena across nodes, and in actual work, if only oil gas analysis or single point temperature rise test is used to evaluate equipment health, it is impossible to understand the multi-link interaction and trend synchronous change, which is particularly obvious in multi-working condition interaction, long-term operation load fluctuation or early hidden fault accumulation, affecting the timely discovery and accurate positioning of fault risks. SUMMARY

[0004] The purpose of the present application is to solve the shortcomings in the prior art, and a transformer component fault diagnosis method based on machine learning is proposed.

[0005] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows: a transformer component fault diagnosis method based on machine learning, comprising the following steps:

[0006] S1: based on the transformer component, analyzing the collected medium data, comparing the dielectric loss factor sequence of the difference components, judging the offset characteristics of the same frequency point signal, and integrating the signal differences under each working condition to obtain an offset time sequence feature set;

[0007] S2: based on the offset time sequence feature set, screening the offset time period, calculating the discharge amplitude change of the transformer component, analyzing the abnormal mutation performance, judging the aggregation of the signal in the feature space, and obtaining an abnormal clustering signal group;

[0008] S3: based on the abnormal clustering signal group, calculate the transformer component temperature rise monitoring information, compare the gas component change, judge the vibration signal center frequency change, summarize the key feature dynamic evolution, combine the historical data, and obtain the trend evolution sequence;

[0009] S4: according to the trend evolution sequence, judge the trend change, analyze the arrangement order of the transformer component in the feature space, compare the main path feature trajectory, optimize the path hierarchical structure, summarize the hierarchical result, and obtain the path main hierarchical structure.

[0010] The application improves that the offset time sequence feature set includes insulation performance index, spectrum change feature, working condition correlation factor, the abnormal clustering signal group includes discharge amplitude identifier, abnormal classification number, signal cluster feature, the trend evolution sequence includes temperature rise trajectory parameter, gas change mode, vibration response feature, and the path main hierarchical structure includes node level relationship, link state set and hierarchical index parameter.

[0011] The application improves that the offset time sequence feature set is obtained by the following steps:

[0012] S111: based on the transformer component, analyze the collected medium data, compare the dielectric loss factor sequence of the same frequency point at the difference time point, judge the offset amplitude of the dielectric loss factor in the time dimension, calculate the change range of each frequency point sequence, identify the frequency point trajectory with offset, and obtain the offset trajectory set;

[0013] S112: based on the offset trajectory set, compare the performance of the dielectric loss factor under each operating condition, analyze the influence of the working condition on the frequency point trajectory change, identify the key feature trajectory of the offset performance under the difference working condition, and integrate the analyzed signal features, and obtain the offset time sequence feature set.

[0014] The application improves that the abnormal clustering signal group is obtained by the following steps:

[0015] S211: based on the offset time sequence feature set, analyze the insulation performance parameter and the spectrum change signal, optimize the continuous change trend of each feature in the difference period, compare the fluctuation range in the time sequence, screen the feature curve mutation section, judge the correlation performance of the signal amplitude rise and the discharge monitoring data, and obtain the discharge mutation amplitude sequence;

[0016] S212: based on the discharge mutation amplitude sequence, compare the spatial distribution of signal mutation in each section, analyze the combination relationship of amplitude change and duration period, calculate the spatial combination difference degree, identify the abnormal signal point of the structure, and obtain the abnormal discharge clustering point.

[0017] S213: Based on the abnormal discharge cluster points, select feature points with large signal amplitude changes, long duration periods and high spatial density, compare the aggregation relationship of each feature point, analyze the spatial structure between representative points, and obtain abnormal cluster signal groups.

[0018] The present invention is improved in that the steps for obtaining the trend evolution sequence are specifically as follows:

[0019] S311: Based on the abnormal clustering signal group, calculate the discharge intensity change trend of each monitoring node in the target time period, compare the gas composition change and temperature rise of each node in the same period, optimize the trend characteristics of each monitoring point, and obtain the discharge trend coefficient.

[0020] S312: Determine the distribution characteristics of the discharge trend coefficient among the various structures, analyze the gas concentration changes and temperature rise amplitude at each monitoring point, identify the offset of the center frequency of the vibration signal at each node, and obtain the structural evolution trend degree.

[0021] S313: Based on the structural evolution trend, analyze the temperature rise trajectory, gas concentration change and frequency shift of each node in the same monitoring period, compare the time series changes in historical state data, and obtain the trend evolution sequence.

[0022] The present invention is improved in that the step of obtaining the hierarchical structure of the path backbone is specifically as follows:

[0023] S411: Based on the trend evolution sequence, determine the changing trend of each characteristic parameter of the transformer component during the operating cycle, compare the trend differences of each parameter in different time periods, and obtain the component change trajectory index.

[0024] S412: Based on the component change trajectory index, analyze the spatial distribution of characteristic points of each node of the transformer under temperature rise, gas and vibration parameters, identify the order and arrangement differences between each node, and obtain the component characteristic arrangement structure;

[0025] S413: Based on the component feature arrangement structure, determine the hierarchical relationship of each node in the path, analyze the spatial coordinate difference, signal direction change and distribution between nodes, and obtain the hierarchical structure of the path backbone.

[0026] The present invention is improved in that the steps further include:

[0027] S5: Based on the hierarchical structure of the main path, analyze each path segment, compare the diagnostic characteristics of temperature rise changes, identify the correlation between the increase in gas component rate and the shift in vibration frequency, adjust the node order, and obtain fault link location data.

[0028] The fault link location data includes the abnormal node number, fault link mapping information, and diagnostic index set.

[0029] The present invention is improved in that the steps for obtaining the fault link location data are specifically as follows:

[0030] S511: Based on the hierarchical structure of the main path, compare the temperature rise monitoring sequence and gas composition change sequence of each path segment, determine the characteristic distribution of each path segment during the temperature rise change process, and combine the actual arrangement of nodes in the hierarchical structure to optimize the correlation performance between path segment diagnostic features and obtain temperature rise gas correlation data.

[0031] S512: Based on the temperature rise gas correlation data, analyze the synchronicity between the gas component change sequence and the vibration frequency shift sequence, screen the behavior of gas component rate increase and vibration main frequency shift synchronizing, and calculate the spatial distribution of the synchronizing behavior between path segment nodes to obtain the synchronization feature aggregation group.

[0032] S513: Based on the aforementioned synchronization feature aggregation group, optimize the node connection order in the main path structure, identify node paths with associated characteristics, analyze the abnormal node numbers and link mapping information of each key path segment, and obtain fault link location data.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] This invention integrates multiple parameters such as dielectric loss, discharge, temperature rise, gas, and vibration. By actively screening for abnormal distributions and capturing real-time trend changes of key nodes, it fuses historical and real-time characteristics under different operating scenarios to efficiently analyze complex link evolution and hidden anomalies. It outputs the fault evolution relationship between nodes in a structured manner, supporting the direct application of diagnostic results in full-link tracing and multi-node trend analysis. This effectively improves the pertinence of fault diagnosis and the ability to identify hidden dangers, thereby enhancing the accuracy and foresight of fault diagnosis. Attached Figure Description

[0035] Figure 1 This is a flowchart of the main steps of the present invention;

[0036] Figure 2 This is a flowchart of the process for obtaining the offset temporal feature set in this invention;

[0037] Figure 3 This is a flowchart illustrating the acquisition of abnormal clustering signal groups in this invention.

[0038] Figure 4 This is a flowchart illustrating the process of obtaining the trend evolution sequence in this invention.

[0039] Figure 5 This is a flowchart illustrating the process of obtaining the hierarchical structure of the path backbone in this invention.

[0040] Figure 6 This is a flowchart illustrating the process of acquiring fault link location data in this invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the 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 merely illustrative and not intended to limit the invention.

[0042] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0043] Example

[0044] Please see Figure 1 This invention provides a technical solution: a machine learning-based method for diagnosing transformer component faults, comprising the following steps:

[0045] S1: Based on transformer components, analyze the collected dielectric data, compare the dielectric loss factor sequences of different transformer components, determine the time-diversion characteristics of signals at the same frequency, optimize the performance differences of the collected signals under various operating conditions, integrate the analyzed signal features, input them into the machine learning process, execute dataset training, and obtain the offset time series feature set.

[0046] S2: Based on the offset time series feature set, the corresponding offset concentrated time period is selected, the discharge amplitude change of the transformer component is calculated, the abnormal sudden change performance during the operation of the component is analyzed, the signal aggregation pattern in the feature space is determined, the spatial distribution between feature points is compared, key abnormal discharge features are aggregated, and abnormal clustering signal groups are obtained.

[0047] S3: Based on the abnormal clustering signal group, calculate the temperature rise monitoring information of the transformer component, compare the changes in the gas composition of the component, determine the trend of the vibration signal center frequency, summarize the dynamic evolution of the key features of the component, and obtain the trend evolution sequence by combining historical state data.

[0048] S4: Based on the trend evolution sequence, determine the trend changes it exhibits, analyze the arrangement order of transformer components in the feature space, compare the characteristic trajectories of the main path, optimize the hierarchical structure of each component path, summarize the hierarchical results, and obtain the hierarchical structure of the main path.

[0049] S5: Based on the hierarchical structure of the main path, analyze each path segment involved, compare the diagnostic characteristics during the temperature rise process, identify the correlation characteristics between the gas component rate increase and vibration frequency shift of the transformer components, adjust the connection order between path nodes, and obtain fault link location data.

[0050] The offset time-series feature set includes insulation performance indicators, spectral change characteristics, and operating condition correlation factors; the abnormal clustering signal group includes discharge amplitude identifier, abnormal classification number, and signal cluster characteristics; the trend evolution sequence includes temperature rise trajectory parameters, gas change patterns, and vibration response characteristics; the path backbone hierarchical structure includes node hierarchy relationships, link status set, and hierarchical index parameters; and the fault link location data includes abnormal node number, fault link mapping information, and diagnostic index set.

[0051] Transformer components in S1 refer to insulation components (such as main insulation paper layers, end insulation pads, and lead wire insulation). The analyzed dielectric data mainly involves the dielectric loss factor of the insulation system, focusing on insulation aging, local losses, and changes in material properties. These are the core data acquisition objects for judging changes in dielectric loss and insulation status. Transformer components in S2 refer to coils, leads, lead-out terminals, and components related to partial discharge monitoring. The focus is on characteristics such as discharge amplitude and abnormal changes. Related signals usually come from areas prone to partial discharge, such as windings (coils), lead-out terminals, and lead-out devices. These are key monitoring points for electrical anomalies. Transformer components in S3 refer to components related to temperature rise, gas, and vibration monitoring points, such as the core, oil ducts, cooling system, and seals. The analysis considers the comprehensive status of temperature rise, gas composition, and vibration signals. The first category, S4, involves the transformer's internal core, oil duct temperature points, gas online monitoring interfaces, and vibration sensing units, reflecting the overall operation and damage trends. The second category, S5, refers to all key operating components (such as functional units, nodes / ports, and path nodes in the main circuit), focusing on all major monitoring objects and their associated paths. It is a comprehensive judgment of the trends and main path structure among all key components, reflecting the global fault chain relationship. The third category, S5, refers to all component nodes participating in the path link (such as components / nodes with significantly abnormal indicators such as temperature rise, gas, and vibration). It targets actions such as path layering, node sorting, and link positioning, actually involving all components with abnormal indicators. It integrates all the aforementioned monitoring points and nodes for tracing and locating the main fault chain.

[0052] In S1, the dielectric loss factor sequence refers to the continuous measurement data sequence of the dielectric loss factor (tgδ) of the transformer insulation medium at different acquisition times or frequency points, reflecting the change of insulation state with time and frequency; the same frequency signal refers to the dielectric loss factor signal obtained at different time points under the same detection frequency, corresponding to the data trajectory of a specific frequency point, which is convenient for comparing the insulation aging trend; the performance difference refers to the difference in the value or trend of the dielectric loss factor or other insulation characteristics under different operating conditions (such as load, voltage, temperature changes, etc.), reflecting the change of insulation characteristics under the influence of operating conditions.

[0053] In S2, the offset concentration period refers to a specific time interval in which signals such as dielectric loss factor show abnormal drift or drastic changes, which usually indicates insulation performance deterioration or fault signs; the discharge amplitude change refers to the amount of change in discharge intensity or amplitude in the electrical signal obtained through partial discharge detection, which can indicate local insulation breakdown, discharge and other anomalies; abnormal mutation manifestation refers to the phenomenon of drastic changes, sharp increases and decreases in monitoring signals such as discharge, current and voltage in a short period of time, which are signal characteristics of local anomalies or initial faults; the clustering pattern refers to the similarity distribution and classification rules of multiple anomaly points (such as discharge mutation points) in the feature space (such as amplitude, duration, shape and other dimensions), which is used to identify anomalies of the same source or similar faults; the key abnormal discharge feature refers to the representative discharge signal features obtained through the above clustering and screening (such as anomalies with the largest amplitude, longest duration and special shape), which can be used as an important input for subsequent fault judgment.

[0054] In S3, gas composition change refers to the change in the concentration of various characteristic gases (such as H2, C2H2, CH4, etc.) in transformer oil over time, which is an important parameter for fault diagnosis in gas analysis in oil (DGA); vibration signal refers to the mechanical vibration data collected by vibration monitoring sensors in transformer structural components, shell, coils, etc., used to reflect abnormalities such as mechanical loosening, partial discharge, or internal impact; dynamic evolution of key features refers to the trajectory and trend of various characteristic parameters such as temperature rise, gas composition, and vibration during operation, which is a complete record of the evolution of equipment health status; historical status data refers to various transformer operating characteristic data collected and archived at different time points and under different operating conditions, used for horizontal or vertical trend comparison and machine learning modeling training.

[0055] In S4, the main path refers to the key fault propagation path identified between components based on signal transmission, fault impact links, physical connections, and other relationships. It is the main chain for source tracing and diagnosis. The characteristic trajectory refers to the change path of characteristic parameters (such as temperature rise, gas, vibration, etc.) at each node along the main path, which is used to analyze the evolution and transmission process of fault signals. The component paths refer to the multiple signal flow or fault propagation paths formed between multiple functional components inside the transformer, which comprehensively reflect the complex physical and electrical link relationships inside the equipment.

[0056] In S5, gas composition refers to the distribution and changes of various characteristic gases (H2, C2H2, CO, CH4, etc.) in the oil at different path nodes, which is the basis for locating fault points and judging deterioration trends; vibration frequency refers to the dominant frequency change of mechanical vibration signals collected by component nodes, reflecting existing loosening, abnormal impact, internal electromagnetic faults, etc.; path nodes refer to specific component nodes (such as iron core, coil, lead wire, cooler, etc.) in the main layered structure, which are important observation and judgment locations on the fault link.

[0057] Please see Figure 2 The specific steps for obtaining the offset temporal feature set are as follows:

[0058] S111: Based on transformer components, analyze the collected dielectric data, compare the dielectric loss factor sequence at the same frequency point under different time points, determine the offset amplitude of the dielectric loss factor in the time dimension, calculate the change range of each frequency point sequence, identify the offset frequency point trajectory, and obtain the offset trajectory set.

[0059] Based on the insulation components in the transformer, such as the main insulation paper layer and lead wire insulation, the dielectric data collected at different time points is first extracted. This data mainly consists of measured values ​​of the dielectric loss factor, generally derived from periodic testing or online monitoring systems. Each test is conducted at multiple frequency points (e.g., 50Hz, 100Hz, 200Hz, etc.). A time series is established for each frequency point. For example, at the 50Hz frequency point, the dielectric loss factor measured in January, February, and March are 0.0082, 0.0091, and 0.0113, respectively. The measured values ​​need to be classified according to frequency points, establishing a two-dimensional frequency-time sequence structure. The difference between each pair of adjacent time points in this sequence is calculated, yielding results of 0.0009 and 0.0022. The offset of the dielectric loss factor is then determined. Using a set offset judgment threshold, such as 0.0015, the change values ​​of each frequency point are compared. If any interval change value exceeds the threshold, such as 0.0022 > 0.0015, then the frequency point is marked as an offset point within this time period. Next, all frequency points are traversed, and the above comparison is repeated in the time series of each frequency point to determine whether there is an abnormal offset. In the time series of each frequency point, if there are three consecutive upward trends in the dielectric loss factor, and the three increments are 0.0010, 0.0013, and 0.0015 respectively, the frequency point is recorded as a continuous offset trajectory. All frequency points with offset records and their change values ​​are combined into a set of data, namely the offset trajectory set. This set is used for subsequent operating condition difference analysis and signal feature extraction.

[0060] S112: Based on the offset trajectory set, compare the performance of dielectric loss factor under various operating conditions, analyze the influence of operating conditions on the frequency point trajectory change, identify the key characteristic trajectories of offset performance under different operating conditions, and integrate the analyzed signal features to obtain the offset timing feature set.

[0061] The offset trajectory set is correlated with the transformer's operating condition data during the corresponding time period. Operating condition data generally includes variables such as operating current, voltage, oil temperature, and ambient temperature. For example, the offset trajectory shows that the dielectric loss factor at frequency point 200Hz increased from 0.0078 to 0.0104 between April and May. Simultaneously, the transformer load current increased from 480A to 760A, and the oil temperature increased from 62℃ to 76℃ during this period. This indicates that the trajectory offset is related to high load and high temperature conditions. All offset frequency point trajectories are grouped for comparison. For example, the dielectric loss factor sequence at the same frequency point under different operating conditions (such as low load and low temperature versus high load and high temperature) is selected, and its average value and fluctuation range are compared. If the average dielectric loss factor is 0.0 under high load... The value of 098 is 0.0081 under low load, with a difference of 0.0017. The standard deviation under high load is 0.0009, which is significantly higher than 0.0004 under low load. Therefore, it can be determined that this frequency point shows significant differences under different operating conditions. Further screening is conducted to identify trajectories with obvious trends, such as those with consecutive increases of 0.0012, 0.0016, and 0.0019 at three different time points, which are considered to be continuously rising trends. These key trajectories are recorded with their frequency point number, operating condition type, trend type, maximum amplitude, mean, and standard deviation, and encapsulated as structured feature records. All such records are integrated to form an offset time series feature set for subsequent cluster analysis and diagnostic path construction.

[0062] Please see Figure 3 The specific steps for obtaining the abnormal clustering signal group are as follows:

[0063] S211: Based on the offset time series feature set, analyze the insulation performance parameters and spectral change signals, optimize the continuous change trend of each feature in different time periods, compare the fluctuation range under the time series, screen the abrupt change segment of the feature curve, judge the correlation between the signal amplitude increase and the discharge monitoring data, and obtain the discharge abrupt change amplitude sequence.

[0064] The insulation performance parameters and their corresponding spectral response signal values ​​are extracted from each characteristic curve. The insulation performance parameters refer to the dielectric loss factor values ​​at the acquisition time points, while the spectral change signal refers to the amplitude information of different frequency components extracted from partial discharge detection or dielectric response testing. During the process, the insulation dielectric signal at each frequency point needs to be analyzed one by one to extract the dielectric loss factor sequence corresponding to the frequency points of 100Hz, 200Hz, and 400Hz. For example, at 100Hz, the time sequence values ​​are 0.0085, 0.0090, 0.0093, and 0.0120, and at 200Hz, they are 0.0079, 0.0081, 0.0084, and 0.0099. After the sequence is sorted, the first-order difference change value of each curve is calculated in time sequence to obtain the continuous change amplitude sequence. Then, the absolute values ​​of the differences between two adjacent time points are summed to determine whether there are obvious fluctuation segments. If two consecutive differences within any segment are greater than 0.0015, it is determined that there is a trend change. Each change point is marked. Subsequently, the amplitude response of the spectral signal within the change segment is compared to determine whether the energy of the main frequency component in the frequency domain within that segment has increased. If the spectral amplitude value within the change segment increases from 0.42 to 0.78, and the monitored discharge intensity voltage amplitude increases from 2.1mV to 5.3mV, then the change segment is recorded as a region with a significant increase in signal amplitude. According to the position of the change in each characteristic curve, the start time point, end time point, duration, initial amplitude value, peak amplitude, and corresponding discharge amplitude value of the change need to be extracted, and a change amplitude sequence needs to be constructed.

[0065] S212: Based on the discharge mutation amplitude sequence, the spatial distribution of signal mutations within each segment is compared, and the combined relationship between amplitude change and duration period is analyzed using the formula:

[0066]

[0067] Calculate the spatial combination difference KG, identify anomalous signal points in the structure, and obtain anomalous discharge cluster points, where M i Let P be the amplitude change at the i-th signal point. i Let Q be the duration of the i-th signal point. i Let R be the interval parameter of the i-th signal point in the spatial distribution. i is the number of time superpositions for the i-th signal point, and n is the total number of signal points;

[0068] To determine whether each signal point belongs to the same type of discharge region structure, analyze its projection relationship in the winding arrangement and sensor array coordinates, and extract the amplitude change M corresponding to each signal point. i Continuous period P i Spatial interval Q i Number of times superimposed with time R iAfter normalization, the values ​​are uniformly weighted and entered into the combined operation model, with the amplitude change M. i The duration P is obtained by detecting the change in the peak amplitude of the partial discharge waveform. i The continuous time length of the mutation segment, and the spatial interval Q. i The number of time stacking times (R) is obtained by converting the three-dimensional distance of the point in the device coordinate system. i The parameters are accumulated based on the number of times the outlier is triggered repeatedly within the sliding time window. Taking four typical signal points as examples, the parameter values ​​are as follows (the values ​​in parentheses are normalized values):

[0069] Signal point 1: M1 = 3.5 (0.538), P1 = 2.0 (0.5), Q1 = 1.2 (0.4), R1 = 2 (0.333);

[0070] Signal point 2: M2 = 4.2 (0.646), P2 = 2.5 (0.625), Q2 = 1.5 (0.5), R2 = 3 (0.5);

[0071] Signal point 3: M3=2.8(0.431), P3=1.8(0.45), Q3=1.1(0.366), R3=2(0.333);

[0072] Signal point 4: M4 = 5.1 (0.785), P4 = 3.0 (0.75), Q4 = 1.3 (0.433), R4 = 4 (0.666);

[0073] After substituting into the formula, the calculation is as follows:

[0074] Molecular part:

[0075] (0.538·0.5-0.4·0.333)=0.269-0.133=0.136;

[0076] (0.646·0.625-0.5·0.5)=0.404-0.25=0.154;

[0077] (0.431·0.45-0.366·0.333)=0.194-0.122=0.072;

[0078] (0.785·0.75-0.433·0.666)=0.589-0.288=0.301;

[0079] The sum of the numerators is:

[0080] 0.136 + 0.154 + 0.072 + 0.301 = 0.663;

[0081] Denominator part:

[0082] (0.538·0.5+0.4·0.333)=0.269+0.133=0.402;

[0083] (0.646·0.625+0.5·0.5)=0.404+0.25=0.654;

[0084] (0.431·0.45+0.366·0.333)=0.194+0.122=0.316;

[0085] (0.785·0.75+0.433·0.666)=0.589+0.288=0.877;

[0086] The sum of the denominators is:

[0087] 0.402 + 0.654 + 0.316 + 0.877 = 2.249;

[0088] The combined dissimilarity is calculated as follows:

[0089]

[0090] The results indicate that the difference between structural factors and signal characteristics in signal mutation behavior is 0.2948, which is lower than the lower limit of 0.35 for combined aggregation difference. Therefore, it can be considered that although there is a mutation phenomenon in the current signal point, it does not have a significant clustering relationship in terms of spatial concentration characteristics. Subsequently, only points with KG exceeding the set threshold are selected for anomaly point aggregation and extraction to construct an abnormal discharge cluster point set. The formula, by introducing the product of normalized discharge signal intensity and time characteristics, as well as the composite parameters of equipment structure distribution and temporal repeatability, can form a two-way comparison mechanism for spatial aggregation, so that spatial anomaly identification is not only based on structural density, but also takes into account the temporal activity of signal mutation, thereby improving the stability and discrimination ability of fault signal aggregation judgment.

[0091] S213: Based on the clustering points of abnormal discharge, feature points with large signal amplitude changes, long duration periods and high spatial density are selected. The aggregation relationship of each feature point is compared and the spatial structure between representative points is analyzed to obtain the abnormal clustering signal group.

[0092] Spatial combination difference degree refers to a quantitative index used in transformer fault diagnosis to measure the degree of combination difference of a group of signal points in spatial dimension structure. It is used to determine whether different signal points (usually discharge mutation signals) exhibit aggregation patterns or discrete characteristics in a space composed of multi-dimensional parameters.

[0093] First, each identified discharge amplitude curve sequence is traversed, and feature points whose amplitude changes exceed a set judgment threshold are extracted. This amplitude change threshold can be set with reference to twice the average discharge amplitude. For example, if the current background average discharge amplitude is 1.8mV, the judgment threshold is set to 3.6mV. If the amplitude of a feature point rises to 4.1mV, its amplitude change is considered to meet the condition. The duration of the abrupt change in the curve containing this feature point is further extracted. If the discharge amplitude continues to exceed the judgment threshold for five consecutive sampling periods (5 seconds per period), this point is marked as a feature point with a long duration. Subsequently, the spatial coordinate information between this feature point and other anomaly points is obtained. For example, the three-dimensional position of its sensor node in the transformer housing coordinate system is recorded as X = 3.5m, Y = 1.2m, Z = 1.2m. 2.0m. By calculating the number of anomalous points within a 1-meter radius around each feature point, if more than 15 anomalous points are detected within this radius, the spatial density of that point meets the high-density judgment condition. Based on the above conditions, all feature points with large amplitude variations, long duration periods, and high spatial density are selected to form a feature point set. All points in the set are compared according to their positional relationships to obtain the Euclidean distance between point pairs and to calculate their average spacing. If the average spacing is less than 1.2 meters, it is considered that there is a spatial aggregation relationship between the points. Then, spatial cluster centers are established based on the average coordinates of each group of aggregation points. The connection relationship between the centers is topologically reconstructed, and the clusters are numbered according to the degree of aggregation. The coordinates of the representative points of each cluster, the number of anomalous points contained therein, and the spatial expansion range are identified. The output is an anomalous clustering signal group.

[0094] Please see Figure 4 The specific steps for obtaining the trend evolution sequence are as follows:

[0095] S311: Based on the abnormal clustering signal group, calculate the discharge intensity change trend of each monitoring node in the target time period, compare the gas composition change and temperature rise of each node in the same period, optimize the trend characteristics of each monitoring point, and obtain the discharge trend coefficient.

[0096] The discharge amplitude data of each corresponding monitoring node in the transformer is retrieved sequentially during the time period when the abnormal signal occurs. A discharge intensity sequence for that node within the selected time interval is constructed. For example, the discharge intensity records for node A at 0 minutes, 5 minutes, 10 minutes, and 15 minutes are 1.8mV, 2.5mV, 3.4mV, and 5.2mV, respectively. The amplitude variation difference sequence of that node within adjacent sampling time intervals is calculated to obtain the discharge change trend of that node in the target time period. By traversing all monitoring nodes with abnormal clustering signals, a discharge change curve for each node is formed. Then, the gas composition change data collected by that node within the same time period is retrieved to analyze the gas concentration, such as hydrogen concentrations of 152ppm, 180ppm, 235ppm, and 312ppm. Simultaneously, temperature rise data is extracted, such as oil temperature measurement points of 64.1℃, 65.3℃, and 66℃. For the three types of data (0.7℃, 68.4℃), time trend extraction was performed. The difference amplitude at each time point was normalized and then sorted. By comparing whether the discharge amplitude trend, gas concentration trend, and temperature rise trend increased synchronously, the coupling characteristics of the abnormal signal evolution at that node were determined. The same process was then performed on all nodes. The discharge change rate, gas growth rate, and temperature rise rate were calculated according to the trend rise slope of each node, and a trend comparison matrix was constructed. Nodes with strong trend consistency were assigned higher trend weight coefficients. The trend weight coefficient range was set from 0 to 1. If the discharge slope of a node is 0.24, the gas slope is 0.21, and the temperature rise slope is 0.22, and the variance of the three is within 0.0002, then a weight coefficient of 0.92 is given. Finally, the nodes were sorted according to their weight coefficients to form a trend level, and the discharge trend coefficient of each node was output.

[0097] S312: Determine the distribution characteristics of the discharge trend coefficient among various structures, analyze the gas concentration changes and temperature rise amplitude at each monitoring point, identify the shift in the center frequency of the vibration signal at each node, and use the following formula:

[0098]

[0099] Obtain the structural evolution trend degree EG Δ Among them, VG ch VG represents the current gas component concentration at monitoring point h. bh The reference gas component concentration at monitoring point h is TG. max,h TG represents the highest temperature during the temperature rise monitoring process at monitoring point h. min,h GQ represents the lowest temperature during the temperature rise monitoring process at monitoring point h. h Let n be the discharge trend coefficient corresponding to the h-th monitoring point. EG Gf represents the number of gas and temperature rise monitoring points. cj Let Gf be the current center frequency of the j-th vibration monitoring point. 0jLet m be the reference center frequency of the j-th vibration monitoring point. EG This refers to the number of vibration monitoring points;

[0100] Based on the discharge trend coefficients of the three monitoring nodes P1, P2, and P3, the current gas concentration VG of each node is obtained by comparing the corresponding gas concentration changes and temperature rise spans. ch Reference gas concentration VG bh The highest temperature rise point (TG) max,h The lowest temperature rise point (TG) min,h and discharge trend coefficient GQ h Substituting into the calculation formula, we further introduce the current center frequency Gf corresponding to vibration monitoring points V1 and V2. cj and reference center frequency Gf 0j Among them, the gas concentration at node P1 increased from 100 μL / L to 125 μL / L, and the temperature rise increased from 65℃ to 78℃, with a discharge trend coefficient of 0.82, which, after normalization, were 0.25, 0.52, and 0.41, respectively; the gas concentration at node P2 increased from 140 μL / L to 150 μL / L, and the temperature rise increased from 70℃ to 85℃, with a discharge trend coefficient of 0.95, which, after normalization, were 0.10, 0.63, and 0.48, respectively; the gas concentration at node P3... As the concentration of particulate matter increased from 160 μL / L to 170 μL / L, the temperature rise was 76℃ to 92℃, with a discharge trend coefficient of 1.10, which, after normalization, were 0.10, 0.73, and 0.52, respectively. The current center frequency of vibration monitoring point V1 was 62.5 Hz, with a reference frequency of 60.0 Hz; the current frequency of V2 was 68.0 Hz, with a reference frequency of 64.0 Hz. The normalized frequency offsets were 0.18 and 0.28, respectively. Substituting these values ​​into the formula, calculations were performed.

[0101] The molecular part is:

[0102] (0.25·0.52·0.41)+(0.10·0.63·0.48)+(0.10·0.73·0.52)

[0103] =0.0533 + 0.0302 + 0.03796 = 0.12146;

[0104] The denominator is:

[0105]

[0106] The degree of structural evolution trend is:

[0107]

[0108] This result indicates that the structural evolution trend degree EG ΔThe value of ≈0.1153 reflects the structural response intensity formed by the combined effects of gas concentration changes, temperature rise range, and discharge trend at multiple monitoring nodes within the current period. After being normalized by the vibration center frequency offset, this value centrally reflects the composite evolution level among multiple physical characteristic indicators. The higher the value, the stronger the synchronous fluctuation of multiple indicators in the structural hierarchical response. The lower the value, the weaker the coupling between various monitoring characteristics. The formula achieves a cross-dimensional integrated expression in the determination of structural trends by simultaneously integrating discharge trend, gas concentration, and temperature rise amplitude, and normalizing them in the frequency perturbation space.

[0109] S313: Based on the structural evolution trend, analyze the temperature rise trajectory, gas concentration change and frequency shift of each node in the same monitoring period, compare the time series changes in historical state data, and obtain the trend evolution sequence.

[0110] Structural evolution trend is an indicator used to measure the comprehensive state change trend of internal structural components of a transformer during operation. It mainly reflects the degree of response change of structural components under the combined action of multiple characteristic parameters. It is the result of comprehensive quantification of the operating state evolution trend of transformer structural components based on the combined performance of gas concentration change, temperature rise amplitude and discharge trend at structural monitoring points, combined with the degree of vibration frequency deviation.

[0111] Temperature rise data sequences, gas concentration data sequences, and vibration frequency data sequences for all monitoring nodes within the same period were extracted. Temperature rise data originated from thermistors distributed near the core, windings, and oil passages. For example, at node B, the oil temperature was recorded as 67.2℃, 68.5℃, 69.3℃, and 70.9℃ in a continuous monitoring period; the gas concentration increased from 15ppm (C₂H₂) to 22ppm, 29ppm, and 40ppm; and the frequency shift changed from 128Hz to 131Hz, 134Hz, and 139Hz. First, the growth rate and average rate of change of the temperature rise sequence were calculated. The threshold for judging temperature rise is set at 2℃ per cycle. If the temperature rise is greater than this threshold for two consecutive cycles, it is marked as an abnormal temperature rise segment. At the same time, it is analyzed whether the gas increase exceeds the baseline value of 6 ppm per day. If the continuous increase is 7 ppm and 8 ppm, the abnormal condition is met. Points in the frequency offset sequence where the change is greater than 3 Hz are marked as abrupt changes. Furthermore, the above-mentioned abnormal points of temperature rise, gas and frequency are marked on the same time axis, and it is analyzed whether the three types of anomalies overlap at certain time points. If the overlap rate of the abnormal time points of the three exceeds 70%, a trend coupling segment is constructed. Next, historical status data is retrieved to examine the trend performance of similar components under the same operating conditions. Monitoring records from three months ago are extracted as a comparison benchmark. For example, the historical temperature rise sequence is 64.2℃, 64.8℃, 65.6℃, and 66.1℃, the gas concentration is stable in the range of 12ppm to 18ppm, and the vibration frequency is maintained between 127Hz and 129Hz. The historical and current values ​​are differentially processed and normalized before being superimposed to form a comparison sequence. If the temperature rise difference exceeds 5℃, the gas concentration increases by more than 15ppm, or the frequency shift exceeds 8Hz, the sequence is identified as a trend evolution abrupt change segment. Finally, the node sequences with significant changes in multiple trend parameters are integrated into a trend evolution sequence.

[0112] Please see Figure 5 The specific steps for obtaining the hierarchical structure of the path backbone are as follows:

[0113] S411: Based on the trend evolution sequence, determine the changing trend of each characteristic parameter of the transformer component during the operating cycle, compare the trend differences of each parameter in different time periods, and obtain the component change trajectory index.

[0114] For each transformer component, time-series data of temperature rise, gas concentration, and vibration frequency are extracted within a specified operating cycle. Each parameter is indexed by time to form an independent sequence structure. For example, the temperature records of a component at 0 minutes, 10 minutes, 20 minutes, and 30 minutes are 63.2℃, 64.5℃, 66.1℃, and 67.8℃, respectively; the acetylene concentration in the gas is 12ppm, 17ppm, 23ppm, and 31ppm, respectively; and the vibration frequency changes are 122Hz, 124Hz, 127Hz, and 131Hz, respectively. Single-point changes within continuous time periods are calculated sequentially for each sequence, and the difference between each pair of adjacent points is recorded as an increase sequence. It is then determined whether this meets the set trend standards. The temperature rise trend standard is set at a temperature rise of no less than 1.2℃ every 10 minutes. If the current three increases are 1.3℃, 1.6℃, and 1.7℃ respectively, then the parameter is considered to be showing a continuous upward trend. The acetylene concentration trend standard is set at an increase of more than 4ppm every 10 minutes. If the increases are 5ppm, 6ppm, and 1.7℃ respectively, then the parameter is considered to be showing a continuous upward trend. A value of 8 ppm satisfies the trend continuity requirement. For vibration frequency, 3 Hz is set as the threshold for abrupt change. If the frequency increase is 2 Hz, 3 Hz, or 4 Hz, then at least the latter two segments meet the abrupt change standard. Subsequently, the trend categories of various parameters are classified and labeled. For example, continuously rising temperature, continuously rising gas concentration, and abrupt changes in vibration frequency are marked as T1, T2, and T3 type features, respectively. The three trends occurring concurrently within the same time period are recorded as joint trend events. At the same time, trend types are compared at different time periods. The mean difference, maximum increase, and trend direction maintenance length of each parameter change within the operating cycle are extracted. The trend stability index is obtained by statistically analyzing the number of trend direction change points. For example, if the temperature rise change direction changes no more than twice in 60 minutes of operation, then the trend is marked as a unidirectional stable trend. Finally, a list of parameter change trends is formed according to the above operations. By integrating the change trajectories of the three types of parameters in different time periods, a sequence index characterizing the change path of the component's operating state is constructed, and the output is the component change trajectory index.

[0115] S412: Based on the component change trajectory index, analyze the spatial distribution of characteristic points of each node of the transformer under temperature rise, gas and vibration parameters, identify the differences in the order and arrangement between each node, and obtain the component characteristic arrangement structure;

[0116] The characteristic peaks of temperature rise, gas concentration, and vibration frequency are extracted from all transformer nodes. The extraction method involves locating the points with the highest rate of change and maximum values ​​in the time series of each parameter as characteristic points. For example, if a node experiences a temperature jump to 69.5℃, a gas concentration of 35ppm, and a vibration frequency increase to 137Hz at the 40th minute of operation, this point is recorded as one of the node's three types of characteristic points. Each node in the spatial structure corresponds to specific coordinate information. Based on the structural layout diagram of the transformer equipment, node A is recorded as being at 2.5 meters on the X-axis, 1.2 meters on the Y-axis, and 3.0 meters on the Z-axis, while node B is recorded as being at 3.0 meters on the X-axis, 1.6 meters on the Y-axis, and 3.1 meters on the Z-axis. Next, the spatial distance between each pair of nodes is calculated, and the order of change of characteristic parameters at corresponding time points is compared. If node A experiences a gas peak at the 30th minute, node A is considered a characteristic point. B experiences a sudden change in vibration frequency at the 35th minute. The order of its changes in the time dimension is determined to be sequential. Combined with its spatial relative position, it is determined whether it is arranged along the main axis of the structure. If the straight line connection direction between nodes is consistent and the time difference of the parameters is less than 10 minutes, a signal propagation path relationship is identified. The arrangement order of the nodes in this path and the spatial connection direction are recorded as an arrangement chain structure. Subsequently, all node combinations are processed sequentially. By extracting the order of parameter changes of each node relative to its preceding and following nodes, as well as the degree of overlap of feature points in the three types of parameters, the arrangement consistency score is calculated. The score threshold is set to 0.8. If the proportion of consistent order in a node sequence is greater than 80%, the sequence is marked as a highly consistent arrangement chain. All highly consistent arrangement chains are classified and numbered to construct a node group arrangement structure information set.

[0117] S413: Based on the component feature arrangement structure, determine the hierarchical relationship of each node in the path, analyze the spatial coordinate difference, signal direction change and distribution between nodes, using the formula:

[0118]

[0119] The hierarchical structure of the main path is obtained, where HB represents the hierarchical coordination level of the main path, reflecting the hierarchical coordination status of each node in the transformer component main path. It is the signal strength direction difference of the z-th node in the θ direction, representing the degree of characteristic change of a specific node in a specified direction, DB. z ΔTB is the spatial coordinate difference between the z-th node and its neighboring nodes, reflecting the physical distribution distance of the nodes in the structural path. z It represents the offset of the z-th node in the trend cycle, reflecting the degree of dynamic evolution of node characteristics over time. HB It is the total number of nodes within the main path, used to determine the range of the summation, BR v is the spatial density distribution of the v-th node, representing the clustering of this node in the main path distribution. The layered coordinate offset of the v-th node in the α direction reflects the change in the node's position in different hierarchical structures.

[0120] The level of hierarchical coordination of the main path is an indicator that measures whether multiple key nodes in the main path have consistency and reasonable hierarchical relationships in terms of space, time and feature evolution. It can help identify potential abnormal combinations, incorrect hierarchical arrangements or disordered transmission of fault signals in the path structure, and is a key parameter for constructing fault chain tracing logic.

[0121] The nodes within the main path of the transformer assembly are numbered and their relative hierarchical position in the structural diagram is determined. The signal propagation direction of each node is then determined based on the path direction. The relative distribution of nodes within the path segment and the direction of change of their corresponding characteristics are analyzed. For each node, three key indicators are obtained: signal strength directional difference, spatial coordinate difference, and trend period offset. Assuming there are four nodes in the main path of the assembly, N1, N2, N3, and N4, the original monitoring data are as follows: Node N1 has a signal strength directional difference of 6.5, a spatial coordinate difference of 2.4, and a trend period offset of 1.3; N2 has 5.2, 3.1, and 1.1; N3 has 7.1, 2.8, and 1.4; and N4 has 6.8, 3.0, and 1. 2. After normalizing all the original data, the normalized signal intensity direction differences are 0.76, 0.58, 0.85, and 0.82; the normalized coordinate differences are 0.60, 0.78, 0.70, and 0.75; and the normalized trend period offsets are 0.68, 0.57, 0.73, and 0.63. Continuing to use the spatial density distribution and hierarchical coordinate offsets for each node, where N1 is 0.90 and 3.0, N2 is 0.85 and 2.8, N3 is 0.92 and 3.2, and N4 is 0.88 and 3.1, the normalized density distributions are 0.72, 0.65, 0.75, and 0.70, and the hierarchical coordinate offsets are 0.81, 0.75, 0.85, and 0.82. Substituting these values ​​into the formula:

[0122] Molecular part:

[0123] N1:

[0124] N2:

[0125] N3:

[0126] N4:

[0127] The summation of the numerators is:

[0128] -0.147 + (-0.386) + (-0.161) + (-0.159) = -0.853;

[0129] The denominator is calculated as follows:

[0130] N1: 0.72·0.81=0.5832;

[0131] N2: 0.65·0.75=0.4875;

[0132] N3: 0.75·0.85=0.6375;

[0133] N4: 0.70·0.82=0.574;

[0134] The sum of the denominators is:

[0135] 0.5832 + 0.4875 + 0.6375 + 0.574 = 2.2822;

[0136] get:

[0137]

[0138] The results indicate that multiple nodes in the current backbone path exhibit collective deviations in signal direction, spatial location, and evolutionary trend, and the calculated HB... p = -0.374 is a negative number, reflecting insufficient coupling and coordination of the nodes in the main path in the spatial and trend dimensions. Further adjustment of the node order or reconstruction of the path structure is needed to obtain the hierarchical structure of the main path. The negative value further indicates that the structure of the path does not have high consistency. Hierarchical optimization should be carried out by adjusting the component order and reorganizing the path segments to build a path with clear hierarchy and continuous structure. Therefore, this value is directly used to identify the node pairs that need to be adjusted in the main path. As an intermediate judgment basis, by comparing with the structural stability reference interval (such as the set reference interval [0.2, 1.5]), the distribution imbalance area can be identified.

[0139] Please see Figure 6 The specific steps for obtaining fault link location data are as follows:

[0140] S511: Based on the hierarchical structure of the main path, compare the temperature rise monitoring sequence and gas composition change sequence of each path segment, determine the characteristic distribution of each path segment during the temperature rise change process, and combine the actual arrangement of nodes in the hierarchical structure to optimize the correlation performance between path segment diagnostic features and obtain temperature rise gas correlation data.

[0141] The temperature rise monitoring sequence and gas composition change sequence corresponding to the nodes involved in each path segment are extracted. The temperature rise monitoring sequence is provided by the temperature sensor at the location of the node. For example, the temperature sequences corresponding to nodes 1 to 3 in path segment A are 67.2℃, 68.5℃, 70.4℃, and 71.3℃, respectively. The gas composition change sequence records the characteristic gas concentrations measured at each time point. For example, acetylene concentrations at the same time point are 12ppm, 17ppm, 25ppm, and 31ppm, respectively. The continuous time difference in the temperature rise curve of each path segment is calculated to obtain the temperature rise slope of each segment and determine whether there is a sudden increase trend. If the temperature rise of a path segment exceeds 1.5℃ for two consecutive sampling periods, it is marked as a high-temperature rise path segment. Similarly, the gas composition curve is processed. If the acetylene concentration increases by 5ppm, 8ppm, and 6ppm for three consecutive periods, respectively, and exceeds the baseline rate of 5ppm, it is recorded as a gas anomaly segment. The process then determines whether the temperature rise and gas change trends overlap within the path segment. If the slopes of the two sequences change in the same direction and the peak time difference does not exceed 10 minutes, they are considered to have trend synchronicity. The actual arrangement of the nodes in the path segment within the hierarchical structure of the main path is then considered, such as node 1 in the upper layer, node 2 in the middle layer, and node 3 in the lower layer. Their vertical arrangement numbers within the structure are recorded. If the heat change develops sequentially from the upper to the lower layer, and the gas concentration rises first at the lower layer nodes before being transmitted to the middle and upper layer nodes, then the temperature rise and gas change in this path segment are considered to have hierarchical offset characteristics. Based on this, a parameter trend matching matrix is ​​constructed for each path segment. A consistency score is calculated based on the time point and intensity of parameter changes corresponding to the nodes. If the score is higher than 0.85, it is recorded as a strongly coupled path segment. The above data is summarized and integrated into a set of path segment-level parameter comparison data, and the output is temperature rise and gas correlation data.

[0142] S512: Based on temperature rise gas correlation data, analyze the synchronicity between gas component change sequence and vibration frequency shift sequence, screen the behavior of gas component rate rise and vibration dominant frequency shift synchronizing, and calculate the spatial distribution of synchronizing behavior among path segment nodes to obtain synchronizing feature aggregation group.

[0143] Gas composition change sequences and vibration frequency shift sequences of key nodes in each path segment are extracted and compared in the time dimension to determine whether they exhibit synchronicity. For example, in path segment B, the acetylene concentration at node X increases from 14 ppm to 21 ppm at minute 20, to 28 ppm at minute 25, and reaches 33 ppm at minute 30, corresponding to vibration frequencies shifting from 129 Hz to 133 Hz, 136 Hz, and 139 Hz, respectively. First, the amplification rate of adjacent sequences within the same time period is calculated, recording the gas change rate as 7 ppm every 5 minutes and the vibration shift rate as 3 Hz every 5 minutes. It is then determined whether they simultaneously exceed their respective reference thresholds, set at 5 ppm for gas rate and 2 Hz for vibration frequency. If the two types of data exceed each other within two or more consecutive time periods... If the threshold is exceeded, it is determined that a synchronization behavior has occurred. The continuity of the synchronization behavior is calculated. If the continuous time is greater than 15 minutes, it is marked as a continuous synchronization behavior. Then, the set of all nodes that have synchronized behavior within the path segment is further counted, and the distribution density of the nodes in the spatial coordinate system is analyzed. The spatial coordinates are read according to the physical location of the nodes. For example, nodes X, Y, and Z are located at (2.0m, 1.2m, 3.0m), (2.4m, 1.3m, 3.2m), and (2.7m, 1.4m, 3.3m) respectively. A spherical region with a radius of 1.0 meter is constructed with each node as the center. The number of synchronized behavior nodes in each region is counted. If a node contains two other synchronized nodes within its radius, then that node is a dense aggregation center. Node groups that meet the requirements of synchronization rate, duration, and spatial density are compiled into synchronization feature aggregation groups.

[0144] S513: Based on the synchronous feature aggregation group, optimize the node connection order in the main path structure, identify the node paths with related characteristics, analyze the abnormal node numbers and link mapping information of each key path segment, and obtain fault link location data.

[0145] Read the connection order of nodes within each aggregation group in the main path structure. If the current order is X→Y→Z, and X, Y, and Z all belong to the same synchronization aggregation group, then the rationality of the connection needs to be re-evaluated. Analyze whether there is a reverse propagation of synchronization behavior, i.e., node Z generates a synchronization event before X in time. If this situation occurs repeatedly in three or more path segments, and the time difference is less than 10 minutes, then the original connection order is considered to be misplaced. Record all node pairs with reverse synchronization time characteristics, extract their connection direction, time difference, spatial distance, and other indicators, and reorder the connection order in the path segments according to time sequence. The node connection relationship is updated to Z→Y→X, and the backbone structure path diagram is corrected. Then, the node where all synchronization feature anomalies first appear is marked in the corrected path diagram and recorded as the anomaly node number. A mapping path between the node and the downstream propagation node is established. For example, if the anomaly first appears at node Z and propagates to X via Y, the link mapping information is recorded as ZYX. At the same time, each link is assigned a number, and the mutation time, mutation amplitude, synchronization strength and other information of the corresponding parameters of each node in the link are summarized to construct a fault signal transmission path model. All mapped links and anomaly node numbers are output as fault link location data.

[0146] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A machine learning-based method for fault diagnosis of transformer components, characterized in that, Includes the following steps: S1: Based on transformer components, analyze the collected dielectric data, compare the dielectric loss factor sequences of different components, determine the offset characteristics of signals at the same frequency, and integrate the signal differences under various operating conditions to obtain the offset time series feature set. S2: Based on the offset time series feature set, filter the offset concentrated time period, calculate the discharge amplitude change of the transformer component, analyze the abnormal sudden change performance, determine the signal aggregation in the feature space, and obtain the abnormal clustering signal group. S3: Based on the abnormal clustering signal group, calculate the temperature rise monitoring information of the transformer components, compare the changes in gas composition, determine the changes in the center frequency of the vibration signal, summarize the dynamic evolution of key features, and combine historical data to obtain the trend evolution sequence. S4: Based on the trend evolution sequence, determine the trend change, analyze the arrangement order of transformer components in the feature space, compare the characteristic trajectories of the main path, optimize the path layering structure, summarize the layering results, and obtain the path main layering structure. S5: Based on the hierarchical structure of the main path, analyze each path segment, compare the diagnostic characteristics of temperature rise changes, identify the correlation between the increase in gas component rate and the shift in vibration frequency, adjust the node order, and obtain fault link location data. The fault link location data includes the abnormal node number, fault link mapping information, and diagnostic index set.

2. The machine learning-based transformer component fault diagnosis method according to claim 1, characterized in that, The offset time-series feature set includes insulation performance indicators, spectral change characteristics, and operating condition correlation factors; the abnormal clustering signal group includes discharge amplitude identifier, abnormal classification number, and signal cluster characteristics; the trend evolution sequence includes temperature rise trajectory parameters, gas change patterns, and vibration response characteristics; and the path backbone hierarchical structure includes node hierarchy relationships, link state set, and hierarchical index parameters.

3. The machine learning-based transformer component fault diagnosis method according to claim 1, characterized in that, The specific steps for obtaining the offset temporal feature set are as follows: S111: Based on transformer components, analyze the collected dielectric data, compare the dielectric loss factor sequence at the same frequency point under different time points, determine the offset amplitude of the dielectric loss factor in the time dimension, calculate the change range of each frequency point sequence, identify the offset frequency point trajectory, and obtain the offset trajectory set. S112: Based on the set of offset trajectories, compare the performance of dielectric loss factor under various operating conditions, analyze the influence of operating conditions on the change of frequency point trajectory, identify the key feature trajectories of offset performance under different operating conditions, and integrate the analyzed signal features to obtain the offset timing feature set.

4. The machine learning-based transformer component fault diagnosis method according to claim 1, characterized in that, The specific steps for obtaining the abnormal clustering signal group are as follows: S211: Based on the offset time series feature set, analyze the insulation performance parameters and spectral change signals, optimize the continuous change trend of each feature in different time periods, compare the fluctuation range under the time series, screen the abrupt change segment of the feature curve, judge the correlation between the signal amplitude increase and the discharge monitoring data, and obtain the discharge abrupt change amplitude sequence. S212: Based on the discharge mutation amplitude sequence, compare the spatial distribution of signal mutations in each segment, analyze the combination relationship between amplitude change and duration period, calculate the spatial combination difference, identify abnormal signal points of the structure, and obtain abnormal discharge cluster points. S213: Based on the abnormal discharge cluster points, select feature points with large signal amplitude changes, long duration periods and high spatial density, compare the aggregation relationship of each feature point, analyze the spatial structure between representative points, and obtain abnormal cluster signal groups.

5. The machine learning-based transformer component fault diagnosis method according to claim 1, characterized in that, The specific steps for obtaining the trend evolution sequence are as follows: S311: Based on the abnormal clustering signal group, calculate the discharge intensity change trend of each monitoring node in the target time period, compare the gas composition change and temperature rise of each node in the same period, optimize the trend characteristics of each monitoring point, and obtain the discharge trend coefficient. S312: Determine the distribution characteristics of the discharge trend coefficient among the various structures, analyze the gas concentration changes and temperature rise amplitude at each monitoring point, identify the offset of the center frequency of the vibration signal at each node, and obtain the structural evolution trend degree. S313: Based on the structural evolution trend, analyze the temperature rise trajectory, gas concentration change and frequency shift of each node in the same monitoring period, compare the time series changes in historical state data, and obtain the trend evolution sequence.

6. The machine learning-based transformer component fault diagnosis method according to claim 1, characterized in that, The specific steps for obtaining the hierarchical structure of the path backbone are as follows: S411: Based on the trend evolution sequence, determine the changing trend of each characteristic parameter of the transformer component during the operating cycle, compare the trend differences of each parameter in different time periods, and obtain the component change trajectory index. S412: Based on the component change trajectory index, analyze the spatial distribution of characteristic points of each node of the transformer under temperature rise, gas and vibration parameters, identify the order and arrangement differences between each node, and obtain the component characteristic arrangement structure; S413: Based on the component feature arrangement structure, determine the hierarchical relationship of each node in the path, analyze the spatial coordinate difference, signal direction change and distribution between nodes, and obtain the hierarchical structure of the path backbone.

7. The machine learning-based transformer component fault diagnosis method according to claim 1, characterized in that, The specific steps for obtaining the fault link location data are as follows: S511: Based on the hierarchical structure of the main path, compare the temperature rise monitoring sequence and gas composition change sequence of each path segment, determine the characteristic distribution of each path segment during the temperature rise change process, and combine the actual arrangement of nodes in the hierarchical structure to optimize the correlation performance between path segment diagnostic features and obtain temperature rise gas correlation data. S512: Based on the temperature rise gas correlation data, analyze the synchronicity between the gas component change sequence and the vibration frequency shift sequence, screen the behavior of gas component rate increase and vibration main frequency shift synchronizing, and calculate the spatial distribution of the synchronizing behavior between path segment nodes to obtain the synchronization feature aggregation group. S513: Based on the aforementioned synchronization feature aggregation group, optimize the node connection order in the main path structure, identify node paths with associated characteristics, analyze the abnormal node numbers and link mapping information of each key path segment, and obtain fault link location data.

Images