Power substation fault detection method
By extracting the correlation differences between baseline and correction characteristic parameters in substations and combining historical and real-time data, accurate identification of fault characteristics and root cause location are achieved. This solves the problems of frequent false alarms and difficulty in locating the root cause in traditional detection methods, and improves operation and maintenance efficiency and the interpretability of detection results.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-12
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional substation fault detection methods cannot effectively distinguish between normal equipment operation fluctuations and real fault signals, resulting in frequent false alarms and an inability to accurately locate the root cause of the fault, making it difficult to meet the timeliness requirements of modern power grid fault handling.
By extracting the correlation differences between the baseline and modified characteristic parameters of power substations, and combining historical fault tracing resources and real-time monitoring data, a hierarchical mining and correlation verification logic is adopted to achieve fault feature identification and root cause localization.
It improves the accuracy of fault feature extraction and the precision of fault root cause localization, enhances the efficiency of substation operation and maintenance and the interpretability of detection results, and is adaptable to fault detection scenarios of substations of different sizes and types.
Smart Images
Figure CN121840905A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of power substation, more particularly, it relates to a power substation fault detection method. BACKGROUND
[0002] In the operation of the power system, the substation as the node of power transmission and distribution, its operation reliability is directly related to the safety and stability of the entire power grid. With the continuous expansion of the power grid scale, the growth of the service life of power equipment, and the influence of complex factors such as external environment and load fluctuation, the probability of substation equipment failure significantly increases. The traditional substation fault detection method mainly relies on single real-time monitoring data or threshold alarm mechanism, which has obvious limitations. It cannot effectively distinguish between normal operation fluctuation and real fault signal, such as environmental temperature rise, temporary increase of power grid load and other non-fault interference factors, which will cause the characteristic parameters to fluctuate similarly to the fault, resulting in frequent false alarms of the detection system, which not only increases the troubleshooting burden of the operation and maintenance personnel, but also may cause unnecessary downtime loss due to over-disposal. At the same time, the traditional method can only identify parameter abnormalities, and cannot locate the fault source. The operation and maintenance personnel often need to rely on experience to speculate the fault root cause, and the troubleshooting process is time-consuming, which is difficult to meet the timeliness requirements of modern power grid fault disposal. SUMMARY
[0003] In view of the deficiencies in the prior art, the purpose of the present application is to provide a power substation fault detection method.
[0004] To achieve the above purpose, the present application provides the following technical scheme: A power substation fault detection method, the method comprising the following steps: Extracting the correlation difference between the corresponding reference characteristic parameters and the corrected characteristic parameters of the power substation in the historical monitoring data to obtain a first correlation difference degree; Filtering the target traceability content related to fault root cause positioning from the historical characteristic parameter fault traceability resource information, and judging the matching degree between the target traceability content and the fault root cause positioning; According to the first correlation difference degree and the matching degree, the standard working condition fault characteristics between the reference characteristic parameters and the corrected characteristic parameters are obtained; Judging the correlation difference between the corresponding target fault characteristic parameters and the historical characteristic parameters of the power substation fault case information to obtain a second correlation difference degree; According to the second correlation difference degree and the matching degree, the standard root cause mining characteristics between the reference characteristic parameters and the target fault characteristic parameters are obtained; Collect current operating characteristic parameters of the power substation reaching the preset operating condition switching threshold, and obtain root cause detection characteristics corresponding to the change trend of the power substation in the current operating condition according to the current operating characteristic parameters and real-time monitoring resource information corresponding to the current operating characteristic parameters; Output the fault root cause detection result according to the root cause detection characteristics, the standard operating condition fault characteristics, and the standard root cause mining characteristics.
[0005] Preferably, the associated differences between the corresponding reference characteristic parameters and the correction characteristic parameters of the power substation in the historical monitoring data are extracted to obtain a first associated difference degree, specifically including the following steps: Mark the characteristic data of the equipment in the stable operating state in the historical monitoring data as the reference characteristic parameters; Mark the characteristic data affected by external environmental changes, equipment operation and maintenance adjustments, and power grid load fluctuations in the historical monitoring data as the correction characteristic parameters; Extract the change difference values of the reference characteristic parameters and the correction characteristic parameters in each characteristic dimension to obtain an initial difference set; Distinguish the invalid differences caused by non-fault interference factors and the valid differences associated with the equipment operating state in the initial difference set to obtain a valid difference set; Calibrate and aggregate the different difference components in the valid difference set to obtain the first associated difference degree between the reference characteristic parameters and the correction characteristic parameters.
[0006] Preferably, the matching degree between the target traceability content and the fault root cause positioning is judged, specifically including the following steps: Determine the target traceability elements related to the fault root cause positioning; Obtain the target judgment elements of the fault root cause positioning; Process and analyze the target judgment elements and the target traceability elements to obtain the matching key points and the matching deviation points; Obtain the matching degree between the target traceability content and the fault root cause positioning based on the matching key points and the matching deviation points.
[0007] Preferably, the matching key points and the matching deviation points are obtained by processing and analyzing the target judgment elements and the target traceability elements, specifically including the following steps: Judge the associated close condition of the target traceability elements and the corresponding target judgment elements; Determine the influence weight of the target traceability elements on the fault root cause positioning based on the associated close condition, and integrate the target traceability elements with different weights to obtain an element association aggregation result; Verify the element association aggregation result to obtain an element association result; The matching status of target source elements and fault root cause location is obtained by evaluating the element association results, and the matching key points and matching deviation points are extracted based on the element matching status.
[0008] Preferably, the standard operating condition fault characteristics between the baseline characteristic parameters and the modified characteristic parameters are obtained based on the first correlation difference degree and the matching fit degree, specifically including the following steps: The first association difference is weighted and calibrated based on the degree of matching fit to obtain the calibrated association difference. Based on the correlation difference degree, the correlation between the benchmark feature parameters and the modified feature parameters is hierarchically analyzed to obtain abnormal correlation deviation data; The abnormal correlation deviation data is compared with the target source content corresponding to the degree of matching fit to obtain the feature correlation status. The standard operating condition fault characteristics between the baseline characteristic parameters and the modified characteristic parameters are obtained by normalizing the characteristic correlation status.
[0009] Preferably, the second correlation difference degree is obtained by determining the correlation difference between the target fault feature parameters and historical feature parameters of the power substation fault case information, specifically including the following steps: Extracting target fault characteristic parameters from power substation fault case information; The target fault feature parameters are compared with historical feature parameters by class, and the fault-related difference information between the target fault feature parameters and historical feature parameters is obtained by combining the target source tracing content corresponding to the degree of matching fit. The difference characterization results are obtained by normalizing the fault-related difference information. The second association difference degree is obtained based on the difference characterization results.
[0010] Preferably, the standard root cause mining features between the baseline feature parameters and the target fault feature parameters are obtained based on the second correlation difference and the matching fit degree, specifically including the following steps: The second correlation difference is calibrated based on the degree of matching fit to obtain the calibrated fault difference. Extract common correlation features related to fault difference degree, target source content and fault root cause, and obtain target difference correlation information pointing to fault root cause based on common correlation features; A root cause-oriented difference correlation system is obtained by hierarchically organizing the target difference correlation information. Based on the differential correlation system, the root cause correlation between the baseline characteristic parameters and the target fault characteristic parameters is clarified, and standard root cause mining features are obtained.
[0011] Preferably, the fault root cause detection results are output based on root cause detection features, standard operating condition fault features, and standard root cause mining features, specifically including the following steps: The root cause detection features are compared with the standard operating condition fault features, and the root cause parameters are mined and corrected by mining the fault root cause features of the current operating feature parameters based on the correlation change trend of the feature parameters of the standard root cause mining features. The root cause detection results are obtained by excavating and correcting the root cause parameters.
[0012] Preferably, the root cause detection features are compared with the standard operating condition fault features, and based on the correlation change trend of the feature parameters of the standard root cause mining features, the current operating feature parameters are used to mine fault root cause features to obtain the mining and correction root cause parameters. Specifically, this includes the following steps: By comparing the root cause detection features with the standard operating condition fault features, the deviations in the feature dimension distribution and correlation trend are obtained. Based on the correlation deviation pattern between the benchmark feature parameter and the modified feature parameter in the first correlation difference degree, the deviation part is calibrated to obtain the effective deviation feature; Based on the effective deviation features, standard root cause mining features are retrieved, and the correlation and change trends of the feature parameters contained in the standard root cause mining features are extracted. Based on the trend of changes in the correlation between effective deviation features and feature parameters, hierarchical root cause feature mining is performed on the current operating feature parameters to obtain the corrected root cause parameters.
[0013] Preferably, obtaining the root cause detection results based on the root cause parameters includes the following steps: Based on the trend of changes in the characteristic parameters, the characteristic changes corresponding to the root causes of the faults are identified, and the potential fault types corresponding to the root cause parameters are clearly identified and corrected. Based on the matching relationship between the target source tracing content and the root cause location of the fault, the source tracing information corresponding to the target source tracing content and the potential fault type is retrieved, and the correlation verification between the root cause mining and correction parameters and the source tracing information is performed to obtain the source tracing verification result. The potential fault types and the source verification results are combined to form the root cause detection results.
[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention improves the accuracy of fault feature extraction by distinguishing between baseline and modified feature parameters. By filtering out effective differences directly related to equipment operating status through a first correlation difference degree, it can identify fault-related features. Deep integration of historical fault cases, maintenance records, and other experience data with real-time monitoring data significantly improves the accuracy of fault root cause localization. Standard operating condition fault features are constructed based on historical normal operating data, reflecting the fault correlation patterns of equipment under stable operating conditions; standard root cause mining features are constructed based on fault case data, reflecting the feature change trends corresponding to different fault root causes. Through a closed-loop logic of hierarchical mining and correlation verification, it achieves full-process quantitative analysis of fault detection from feature identification to root cause localization, greatly improving the interpretability of detection results and the efficiency of maintenance decisions. This enhances the overall efficiency of substation maintenance and is adaptable to fault detection scenarios of substations of different sizes and types. Attached Figure Description
[0015] Fig. 1 This is a schematic diagram illustrating the steps of a power substation fault detection method according to an embodiment of the present invention; Fig. 2 This is a schematic diagram illustrating the steps of obtaining the matching degree in a power substation fault detection method according to an embodiment of the present invention. Detailed Implementation
[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0018] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.
[0019] Reference Figs. 1-2 As shown.
[0020] The embodiments further illustrate the fault detection method for power substations proposed in this invention.
[0021] A method for fault detection in a power substation, comprising the following steps: The correlation difference between the baseline characteristic parameters and the modified characteristic parameters of the power substation is extracted from historical monitoring data to obtain the first correlation difference degree. Target source tracing content related to fault root cause localization is obtained from the fault source tracing resource information of historical feature parameters, and the degree of matching between the target source tracing content and the fault root cause localization is judged. Based on the first correlation difference degree and the matching fit degree, the standard operating condition fault characteristics between the baseline characteristic parameters and the modified characteristic parameters are obtained; The second correlation difference degree is obtained by judging the correlation difference between the target fault characteristic parameters and historical characteristic parameters of power substation fault case information. Standard root cause mining features between baseline feature parameters and target fault feature parameters are obtained based on the second correlation difference degree and matching fit degree. Collect the current operating characteristic parameters of the power substation when it reaches the preset operating condition switching threshold, and obtain the root cause detection characteristics of the power substation under the current operating condition based on the current operating characteristic parameters and the corresponding real-time monitoring resource information. First, a switching threshold is preset. This threshold is determined based on the substation's historical operating data and maintenance experience, and is used to determine whether equipment should enter a new operating condition. For example, for a 110kV substation, a switching condition is triggered when the grid load exceeds 85% of its rated value for 15 consecutive minutes, or when the ambient temperature exceeds 35 degrees Celsius for 30 consecutive minutes.
[0022] The current operating characteristic parameters cover the core operating indicators of the equipment, such as the transformer winding temperature, oil level, vibration amplitude, cooling system fan speed, as well as the partial discharge, bus voltage, and current of the switchgear. Taking the transformer as an example, when the grid load reaches the threshold of 85%, the real-time collected winding temperature is 72 degrees Celsius, the oil level is 52% of the oil conservator scale, the vibration amplitude is 0.04 mm, and the fan speed is 90% of the rated value.
[0023] Real-time monitoring resource information includes the location of monitoring points, the accuracy level of sensors, data sampling frequency, and transmission delay. For example, a winding temperature monitoring sensor is installed on the top of the high-voltage winding of a transformer, with an accuracy of ±0.5 degrees Celsius and a sampling frequency of 1 time per minute; a vibration sensor is installed on the transformer base, with an accuracy of ±0.001 millimeters and a sampling frequency of 10 times per minute. This monitoring resource information can assess the reliability of current operating characteristic parameters. If the historical data transmission delay of a sensor exceeds 2 seconds, the real-time data of that sensor is assigned a lower confidence weight.
[0024] After obtaining the current operating characteristic parameters and corresponding monitoring resource information, the data undergoes multi-dimensional preprocessing, including data cleaning, normalization, and trend extraction. Data cleaning removes outliers caused by temporary sensor malfunctions or signal interference, such as invalid data where the winding temperature jumps to 100 degrees Celsius instantaneously. Normalization unifies characteristic parameters from different dimensions to the same numerical range. For example, the normalized calculation formula for winding temperature is: Normalized Temperature = (Current Temperature - Historical Lowest Temperature) / (Historical Highest Temperature - Historical Lowest Temperature). Assuming the historical lowest temperature is 40 degrees Celsius, the historical highest temperature is 90 degrees Celsius, and the current temperature is 72 degrees Celsius, then the normalized temperature = (72 - 40) / (90 - 40) = 32 / 50 = 0.64. Trend extraction calculates the rate of change of each parameter using a sliding window algorithm. For example, the average rate of increase of winding temperature over the past 10 minutes is 0.3 degrees Celsius per minute.
[0025] The preprocessed current operating characteristic parameters are weighted and fused with the confidence weights of the monitoring resource information to generate root cause detection features. For example, winding temperature is assigned a weight of 0.4, oil level a weight of 0.2, vibration amplitude a weight of 0.3, and fan speed a weight of 0.1. Root cause detection feature = (normalized temperature × 0.4 + normalized oil level × 0.2 + normalized vibration amplitude × 0.3 + normalized fan speed × 0.1) × rate of change weighting coefficient. Assuming normalized oil level is 0.56, normalized vibration amplitude is 0.6, normalized fan speed is 0.9, and the rate of change weighting coefficient is 1.2, then root cause detection feature = (0.64 × 0.4 + 0.56 × 0.2 + 0.6 × 0.3 + 0.9 × 0.1) × 1.2 ≈ 0.766. It not only reflects the degree of abnormality of the current operating parameters, but also includes the changing trend of the root cause of the fault, providing key real-time basis for subsequent comparison with the fault characteristics of standard operating conditions and the characteristics of standard root cause mining.
[0026] The fault root cause detection results are output based on the root cause detection features, standard operating condition fault features, and standard root cause mining features.
[0027] Extracting the correlation difference between the baseline characteristic parameters and the modified characteristic parameters of power substations from historical monitoring data to obtain the first correlation difference degree, specifically includes the following steps: Mark the characteristic data of equipment in stable operating state in historical monitoring data as the baseline characteristic parameters; Mark the characteristic data in historical monitoring data that are affected by changes in the external environment, equipment operation and maintenance adjustments, and power grid load fluctuations as correction characteristic parameters; Extract the differences between the baseline feature parameters and the modified feature parameters in each feature dimension to obtain an initial set of differences; The effective difference set is obtained by distinguishing between invalid differences caused by non-fault interference factors and valid differences related to equipment operating status in the initial difference set; The first correlation difference degree between the baseline characteristic parameter and the modified characteristic parameter is obtained by calibrating and summarizing the different difference components in the effective difference set.
[0028] Characteristic data indicating stable equipment operation are selected from historical monitoring data of power substations and marked as baseline characteristic parameters. These represent the baseline for normal operation of the equipment under fault-free and interference-free conditions. For example, when a transformer operates at an ambient temperature of 25 degrees Celsius, the winding temperature is stable between 60 and 62 degrees Celsius, the oil level of an oil-immersed transformer is maintained at 45% to 50% of the oil conservator scale, and the equipment vibration amplitude is consistently within the range of 0.02 to 0.03 millimeters. These values constitute the baseline characteristic parameters of the equipment.
[0029] Characteristic data affected by changes in the external environment, equipment operation and maintenance adjustments, and power grid load fluctuations are selected from historical monitoring data and marked as correction characteristic parameters. These parameters reflect the operating status of equipment under normal disturbances. For example, when the ambient temperature rises to 35 degrees Celsius in summer, the winding temperature of the same transformer rises to 68 to 70 degrees Celsius; when the power grid load temporarily increases to 95% of its rated value, the winding temperature rises to 72 to 74 degrees Celsius; during equipment inspection and maintenance adjustments, the oil level is adjusted to 55% of the oil tank scale. These characteristic data, deviating from the baseline value due to external factors, are marked as correction characteristic parameters.
[0030] An initial set of differences is obtained by calculating the variations between the baseline and corrected characteristic parameters across each characteristic dimension. Taking winding temperature as an example, the mean of the baseline characteristic parameter is 61 degrees Celsius. When the ambient temperature is 35 degrees Celsius, the mean of the corrected characteristic parameter is 69 degrees Celsius, so the variation in this dimension is 69 - 61 = 8 degrees Celsius. When the grid load increases to 95%, the mean of the corrected characteristic parameter is 73 degrees Celsius, so the variation in this dimension is 73 - 61 = 12 degrees Celsius. The baseline value for oil level is 47.5%, and the adjusted corrected value is 55%, so the variation in this dimension is 55 - 47.5 = 7.5%. These variations across different dimensions form the initial set of differences.
[0031] The initial set of differences is filtered to distinguish between invalid differences caused by non-fault interference factors and valid differences directly related to the equipment's operating status. For example, changes in winding temperature due to fluctuations in ambient temperature and grid load, and changes in oil level due to maintenance adjustments, are all non-fault interferences and are invalid differences. If the winding temperature suddenly rises by an abnormal 15 degrees Celsius without significant changes in ambient temperature and load, and this is not caused by external interference factors, then this difference is directly related to the equipment's operating status and is considered a valid difference. This process of identifying valid differences forms the set of valid differences.
[0032] The different difference components of the effective difference set are calibrated and summarized. A weight is assigned to the effective difference of each feature dimension, with the weight set according to the degree of influence of that dimension on equipment failure. For example, the weight of winding temperature is set to 0.4, oil level to 0.2, and vibration amplitude to 0.4. Assuming the effective difference set includes a winding temperature difference of 15 degrees Celsius, an oil level difference of 5%, and a vibration amplitude difference of 0.05 mm, the first correlation difference degree is calculated by weighted summation: First correlation difference degree = (15 × 0.4) + (5 × 0.2) + (0.05 × 0.4) = 7.02. The first correlation difference degree reflects the degree of effective correlation difference between the baseline feature parameter and the modified feature parameter.
[0033] Determining the degree of matching between the target source tracing content and the fault root cause localization includes the following steps: Determine the target tracing content and target tracing elements related to fault root cause localization; Obtain the target judgment elements for fault root cause localization; The target judgment elements and target tracing elements are processed and analyzed to obtain the key matching points and matching deviation points, which specifically includes the following steps: Determine the closeness of the correlation between the target tracing elements and the corresponding target judgment elements; Based on the close correlation, the influence weight of the target tracing elements on the root cause location of the fault is determined, and the target tracing elements with different weights are integrated to obtain the element correlation summary result. The element association results are obtained by validating the summarized results of element association. The matching status of target source elements and fault root cause location is obtained by evaluating the element association results, and the matching key points and matching deviation points are extracted based on the element matching status.
[0034] The degree of matching between the target source content and the root cause location of the fault is obtained based on the matching key points and matching deviation points.
[0035] First, identify the target tracing elements and target judgment elements. Filter the fault tracing resource information to obtain target tracing elements directly related to the current fault root cause location. For example, when locating a transformer overheating fault, target tracing elements include historical monitoring records of abnormal winding temperature fluctuations, equipment maintenance and repair records for the past three months, and typical fault cases of the same model of equipment. Next, it is necessary to clarify the target judgment elements for fault root cause location. Target judgment elements are the basis for determining the fault root cause. For transformer overheating faults, target judgment elements are divided into equipment-related elements such as winding insulation aging and cooling system failure; external environmental elements such as sudden rise in ambient temperature and blocked ventilation channels; and power grid operation elements such as overload and harmonic interference.
[0036] The correlation between the target traceability element and the corresponding target judgment element is determined, and the strength of the correlation between each target traceability element and the target judgment element is assessed one by one. For example, the continuous abnormal fluctuation of winding temperature in historical monitoring is directly related to the winding insulation aging in the target judgment element, and the correlation between the two is relatively strong; while the operation and maintenance records of the equipment in the past three months include the cleaning and maintenance of the cooling system, and the correlation between the cooling system failure in the target judgment element is moderate; if the typical failure cases of the same model of equipment include overheating caused by excessively high ambient temperature, the correlation between the sudden rise in ambient temperature in the target judgment element is relatively weak.
[0037] Determine the influence weights and integrate the results of the element association summary. Assign a corresponding influence weight to each target traceability element based on the strength of the association; the stronger the association, the higher the weight value. Taking transformer overheating faults as an example, assign a weight of 0.4 to records of abnormal winding temperature fluctuations, a weight of 0.3 to records of cooling system operation and maintenance, a weight of 0.2 to cases of faults related to ambient temperature of the same type of equipment, and a weight of 0.1 to other traceability elements with weaker associations. Weighted integration of these target traceability elements with different weights yields the element association summary result: Element association summary result = 0.4 × Records of abnormal winding temperature fluctuations + 0.3 × Records of cooling system operation and maintenance + 0.2 × Cases of faults related to ambient temperature of the same type of equipment + 0.1 × Other traceability elements.
[0038] After obtaining the summary results of element associations, they are validated to obtain the final element association results. The validation process typically combines multi-dimensional actual operating data, such as comparing the real-time load data of the current transformer, ambient temperature data, and infrared thermal imaging detection results of the equipment, to verify the rationality of the summary results of element associations. If it is found that the weight of cooling system failure in the summary results does not match the actual detected normal cooling fan speed, the weights are adjusted to ensure that the element association results can reflect the correlation between each traceable element and the root cause of the failure.
[0039] Based on the element correlation results, the matching degree between the target tracing elements and the elements for fault root cause localization is evaluated, and key matching points and matching deviation points are extracted. Key matching points refer to tracing elements that are highly matched with the target judgment elements. For example, a high correlation between abnormal winding temperature fluctuation records and winding insulation aging is a key matching point. Matching deviation points refer to tracing elements that are weakly correlated with the target judgment elements or that are contradictory. For example, a failure case related to ambient temperature of the same model of equipment does not match the situation where the ambient temperature was normal when the current failure occurred.
[0040] The degree of matching between the target source tracing content and the root cause location is obtained based on the matching key points and matching deviation points. The more matching key points and the higher their weights, the higher the degree of matching; conversely, the more matching deviation points and the higher their weights, the lower the degree of matching. The degree of matching = (sum of weights of matching key points - sum of weights of matching deviation points) / sum of weights of all source tracing elements × 100%. Taking a transformer overheating fault as an example, if the sum of weights of matching key points is 0.7, the sum of weights of matching deviation points is 0.1, and the sum of weights of all source tracing elements is 1, then the degree of matching = (0.7 - 0.1) / 1 × 100% = 60%. The degree of matching reflects the degree to which the current source tracing content supports the root cause location of the fault.
[0041] Based on the first correlation difference and matching fit, the standard operating condition fault characteristics between the baseline characteristic parameters and the modified characteristic parameters are obtained, specifically including the following steps: The first association difference is weighted and calibrated based on the degree of matching fit to obtain the calibrated association difference. Based on the correlation difference degree, the correlation between the benchmark feature parameters and the modified feature parameters is hierarchically analyzed to obtain abnormal correlation deviation data; The abnormal correlation deviation data is compared with the target source content corresponding to the degree of matching fit to obtain the feature correlation status. The standard operating condition fault characteristics between the baseline characteristic parameters and the modified characteristic parameters are obtained by normalizing the characteristic correlation status.
[0042] The weighted correlation difference is adjusted based on the matching fit to obtain the adjusted correlation difference. The first correlation difference reflects the effective difference between the baseline feature parameters and the modified feature parameters, depending on the degree to which the fault tracing content supports root cause localization. A higher matching fit indicates stronger support for root cause localization from the current tracing content; conversely, its weight needs to be reduced. If the initial value of the first correlation difference is 7.02, corresponding to a matching fit of 60%, the adjusted correlation difference is calculated using weighted average: Adjusted correlation difference = First correlation difference × Matching fit = 7.02 × 0.6 = 4.212. This effectively reflects the correlation difference in faults.
[0043] Based on the calibrated correlation difference, the correlation between the baseline feature parameters and the corrected feature parameters is hierarchically analyzed to obtain abnormal correlation deviation data. The hierarchical analysis is conducted according to the importance and degree of difference of the feature dimensions. For example, for transformer monitoring data, the correlation between winding temperature, oil level, and vibration amplitude is prioritized. A difference threshold of 3.0 is set. When the correlation difference of a certain feature dimension exceeds this threshold, it is identified as abnormal correlation deviation data. For example, the correlation difference for the winding temperature dimension is 3.5, exceeding the threshold, and is therefore included in the abnormal correlation deviation data, while the correlation difference for the oil level dimension is 1.2, not exceeding the threshold, and is not included in the abnormal correlation deviation data.
[0044] The abnormal correlation deviation data is compared with the target traceability content corresponding to the degree of matching to obtain the feature correlation status. The target traceability content includes historical cases and operation and maintenance records related to the root cause of the fault. The abnormal correlation deviation data is matched with these contents one by one to determine the correspondence between the two. For example, the abnormal correlation deviation data of winding temperature is compared with the fault cases of winding insulation aging and cooling system failure recorded in the target traceability content. If the abnormal deviation data is found to be highly consistent with multiple sets of winding insulation aging fault cases, then a strong correlation is determined; if it is only related to a few cooling system failure cases, the correlation is weak. Through comparison, the potential root cause of the fault corresponding to each abnormal correlation deviation data can be identified, thus forming the feature correlation status.
[0045] The standard operating condition fault characteristics are obtained by normalizing the feature correlation status, which involves comparing the baseline and modified feature parameters. This normalization process transforms the feature correlation status into a standardized fault characteristic model, including feature dimensions, difference thresholds, and associated root causes. For example, the abnormal correlation deviation threshold for the winding temperature dimension is set to 3.0; when the correlation difference in this dimension exceeds this threshold, it corresponds to the root cause of winding insulation aging. The abnormal correlation deviation threshold for the oil level dimension is set to 2.0; when this threshold is exceeded, it corresponds to the root cause of oil tank leakage. These standardized components together constitute the standard operating condition fault characteristics. If the correlation difference between the current operating feature parameters and the baseline feature parameters triggers the corresponding threshold, the potential root cause of the fault can be located.
[0046] The second correlation difference degree is obtained by determining the correlation difference between the target fault characteristic parameters and historical characteristic parameters of power substation fault case information. This includes the following steps: Extracting target fault characteristic parameters from power substation fault case information; The target fault feature parameters are compared with historical feature parameters by class, and the fault-related difference information between the target fault feature parameters and historical feature parameters is obtained by combining the target source tracing content corresponding to the degree of matching fit. The difference characterization results are obtained by normalizing the fault-related difference information. The second association difference degree is obtained based on the difference characterization results.
[0047] First, target fault characteristic parameters are extracted. Characteristic parameters directly related to the fault are screened and extracted from existing fault case information of power substations. For example, in a transformer overheating fault case, target fault characteristic parameters include a sudden rise in winding temperature to 92 degrees Celsius, an oil level drop to 20% of the oil tank scale, an increase in vibration amplitude to 0.08 mm, and a drop in cooling system fan speed to 50% of the rated value when the fault occurs.
[0048] The system correlates and compares fault-related differences to obtain relevant information. The extracted target fault feature parameters are compared and correlated with historical feature parameters, which are normal data accumulated during long-term stable operation of the equipment. Based on the degree of matching and the target source tracing content, fault-related differences are obtained. For example, comparing the winding temperature of 92 degrees Celsius in the fault case with the normal range of 60 to 62 degrees Celsius in the historical feature parameters yields a temperature difference of 30 to 32 degrees Celsius; comparing the oil level of 20% with the normal range of 45 to 50% in the historical feature parameters yields an oil level difference of 25 to 30%; and comparing the vibration amplitude of 0.08 mm with the normal range of 0.02 to 0.03 mm in the historical feature parameters yields a vibration difference of 0.05 to 0.06 mm. Combined with information from the target source tracing content regarding cooling system failure causing increased winding temperature, it can be confirmed that these differences are directly related to the fault, thus forming a set of fault-related difference information.
[0049] The difference information related to the fault is normalized to obtain the difference characterization result. Normalization transforms the scattered difference information into a structured and quantifiable representation, facilitating subsequent calculations. For example, the difference information of each feature dimension is unified into a relative difference rate, where relative difference rate = (fault feature parameter value - historical feature parameter mean) / historical feature parameter mean × 100%. Taking winding temperature as an example, the historical feature parameter mean is 61 degrees Celsius, and the fault feature parameter value is 92 degrees Celsius, so the relative difference rate is (92-61) / 61×100%≈50.8%; the historical average oil level is 47.5%, and the fault value is 20%, so the relative difference rate is (20-47.5) / 47.5×100%≈-57.9%; the historical average vibration amplitude is 0.025 mm, and the fault value is 0.08 mm, so the relative difference rate is (0.08-0.025) / 0.025×100%=220%. The set of these relative difference rates constitutes the difference characterization result.
[0050] The second correlation difference degree is obtained based on the difference characterization results. A weight is assigned to the relative difference rate of each feature dimension, with the weight set according to the degree of influence of that dimension on the fault. For example, the weight of winding temperature is set to 0.4, oil level to 0.2, vibration amplitude to 0.3, and cooling system fan speed to 0.1. The second correlation difference degree is calculated by weighted summation: Second Correlation Difference Degree = (Relative Difference Rate of Winding Temperature × 0.4) + (Relative Difference Rate of Oil Level × 0.2) + (Relative Difference Rate of Vibration Amplitude × 0.3) + (Relative Difference Rate of Fan Speed × 0.1). Assuming the relative difference rate of fan speed is -50%, then the second correlation difference degree = (50.8% × 0.4) + (-57.9% × 0.2) + (220% × 0.3) + (-50% × 0.1) = 69.74%. This comprehensively reflects the degree of correlation difference between the feature parameters of the fault case and historical normal parameters.
[0051] Based on the second correlation difference and matching fit, the standard root cause mining features between the baseline feature parameters and the target fault feature parameters are obtained, specifically including the following steps: The second correlation difference is calibrated based on the degree of matching fit to obtain the calibrated fault difference. Extract common correlation features related to fault difference degree, target source content and fault root cause, and obtain target difference correlation information pointing to fault root cause based on common correlation features; A root cause-oriented difference correlation system is obtained by hierarchically organizing the target difference correlation information. Based on the differential correlation system, the root cause correlation between the baseline characteristic parameters and the target fault characteristic parameters is clarified, and standard root cause mining features are obtained.
[0052] First, the second correlation difference is calibrated based on the degree of matching fit, resulting in the calibrated fault difference. The second correlation difference reflects the correlation difference between fault case characteristics and historical normal characteristics. A higher degree of matching fit indicates stronger support for root cause localization from the current source tracing content; conversely, its weight needs to be reduced. For example, if the initial value of the second correlation difference is 69.74%, corresponding to a matching fit of 75%, the calibrated fault difference is calculated through weighted average: calibrated fault difference = second correlation difference × matching fit = 69.74% × 0.75 ≈ 52.31%. This reflects the degree of difference directly related to the root cause of the fault.
[0053] Common correlation features related to fault difference degree, target source tracing content, and fault root cause are extracted. Based on these common correlation features, target difference correlation information pointing to the fault root cause is obtained. Target source tracing content includes historical fault cases, maintenance records, and root cause-related information. For example, in the scenario of transformer overheating faults, the calibrated fault difference degree is 52.31%. The target source tracing content records multiple overheating faults caused by winding insulation aging. The common correlation features of these fault cases are a high relative difference rate of winding temperature and abnormal fluctuations in vibration amplitude. Matching the relative difference rate of winding temperature (50.8%) and relative difference rate of vibration amplitude (220%) of the fault difference degree with these common features confirms a high degree of agreement, thus obtaining target difference correlation information pointing to the root cause of winding insulation aging faults, including abnormal difference data of winding temperature and vibration amplitude.
[0054] A root cause-oriented difference correlation system is obtained by hierarchically organizing the target difference correlation information. The hierarchical organization is based on the type and importance of the characteristic dimensions of the fault root cause. For example, for the root cause of winding insulation aging, the target difference correlation information is divided into a core characteristic layer and an auxiliary characteristic layer. The core characteristic layer includes features that directly reflect the root cause, such as the relative difference rate of winding temperature and the relative difference rate of vibration amplitude; the auxiliary characteristic layer includes features that are indirectly related, such as the relative difference rate of oil level and the relative difference rate of cooling system fan speed. Difference thresholds are set for each feature; for example, the threshold for the relative difference rate of winding temperature in the core characteristic layer is 40%, and the threshold for the relative difference rate of vibration amplitude is 150%; the threshold for the relative difference rate of oil level in the auxiliary characteristic layer is -40%, and the threshold for the relative difference rate of fan speed is -30%. This hierarchical organization constructs a root cause-oriented difference correlation system, clarifying the correlation levels and judgment criteria between different features and the fault root cause.
[0055] The root cause correlation between baseline characteristic parameters and target fault characteristic parameters is clarified based on the differential correlation system, thus obtaining standard root cause mining features. The correlation between baseline characteristic parameters (normal operation data) and target fault characteristic parameters (fault case data), and the corresponding root causes of these relationships, are analyzed based on the root cause-oriented differential correlation system. When the relative difference rate of winding temperature exceeds 40% and the relative difference rate of vibration amplitude exceeds 150%, the root cause of the fault is determined to be winding insulation aging; when the relative difference rate of oil level is below -40% and the relative difference rate of fan speed is below -30%, the root cause of the fault is determined to be cooling system failure. These clearly defined correlation judgment rules together constitute the standard root cause mining features.
[0056] The fault root cause detection results are output based on root cause detection features, standard operating condition fault features, and standard root cause mining features. Specifically, the steps include: The root cause detection features are compared with the standard operating condition fault features, and the root cause parameters are mined and corrected by mining the fault root cause features of the current operating feature parameters based on the correlation change trend of the feature parameters of the standard root cause mining features. The root cause detection results are obtained by excavating and correcting the root cause parameters.
[0057] The root cause detection features are compared with the standard operating condition fault features. Based on the correlation and change trends of the feature parameters of the standard root cause mining features, the current operating feature parameters are used to mine fault root cause features to obtain the root cause mining and correction parameters. The specific steps include: By comparing the root cause detection features with the standard operating condition fault features, the deviations in the feature dimension distribution and correlation trend are obtained. Based on the correlation deviation pattern between the benchmark feature parameter and the modified feature parameter in the first correlation difference degree, the deviation part is calibrated to obtain the effective deviation feature; Based on the effective deviation features, standard root cause mining features are retrieved, and the correlation and change trends of the feature parameters contained in the standard root cause mining features are extracted. Based on the trend of changes in the correlation between effective deviation features and feature parameters, hierarchical root cause feature mining is performed on the current operating feature parameters to obtain the corrected root cause parameters.
[0058] By comparing root cause detection features with standard operating condition fault features, the deviations in feature dimension distribution and correlation trends are obtained. Root cause detection features are derived from real-time monitoring of the substation's current operating status, including real-time data on winding temperature, oil level, and vibration amplitude. Standard operating condition fault features are based on a fault baseline model constructed from historical data, including the normal fluctuation range and fault correlation thresholds for each feature dimension. Taking transformer monitoring as an example, assuming the normal fluctuation threshold for winding temperature in standard operating condition fault features is 60 to 62 degrees Celsius, and the normal threshold for vibration amplitude is 0.02 to 0.03 mm, while the current root cause detection features show a winding temperature of 75 degrees Celsius and a vibration amplitude of 0.06 mm, then the winding temperature deviates from the normal threshold by 13 degrees Celsius, and the vibration amplitude deviates by 0.03 mm. These deviations in value and dimension constitute the deviation portion. The correlation trend between features is then determined. For example, under standard operating conditions, winding temperature and ambient temperature are linearly positively correlated. If the increase in winding temperature in the current data is greater than the change in ambient temperature, then the abnormal trend in the correlation is included in the deviation portion.
[0059] Based on the correlation deviation pattern between the baseline characteristic parameters and the corrected characteristic parameters in the first correlation difference degree, the deviation portion is calibrated to obtain the effective deviation feature. The first correlation difference degree records the deviation pattern caused by normal interference factors in historical data. For example, for every 10 degrees Celsius increase in ambient temperature, the winding temperature increases by 5 degrees Celsius; for every 10% increase in grid load, the winding temperature increases by 3 degrees Celsius. If the current ambient temperature is 8 degrees Celsius higher than the baseline state and the grid load increases by 5%, the winding temperature deviation caused by normal interference is calculated to be 8×(5 / 10)+5×(3 / 10)=4+1.5=5.5 degrees Celsius according to the correlation deviation pattern. Subtracting this normal deviation from the actual deviation of 13 degrees Celsius yields an effective deviation value of 13-5.5=7.5 degrees Celsius. This value belongs to the effective deviation feature, and abnormal deviations directly related to equipment failure are retained. If the fluctuation of the power grid load in the historical data causes a normal deviation of 0.01 mm in the vibration amplitude, then the effective deviation of the current vibration amplitude is 0.03-0.01=0.02 mm, and thus it is included in the effective deviation feature.
[0060] Based on the effective deviation features, the corresponding standard root cause mining features are retrieved, and the correlation trends of the feature parameters contained therein are extracted. The standard root cause mining features are root cause correlation models of historical fault cases, containing the feature change patterns corresponding to different fault root causes. When the effective deviation features show that the winding temperature deviates by 7.5 degrees Celsius and the vibration amplitude deviates by 0.02 mm, the standard root cause mining features pointing to winding insulation aging are automatically retrieved. These features record the correlation trends of the feature parameters of winding insulation aging, specifically, the vibration amplitude fluctuates periodically when the winding temperature continues to rise, and the rise in both is positively correlated.
[0061] Based on the correlation trends between effective deviation characteristics and characteristic parameters, hierarchical root cause feature mining is performed on the current operating characteristic parameters to obtain corrective root cause parameters. Hierarchical mining is divided into a core layer and an auxiliary layer according to the degree of correlation between features and root causes. The core layer includes features that directly reflect the root cause, such as winding temperature and vibration amplitude, while the auxiliary layer includes features that are indirectly related, such as oil level and cooling system fan speed. The correlation trends of characteristic parameters are combined to judge the current operating characteristic parameters. For example, when the correlation trend of winding insulation aging is observed, the rate of increase of the current winding temperature and the fluctuation period of the vibration amplitude are mined. If it is found that the current winding temperature is rising at a rate of 2 degrees Celsius per hour, and the fluctuation period of the vibration amplitude perfectly matches the typical period of winding insulation aging, these features constitute the corrective root cause parameters. These parameters not only include the quantitative values of effective deviation but also the dynamic change characteristics directly related to the root cause of the fault, providing a basis for fault root cause detection.
[0062] The root cause detection results are obtained by mining and correcting the root cause parameters, specifically including the following steps: Based on the trend of changes in the characteristic parameters, the characteristic changes corresponding to the root causes of the faults are identified, and the potential fault types corresponding to the root cause parameters are clearly identified and corrected. Based on the matching relationship between the target source tracing content and the root cause location of the fault, the source tracing information corresponding to the target source tracing content and the potential fault type is retrieved, and the correlation verification between the root cause mining and correction parameters and the source tracing information is performed to obtain the source tracing verification result. The potential fault types and the source verification results are combined to form the root cause detection results.
[0063] First, based on the correlation trends of characteristic parameter changes corresponding to the characteristic changes of the root cause of the fault, the potential fault type corresponding to the root cause parameters is identified. The correlation trends of characteristic parameter changes are extracted from the standard root cause mining features; different fault root causes correspond to different characteristic change patterns. Taking transformer monitoring as an example, if the root cause mining parameters show that the winding temperature is continuously rising at a rate of 2 degrees Celsius per hour, while the vibration amplitude fluctuates in a manner consistent with the typical cycle of insulation aging, this is a characteristic change trend corresponding to winding insulation aging, and the current potential fault type can be clearly identified as winding insulation aging. If the root cause mining parameters show that the oil level is continuously decreasing and the cooling system fan speed is abnormally low, then this characteristic change trend points to cooling system failure, and the potential fault type is thus determined to be cooling system failure.
[0064] Based on the compatibility between the target source tracing content and the root cause localization, source tracing information corresponding to the potential fault type is retrieved from the target source tracing content. The correlation between the root cause mining and correction parameters and the source tracing information is then verified to obtain the source tracing verification results. The target source tracing content includes historical fault cases, maintenance records, and equipment repair reports. This information has a clear compatibility with different fault root causes. For example, for the potential fault type of winding insulation aging, source tracing information such as historical cases of overheating caused by insulation aging in transformers of the same model, insulation inspection reports of the equipment over the past five years, and recent partial discharge monitoring data are retrieved from the target source tracing content. The root cause parameters are compared and verified with this source tracing information. If the current winding temperature rise rate and vibration amplitude fluctuation period are found to be completely consistent with the characteristics in historical cases, and the equipment insulation test report shows that the insulation performance has been declining year by year, and the partial discharge monitoring data has recently shown abnormalities, then the source tracing verification result is obtained as passed. If the root cause parameters are found to be inconsistent with the source tracing information, such as the current ambient temperature being much lower than the fault trigger temperature in historical cases, then the source tracing verification result is shown as doubtful and further investigation is required.
[0065] By standardizing and combining potential fault types and source tracing verification results, a complete root cause detection result is formed. This result not only includes a clear fault type determination but also supporting evidence for source tracing verification, possessing high reliability and traceability. For example, the final root cause detection result clearly identifies the potential fault type as winding insulation aging, and includes source tracing verification results. This indicates that the currently identified and corrected root cause parameters highly match the characteristics of historical insulation aging fault cases, and both the equipment insulation test report and partial discharge data support this determination. This detection result can directly provide maintenance personnel with fault location and decision-making basis, thereby guiding repair and maintenance work.
[0066] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.
[0067] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.
[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for fault detection in a power substation, characterized in that, The method includes the following steps: The correlation difference between the baseline characteristic parameters and the modified characteristic parameters of the power substation is extracted from historical monitoring data to obtain the first correlation difference degree. Target source tracing content related to fault root cause localization is obtained from the fault source tracing resource information of historical feature parameters, and the degree of matching between the target source tracing content and the fault root cause localization is judged. Based on the first correlation difference degree and the matching fit degree, the standard operating condition fault characteristics between the baseline characteristic parameters and the modified characteristic parameters are obtained; The second correlation difference degree is obtained by judging the correlation difference between the target fault characteristic parameters and historical characteristic parameters of power substation fault case information. Standard root cause mining features between baseline feature parameters and target fault feature parameters are obtained based on the second correlation difference degree and matching fit degree. Collect the current operating characteristic parameters of the power substation when it reaches the preset operating condition switching threshold, and obtain the root cause detection characteristics of the power substation under the current operating condition based on the current operating characteristic parameters and the corresponding real-time monitoring resource information. The fault root cause detection results are output based on the root cause detection features, standard operating condition fault features, and standard root cause mining features.
2. The method for detecting faults in a power substation according to claim 1, characterized in that, Extracting the correlation difference between the baseline characteristic parameters and the modified characteristic parameters of power substations from historical monitoring data to obtain the first correlation difference degree, specifically includes the following steps: Mark the characteristic data of equipment in stable operating state in historical monitoring data as the baseline characteristic parameters; Mark the characteristic data in historical monitoring data that are affected by changes in the external environment, equipment operation and maintenance adjustments, and power grid load fluctuations as correction characteristic parameters; Extract the differences between the baseline feature parameters and the modified feature parameters in each feature dimension to obtain an initial set of differences; The effective difference set is obtained by distinguishing between invalid differences caused by non-fault interference factors and valid differences related to equipment operating status in the initial difference set; The first correlation difference degree between the baseline characteristic parameter and the modified characteristic parameter is obtained by calibrating and summarizing the different difference components in the effective difference set.
3. The method for detecting faults in a power substation according to claim 1, characterized in that, Determining the degree of matching between the target source tracing content and the fault root cause localization includes the following steps: Determine the target tracing content and target tracing elements related to fault root cause localization; Obtain the target judgment elements for fault root cause localization; By processing and analyzing the target judgment elements and target tracing elements, we can obtain the key matching points and the matching deviation points. The degree of matching between the target source content and the root cause location of the fault is obtained based on the matching key points and matching deviation points.
4. The method for detecting faults in a power substation according to claim 3, characterized in that, The target judgment elements and target tracing elements are processed and analyzed to obtain the key matching points and matching deviation points, which specifically includes the following steps: Determine the closeness of the correlation between the target tracing elements and the corresponding target judgment elements; Based on the close correlation, the influence weight of the target tracing elements on the root cause location of the fault is determined, and the target tracing elements with different weights are integrated to obtain the element correlation summary result. The element association results are obtained by validating the summarized results of element association. The matching status of target source elements and fault root cause location is obtained by evaluating the element association results, and the matching key points and matching deviation points are extracted based on the element matching status.
5. The method for detecting faults in a power substation according to claim 1, characterized in that, Based on the first correlation difference and matching fit, the standard operating condition fault characteristics between the baseline characteristic parameters and the modified characteristic parameters are obtained, specifically including the following steps: The first association difference is weighted and calibrated based on the degree of matching fit to obtain the calibrated association difference. Based on the correlation difference degree, the correlation between the benchmark feature parameters and the modified feature parameters is hierarchically analyzed to obtain abnormal correlation deviation data; The abnormal correlation deviation data is compared with the target source content corresponding to the degree of matching fit to obtain the feature correlation status. The standard operating condition fault characteristics between the baseline characteristic parameters and the modified characteristic parameters are obtained by normalizing the characteristic correlation status.
6. The method for fault detection in a power substation according to claim 1, characterized in that, The second correlation difference degree is obtained by determining the correlation difference between the target fault characteristic parameters and historical characteristic parameters of power substation fault case information. This includes the following steps: Extracting target fault characteristic parameters from power substation fault case information; The target fault feature parameters are compared with historical feature parameters by class, and the fault-related difference information between the target fault feature parameters and historical feature parameters is obtained by combining the target source tracing content corresponding to the degree of matching fit. The difference characterization results are obtained by normalizing the fault-related difference information. The second association difference degree is obtained based on the difference characterization results.
7. The method for detecting faults in a power substation according to claim 1, characterized in that, Based on the second correlation difference and matching fit, the standard root cause mining features between the baseline feature parameters and the target fault feature parameters are obtained, specifically including the following steps: The second correlation difference is calibrated based on the degree of matching fit to obtain the calibrated fault difference. Extract common correlation features related to fault difference degree, target source content and fault root cause, and obtain target difference correlation information pointing to fault root cause based on common correlation features; A root cause-oriented difference correlation system is obtained by hierarchically organizing the target difference correlation information. Based on the differential correlation system, the root cause correlation between the baseline characteristic parameters and the target fault characteristic parameters is clarified, and standard root cause mining features are obtained.
8. The method for detecting faults in a power substation according to claim 1, characterized in that, The fault root cause detection results are output based on root cause detection features, standard operating condition fault features, and standard root cause mining features. Specifically, the steps include: The root cause detection features are compared with the standard operating condition fault features, and the root cause parameters are mined and corrected by mining the fault root cause features of the current operating feature parameters based on the correlation change trend of the feature parameters of the standard root cause mining features. The root cause detection results are obtained by excavating and correcting the root cause parameters.
9. A method for detecting faults in a power substation according to claim 8, characterized in that, The root cause detection features are compared with the standard operating condition fault features. Based on the correlation and change trends of the feature parameters of the standard root cause mining features, the current operating feature parameters are used to mine fault root cause features to obtain the root cause mining and correction parameters. The specific steps include: By comparing the root cause detection features with the standard operating condition fault features, the deviations in the feature dimension distribution and correlation trend are obtained. Based on the correlation deviation pattern between the benchmark feature parameter and the modified feature parameter in the first correlation difference degree, the deviation part is calibrated to obtain the effective deviation feature; Based on the effective deviation features, standard root cause mining features are retrieved, and the correlation and change trends of the feature parameters contained in the standard root cause mining features are extracted. Based on the trend of changes in the correlation between effective deviation features and feature parameters, hierarchical root cause feature mining is performed on the current operating feature parameters to obtain the corrected root cause parameters.
10. A method for detecting faults in a power substation according to claim 9, characterized in that, The root cause detection results are obtained by mining and correcting the root cause parameters, specifically including the following steps: Based on the trend of changes in the characteristic parameters, the characteristic changes corresponding to the root causes of the faults are identified, and the potential fault types corresponding to the root cause parameters are clearly identified and corrected. Based on the matching relationship between the target source tracing content and the root cause location of the fault, the source tracing information corresponding to the target source tracing content and the potential fault type is retrieved, and the correlation verification between the root cause mining and correction parameters and the source tracing information is performed to obtain the source tracing verification result. The potential fault types and the source verification results are combined to form the root cause detection results.