Distribution transformer small-size defect identification method and device based on large electric power model and storage medium

Abstract

The invention discloses a distribution transformer small-size defect identification method and device based on a large electric power model and a storage medium, and belongs to the field of distribution transformer defect identification. The method comprises the steps that electric field data, voltage data and current data of a distribution transformer are collected, and the electric field distortion degree is calculated based on the electric field data; calculating a dielectric loss angle increment based on the voltage data and the current data; constructing a power operation semantic network through the distribution transformer image, and performing electric field distortion detection in combination with the electric field distortion degree to obtain an electric field distortion detection result; performing fault type identification based on the dielectric loss angle increment and dielectric loss angle parameters in the power operation semantic network to obtain an abnormal fault type; and matching in a preset small-size defect identification library based on the electric field distortion detection result and the abnormal fault type to obtain the small-size defect type of the distribution transformer. Therefore, by implementing the method, accurate identification of the small-size defect type of the distribution transformer can be realized.

Description

Technical Field

[0001] This invention relates to the field of distribution transformer defect identification, and in particular to a method, device, and storage medium for identifying small-sized defects in distribution transformers based on a large power model. Background Technology

[0002] In the field of distribution transformer operation and maintenance, traditional methods for identifying small-sized defects mainly rely on single electrical parameter detection or manual inspection. Single electrical parameter detection, such as judging defects solely by measuring the partial discharge of the transformer, suffers from significant errors in detection results due to the weak partial discharge signals caused by small-sized defects in their early stages, which are easily affected by external electromagnetic interference. Manual inspection, on the other hand, is inefficient and highly subjective; the experience and condition of the inspectors can significantly influence the results, and it is difficult to detect small-sized defects inside the equipment. Therefore, existing methods for identifying small-sized defects in distribution transformers cannot be identified in a timely and accurate manner, seriously affecting the safe and stable operation of distribution transformers and increasing the risk of power system failures. Summary of the Invention

[0003] This invention provides a method, device, and storage medium for identifying small-sized defects in distribution transformers based on a large power model, which can accurately identify the types of small-sized defects in distribution transformers.

[0004] This invention provides a method for identifying small-size defects in distribution transformers based on a large power model, including: Collect electric field data, voltage data, and current data of the distribution transformer; calculate the degree of electric field distortion based on the electric field data; and calculate the dielectric loss angle increment based on the voltage data and current data. Based on the degree of electric field distortion and the electric field parameters in the pre-constructed power operation semantic network, electric field distortion detection is performed to obtain the electric field distortion detection result. Based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, fault type identification is performed to obtain the abnormal fault type; Based on the electric field distortion detection results and the abnormal fault type, a matching is performed in a preset small-size defect identification library to obtain the small-size defect type of the distribution transformer.

[0005] This invention, through the acquisition of electric field, voltage, and current data, and the calculation of the degree of electric field distortion and the increment of dielectric loss angle, can obtain key parameters reflecting changes in the transformer's electric field and internal electrical characteristics, providing a data foundation for subsequent anomaly detection. By combining the degree of electric field distortion with the correlation of electric field parameters in a semantic network, potential problems affecting the electric field can be identified. By combining the increment of dielectric loss angle with the correlation of dielectric loss angle parameters in a semantic network, the types of abnormal faults that may exist inside the distribution transformer can be clearly identified. By matching the electric field distortion detection results and abnormal fault types with a defect database, small-sized defect types in the distribution transformer can be identified. Compared to existing technologies that rely on single-parameter detection and are susceptible to interference, this application can achieve accurate identification of small-sized defect types in distribution transformers.

[0006] Further, the electric field distortion detection based on the degree of electric field distortion and the electric field parameters in the pre-constructed power operation semantic network, to obtain the electric field distortion detection result, includes: Based on the comparison between the electric field strength reflected by the degree of electric field distortion and the range of electric field strength in the power operation semantic network, the electric field strength deviation value is obtained. Based on the comparison between the electric field distribution reflected by the degree of electric field distortion and the electric field distribution uniformity index in the power operation semantic network, the electric field distribution uniformity difference value is obtained. By comparing the spatial variation of electric field intensity reflected by the degree of electric field distortion with the curve of electric field intensity variation with spatial location in the power operation semantic network point by point, the deviation curve of spatial variation of electric field intensity is obtained. The deviation value of electric field intensity and the difference value of electric field distribution uniformity are used as numerical features. The spatial coordinate points in the deviation curve of electric field intensity with deviation values ​​greater than a preset threshold and the deviation values ​​corresponding to the spatial coordinate points are used as spatial features to obtain multiple abnormal electric field feature sets. Electric field distortion detection is performed based on the aforementioned abnormal electric field feature sets to obtain electric field distortion detection results.

[0007] This invention compares electric field strength, distribution uniformity, and spatial variation patterns with standard parameters in a semantic network, and extracts numerical and spatial features to construct an abnormal electric field feature set. This comprehensively describes abnormal electric field conditions and provides accurate input for subsequent electric field distortion detection.

[0008] Further, the step of detecting electric field distortion based on each of the abnormal electric field feature sets to obtain electric field distortion detection results includes: Based on the causal relationship between the electric field parameters in each of the abnormal electric field feature sets, the first causal relationship path corresponding to each of the abnormal electric field feature sets is determined. Based on the target node associated with electric field distortion, each of the first causal paths is filtered to obtain multiple second causal paths; Determine the transmission path length from the starting node to the target node of each of the second causal association paths; Based on the length of each of the aforementioned transmission paths, each of the second causal association paths is sorted to obtain a causal association path sequence, and the first second causal association path in the causal association path sequence is taken as an abnormal association path. The mapping information of the abnormal association path in the power operation semantic network is determined as the electric field distortion detection result.

[0009] This invention, through mining the causal relationships within the abnormal electric field feature set, filters and sorts the causal paths most likely to cause electric field distortion, and uses the mapping information of this path in the semantic network as the detection result, can accurately locate the specific cause and fault information of electric field distortion, thereby improving the accuracy of detection.

[0010] Furthermore, the fault type identification based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, to obtain the abnormal fault type, includes: Based on the comparison between the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, the dielectric loss angle deviation interval is obtained; The load rate difference value is obtained by calculating the difference between the load rate of the distribution transformer under the current operating conditions and the preset standard load rate. The change pattern is determined based on the deviation range of the dielectric loss angle and the difference in load rate; Based on the change patterns and the correspondence between preset change patterns and fault types, multiple fault types are determined; Based on the correspondence between each of the fault types and the historical fault types and fault occurrence probabilities in the power operation semantic network, the occurrence probability of each of the fault types is determined. Based on the occurrence probability of each fault type, the fault types are sorted to obtain a fault type sequence, and the first fault type in the fault type sequence is determined as the abnormal fault type.

[0011] This invention determines the change pattern by combining the dielectric loss angle deviation range and the load rate difference value, and sorts the potential fault types according to this pattern and historical fault probabilities, thereby accurately identifying the most likely abnormal fault type at present.

[0012] Furthermore, determining the change pattern based on the deviation range of the dielectric loss angle and the difference in load rate includes: If the deviation range of the dielectric loss angle and the difference value of the load rate change in the same direction, then based on the correlation between the load rate and the increment of the dielectric loss angle in the power operation semantic network, the first rate of change of the increment of the dielectric loss angle under the load rate is determined. The rate of change of the dielectric loss angle increment is determined based on the ratio of the first rate of change to the second rate of change of the dielectric loss angle increment under the load rate. If the rate of change is greater than a preset ratio threshold, the change mode is determined as the first abnormal mode; wherein, the first abnormal mode represents a mode in which the abnormal growth rate is greater than the preset growth rate threshold. If the rate change ratio is less than or equal to a preset ratio threshold, the change mode is determined as the second abnormal mode; wherein, the second abnormal mode represents a mode in which the abnormal growth rate is less than or equal to the preset growth rate threshold.

[0013] This invention, by comparing the theoretical rate of change with the actual rate of change when the dielectric loss angle and the load rate change in the same direction, can determine the abnormal growth pattern and accurately judge whether the change in dielectric loss angle is abnormal and the degree of abnormality.

[0014] Furthermore, the step of determining the change pattern based on the deviation range of the dielectric loss angle and the difference in load rate also includes: If the deviation range of the dielectric loss angle and the difference value of the load rate change in opposite directions, then the set of fluctuation standard deviations of the dielectric loss angle increment under different load rates is obtained from the power operation semantic network; Based on the load rate, the target fluctuation standard deviation is obtained by traversing the set of fluctuation standard deviations to obtain the target load rate whose difference from the load rate is less than a preset difference threshold. A standard score is calculated based on the mean of the dielectric loss angle increment under the load rate, the dielectric loss angle increment, and the standard deviation of the target fluctuation. If the standard score is greater than the preset score threshold, the change pattern is determined as the first abnormal pattern. If the standard score is less than or equal to the preset score threshold, the change mode is determined as the second abnormal mode.

[0015] This invention, through the calculation of standard scores using the standard deviation of the load rate fluctuation when the dielectric loss angle and the load rate change in opposite directions, determines the abnormal mode, enabling accurate judgment of the degree and mode of abnormality in cases of inverse changes.

[0016] Further, the matching of the electric field distortion detection results and the abnormal fault type in a preset small-size defect identification database to obtain the small-size defect type of the distribution transformer includes: In the small-size defect identification library, at least one of the small-size defect types that matches the electric field distortion detection result or the abnormal fault type is selected to obtain a candidate defect set. The electric field distortion results of each small-sized defect type in the candidate defect set are intersected with the electric field distortion detection results to obtain multiple electric field distortion intersection features; The intersection of the fault types of each small-sized defect type in the candidate defect set with the abnormal fault type yields multiple fault type intersection features. The common intersection features of the candidate defect set are obtained by intersecting the electric field distortion intersection features and the fault type intersection features of each of the small-size defect types. Based on the electric field distortion results and fault types of each small-size defect type, a correlation analysis is performed with the common intersection features to obtain the correlation degree of the common features of each small-size defect type. Based on the correlation of the common features, the defect types are screened to obtain the small-size defect types of the distribution transformer.

[0017] The embodiments of the present invention first obtain a candidate set by preliminary screening based on electric field or fault characteristics, and then perform multi-level intersection operations and correlation analysis, which can effectively eliminate interference information and quantify the degree of matching between each candidate defect and the actual detection results.

[0018] Furthermore, the defect type screening based on the correlation degree of each common feature to obtain the small-size defect types of the distribution transformer includes: The small-size defect type corresponding to the common feature correlation degree with the largest value among the common feature correlation degrees is taken as the target small-size defect type. The difference between the electric field distortion result and fault type of each target small-size defect type and the electric field distortion detection result and abnormal fault type is determined. Based on the aforementioned difference degree, the target small-size defect types are sorted to obtain a sequence of small-size defect types arranged in ascending order of difference degree value; According to the order of the small-size defect type sequence, the target small-size defect types are selected sequentially to construct a feature comparison matrix for judgment; wherein, the rows in the feature comparison matrix represent the electric field distortion detection results and abnormal fault types, the columns represent the electric field distortion results and fault types of the target small-size defect types, and the matrix elements represent the matching status of the corresponding features in the rows and columns; When all matrix elements in the feature comparison matrix match for the first time, the target small-size defect type corresponding to the feature comparison matrix is ​​determined as the small-size defect type of the distribution transformer.

[0019] The embodiments of the present invention rank candidate defects based on the correlation and difference of common features, and construct a feature comparison matrix for matching and verification one by one. This ensures that the finally determined defect type is highly consistent with the actual detection results at the feature level, and avoids misjudgment caused by partial feature matching.

[0020] Another embodiment of the present invention provides a small-size defect identification device for distribution transformers based on a large power model, comprising: a data acquisition module, a first anomaly detection module, a second anomaly detection module, and a defect matching and identification module; The data acquisition module is used to collect electric field data, voltage data and current data of the distribution transformer, calculate the degree of electric field distortion based on the electric field data, and calculate the dielectric loss angle increment based on the voltage data and current data. The first anomaly detection module is used to perform electric field distortion detection based on the degree of electric field distortion and each electric field parameter in the pre-constructed power operation semantic network, and obtain the electric field distortion detection result; The second anomaly detection module is used to identify the fault type based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, and obtain the abnormal fault type; The defect matching and identification module is used to match the electric field distortion detection results and the abnormal fault type in a preset small-size defect identification library to obtain the small-size defect type of the distribution transformer.

[0021] Another embodiment of the present invention also provides a computer-readable storage medium item, including: a stored computer program, which, when the computer program is running, controls the device where the computer-readable storage medium is located to perform the steps of a method for identifying small-size defects in distribution transformers based on a large power model according to the present invention. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating an embodiment of the method for identifying small-size defects in distribution transformers based on a large power model provided by the present invention. Figure 2 This is a schematic diagram of a structure of an embodiment of the small-size defect identification device for distribution transformers based on a large power model provided by the present invention. Figure 3 A schematic diagram of the structure of an embodiment of the electronic device provided by the present invention; Figure 4 A schematic diagram of the structure of the computer-readable storage medium provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0025] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0026] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0027] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0028] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0029] See Figure 1 To address the problem of reliance on single-parameter detection in existing technologies being susceptible to interference, an embodiment of the present invention provides a method for identifying small-size defects in distribution transformers based on a large power model, including steps S101 to S104: Step S101: Collect electric field data, voltage data and current data of the distribution transformer, calculate the degree of electric field distortion based on the electric field data, and calculate the dielectric loss angle increment based on the voltage data and current data.

[0030] It should be noted that the large-scale power model is a machine learning model trained on massive amounts of power data. Through algorithm optimization, it achieves intelligent decision-making functions such as power load forecasting, grid dispatching, and equipment fault early warning. Its core advantage lies in processing complex, unstructured data and possessing strong generalization capabilities. Specifically, the large-scale power model described in this invention is a large-scale artificial intelligence model pre-trained on multimodal data such as distribution transformer images and power operation information, specifically designed for intelligent analysis and decision-making in power systems.

[0031] It should be noted that the electric field data is collected by electric field sensors installed at preset locations on the distribution transformer; the electric field sensor is a device that can convert the electric field signals generated during the operation of the distribution transformer into processable electrical signals, and it obtains corresponding data by sensing changes in the electric field strength.

[0032] Furthermore, the collected electric field data can be analyzed and processed using a pre-defined electric field distortion calculation model. This model is based on a large number of electric field data samples under normal and abnormal operating conditions, and is trained using data mining and machine learning algorithms. Its core principle is to compare the differences between the characteristics of actual electric field data and normal electric field data.

[0033] It should be noted that the voltage and current data are collected by the voltage transformer and current transformer in the measurement circuit of the distribution transformer, respectively. The voltage transformer is a device that transforms high voltage into low voltage proportionally based on the principle of electromagnetic induction, and is used for measurement and protection. The current transformer is a device that transforms large current into small current proportionally, and is also used for measurement and protection.

[0034] Furthermore, based on voltage and current data, the current dielectric loss angle can be determined using a preset algorithm for calculating the dielectric loss angle. This angle is then compared with the dielectric loss angle under historical normal operating conditions to obtain the increment of the current dielectric loss angle. The algorithm for calculating the dielectric loss angle is designed based on the principle of phase difference between voltage and current in AC circuits.

[0035] Step S102: Based on the degree of electric field distortion and the electric field parameters in the pre-constructed power operation semantic network, electric field distortion detection is performed to obtain the electric field distortion detection result.

[0036] It should be noted that the power operation semantic network is a structured knowledge base built on knowledge representation and semantic network technology. It stores the correlation between various electrical parameters (including electric field parameters) and equipment status during normal operation of distribution transformers, the variation range of dielectric loss angle parameters during normal operation of distribution transformers, the correlation between different dielectric loss angle increments and various abnormal faults, the mapping information of each path, and the correspondence between historical fault types and fault occurrence probabilities. Among these, the correlation and correspondence are obtained by integrating and analyzing distribution transformer operation data, fault cases, and expert experience.

[0037] It should be noted that electric field distortion detection based on the stated degree of electric field distortion and the electric field parameters in the pre-constructed power operation semantic network, yielding the electric field distortion detection result, means searching the power operation semantic network for descriptions of normal ranges and possible abnormal situations corresponding to the current degree of electric field distortion. If the current degree of electric field distortion exceeds the normal operating range specified in the power operation semantic network, an electric field distortion anomaly is determined, and an electric field distortion detection result is given based on the relevant information recorded in the power operation semantic network.

[0038] Preferably, the electric field distortion detection based on the degree of electric field distortion and the electric field parameters in the pre-constructed power operation semantic network, to obtain the electric field distortion detection result, includes: Step S1021: Based on the comparison between the electric field strength reflected by the degree of electric field distortion and the range of electric field strength in the power operation semantic network, the electric field strength deviation value is obtained.

[0039] Specifically, the electric field strength data reflected by the degree of electric field distortion, i.e., the actual electric field strength value, is obtained and compared with the standard range of electric field strength at the corresponding location and under the corresponding operating condition in the power operation semantic network. By comparing the actual electric field strength value with the standard range boundary value, the deviation of the electric field strength is determined. If the actual electric field strength value is... ,when When, the electric field strength deviates from the value ;when When, the electric field strength deviates from the value ;when When, the electric field strength deviates from the value .

[0040] In one embodiment, the obtained electric field strength value at a certain moment is ,in, This represents the total number of electric field strength measurement points at the same time. Assuming that in the power operation semantic network, the standard range of electric field strength at the corresponding location and under the corresponding operating condition of this distribution transformer is... .in, Therefore, for electric field strength deviation Other electric field strength values ​​were calculated similarly, ultimately yielding a set of electric field strength deviation values. .

[0041] Step S1022: Based on the comparison between the electric field distribution reflected by the degree of electric field distortion and the electric field distribution uniformity index in the power operation semantic network, the electric field distribution uniformity difference value is obtained.

[0042] Specifically, the uniformity index of electric field distribution measures uniformity by calculating the spatial dispersion of the electric field intensity during normal operation, and the standard deviation can be used. Therefore, the uniformity index of the current electric field distribution can be calculated based on the electric field distribution data reflected by the degree of electric field distortion, also using the standard deviation. The calculation uses the standard deviation formula. It calculates the difference between the current electric field distribution uniformity index and the uniformity index during normal operation in the power operation semantic network, i.e., the electric field distribution uniformity difference value. .

[0043] In one embodiment, the obtained electric field strength value at a certain moment is Assuming the standard deviation of the current electric field distribution is calculated The standard deviation of the electric field distribution of this distribution transformer during normal operation in the power operation semantic network. The difference in the uniformity of the electric field distribution .

[0044] Step S1023: The spatial variation of electric field strength reflected by the degree of electric field distortion is compared point by point with the curve of electric field strength variation with spatial location in the power operation semantic network to obtain the deviation curve of spatial variation of electric field strength.

[0045] Specifically, the curve showing the variation of electric field strength with spatial location is obtained by using a curve fitting algorithm after collecting a large amount of normal operation data from sensors placed at different spatial locations on the distribution transformer. This curve reflects the spatial variation trend of the electric field strength under normal conditions. Therefore, by obtaining the spatial distribution data of the electric field strength, which reflects the degree of electric field distortion, the actual electric field strength value at each spatial location is compared with the corresponding electric field strength value on the curve in the power operation semantic network. For example, let the spatial location coordinates be... The electric field strength at the corresponding position on the curve during normal operation is... The actual electric field strength value is Calculate the deviation value at each position. By connecting the deviation values ​​of all spatial locations, we obtain the deviation curve of the spatial variation of electric field intensity.

[0046] In one embodiment, a coordinate system is established at a certain cross-section of the distribution transformer, with the center of the distribution transformer as the origin. The position is then determined on the curve of the semantic network. electric field strength at The actual measured electric field strength value at that location... Then the deviation value at that position The same calculation is performed at all measurement locations on this cross section, and these deviation values ​​are connected in spatial order to obtain the spatial variation deviation curve of the electric field intensity on this cross section.

[0047] Step S1024: The electric field intensity deviation value and the electric field distribution uniformity difference value are used as numerical features, and the spatial coordinate points in the electric field intensity spatial variation deviation curve with deviation values ​​greater than a preset threshold and the deviation values ​​corresponding to the spatial coordinate points are used as spatial features to obtain multiple abnormal electric field feature sets.

[0048] Specifically, the preset threshold is determined based on historical data of distribution transformer operation and expert experience, and is used to judge whether the deviation of the spatial variation of electric field intensity reaches an abnormal level. Therefore, the deviation value of electric field intensity and the difference value of electric field distribution uniformity are used as numerical features, and spatial coordinate points with deviation values ​​greater than the preset threshold are selected from the deviation curve of spatial variation of electric field intensity. and their corresponding deviation values As spatial features, these numerical and spatial features are integrated to obtain the anomalous electric field feature set, which comprehensively describes the anomalous situation of the current electric field.

[0049] In one embodiment, the preset threshold is When the electric field strength deviates from the value In the middle, if , Equal to or greater than a preset threshold; electric field distribution uniformity difference value The spatial variation of electric field intensity deviates from the curve at different locations. , If the deviation value is greater than a preset threshold, then the abnormal electric field feature set is: numerical features include electric field intensity deviation value. , The difference in electric field distribution uniformity is 25; spatial characteristics include spatial coordinate points. and their corresponding deviation values , and their corresponding deviation values A set composed of, etc.

[0050] Step S1025: Based on each of the aforementioned abnormal electric field feature sets, perform electric field distortion detection to obtain electric field distortion detection results, including: Step S10251: Based on the causal relationship between the electric field parameters in each of the abnormal electric field feature sets, determine the first causal relationship path corresponding to each of the abnormal electric field feature sets.

[0051] Specifically, for each abnormal electric field feature set, the nodes corresponding to each electric field parameter are searched in the power operation semantic network, and based on the causal relationships recorded in the power operation semantic network, all possible paths from one parameter node to another are sorted out to obtain the first causal relationship path corresponding to each abnormal electric field feature set.

[0052] Step S10252: Based on the target node associated with the electric field distortion, each of the first causal paths is filtered to obtain multiple second causal paths.

[0053] Specifically, a target node refers to a node in the power operation semantic network that is explicitly marked as having a direct association with the occurrence of electric field distortion. Target nodes typically correspond to key factors or fault points that may lead to electric field distortion. Therefore, each first causal path is checked to determine whether it contains a target node. Only first causal paths that contain target nodes are retained as second causal paths.

[0054] In one embodiment, nodes such as "abnormally high electric field strength," "non-uniform electric field distribution," and "abrupt change in electric field strength at a specific location" are associated with electric field distortion. "Abnormally high electric field strength" and "non-uniform electric field distribution" are marked as target nodes. The nodes corresponding to these features are searched in the power operation semantic network, and a first causal relationship path is identified. For example, from the node "high measured electric field strength at a certain location" to the node "abnormally high electric field strength," and then to the node "electric field distortion"; from the node "increased standard deviation of electric field distribution" to the node "non-uniform electric field distribution," and then to the node "electric field distortion," etc. After filtering, paths containing the target nodes "abnormally high electric field strength" and "non-uniform electric field distribution" are retained, resulting in a second causal relationship path, such as "increased standard deviation of electric field distribution - non-uniform electric field distribution - electric field distortion" and "high measured electric field strength at a certain location - abnormally high electric field strength - electric field distortion."

[0055] Step S10253: Determine the transmission path length from the starting node to the target node of each of the second causal association paths.

[0056] Specifically, in the second causal path, nodes are connected by edges, and each edge represents a causal transmission. The transmission path length refers to the number of edges traversed from the starting node of the second causal path to the target node. Therefore, by traversing each second causal path, starting from the starting node, counting the edges traversed in the order of connection along the path until the target node is reached, the counted number of edges is the transmission path length from the starting node to the target node.

[0057] In one embodiment, for the second causal path "increased standard deviation of electric field distribution - uneven electric field distribution - electric field distortion", the path from the starting node "increased standard deviation of electric field distribution" to the target node "uneven electric field distribution" passes through 1 edge, so the path length from the starting node to the target node is 1; for the path "high measured value of electric field strength at a certain location - abnormal increase in electric field strength - electric field distortion", the path from the starting node "high measured value of electric field strength at a certain location" to the target node "abnormal increase in electric field strength" passes through 1 edge, and its path length is also 1.

[0058] Step S10254: Sort each second causal path based on the length of each transmission path to obtain a causal path sequence, and take the first second causal path in the causal path sequence as an abnormal path.

[0059] Specifically, the second causal paths are sorted according to their transmission path length to obtain a causal path sequence. The shorter the transmission path length, the more direct the causal transmission from the initial factor to the key factor causing the electric field distortion, and the higher the probability that the causal relationship corresponding to this second causal path can explain the electric field distortion phenomenon. Furthermore, the first second causal path in the sorted causal path sequence is identified as the anomalous path; this path is considered the most likely causal relationship path to cause the current electric field distortion.

[0060] Step S10255: The mapping information of the abnormal association path in the power operation semantic network is determined as the electric field distortion detection result.

[0061] Specifically, the mapping information corresponding to the abnormal correlation path is searched in the power operation semantic network. The mapping information includes detailed fault descriptions and possible impact ranges for each node in the path, and the final electric field distortion detection result is obtained.

[0062] In one embodiment, in addition to the two second causal paths mentioned above, there is another second causal path: "local overheating of transformer winding - change in gas composition in oil - change in electric field distribution - uneven electric field distribution - electric field distortion," whose transmission path length from the starting node "local overheating of transformer winding" to the target node "uneven electric field distribution" is 3. Sorting the three second causal paths according to their transmission path lengths yields the following sorted causal path sequence: 1. "Increased standard deviation of electric field distribution - uneven electric field distribution - electric field distortion" (transmission path length is 1); 2. "Electric field strength measurement at a certain location is too high - electric field strength is abnormally increased - electric field distortion" (transmission path length is 1); 3. "Local overheating of transformer windings - change in gas composition in oil - change in electric field distribution - uneven electric field distribution - electric field distortion" (transmission path length is 3).

[0063] Since the first two paths have the same path length, the first and second causal paths can be further determined according to alphabetical order or other preset rules. Let's assume that "increased standard deviation of electric field distribution - uneven electric field distribution - electric field distortion" is determined as the first and second causal path, i.e., the abnormal correlation path. In the power operation semantic network, the mapping information corresponding to this path shows: the fault type is electric field distortion caused by uneven electric field distribution, which may be due to a decline in the performance of the insulating medium; the scope of influence is the overall electric field environment inside the transformer. This information is determined as the final electric field distortion detection result.

[0064] Step S103: Based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, fault type identification is performed to obtain the abnormal fault type.

[0065] It should be noted that, based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, the abnormal fault type is obtained by searching for abnormal fault types and related descriptions that match the current dielectric loss angle increment in the power operation semantic network. If the current dielectric loss angle increment exceeds the normal operating range specified in the power operation semantic network, the possible abnormal fault type will be determined according to the corresponding relationship recorded in the power operation semantic network.

[0066] Preferably, the fault type identification based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network to obtain the abnormal fault type includes: Step S1031: Based on the comparison results between the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, the dielectric loss angle deviation interval is obtained.

[0067] Specifically, the current dielectric loss angle increment is obtained and compared with the normal parameter range of dielectric loss angle under the corresponding operating condition in the power operation semantic network. By comparison, the dielectric loss angle deviation interval is obtained. The dielectric loss angle deviation interval includes a first deviation interval and a second deviation interval. The first deviation interval indicates that the current increase in dielectric loss angle is less than the minimum value of the dielectric loss angle parameter during normal operation, and the second deviation interval indicates that the current increase in dielectric loss angle is greater than the maximum value of the dielectric loss angle parameter during normal operation. Therefore, if Less than Then the angle of dielectric loss deviates from the interval. That is, the first deviation interval, if the current dielectric loss angle increment Greater than Then the angle of dielectric loss deviates from the interval. That is, the second deviation interval.

[0068] Step S1032: Calculate the difference between the load rate of the distribution transformer under the current operating conditions and the preset standard load rate to obtain the load rate difference value.

[0069] Specifically, obtain the load factor of the distribution transformer under the current operating conditions. Load factor refers to the ratio of the actual output power of a transformer to its rated power. Preset standard load factor. It is a reference load rate value set according to the transformer's design parameters and normal operating requirements, and the difference between the two is calculated. The load rate difference value is obtained.

[0070] Step S1033, determining the change pattern based on the dielectric loss angle deviation range and the load rate difference value, includes: Step S10331: If the deviation range of the dielectric loss angle and the difference value of the load rate change in the same direction, then based on the correlation between the load rate and the increment of the dielectric loss angle in the power operation semantic network, determine the first rate of change of the increment of the dielectric loss angle under the load rate.

[0071] Specifically, the preset difference threshold is a reference value set based on the reasonable range of load rate fluctuations during normal operation of the distribution transformer, used to determine the magnitude of the load rate difference. Therefore, if the dielectric loss angle deviation range is within the first deviation range and the load rate difference is less than the preset difference threshold, or if the dielectric loss angle deviation range is within the second deviation range and the load rate difference is greater than the preset difference threshold, it is considered a change in the same direction. Here, the first deviation range indicates that the current dielectric loss angle increment is less than the minimum value of the dielectric loss angle parameter during normal operation, and the second deviation range indicates that the current dielectric loss angle increment is greater than the maximum value of the dielectric loss angle parameter during normal operation.

[0072] Furthermore, once the change is determined to be in the same direction, based on the correlation between load factor and dielectric loss angle increment in the power operation semantic network, i.e., the load during normal operation... With the increase of the dielectric loss angle The mathematical model incorporates the load factor to calculate the first rate of change of the current dielectric loss angle increment under the load factor and according to normal operating conditions. Among them, the load rate during normal operation in the power operation semantic network. With the increase of the dielectric loss angle The mathematical model is as follows: .

[0073] In one embodiment, the standard range of electric field strength The current dielectric loss angle increment is Therefore, the current increase in dielectric loss angle is within the dielectric loss angle deviation range. It falls within the second deviation range; assuming the load factor difference is... The preset difference threshold is 0.1, because The conditions are met: the dielectric loss angle deviation range is within the second deviation range and the load rate difference is greater than the preset difference threshold. Therefore, the dielectric loss angle deviation range and the load rate difference change in the same direction. The load rate... Substituting the values, we obtain the first rate of change of the current dielectric loss angle increment under the load rate. ( / Unit load rate).

[0074] Step S10332: Determine the rate change ratio of the dielectric loss angle increment based on the ratio of the first rate of change to the second rate of change of the dielectric loss angle increment under the load rate.

[0075] Specifically, the second rate of change This refers to the rate of increase in the actual dielectric loss angle relative to the change in the load rate at the current load rate. The calculation method is as follows: (when (time), among which, This represents the current increment of the dielectric loss angle. This represents the load factor difference. Further, the first rate of change is calculated. With the second rate of change The ratio of the two values ​​is used to obtain the rate change ratio. The calculation formula is: The ratio of the rate of change reflects the degree of deviation between the actual rate of change of dielectric loss angle increment and the theoretical rate of change during normal operation.

[0076] In one embodiment, , Then the second rate of change / Unit load rate, second rate of change / unit load rate, therefore, rate of change ratio .

[0077] Step S10333: If the rate change ratio is greater than a preset ratio threshold, the change mode is determined as a first abnormal mode; wherein, the first abnormal mode represents a mode in which the abnormal growth rate is greater than a preset growth rate threshold.

[0078] Specifically, the preset ratio threshold is a critical value set based on the statistical analysis results of the deviation between the actual and theoretical rates of change of the dielectric loss angle increment under different fault scenarios of the distribution transformer. It is used to distinguish whether the change in the dielectric loss angle increment belongs to an abnormally rapid growth mode. Therefore, the rate change ratio... Compared with the preset ratio threshold If a comparison is made, If the change in the current dielectric loss angle increment shows an abnormally rapid growth characteristic, the change pattern is determined to be the first abnormal pattern, that is, the pattern in which the abnormal growth rate is greater than the preset growth rate threshold.

[0079] In one embodiment, a preset ratio threshold is used. rate of change ,because Therefore, the change pattern was determined to be the first abnormal pattern.

[0080] Step S10334: If the rate change ratio is less than or equal to a preset ratio threshold, the change mode is determined as the second abnormal mode; wherein, the second abnormal mode represents a mode in which the abnormal growth rate is less than or equal to the preset growth rate threshold.

[0081] Specifically, when judging the rate of change ratio Less than or equal to the preset ratio threshold When the actual rate of change of the current dielectric loss angle increment is small compared with the theoretical rate of change during normal operation, it does not reach the standard of abnormal rapid growth. At this time, the defect identification system determines the change mode as the second abnormal mode, that is, the mode in which the abnormal growth rate is less than or equal to the preset growth rate threshold.

[0082] In one embodiment, if the preset ratio threshold is adjusted to rate of change ,because Therefore, the change pattern was determined to be the second abnormal pattern.

[0083] Step S10335: If the deviation range of the dielectric loss angle and the difference value of the load rate change in opposite directions, then obtain the set of fluctuation standard deviations of the dielectric loss angle increment under different load rates from the power operation semantic network.

[0084] Specifically, if the dielectric loss angle deviation interval and the load rate difference value are opposite, that is, the dielectric loss angle deviation interval is in the first deviation interval and the load rate difference value is greater than the preset difference threshold, or the dielectric loss angle deviation interval is in the second deviation interval and the load rate difference value is less than the preset difference threshold, the dielectric loss angle increment under different load rates during normal operation is obtained from the power operation semantic network, and standard deviation analysis is performed on these dielectric loss angle increments to obtain the fluctuation standard deviation of the dielectric loss angle increment corresponding to each load rate. All these standard deviations constitute the fluctuation standard deviation set.

[0085] In one embodiment, the known dielectric loss angle deviation range is... This falls within the second deviation range. Assume the load factor difference becomes... (That is, the actual load rate is lower than the preset standard load rate), the preset difference threshold is 0.1, because The conditions are met: the dielectric loss angle deviation interval is within the second deviation interval and the load rate difference is less than a preset difference threshold. Therefore, the dielectric loss angle deviation interval and the load rate difference are inversely related. At this point, the standard deviation set of fluctuations in the dielectric loss angle increment under different load rates during normal operation is obtained from the power operation semantic network, for example... Each value in the standard deviation set of fluctuations corresponds to a specific load rate.

[0086] Step S10336: Based on the load rate, traverse the set of standard deviations of fluctuation to obtain the target standard deviation of fluctuation corresponding to the target load rate whose difference from the load rate is less than a preset difference threshold.

[0087] Specifically, the actual load rate is compared one by one with each load rate in the standard deviation set of fluctuations, and the difference between them is calculated. The preset difference threshold is a value used to measure how close the load rates are. When the difference between a certain load rate and the actual load rate is less than the preset difference threshold, that load rate is the target load rate, and its corresponding standard deviation of fluctuation is the target standard deviation of fluctuation.

[0088] In one embodiment, the load rate is known. The standard deviation set of fluctuations is (Here, for each data set, the first element represents the load factor, and the second represents the corresponding standard deviation of fluctuation.) The preset difference threshold is 0.05. Traversing this set, it is found that the difference between a load factor of 0.8 and the actual load factor of 0.8 is 0. Therefore, the target load factor is 0.8, and its corresponding target standard deviation of fluctuation is... .

[0089] Step S10337: Calculate the standard score based on the mean of the dielectric loss angle increment under the load rate, the dielectric loss angle increment, and the target fluctuation standard deviation.

[0090] Specifically, the mean of the dielectric loss angle increment under load rate is the average value of the dielectric loss angle increment recorded in the power operation semantic network when operating normally near that load rate, reflecting the normal level of the dielectric loss angle increment under that load rate. The standard score (Z-score) is a statistical method used to measure the relative position of a data point to the mean; its calculation formula is as follows: ,in, The standard score, This represents the current increment of the dielectric loss angle. This represents the average value of the dielectric loss angle increment under load conditions. The target is the standard deviation of the fluctuation. Therefore, by substituting the relevant data into the formula, the standard score is calculated. The standard score represents the degree of deviation of the current dielectric loss angle increment from the normal level.

[0091] In one embodiment, at the load rate Below, the mean value of the dielectric loss angle increment recorded in the power operation semantic network. Current dielectric loss angle increment Target fluctuation standard deviation According to the standard score calculation formula, we can obtain: That is, the standard score is 6.

[0092] Step S10338: If the standard score is greater than the preset score threshold, the change mode is determined as the first abnormal mode.

[0093] Step S10339: If the standard score is less than or equal to the preset score threshold, the change mode is determined as the second abnormal mode.

[0094] Specifically, the preset score threshold is a critical value set based on statistical analysis of the deviation of the dielectric loss angle increment under different fault scenarios of the distribution transformer. It is used to determine whether the change in the dielectric loss angle increment belongs to an abnormally rapid growth mode. Therefore, the standard score is compared with the preset score threshold. If the standard score is greater than the preset score threshold, it indicates that the current dielectric loss angle increment deviates significantly from the normal level, exhibiting abnormal change characteristics. At this time, the change mode is determined to be the first abnormal mode. If the standard score is less than or equal to the preset score threshold, it indicates that the current change in the dielectric loss angle increment is within the normal fluctuation range or the deviation is small. The change mode is determined to be the second abnormal mode.

[0095] In one embodiment, it is assumed that the preset score threshold is 3 and the standard score is 6, because Therefore, the change pattern is determined to be the first abnormal pattern. If the preset score threshold is adjusted to 7, because If so, the change pattern is determined to be the second abnormal pattern.

[0096] Step S1034: Based on the change pattern and the correspondence between the preset change pattern and the fault type, determine multiple fault types.

[0097] Specifically, the correspondence between preset change patterns and fault types is an empirical mapping relationship obtained through in-depth mining and analysis of a large amount of distribution transformer fault data, and is stored in the system's database. Therefore, once a change pattern is determined, the corresponding fault type is searched in the preset change pattern-fault type correspondence table to obtain the fault type set.

[0098] Step S1035: Determine the occurrence probability of each fault type based on the correspondence between each fault type and the historical fault types and fault occurrence probabilities in the power operation semantic network.

[0099] Specifically, the power operation semantic network records the occurrence probabilities of various fault types under different operating conditions throughout history. This probability data is derived from long-term fault statistics and analysis. Therefore, for each fault type, the probability of occurrence of that fault type under the corresponding operating condition is searched within the power operation semantic network. ,in, The serial number indicates the type of fault.

[0100] In one embodiment, for example, the change mode is a first abnormal mode. In a preset correspondence table between change modes and fault types, the fault types corresponding to the first abnormal mode are "accelerated insulation degradation due to moisture" and "partial discharge of capacitive equipment". In the power operation semantic network, the probability of "accelerated insulation degradation due to moisture" occurring under the current operating condition is queried. The probability of partial discharge in capacitive devices .

[0101] Step S1036: Sort each fault type according to the occurrence probability to obtain a fault type sequence, and determine the first fault type in the fault type sequence as the abnormal fault type.

[0102] Specifically, each fault type and its corresponding probability of occurrence are sorted in descending order of probability to obtain a fault type sequence. Since the fault type with the highest probability of occurrence under the current operating conditions is identified as the abnormal fault type, the fault type that appears first in the sorted sequence is determined to be the abnormal fault type.

[0103] In one embodiment, the probability of occurrence of the fault type "insulation moisture accelerates degradation" is... Probability of partial discharge in capacitive devices Sorting the fault types from highest to lowest probability, the sequence is: 1. "Insulation moisture accelerates deterioration"; 2. "Partial discharge in capacitive equipment". Therefore, "Insulation moisture accelerates deterioration", which is at the top, is identified as the abnormal fault type.

[0104] Step S104: Based on the electric field distortion detection results and the abnormal fault type, match them in a preset small-size defect identification library to obtain the small-size defect type of the distribution transformer.

[0105] It should be noted that the pre-set small-size defect identification database is a database built through extensive verification of practical cases and data accumulation. It stores the correspondence between various electric field distortion detection results, abnormal fault types, and small-size defect types. This correspondence was derived through the collation and summarization of experimental research, fault analysis, and on-site detection data of small-size defects in distribution transformers. Therefore, using electric field distortion detection results and abnormal fault types as search criteria, a matching query is performed in the small-size defect identification database. Specifically, records matching these two search criteria are found in the database, and the final small-size defect type is determined based on the matching results.

[0106] Preferably, the matching of the electric field distortion detection results and the abnormal fault type in a preset small-size defect identification database to obtain the small-size defect type of the distribution transformer includes: Step S1041: In the small-size defect identification library, select at least one of the small-size defect types that matches the electric field distortion detection result or the abnormal fault type to obtain a candidate defect set.

[0107] Specifically, the search is conducted in the small-size defect identification database using electric field distortion detection results and abnormal fault types as search criteria. During the search, a combination of exact matching and fuzzy matching can be used: exact matching is performed for clearly defined fault type names and electric field distortion characteristic keywords; fuzzy matching is performed for cases with similar semantics or features. Small-size defect types that match the electric field distortion detection results, the abnormal fault types, or both, are identified as the candidate defect set.

[0108] In one embodiment, the electric field distortion detection result is "the distribution transformer has an internal partial discharge fault, with a fault severity of moderate," and the abnormal fault type is "insulation moisture accelerates deterioration." The small-size defect identification database stores the following data: Defect type A corresponds to the electric field distortion result of "internal partial discharge, mild," and the fault type is "insulation aging"; defect type B corresponds to the electric field distortion result of "internal partial discharge, moderate," and the fault type is "insulation moisture"; defect type C corresponds to the electric field distortion result of "abnormal electric field distribution," and the fault type is "winding short circuit." After searching, it is found that defect type A partially matches the electric field distortion detection result (both have internal partial discharge), and defect type B completely matches the electric field distortion detection result and partially matches the abnormal fault type. Therefore, defect types A and B are selected to obtain the candidate defect set {defect type A, defect type B}.

[0109] Step S1042: Intersect the electric field distortion results of each small-sized defect type in the candidate defect set with the electric field distortion detection results to obtain multiple electric field distortion intersection features.

[0110] Step S1043: Intersect the fault types of each small-size defect type in the candidate defect set with the abnormal fault types to obtain multiple fault type intersection features.

[0111] Specifically, for each candidate small-size defect type in the candidate defect set, the intersection operation is performed between its corresponding electric field distortion result and the electric field distortion detection result to find the common features, thus obtaining the electric field distortion intersection feature; similarly, the intersection operation is performed between its corresponding fault type and the abnormal fault type to find the common features, thus obtaining the fault type intersection feature. The rules for the intersection operation are as follows: for textual description features, keywords are extracted and compared; identical keywords form the intersection; for numerical or range-based features (such as the severity level of a fault), it is determined whether they fall within the same range or have the same value.

[0112] Step S1044: Based on the intersection features of electric field distortion and fault type of each small-size defect type, the intersection features of the candidate defect set are obtained.

[0113] Specifically, the intersection features of electric field distortion and fault type obtained for each candidate small-size defect type are intersected again to obtain the common intersection features of the entire candidate defect set. The common intersection features reflect the common correlation between each defect type in the candidate defect set and the actual detection results in terms of electric field distortion and fault type.

[0114] In one embodiment, for a candidate defect set {defect type A, defect type B}, the electric field distortion result of defect type A is "internal partial discharge, mild". This is intersected with the electric field distortion detection result "the distribution transformer has an internal partial discharge fault, the fault severity is moderate", and the keyword "internal partial discharge" is extracted, resulting in the electric field distortion intersection feature "internal partial discharge". The intersection of its fault type "insulation aging" and the abnormal fault type "insulation moisture accelerates deterioration" is empty. The electric field distortion result of defect type B is "internal partial discharge, moderate". This is intersected with the electric field distortion detection result, resulting in the electric field distortion intersection feature "internal partial discharge, moderate". The intersection of its fault type "insulation moisture" and the abnormal fault type "insulation moisture accelerates deterioration" yields the fault type intersection feature "insulation moisture". The intersection features of the electric field distortion and fault type features of defect types A and B are then intersected to obtain the common intersection feature of the candidate defect set: {"internal partial discharge", "insulation moisture"}.

[0115] Step S1045: Based on the electric field distortion results and fault types of each small-size defect type, a correlation analysis is performed with the common intersection features to obtain the correlation degree of the common features of each small-size defect type.

[0116] Specifically, correlation analysis determines the degree of correlation between the electric field distortion results and fault types of each small-sized defect type and the common intersection features by calculating the proportion or degree of such features. The specific calculation method is as follows: count the number of common intersection features between the electric field distortion results and fault types of each small-sized defect type. And the total number of features inherent to this small-sized defect type. Through formula Calculate the correlation degree of common features ,in, The index represents the type of small-sized defect. The higher the correlation value, the higher the degree of feature matching between the small-sized defect type and the actual detection result.

[0117] In one embodiment, for defect type A: it has the characteristics of {"internal partial discharge, mild", "insulation aging"}, totaling... The features include common intersection features {"internal partial discharge", "insulation moisture"}. Each feature is used to calculate the common feature correlation degree according to the formula. For defect type B: its characteristics are {"internal partial discharge, moderate", "insulation dampness"}, totaling... The features include common intersection features {"internal partial discharge", "insulation moisture"}. Calculate the correlation degree of common features for each feature. .

[0118] Step S1046: Based on the correlation degree of each common feature, defect type screening is performed to obtain the small-size defect types of the distribution transformer, including: Step S10461: Take the small-size defect type corresponding to the largest common feature correlation degree among the common feature correlation degrees as the target small-size defect type, and determine the degree of difference between the electric field distortion result and fault type of each target small-size defect type and the electric field distortion detection result and abnormal fault type.

[0119] Specifically, among the common feature correlations obtained for each small-size defect type, the one with the largest common feature correlation is identified, and the small-size defect type with the largest common feature correlation is determined as the target small-size defect type. It should be noted that there may be multiple small-size defect types with the same maximum common feature correlation; in this case, all of these small-size defect types are identified as the target small-size defect type.

[0120] Furthermore, for each target small-size defect type, its electric field distortion results are compared with the electric field distortion detection results, and its fault type is compared with the abnormal fault type. The degree of difference is determined by calculating the proportion of the number of different features between the two to the total number of features. Specifically, the degree of difference is calculated as follows: Let the electric field distortion results and the total number of fault features for the target small-size defect type be... The number of features that differ from the electric field distortion detection results and abnormal fault types is Then the degree of difference ,in, The index indicates the type of small-sized defect in the target. The smaller the difference value, the smaller the difference between the target small-sized defect type and the actual detection result.

[0121] In one embodiment, the common feature correlation of defect type A is known. The correlation of common features of defect type B ,because Therefore, defect type B is identified as the target small-size defect type. The electric field distortion result for defect type B is "internal partial discharge, moderate," and the fault type is "insulation dampness." The electric field distortion detection result is "internal partial discharge fault exists in the distribution transformer, fault severity is moderate," and the abnormal fault type is "insulation dampness accelerates deterioration." The electric field distortion result and the total number of fault type features for defect type B are listed below. (That is, the two characteristics of "internal partial discharge, moderate" and "insulation dampness"). The electric field distortion results are completely identical to the electric field distortion detection results; however, the number of different characteristics between the fault type "insulation dampness" and the abnormal fault type "insulation dampness accelerates deterioration" is significant. (The characteristic of "accelerated degradation" is different), the degree of difference is calculated according to the formula. .

[0122] Step S10462: Based on the difference degree, sort the target small-size defect types to obtain a sequence of small-size defect types arranged in ascending order of difference degree value.

[0123] Step S10463: Select target small-size defect types sequentially according to the order of the small-size defect type sequence to construct a feature comparison matrix for judgment; wherein, the rows in the feature comparison matrix represent the electric field distortion detection result and abnormal fault type, the columns represent the electric field distortion result and fault type of the target small-size defect type, and the matrix elements represent the matching status of the corresponding features in the rows and columns.

[0124] Specifically, among the obtained differences for each target small-size defect type, the smallest difference is identified. A feature comparison matrix is ​​constructed based on the target small-size defect type with the smallest difference. The rows of the matrix correspond to the feature elements of the electric field distortion detection result and the feature elements of the abnormal fault type, respectively. The columns of the matrix correspond to the feature elements of the electric field distortion result and the feature elements of the fault type, respectively. For each element at the intersection of a row and a column in the matrix, if the feature element corresponding to the row is the same as the feature element corresponding to the column, the matrix element is assigned a value of 1, indicating a match; otherwise, the matrix element is assigned a value of 0, indicating a mismatch.

[0125] In one embodiment, in addition to defect type B, there is also defect type D, whose common feature correlation degree is also 1, and it is identified as the target small-size defect type, and its difference degree calculation result is... The degree of difference of defect type B ,because Therefore, a feature comparison matrix is ​​constructed based on defect type B, as shown below. Step S10464: When all matrix elements in the feature comparison matrix match for the first time, the target small-size defect type corresponding to the feature comparison matrix is ​​determined as the small-size defect type of the distribution transformer.

[0126] Specifically, the constructed feature comparison matrix is ​​checked to see if all elements in the matrix are 1 (i.e., whether the features corresponding to all rows and columns match). If all matrix elements are 1, it means that the electric field distortion result and fault type of the target small-size defect type completely match the electric field distortion detection result and abnormal fault type, and the target small-size defect type is determined as the final small-size defect type.

[0127] Furthermore, if there are elements of 0 in the feature comparison matrix (i.e., there are mismatched features), then in the sequence of small-sized defect types, the target small-sized defect type with the second smallest difference is selected. A feature comparison matrix is ​​constructed based on this target small-sized defect type, and the judgment is made again. This process is repeated until a target small-sized defect type is found whose corresponding feature comparison matrix has all elements of 1. This target small-sized defect type is then determined as the final small-sized defect type.

[0128] In one embodiment, the presence of an element with a value of 0 in the constructed feature comparison matrix indicates that the features of defect type B do not perfectly match the actual detection results. In the small-size defect type sequence, defect type D has the second smallest degree of difference. Assuming the electric field distortion result of defect type D is "internal partial discharge, moderate" and the fault type is "insulation moisture-induced accelerated aging," the feature comparison matrix is ​​reconstructed as follows: If the feature comparison matrix still contains mismatched elements, continue to select the next target small-size defect type with the second smallest difference for construction and judgment, until a target small-size defect type with all elements of the feature comparison matrix being found is found, and this type is determined as the final small-size defect type.

[0129] This invention, through the acquisition of electric field, voltage, and current data, and the calculation of the degree of electric field distortion and the increment of dielectric loss angle, can obtain key parameters reflecting changes in the electric field and internal electrical characteristics of the transformer, providing a data foundation for subsequent anomaly detection. By combining the degree of electric field distortion with the correlation of electric field parameters in the semantic network, potential problems affecting the electric field can be identified. By combining the increment of dielectric loss angle with the correlation of dielectric loss angle parameters in the semantic network, the types of abnormal faults that may exist inside the distribution transformer can be clearly identified. By matching the electric field distortion detection results and abnormal fault types with a defect database, the types of small-sized defects in the distribution transformer can be identified.

[0130] Optionally, in this embodiment of the invention, the step of detecting electric field distortion based on the degree of electric field distortion and the electric field parameters in the pre-constructed power operation semantic network to obtain electric field distortion detection results includes: Based on the comparison between the electric field strength reflected by the degree of electric field distortion and the range of electric field strength in the power operation semantic network, the electric field strength deviation value is obtained. Based on the comparison between the electric field distribution reflected by the degree of electric field distortion and the electric field distribution uniformity index in the power operation semantic network, the electric field distribution uniformity difference value is obtained. By comparing the spatial variation of electric field intensity reflected by the degree of electric field distortion with the curve of electric field intensity variation with spatial location in the power operation semantic network point by point, the deviation curve of spatial variation of electric field intensity is obtained. The deviation value of electric field intensity and the difference value of electric field distribution uniformity are used as numerical features. The spatial coordinate points in the deviation curve of electric field intensity with deviation values ​​greater than a preset threshold and the deviation values ​​corresponding to the spatial coordinate points are used as spatial features to obtain multiple abnormal electric field feature sets. Electric field distortion detection is performed based on the aforementioned abnormal electric field feature sets to obtain electric field distortion detection results.

[0131] This invention compares electric field strength, distribution uniformity, and spatial variation patterns with standard parameters in a semantic network, and extracts numerical and spatial features to construct an abnormal electric field feature set. This comprehensively describes abnormal electric field conditions and provides accurate input for subsequent electric field distortion detection.

[0132] Optionally, in this embodiment of the invention, the step of detecting electric field distortion based on each of the abnormal electric field feature sets to obtain electric field distortion detection results includes: Based on the causal relationship between the electric field parameters in each of the abnormal electric field feature sets, the first causal relationship path corresponding to each of the abnormal electric field feature sets is determined. Based on the target node associated with electric field distortion, each of the first causal correlation paths is filtered to obtain multiple second causal correlation paths; Determine the transmission path length from the starting node to the target node of each of the second causal association paths; Based on the length of each of the aforementioned transmission paths, each of the second causal association paths is sorted to obtain a causal association path sequence, and the first second causal association path in the causal association path sequence is taken as an abnormal association path. The mapping information of the abnormal association path in the power operation semantic network is determined as the electric field distortion detection result.

[0133] This invention, through mining the causal relationships within the abnormal electric field feature set, filters and sorts the causal paths most likely to cause electric field distortion, and uses the mapping information of this path in the semantic network as the detection result, can accurately locate the specific cause and fault information of electric field distortion, thereby improving the accuracy of detection.

[0134] Optionally, in this embodiment of the invention, the step of identifying the fault type based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network to obtain the abnormal fault type includes: Based on the comparison between the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, the dielectric loss angle deviation interval is obtained; The load rate difference value is obtained by calculating the difference between the load rate of the distribution transformer under the current operating conditions and the preset standard load rate. The change pattern is determined based on the deviation range of the dielectric loss angle and the difference in load rate; Based on the change patterns and the correspondence between preset change patterns and fault types, multiple fault types are determined; Based on the correspondence between each of the fault types and the historical fault types and fault occurrence probabilities in the power operation semantic network, the occurrence probability of each of the fault types is determined. Based on the occurrence probability of each fault type, the fault types are sorted to obtain a fault type sequence, and the first fault type in the fault type sequence is determined as the abnormal fault type.

[0135] This invention determines the change pattern by combining the dielectric loss angle deviation range and the load rate difference value, and sorts the potential fault types according to this pattern and historical fault probabilities, thereby accurately identifying the most likely abnormal fault type at present.

[0136] Optionally, in this embodiment of the invention, determining the change pattern based on the deviation range of the dielectric loss angle and the difference in load rate includes: If the deviation range of the dielectric loss angle and the difference value of the load rate change in the same direction, then based on the correlation between the load rate and the increment of the dielectric loss angle in the power operation semantic network, the first rate of change of the increment of the dielectric loss angle under the load rate is determined. The rate of change of the dielectric loss angle increment is determined based on the ratio of the first rate of change to the second rate of change of the dielectric loss angle increment under the load rate. If the rate of change is greater than a preset ratio threshold, the change mode is determined as the first abnormal mode; wherein, the first abnormal mode represents a mode in which the abnormal growth rate is greater than the preset growth rate threshold. If the rate change ratio is less than or equal to a preset ratio threshold, the change mode is determined as the second abnormal mode; wherein, the second abnormal mode represents a mode in which the abnormal growth rate is less than or equal to the preset growth rate threshold.

[0137] This invention, by comparing the theoretical rate of change with the actual rate of change when the dielectric loss angle and the load rate change in the same direction, can determine the abnormal growth pattern and accurately judge whether the change in dielectric loss angle is abnormal and the degree of abnormality.

[0138] Optionally, in this embodiment of the invention, determining the change pattern based on the deviation range of the dielectric loss angle and the difference in load rate further includes: If the deviation range of the dielectric loss angle and the difference value of the load rate change in opposite directions, then the set of fluctuation standard deviations of the dielectric loss angle increment under different load rates is obtained from the power operation semantic network; Based on the load rate, the target fluctuation standard deviation is obtained by traversing the set of fluctuation standard deviations to obtain the target load rate whose difference from the load rate is less than a preset difference threshold. A standard score is calculated based on the mean of the dielectric loss angle increment under the load rate, the dielectric loss angle increment, and the standard deviation of the target fluctuation. If the standard score is greater than the preset score threshold, the change pattern is determined as the first abnormal pattern. If the standard score is less than or equal to the preset score threshold, the change mode is determined as the second abnormal mode.

[0139] This invention, through the calculation of standard scores using the standard deviation of the load rate fluctuation when the dielectric loss angle and the load rate change in opposite directions, determines the abnormal mode, enabling accurate judgment of the degree and mode of abnormality in cases of inverse changes.

[0140] Optionally, in this embodiment of the invention, the step of matching the electric field distortion detection result and the abnormal fault type in a preset small-size defect identification database to obtain the small-size defect type of the distribution transformer includes: In the small-size defect identification library, at least one of the small-size defect types that matches the electric field distortion detection result or the abnormal fault type is selected to obtain a candidate defect set. The electric field distortion results of each small-sized defect type in the candidate defect set are intersected with the electric field distortion detection results to obtain multiple electric field distortion intersection features; The intersection of the fault types of each small-sized defect type in the candidate defect set with the abnormal fault type yields multiple fault type intersection features. The common intersection features of the candidate defect set are obtained by intersecting the electric field distortion intersection features and the fault type intersection features of each of the small-size defect types. Based on the electric field distortion results and fault types of each small-size defect type, a correlation analysis is performed with the common intersection features to obtain the correlation degree of the common features of each small-size defect type. Based on the correlation of the common features, the defect types are screened to obtain the small-size defect types of the distribution transformer.

[0141] The embodiments of the present invention first obtain a candidate set by preliminary screening based on electric field or fault characteristics, and then perform multi-level intersection operations and correlation analysis, which can effectively eliminate interference information and quantify the degree of matching between each candidate defect and the actual detection results.

[0142] Optionally, in this embodiment of the invention, the defect type screening based on the correlation degree of each common feature to obtain the small-size defect type of the distribution transformer includes: The small-size defect type corresponding to the common feature correlation degree with the largest value among the common feature correlation degrees is taken as the target small-size defect type. The difference between the electric field distortion result and fault type of each target small-size defect type and the electric field distortion detection result and abnormal fault type is determined. Based on the aforementioned difference degree, the target small-size defect types are sorted to obtain a sequence of small-size defect types arranged in ascending order of difference degree value; According to the order of the small-size defect type sequence, the target small-size defect types are selected sequentially to construct a feature comparison matrix for judgment; wherein, the rows in the feature comparison matrix represent the electric field distortion detection results and abnormal fault types, the columns represent the electric field distortion results and fault types of the target small-size defect types, and the matrix elements represent the matching status of the corresponding features in the rows and columns; When all matrix elements in the feature comparison matrix match for the first time, the target small-size defect type corresponding to the feature comparison matrix is ​​determined as the small-size defect type of the distribution transformer.

[0143] The embodiments of the present invention rank candidate defects based on the correlation and difference of common features, and construct a feature comparison matrix for matching and verification one by one. This ensures that the finally determined defect type is highly consistent with the actual detection results at the feature level, and avoids misjudgment caused by partial feature matching.

[0144] like Figure 2 As shown, based on the above method embodiments, corresponding apparatus embodiments are provided; An embodiment of the present invention provides a small-size defect identification device for distribution transformers based on a large power model, comprising: a data acquisition module 201, a first anomaly detection module 202, a second anomaly detection module 203, and a defect matching and identification module 204; The data acquisition module 201 is used to acquire electric field data, voltage data and current data of the distribution transformer, calculate the degree of electric field distortion based on the electric field data, and calculate the dielectric loss angle increment based on the voltage data and current data. The first anomaly detection module 202 is used to perform electric field distortion detection based on the degree of electric field distortion and each electric field parameter in the pre-constructed power operation semantic network, and obtain electric field distortion detection results; The second anomaly detection module 203 is used to identify the fault type based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, and obtain the abnormal fault type; The defect matching and identification module 204 is used to match the electric field distortion detection results and the abnormal fault type in a preset small-size defect identification library to obtain the small-size defect type of the distribution transformer.

[0145] Optionally, in this embodiment of the invention, the first anomaly detection module 202 includes: a first comparison submodule, a second comparison submodule, a third comparison submodule, an abnormal electric field feature set submodule, and an electric field distortion detection submodule; The first comparison submodule is used to obtain the electric field strength deviation value based on the comparison result between the electric field strength reflected by the degree of electric field distortion and the range of electric field strength in the power operation semantic network; The second comparison submodule is used to obtain the electric field distribution uniformity difference value based on the comparison result between the electric field distribution reflected by the degree of electric field distortion and the electric field distribution uniformity index in the power operation semantic network. The third comparison submodule is used to compare the spatial change of electric field intensity reflected by the degree of electric field distortion with the curve of electric field intensity changing with spatial location in the power operation semantic network point by point to obtain the deviation curve of electric field intensity spatial change. The abnormal electric field feature set submodule is used to take the electric field intensity deviation value and the electric field distribution uniformity difference value as numerical features, and take the spatial coordinate points in the electric field intensity spatial change deviation curve with deviation values ​​greater than a preset threshold and the deviation values ​​corresponding to the spatial coordinate points as spatial features, so as to obtain multiple abnormal electric field feature sets. The electric field distortion detection submodule is used to perform electric field distortion detection based on each of the abnormal electric field feature sets, and obtain the electric field distortion detection result.

[0146] This invention compares electric field strength, distribution uniformity, and spatial variation patterns with standard parameters in a semantic network, and extracts numerical and spatial features to construct an abnormal electric field feature set. This comprehensively describes abnormal electric field conditions and provides accurate input for subsequent electric field distortion detection.

[0147] Optionally, in this embodiment of the invention, the electric field distortion detection submodule includes: a first causal association path unit, a second causal association path unit, a transmission path length unit, an abnormal association path unit, and an electric field distortion detection unit; The first causal association path unit is used to determine the first causal association path corresponding to each of the abnormal electric field feature sets based on the causal association relationship between each electric field parameter in each of the abnormal electric field feature sets. The second causal path unit is used to filter each of the first causal paths based on the target node associated with the electric field distortion to obtain multiple second causal paths; The transmission path length unit is used to determine the transmission path length from the starting node to the target node of each second causal relationship path; The abnormal association path unit is used to sort each second causal association path based on the length of each transmission path to obtain a causal association path sequence, and to take the first second causal association path in the causal association path sequence as the abnormal association path. The electric field distortion detection unit is used to determine the mapping information of the abnormal association path in the power operation semantic network as the electric field distortion detection result.

[0148] This invention, through mining the causal relationships within the abnormal electric field feature set, filters and sorts the causal paths most likely to cause electric field distortion, and uses the mapping information of this path in the semantic network as the detection result, can accurately locate the specific cause and fault information of electric field distortion, thereby improving the accuracy of detection.

[0149] Optionally, in this embodiment of the invention, the second anomaly detection module 203 includes: a dielectric loss angle deviation range submodule, a load rate difference value submodule, a change mode submodule, a fault type submodule, an occurrence probability submodule, and an abnormal fault type submodule; The dielectric loss angle deviation interval submodule is used to obtain the dielectric loss angle deviation interval based on the comparison result between the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network; The load rate difference value submodule is used to calculate the difference between the load rate of the distribution transformer under the current operating conditions and the preset standard load rate to obtain the load rate difference value. The change mode submodule is used to determine the change mode based on the dielectric loss angle deviation range and the load rate difference value; The fault type submodule is used to determine multiple fault types based on the change pattern and the preset correspondence between the change pattern and the fault type; The occurrence probability submodule is used to determine the occurrence probability of each fault type based on the correspondence between each fault type and the historical fault types and fault occurrence probabilities in the power operation semantic network. The abnormal fault type submodule is used to sort each fault type based on the occurrence probability to obtain a fault type sequence, and to determine the first fault type in the fault type sequence as the abnormal fault type.

[0150] This invention determines the change pattern by combining the dielectric loss angle deviation range and the load rate difference value, and sorts the potential fault types according to this pattern and historical fault probabilities, thereby accurately identifying the most likely abnormal fault type at present.

[0151] Optionally, in this embodiment of the invention, the change mode submodule includes: a same-direction change unit, a rate change ratio unit, a first change mode unit, and a second change mode unit; The same-direction change unit is used to determine the first rate of change of the dielectric loss angle increment under the load rate if the difference between the dielectric loss angle deviation range and the load rate is the same-direction change, based on the correlation between the load rate and the dielectric loss angle increment in the power operation semantic network. The rate change ratio unit is used to determine the rate change ratio of the dielectric loss angle increment based on the ratio of the first change rate to the second change rate of the dielectric loss angle increment under the load rate. The first change mode unit is used to determine the change mode as a first abnormal mode if the rate change ratio is greater than a preset ratio threshold; wherein, the first abnormal mode represents a mode in which the abnormal growth rate is greater than a preset growth rate threshold. The second change mode unit is used to determine the change mode as a second abnormal mode if the rate change ratio is less than or equal to a preset ratio threshold; wherein, the second abnormal mode represents a mode in which the abnormal growth rate is less than or equal to a preset growth rate threshold.

[0152] This invention, by comparing the theoretical rate of change with the actual rate of change when the dielectric loss angle and the load rate change in the same direction, can determine the abnormal growth pattern and accurately judge whether the change in dielectric loss angle is abnormal and the degree of abnormality.

[0153] Optionally, in this embodiment of the invention, the change mode submodule further includes: a reverse change unit, a target fluctuation standard deviation unit, a standard score unit, a third change mode unit, and a fourth change mode unit; The reverse change unit is used to obtain the set of fluctuation standard deviations of the dielectric loss angle increment under different load rates from the power operation semantic network if the difference between the dielectric loss angle deviation range and the load rate is reversed. The target fluctuation standard deviation unit is used to traverse the fluctuation standard deviation set based on the load rate to obtain the target fluctuation standard deviation corresponding to the target load rate whose difference from the load rate is less than a preset difference threshold. The standard score unit is used to calculate a standard score based on the mean of the dielectric loss angle increment under the load rate, the dielectric loss angle increment, and the target fluctuation standard deviation. The third change mode unit is used to determine the change mode as the first abnormal mode if the standard score is greater than the preset score threshold. The fourth change mode unit is used to determine the change mode as the second abnormal mode if the standard score is less than or equal to a preset score threshold.

[0154] This invention, through the calculation of standard scores using the standard deviation of the load rate fluctuation when the dielectric loss angle and the load rate change in opposite directions, determines the abnormal mode, enabling accurate judgment of the degree and mode of abnormality in cases of inverse changes.

[0155] Optionally, in this embodiment of the invention, the defect matching and identification module 204 includes: a candidate defect set submodule, an electric field distortion intersection feature submodule, a fault type intersection feature submodule, a common intersection feature submodule, a common feature correlation degree submodule, and a small-size defect type submodule; The candidate defect set submodule is used to filter out small-size defect types that match at least one of the electric field distortion detection results or the abnormal fault types from the small-size defect identification library to obtain a candidate defect set. The electric field distortion intersection feature submodule is used to intersect the electric field distortion results of each small-sized defect type in the candidate defect set with the electric field distortion detection results to obtain multiple electric field distortion intersection features. The fault type intersection feature submodule is used to intersect the fault types of each small-size defect type in the candidate defect set with the abnormal fault type to obtain multiple fault type intersection features. The common intersection feature submodule is used to intersect the electric field distortion intersection features and fault type intersection features of each of the small-size defect types to obtain the common intersection features of the candidate defect set. The common feature correlation submodule is used to perform correlation analysis based on the electric field distortion results and fault types of each small-size defect type and the common intersection features to obtain the common feature correlation degree of each small-size defect type. The small-size defect type submodule is used to filter defect types based on the correlation degree of each common feature to obtain the small-size defect type of the distribution transformer.

[0156] The embodiments of the present invention first obtain a candidate set by preliminary screening based on electric field or fault characteristics, and then perform multi-level intersection operations and correlation analysis, which can effectively eliminate interference information and quantify the degree of matching between each candidate defect and the actual detection results.

[0157] Optionally, in this embodiment of the invention, the small-size defect type submodule includes: a difference degree unit, a small-size defect type sequence unit, a feature comparison matrix unit, and a small-size defect type unit; The difference unit is used to take the small-size defect type corresponding to the largest common feature correlation degree among the common feature correlation degrees as the target small-size defect type, and determine the difference degree between the electric field distortion result and fault type of each target small-size defect type and the electric field distortion detection result and abnormal fault type. The small-size defect type sequence unit is used to sort each target small-size defect type based on each difference degree, so as to obtain a small-size defect type sequence arranged in ascending order of difference degree value; The feature comparison matrix unit is used to select target small-size defect types sequentially according to the order of the small-size defect type sequence to construct a feature comparison matrix for judgment; wherein, the rows in the feature comparison matrix represent the electric field distortion detection results and abnormal fault types, the columns represent the electric field distortion results and fault types of the target small-size defect types, and the matrix elements represent the matching status of the corresponding features in the rows and columns; The small-size defect type unit is used to determine the target small-size defect type corresponding to the feature comparison matrix as the small-size defect type of the distribution transformer when all matrix elements in the feature comparison matrix are matched for the first time.

[0158] The embodiments of the present invention rank candidate defects based on the correlation and difference of common features, and construct a feature comparison matrix for matching and verification one by one. This ensures that the finally determined defect type is highly consistent with the actual detection results at the feature level, and avoids misjudgment caused by partial feature matching.

[0159] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can realize the method for identifying small-size defects in distribution transformers based on a large power model provided by any of the above-described method embodiments of the present invention.

[0160] In this embodiment of the invention, the data acquisition module 201 collects electric field, voltage, and current data and calculates the degree of electric field distortion and the increment of dielectric loss angle. This allows for the acquisition of key parameters reflecting changes in the transformer's electric field and internal electrical characteristics, providing a data foundation for subsequent anomaly detection. The first anomaly detection module 202 detects potential problems affecting the electric field by combining the degree of electric field distortion with the correlation between electric field parameters in the semantic network. The second anomaly detection module 203 identifies possible abnormal fault types within the distribution transformer by combining the increment of dielectric loss angle with the correlation between dielectric loss angle parameters in the semantic network. Finally, the defect matching and identification module 204 matches the electric field distortion detection results and abnormal fault types with a defect database, enabling the identification of small-sized defect types in the distribution transformer.

[0161] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0162] like Figure 3 As shown, based on the above embodiment of a method for identifying small-size defects in distribution transformers based on a large power model, another embodiment of the present invention provides an electronic device 300. The electronic device 300 includes a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and configured to be executed by the processor 320. When the processor 320 executes the computer program 311, it implements a method for identifying small-size defects in distribution transformers based on a large power model according to any embodiment of the present invention.

[0163] For example, in this embodiment, the computer program 311 can be divided into one or more modules, which are stored in the memory 310 and executed by the processor 320 to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program 311 in the electronic device 300.

[0164] The electronic device 300 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. The electronic device 300 may include, but is not limited to, a processor 320 and a memory 310.

[0165] The processor 320 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The processor 320 is the control center of the electronic device 300, connecting various parts of the electronic device through various interfaces and lines.

[0166] like Figure 4 As shown, based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium 400, including a stored computer program 311, wherein, when the computer program 311 is running, it controls the device where the computer-readable storage medium 400 is located to execute the method for identifying small-size defects in distribution transformers based on a large power model as described in any of the above-described method embodiments of the present invention.

[0167] The modules / units integrated in the aforementioned device / electronic device, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0168] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for identifying small-size defects in distribution transformers based on a large power model, characterized in that, include: Collect electric field data, voltage data, and current data of the distribution transformer; calculate the degree of electric field distortion based on the electric field data; and calculate the dielectric loss angle increment based on the voltage data and current data. Based on the degree of electric field distortion and the electric field parameters in the pre-constructed power operation semantic network, electric field distortion detection is performed to obtain the electric field distortion detection result. Based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, fault type identification is performed to obtain the abnormal fault type; Based on the electric field distortion detection results and the abnormal fault type, the small-size defect type of the distribution transformer is obtained by matching it in a preset small-size defect identification library.

2. The method for identifying small-size defects in distribution transformers based on a large power model as described in claim 1, characterized in that, The electric field distortion detection is performed based on the degree of electric field distortion and the electric field parameters in the pre-constructed power operation semantic network to obtain electric field distortion detection results, including: Based on the comparison between the electric field strength reflected by the degree of electric field distortion and the range of electric field strength in the power operation semantic network, the electric field strength deviation value is obtained. Based on the comparison between the electric field distribution reflected by the degree of electric field distortion and the electric field distribution uniformity index in the power operation semantic network, the electric field distribution uniformity difference value is obtained. By comparing the spatial variation of electric field intensity reflected by the degree of electric field distortion with the curve of electric field intensity variation with spatial location in the power operation semantic network point by point, the deviation curve of spatial variation of electric field intensity is obtained. The deviation value of electric field intensity and the difference value of electric field distribution uniformity are used as numerical features. The spatial coordinate points in the deviation curve of electric field intensity with deviation values ​​greater than a preset threshold and the deviation values ​​corresponding to the spatial coordinate points are used as spatial features to obtain multiple abnormal electric field feature sets. Electric field distortion detection is performed based on the aforementioned abnormal electric field feature sets to obtain electric field distortion detection results.

3. The method for identifying small-size defects in distribution transformers based on a large power model as described in claim 2, characterized in that, The electric field distortion detection based on each of the aforementioned abnormal electric field feature sets, to obtain the electric field distortion detection result, includes: Based on the causal relationship between the electric field parameters in each of the abnormal electric field feature sets, the first causal relationship path corresponding to each of the abnormal electric field feature sets is determined. Based on the target node associated with electric field distortion, each of the first causal correlation paths is filtered to obtain multiple second causal correlation paths; Determine the transmission path length from the starting node to the target node of each of the second causal association paths; Based on the length of each of the aforementioned transmission paths, each of the second causal association paths is sorted to obtain a causal association path sequence, and the first second causal association path in the causal association path sequence is taken as an abnormal association path. The mapping information of the abnormal association path in the power operation semantic network is determined as the electric field distortion detection result.

4. The method for identifying small-size defects in distribution transformers based on a large power model as described in claim 1, characterized in that, The fault type identification based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network yields abnormal fault types, including: Based on the comparison between the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, the dielectric loss angle deviation interval is obtained; The load rate difference value is obtained by calculating the difference between the load rate of the distribution transformer under the current operating conditions and the preset standard load rate. The change pattern is determined based on the deviation range of the dielectric loss angle and the difference in load rate; Based on the change patterns and the correspondence between preset change patterns and fault types, multiple fault types are determined; Based on the correspondence between each of the fault types and the historical fault types and fault occurrence probabilities in the power operation semantic network, the occurrence probability of each of the fault types is determined. Based on the occurrence probability of each fault type, the fault types are sorted to obtain a fault type sequence, and the first fault type in the fault type sequence is determined as the abnormal fault type.

5. The method for identifying small-size defects in distribution transformers based on a large power model as described in claim 4, characterized in that, The determination of the change pattern based on the deviation range of the dielectric loss angle and the difference in load rate includes: If the deviation range of the dielectric loss angle and the difference value of the load rate change in the same direction, then based on the correlation between the load rate and the increment of the dielectric loss angle in the power operation semantic network, the first rate of change of the increment of the dielectric loss angle under the load rate is determined. The rate of change of the dielectric loss angle increment is determined based on the ratio of the first rate of change to the second rate of change of the dielectric loss angle increment under the load rate. If the rate of change is greater than a preset ratio threshold, the change mode is determined as the first abnormal mode; wherein, the first abnormal mode represents a mode in which the abnormal growth rate is greater than the preset growth rate threshold. If the rate change ratio is less than or equal to a preset ratio threshold, the change mode is determined as the second abnormal mode; wherein, the second abnormal mode represents a mode in which the abnormal growth rate is less than or equal to the preset growth rate threshold.

6. The method for identifying small-size defects in distribution transformers based on a large power model as described in claim 5, characterized in that, The method of determining the change pattern based on the deviation range of the dielectric loss angle and the difference in load rate also includes: If the deviation range of the dielectric loss angle and the difference value of the load rate change in opposite directions, then the set of fluctuation standard deviations of the dielectric loss angle increment under different load rates is obtained from the power operation semantic network; Based on the load rate, the target fluctuation standard deviation is obtained by traversing the set of fluctuation standard deviations to obtain the target load rate whose difference from the load rate is less than a preset difference threshold. A standard score is calculated based on the mean of the dielectric loss angle increment under the load rate, the dielectric loss angle increment, and the standard deviation of the target fluctuation. If the standard score is greater than the preset score threshold, the change pattern is determined as the first abnormal pattern. If the standard score is less than or equal to the preset score threshold, the change mode is determined as the second abnormal mode.

7. The method for identifying small-size defects in distribution transformers based on a large power model as described in claim 1, characterized in that, The matching of the electric field distortion detection results and the abnormal fault type in a preset small-size defect identification database yields the small-size defect types of the distribution transformer, including: In the small-size defect identification library, at least one of the small-size defect types that matches the electric field distortion detection result or the abnormal fault type is selected to obtain a candidate defect set. The electric field distortion results of each small-sized defect type in the candidate defect set are intersected with the electric field distortion detection results to obtain multiple electric field distortion intersection features; The intersection of the fault types of each small-sized defect type in the candidate defect set with the abnormal fault type yields multiple fault type intersection features. The common intersection features of the candidate defect set are obtained by intersecting the electric field distortion intersection features and the fault type intersection features of each of the small-size defect types. Based on the electric field distortion results and fault types of each small-size defect type, a correlation analysis is performed with the common intersection features to obtain the correlation degree of the common features of each small-size defect type. Based on the correlation of the common features, the defect types are screened to obtain the small-size defect types of the distribution transformer.

8. The method for identifying small-size defects in distribution transformers based on a large power model as described in claim 7, characterized in that, The defect type screening based on the correlation of the common features yields the small-size defect types of the distribution transformer, including: The small-size defect type corresponding to the common feature correlation degree with the largest value among the common feature correlation degrees is taken as the target small-size defect type. The difference between the electric field distortion result and fault type of each target small-size defect type and the electric field distortion detection result and abnormal fault type is determined. Based on the aforementioned difference degree, the target small-size defect types are sorted to obtain a sequence of small-size defect types arranged in ascending order of difference degree value; According to the order of the small-size defect type sequence, the target small-size defect types are selected sequentially to construct a feature comparison matrix for judgment; wherein, the rows in the feature comparison matrix represent the electric field distortion detection results and abnormal fault types, the columns represent the electric field distortion results and fault types of the target small-size defect types, and the matrix elements represent the matching status of the corresponding features in the rows and columns; When all matrix elements in the feature comparison matrix match for the first time, the target small-size defect type corresponding to the feature comparison matrix is ​​determined as the small-size defect type of the distribution transformer.

9. A device for identifying small-sized defects in distribution transformers based on a large power model, characterized in that, include: The system includes a data acquisition module, a first anomaly detection module, a second anomaly detection module, and a defect matching and identification module. The data acquisition module is used to collect electric field data, voltage data and current data of the distribution transformer, calculate the degree of electric field distortion based on the electric field data, and calculate the dielectric loss angle increment based on the voltage data and current data. The first anomaly detection module is used to perform electric field distortion detection based on the degree of electric field distortion and each electric field parameter in the pre-constructed power operation semantic network, and obtain the electric field distortion detection result; The second anomaly detection module is used to identify the fault type based on the dielectric loss angle increment and the dielectric loss angle parameter in the power operation semantic network, and obtain the abnormal fault type; The defect matching and identification module is used to match the electric field distortion detection results and the abnormal fault type in a preset small-size defect identification library to obtain the small-size defect type of the distribution transformer.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, the device containing the computer-readable storage medium is controlled to perform a method for identifying small-size defects in distribution transformers based on a large power model as described in any one of claims 1-7.

Images