Distribution network fault prediction method based on digital twin

Abstract

The present invention discloses a distribution network fault prediction method based on digital twins, and relates to the field of data processing technology. The method comprises: calling a digital twin model of a target distribution network to obtain a plurality of fault update records; performing explicit indicator feature extraction on the plurality of fault update records to obtain a plurality of explicit indicator feature sets; performing fault aggregation of the same type, and performing attenuation cumulative analysis to generate K aggregated explicit indicator sets and K aggregated explicit indicator attenuation factor sets; performing equipment implicit association identification to obtain K implicit indicator sets, and constructing a mapping relationship to generate an explicit-implicit indicator feature mapping relationship; obtaining a real-time explicit indicator set and a real-time explicit indicator feature set; performing real-time fault prediction of the target distribution network to obtain a fault prediction result. The method solves the technical problem that the distribution network fault prediction in the prior art is not accurate and timely enough, and achieves the technical effect of improving the accuracy and real-time performance of the distribution network fault prediction.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a distribution network fault prediction method based on digital twins. Background Art

[0002] In the context of the rapid development of smart grids today, the stability and reliability of distribution networks, as a key link in power transmission and distribution, are directly related to the normal operation of the social economy and the quality of life of residents. However, with the increasing scale and complexity of power grids, the risk of failure faced by distribution networks has also increased. Traditional fault prediction methods often rely on statistical analysis of historical data and expert experience and judgment, which makes it difficult to fully and real-timely capture the dynamic changes and potential fault factors in the operation of distribution networks, resulting in limited accuracy and timeliness of fault prediction. Summary of the invention

[0003] The present application provides a distribution network fault prediction method based on digital twins, which solves the technical problem that distribution network fault prediction in the prior art is not accurate and timely enough.

[0004] In view of the above problems, the present application provides a distribution network fault prediction method based on digital twins.

[0005] The present application provides a distribution network fault prediction method based on digital twins, the method comprising:

[0006] Retrieve the digital twin model of the target distribution network, interact with the digital twin model to extract fault update records, and obtain multiple fault update records, wherein the multiple fault update records include multiple update times; use explicit identification branches to extract explicit indicator features of the multiple fault update records to obtain multiple explicit indicator feature sets; perform fault-like aggregation on the multiple explicit indicator feature sets, and perform attenuation cumulative analysis on the aggregation results according to the multiple update times to generate K aggregated explicit indicator sets and K aggregated explicit indicator attenuation factor sets; perform implicit device association identification on the K aggregated explicit indicator sets and the device configuration information of the target distribution network to obtain K implicit indicator sets. The method comprises the steps of: obtaining a real-time operating data set of the target distribution network, identifying outliers of the real-time operating data set, and obtaining a real-time explicit indicator set and a real-time explicit indicator feature set based on the identification result; performing real-time fault prediction of the target distribution network based on the explicit-implicit indicator feature mapping relationship, the K aggregated explicit indicator sets and the K aggregated explicit indicator attenuation factor sets, and the real-time explicit indicator set and the real-time explicit indicator feature set, and obtaining a fault prediction result.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] By calling the digital twin model of the target distribution network, the interactive digital twin model is used to extract fault update records to obtain multiple fault update records, wherein the multiple fault update records include multiple update times. Then, the explicit indicator feature extraction is performed on the multiple fault update records using the explicit identification branch to obtain multiple explicit indicator feature sets. Further, multiple explicit indicator feature sets are aggregated according to the same fault type, and the aggregation results are analyzed by attenuation accumulation according to multiple update times to generate K aggregated explicit indicator sets and K aggregated explicit indicator attenuation factor sets. Next, the implicit association of the K aggregated explicit indicator sets and the equipment configuration information of the target distribution network is identified to obtain K implicit indicator sets, and the mapping relationship between the K aggregated explicit indicator sets and the K implicit indicator sets is constructed to generate an explicit-implicit indicator feature mapping relationship. By interacting with the digital twin model at the current moment, the real-time operation data set of the target distribution network is obtained, and the outlier identification is performed on the real-time operation data set. According to the identification results, the real-time explicit indicator set and the real-time explicit indicator feature set are obtained. Finally, based on the explicit-implicit indicator feature mapping relationship, K aggregated explicit indicator sets and K aggregated explicit indicator attenuation factor sets, as well as the real-time explicit indicator set and the real-time explicit indicator feature set, the real-time fault prediction of the target distribution network is performed to obtain the fault prediction result. This solves the technical problem that the distribution network fault prediction in the prior art is not accurate and timely enough, and achieves the technical effect of improving the accuracy and real-time performance of the distribution network fault prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of a flow chart of a distribution network fault prediction method based on digital twins provided in an embodiment of the present application;

[0011] Figure 2 A schematic diagram of the process of performing fault-type aggregation in the distribution network fault prediction method based on digital twins provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The present application solves the technical problem that distribution network fault prediction is not accurate and timely enough in the prior art by providing a distribution network fault prediction method based on digital twins.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that the terms "including" and "having" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0015] Examples, such as Figure 1 As shown, the embodiment of the present application provides a distribution network fault prediction method based on digital twins, wherein the method includes:

[0016] The digital twin model of the target distribution network is retrieved, and the digital twin model is interacted with to extract fault update records to obtain multiple fault update records, wherein the multiple fault update records include multiple update times.

[0017] The digital twin model of the target distribution network is a model based on digital twin technology that reflects the real situation of the target distribution network. It contains information on the structure, equipment parameters, and operation data of the distribution network. By calling the digital twin model of the target distribution network and interacting with the digital twin model to extract fault update records, multiple fault update records are obtained. The fault update records include fault description (fault type and impact range), fault time (fault occurrence event), fault location (fault location in the distribution network), fault handling (measures taken for the fault and results), and update time (the last update time of the fault record).

[0018] Explicit indicator feature extraction is performed on the multiple fault update records using explicit identification branches to obtain multiple explicit indicator feature sets.

[0019] Explicit identification branches are set based on multiple dimensions such as fault type, fault location, fault impact range, fault occurrence time, and repair time. For example, the fault type branch is set to distinguish different types of faults (such as short circuit, open circuit, overload, etc.), and the fault impact range branch can also be set to evaluate the impact of the fault on the overall operation of the power grid. Explicit indicator features refer to indicative indicators of faults, such as voltage fluctuation indicators of distribution networks, which can directly reflect the characteristics of faults. Explicit identification branches are used to extract explicit indicator features from multiple fault update records to obtain multiple explicit indicator feature sets.

[0020] The multiple explicit indicator feature sets are aggregated according to the same type of faults, and the aggregation results are analyzed by attenuation accumulation according to the multiple update times to generate K aggregated explicit indicator sets and K aggregated explicit indicator attenuation factor sets.

[0021] Multiple explicit indicator feature sets are aggregated into the same fault category, that is, multiple records are grouped into different fault categories according to the fault type or fault feature similarity in the explicit indicator feature set, and K aggregated explicit indicator sets are obtained. The aggregation results are subjected to attenuation cumulative analysis according to multiple update times, that is, based on the aggregation of faults of the same type, the degree of change of the indicator response data at different time points is further analyzed, and then K aggregated explicit indicator attenuation factor sets are obtained.

[0022] Furthermore, if Figure 2 As shown, the multiple explicit indicator feature sets are aggregated into the same type of faults, including:

[0023] Taking the fault description as an index, extract data from the multiple fault update records to obtain multiple fault description data; randomly extract a first explicit indicator feature set from the multiple explicit indicator feature sets, and use the corresponding first fault description data as the first aggregation node of the fault aggregation tree; randomly extract a second explicit indicator feature set from the multiple explicit indicator feature sets again, and use the corresponding second fault description data as the second aggregation node of the fault aggregation tree; randomly extract a K-1th explicit indicator feature set from the multiple explicit indicator feature sets, and use the corresponding K-1th fault description data as the K-1th aggregation node of the fault aggregation tree, wherein the data similarity between the K-1th fault description data and the first fault description data and the data similarity between the K-1th fault description data and the second fault description data are lower than a preset minimum similarity; generate the fault aggregation tree according to the first aggregation node, the second aggregation node and the K-1th aggregation node; input the multiple explicit indicator feature sets into the fault aggregation tree to generate an aggregation result, wherein the aggregation result includes K aggregated explicit indicator feature clusters.

[0024] Preferably, with the fault description as an index, corresponding fault description data is extracted from multiple fault update records, and each fault description data is associated with its corresponding explicit indicator feature set; an explicit indicator feature set is randomly extracted from multiple explicit indicator feature sets as a first explicit indicator feature set, and the first fault description data corresponding to the first explicit indicator feature set is used as a first aggregation node of the fault aggregation tree; an explicit indicator feature set is randomly extracted from multiple explicit indicator feature sets as a second explicit indicator feature set, and the second fault description data corresponding to the second explicit indicator feature set is used as a second aggregation node of the fault aggregation tree; Take the K-1th explicit indicator feature set, and use the K-1th fault description data corresponding to the K-1th explicit indicator feature set as the K-1th aggregation node of the fault aggregation tree, wherein the data similarity between the K-1th fault description data and the first fault description data and the data similarity between the K-1th fault description data and the second fault description data are lower than a preset minimum similarity; generate a fault aggregation tree according to the first aggregation node, the second aggregation node and the K-1th aggregation node; input multiple explicit indicator feature sets into the fault aggregation tree to generate an aggregation result, specifically, by traversing the fault aggregation tree, aggregate explicit indicator feature sets with similar characteristics together to form K aggregated explicit indicator feature clusters.

[0025] Furthermore, the data similarity between the first fault description data and the second fault description data is lower than a preset minimum similarity.

[0026] In the process of constructing the fault aggregation tree, ensure that the similarity between the fault description data at different levels or branches is lower than the preset minimum similarity, so that the hierarchical structure of the aggregation tree can effectively reflect the differences between different fault types or fault modes. The data similarity between the first fault description data and the second fault description data is lower than the preset minimum similarity, which means that in the process of constructing the fault aggregation tree, the similarity between the newly added fault description data (such as the second fault description data) and the existing nodes (such as the nodes corresponding to the first fault description data) will be checked to ensure that the difference between them is large enough to avoid aggregating them under the same node. The similarity between the first fault description data and the second fault description data can be calculated using a similarity measurement method (such as cosine similarity, Jaccard similarity, etc.).

[0027] Further, including:

[0028] The multiple explicit indicator feature sets are input into the fault aggregation tree, and the data similarity between the multiple fault description data and the first fault description data in the first aggregation node is calculated to obtain multiple data similarities; when the multiple data similarities meet the preset similarity interval, the corresponding multiple explicit indicator feature sets are stored in the first internal node of the first aggregation node; when the multiple data similarities do not meet the preset similarity interval, the corresponding multiple explicit indicator feature sets are added to the first iteration set; the first iteration set is aggregated by the second aggregation node to obtain a second internal node and a second iteration set; the multiple explicit indicator feature sets in the K-2 iteration set are aggregated by the K-1 aggregation node to obtain a K-1 internal node and a K-1 iteration set; according to the first internal node, the second internal node, the K-1 internal node and the K-1 iteration set, the K aggregated explicit indicator feature clusters are obtained, wherein each aggregated explicit indicator feature cluster includes multiple aggregated explicit indicator feature sets with different update times.

[0029] Preferably, multiple explicit indicator feature sets are input into the fault aggregation tree, and the data similarity between the fault description data in these sets and the first fault description data in the first aggregation node is calculated to obtain multiple data similarities; when multiple data similarities meet the preset similarity interval, it is considered that the corresponding multiple explicit indicator feature sets are similar to the fault description data in the first aggregation node, and therefore they are stored in the first internal node of the first aggregation node; if the data similarity does not meet the preset similarity interval, these sets are added to the first iteration set for subsequent iterative aggregation; the first iteration set is aggregated in the same category using the second aggregation node, that is, under the standard of the second aggregation node, the first iteration set is aggregated into the same category. The sets in an iteration set are grouped according to similarity to form a second internal node and a second iteration set; the above process is repeated, and the multiple explicit indicator feature sets in the K-2 iteration set are aggregated in the same category using the K-1th aggregation node to obtain the K-1th internal node and the K-1th iteration set; after multiple iterations, K aggregated explicit indicator feature clusters are formed according to the first internal node, the second internal node, the K-1th internal node and the final K-1th iteration set, each aggregated explicit indicator feature cluster includes multiple aggregated explicit indicator feature sets with different update times, that is, the sets in the same cluster may come from different time points, but they have similar fault characteristics to some extent.

[0030] Further, including:

[0031] Indicators are extracted according to the K aggregated explicit indicator feature clusters to obtain K aggregated explicit indicator sets; the K aggregated explicit indicator feature clusters are sorted in chronological order based on the multiple update times to obtain K aggregated explicit indicator feature set sequences; with the aggregated explicit indicators as indexes, indicator attenuation cumulative analysis is performed on the K aggregated explicit indicator feature set sequences to generate the K aggregated explicit indicator attenuation factor sets.

[0032] Preferably, for each aggregated explicit indicator feature cluster, a representative explicit indicator is extracted as the aggregated explicit indicator of the cluster, and for K aggregated explicit indicator feature clusters, K aggregated explicit indicators are extracted respectively to form K aggregated explicit indicator sets; for each aggregated explicit indicator feature cluster, the explicit indicator feature sets contained therein are sorted according to the update time, that is, they are sorted in chronological order to obtain K aggregated explicit indicator feature set sequences, each sequence represents the change of the corresponding aggregated explicit indicator over time; with the aggregated explicit indicator as the index, each aggregated explicit indicator feature set sequence is subjected to attenuation cumulative analysis, by calculating the attenuation of the indicator value at each time point (that is, the decrease compared with the previous time point), and then these attenuations are accumulated to generate the attenuation factor sequence of the indicator; for each aggregated explicit indicator, its attenuation factors at all time points are combined to form the aggregated explicit indicator attenuation factor set of the indicator, and the aggregated explicit indicator attenuation factor set reflects the attenuation trend and degree of the indicator over time.

[0033] The K aggregated explicit indicator sets and the equipment configuration information of the target power distribution network are identified for implicit equipment association, K implicit indicator sets are obtained, and a mapping relationship between the K aggregated explicit indicator sets and the K implicit indicator sets is constructed to generate an explicit-implicit indicator feature mapping relationship.

[0034] The equipment configuration information of the target distribution network, including the equipment model, specifications, installation location, operating parameters, etc., uses data analysis techniques (such as association rule mining, causal reasoning, machine learning, etc.) to explore the implicit relationship between the aggregated explicit indicator set and the equipment configuration information. These relationships may be manifested as implicit associations between changes in certain explicit indicators and specific equipment states or performance parameters. Based on the results of association analysis, those implicit indicators that are closely related to the equipment state or performance are extracted to obtain K implicit indicator sets. For each aggregated explicit indicator set, a set of implicit indicators associated with it can be found to construct a mapping relationship between K aggregated explicit indicator sets and K implicit indicator sets, and integrate the mapping relationships between all aggregated explicit indicator sets and implicit indicator sets to form a complete explicit-implicit indicator feature mapping relationship.

[0035] The digital twin model of the current moment is interacted to obtain the real-time operation data set of the target distribution network, and outlier identification is performed on the real-time operation data set, and a real-time explicit indicator set and a real-time explicit indicator feature set are obtained according to the identification results.

[0036] By interacting with the digital twin model at the current moment, the real-time operating data set of the target distribution network is obtained, including parameters such as voltage, current, and power. Statistical methods (such as Z-score, IQR, etc.), machine learning algorithms (such as isolation forest, support vector machine, etc.) or rule-based methods are used to detect outliers in the data. Outliers may be caused by sensor failure, data transmission errors, or abnormal conditions in the distribution network. The detected outliers are processed according to the specific situation, including deleting outliers, replacing them with mean values ​​or medians, filling them with interpolation methods, etc. According to the recognition results, a real-time explicit indicator set and a real-time explicit indicator feature set are obtained.

[0037] Based on the explicit-implicit indicator feature mapping relationship, the K aggregated explicit indicator sets and the K aggregated explicit indicator attenuation factor sets, as well as the real-time explicit indicator set and the real-time explicit indicator feature set, real-time fault prediction of the target distribution network is performed to obtain a fault prediction result.

[0038] The explicit-implicit indicator feature mapping relationship, aggregated explicit indicators and their attenuation factors, and real-time explicit indicators and feature sets are used to perform real-time fault prediction on the target distribution network and obtain the fault prediction results.

[0039] Further, including:

[0040] The real-time explicit indicator set and the K aggregated explicit indicator sets are matched to obtain a matching explicit indicator set and a matching explicit indicator attenuation factor set, and a search is performed according to the explicit-implicit indicator mapping relationship to obtain a matching implicit indicator set; based on the matching implicit indicator set, indicator features of the digital twin model at the current moment are extracted to obtain a matching implicit indicator feature set; fault prediction is performed on the real-time explicit indicator feature set, the matching explicit indicator attenuation factor set and the matching implicit indicator feature set to obtain a fault prediction result of the target distribution network.

[0041] Preferably, the real-time explicit indicator set is compared with the K aggregated explicit indicator sets to find those explicit indicators that exist in both sets to form a matching explicit indicator set; for each indicator in the matching explicit indicator set, the corresponding attenuation factor is retrieved from the K aggregated explicit indicator attenuation factor sets to form a matching explicit indicator attenuation factor set; using the explicit-implicit indicator mapping relationship, each indicator in the matching explicit indicator set is retrieved to find the implicit indicator associated with it to form a matching implicit indicator set; interacting with the digital twin model at the current moment, requesting to extract the current eigenvalue of each implicit indicator in the matching implicit indicator set, and organizing the eigenvalues ​​obtained from the digital twin model into a matching implicit indicator feature set; the real-time explicit indicator feature set, the matching explicit indicator attenuation factor set and the matching implicit indicator feature set are fused to form a comprehensive data set for fault prediction, and the comprehensive data set is processed using a pre-trained fault prediction model (such as a machine learning model, a deep learning model, etc.) to calculate the fault prediction result of the target distribution network.

[0042] Further, including:

[0043] Constructing a fault prediction network layer; inputting the real-time explicit indicator feature set, the matching explicit indicator attenuation factor set and the matching implicit indicator feature set into the fault prediction network layer for analysis to obtain the fault prediction result.

[0044] Preferably, a fault prediction network layer is constructed based on a deep learning framework, and the prediction network layer includes an input layer, a hidden layer, an activation function, and an output layer, wherein the input layer is used to receive a real-time explicit indicator feature set, a matching explicit indicator attenuation factor set, and a matching implicit indicator feature set, and the output layer is used to output a fault prediction result. Specifically, the fault prediction network is trained by using historical data (including data of known faults and data of normal operation), and a fault prediction network layer is obtained after the training is completed; the real-time explicit indicator feature set, the matching explicit indicator attenuation factor set, and the matching implicit indicator feature set are taken as input and sent to the fault prediction network layer, and the output of the network layer is calculated through forward propagation to obtain a fault prediction result.

[0045] Further, including:

[0046] Using the K aggregated explicit indicator sets as indexes, respectively, extracting contemporaneous indicators from the maintenance records in the multiple fault update records to obtain K contemporaneous indicator sets, wherein each contemporaneous indicator has an occurrence frequency identifier; retaining the contemporaneous indicators whose frequency identifiers are in the first m positions in the K contemporaneous indicator sets to obtain K target contemporaneous indicator sets; performing equipment matching according to the equipment configuration information and the K target contemporaneous indicator sets to obtain K contemporaneously associated equipment; and using the K equipment indicator sets of the K contemporaneously associated equipment as the K implicit indicator sets.

[0047] Preferably, each indicator set in the K aggregated explicit indicator sets is used as an index. For each index (i.e., an aggregated explicit indicator set), maintenance records related to it are retrieved from multiple fault update records, and other indicators that are contemporaneous with the index indicator are extracted from the retrieved maintenance records to form K contemporaneous indicator sets. The indicators in each set should have an occurrence frequency identifier, indicating the number of times the indicator appears in the same period; each set in the K contemporaneous indicator sets is sorted according to the frequency identifiers of its internal indicators, and the contemporaneous indicators with frequency identifiers in the first m positions are retained from each set, where m is a preset threshold value for controlling the number of retained indicators; equipment matching is performed based on equipment configuration information and K target contemporaneous indicator sets, so as to associate the indicators in each target contemporaneous indicator set with specific equipment to obtain K contemporaneous associated equipment; for each of the K contemporaneous associated equipment, the current equipment indicator set is extracted from its corresponding digital twin model, and the K equipment indicator sets of the K contemporaneous associated equipment are used as K implicit indicator sets.

[0048] In summary, the embodiments of the present application have at least the following technical effects:

[0049] By calling the digital twin model of the target distribution network, the interactive digital twin model is used to extract fault update records to obtain multiple fault update records, wherein the multiple fault update records include multiple update times. Then, the explicit indicator feature extraction is performed on the multiple fault update records using the explicit identification branch to obtain multiple explicit indicator feature sets. Further, multiple explicit indicator feature sets are aggregated according to the same fault type, and the aggregation results are analyzed by attenuation accumulation according to multiple update times to generate K aggregated explicit indicator sets and K aggregated explicit indicator attenuation factor sets. Next, the implicit association of the K aggregated explicit indicator sets and the equipment configuration information of the target distribution network is identified to obtain K implicit indicator sets, and the mapping relationship between the K aggregated explicit indicator sets and the K implicit indicator sets is constructed to generate an explicit-implicit indicator feature mapping relationship. By interacting with the digital twin model at the current moment, the real-time operation data set of the target distribution network is obtained, and the outlier identification is performed on the real-time operation data set. According to the identification results, the real-time explicit indicator set and the real-time explicit indicator feature set are obtained. Finally, based on the explicit-implicit indicator feature mapping relationship, K aggregated explicit indicator sets and K aggregated explicit indicator attenuation factor sets, as well as the real-time explicit indicator set and the real-time explicit indicator feature set, the real-time fault prediction of the target distribution network is performed to obtain the fault prediction result. This solves the technical problem that the distribution network fault prediction in the prior art is not accurate and timely enough, and achieves the technical effect of improving the accuracy and real-time performance of the distribution network fault prediction.

[0050] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0051] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0052] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A distribution network fault prediction method based on digital twins, characterized in that: The method comprises: Retrieving a digital twin model of a target distribution network, interacting with the digital twin model to extract fault update records, and obtaining a plurality of fault update records, wherein the plurality of fault update records include a plurality of update times; Utilizing explicit identification branches to extract explicit indicator features from the multiple fault update records, and obtaining multiple explicit indicator feature sets, wherein the display indicator features are indicative indicators that directly reflect the occurrence of a fault, and the display identification branches are set based on the fault type, fault location, fault impact range, fault occurrence time, and repair time; Aggregate the multiple explicit indicator feature sets according to the same type of faults, and perform attenuation accumulation analysis on the aggregation results according to the multiple update times to generate K aggregated explicit indicator sets and K aggregated explicit indicator attenuation factor sets; Performing device implicit association identification on the K aggregated explicit indicator sets and the device configuration information of the target distribution network to obtain K implicit indicator sets, and constructing a mapping relationship between the K aggregated explicit indicator sets and the K implicit indicator sets to generate an explicit-implicit indicator feature mapping relationship; Interacting with the digital twin model at the current moment, obtaining a real-time operation data set of the target distribution network, and identifying outliers on the real-time operation data set, and obtaining a real-time explicit indicator set and a real-time explicit indicator feature set according to the identification result; Based on the explicit-implicit indicator feature mapping relationship, the K aggregated explicit indicator sets and the K aggregated explicit indicator attenuation factor sets, and the real-time explicit indicator set and the real-time explicit indicator feature set, real-time fault prediction of the target distribution network is performed to obtain a fault prediction result; The generation of K aggregate explicit indicator sets and K aggregate explicit indicator attenuation factor sets includes: Indicators are extracted according to K aggregated explicit indicator feature clusters to obtain K aggregated explicit indicator sets; Based on the multiple update times, the K aggregated explicit indicator feature clusters are sorted in chronological order to obtain K aggregated explicit indicator feature set sequences; Taking the aggregated explicit indicator as an index, the K aggregated explicit indicator feature set sequences are respectively subjected to an indicator attenuation cumulative analysis to generate the K aggregated explicit indicator attenuation factor sets.

2. The distribution network fault prediction method based on digital twin according to claim 1, characterized in that: include: Taking the fault description as an index, extracting data from the multiple fault update records to obtain multiple fault description data; Randomly extracting a first explicit indicator feature set from the multiple explicit indicator feature sets, and using the first fault description data corresponding to the first explicit indicator feature set as the first aggregation node of the fault aggregation tree; randomly extracting a second explicit indicator feature set from the multiple explicit indicator feature sets again, and using the second fault description data corresponding to the second explicit indicator feature set as a second aggregation node of the fault aggregation tree; randomly extracting a K-1th explicit indicator feature set from the multiple explicit indicator feature sets, and using the K-1th fault description data corresponding thereto as the K-1th aggregation node of the fault aggregation tree, wherein the data similarity between the K-1th fault description data and the first fault description data and the data similarity between the K-1th fault description data and the second fault description data are lower than a preset minimum similarity; Generate the fault aggregation tree according to the first aggregation node, the second aggregation node and the K-1th aggregation node; The multiple explicit indicator feature sets are input into the fault aggregation tree to generate an aggregation result, wherein the aggregation result includes K aggregated explicit indicator feature clusters.

3. The distribution network fault prediction method based on digital twin according to claim 2, characterized in that: The data similarity between the first fault description data and the second fault description data is lower than a preset minimum similarity.

4. The distribution network fault prediction method based on digital twin according to claim 2, characterized in that: include: Inputting the multiple explicit indicator feature sets into the fault aggregation tree, calculating data similarities between the multiple fault description data and the first fault description data in the first aggregation node, and obtaining multiple data similarities; When the multiple data similarities satisfy a preset similarity interval, storing the corresponding multiple explicit indicator feature sets in a first internal node of the first aggregation node; When the multiple data similarities do not satisfy the preset similarity interval, adding the corresponding multiple explicit indicator feature sets into the first iteration set; Using the second aggregation node to perform homogeneous aggregation on the first iteration set to obtain a second internal node and a second iteration set; The K-1th aggregation node is used again to perform similar aggregation on multiple explicit indicator feature sets in the K-2th iteration set to obtain the K-1th internal node and the K-1th iteration set; The K aggregated explicit indicator feature clusters are obtained according to the first internal node, the second internal node, the K-1th internal node and the K-1th iteration set, wherein each aggregated explicit indicator feature cluster includes multiple aggregated explicit indicator feature sets with different update times.

5. The distribution network fault prediction method based on digital twin according to claim 1, characterized in that: include: Matching the real-time explicit indicator set with the K aggregated explicit indicator sets to obtain a matching explicit indicator set and a matching explicit indicator attenuation factor set, and performing a search based on the explicit-implicit indicator mapping relationship to obtain a matching implicit indicator set; Extracting indicator features of the digital twin model at the current moment based on the matching implicit indicator set to obtain a matching implicit indicator feature set; Fault prediction is performed on the real-time explicit indicator feature set, the matching explicit indicator attenuation factor set and the matching implicit indicator feature set to obtain a fault prediction result of the target distribution network.

6. The distribution network fault prediction method based on digital twin according to claim 5, characterized in that: include: Build a fault prediction network layer; The real-time explicit indicator feature set, the matching explicit indicator attenuation factor set and the matching implicit indicator feature set are input into the fault prediction network layer for analysis to obtain the fault prediction result.

7. The distribution network fault prediction method based on digital twin according to claim 5, characterized in that: include: Using the K aggregate explicit indicator sets as indexes respectively, extracting contemporaneous indicators from the maintenance records in the multiple fault update records to obtain K contemporaneous indicator sets, wherein each contemporaneous indicator has an occurrence frequency identifier; The contemporaneous indicators whose frequency identifiers are in the first m positions in the K contemporaneous indicator sets are retained to obtain K target contemporaneous indicator sets; Perform device matching according to the device configuration information and the K target synchronous indicator sets to obtain K synchronous associated devices; The K device indicator sets of the K concurrently associated devices are used as the K implicit indicator sets.

Images