A power distribution network planning operation typical problem library construction method, system, device and medium

Abstract

The application discloses a power distribution network planning operation typical problem library construction method, system, equipment and medium, and belongs to the technical field of power distribution network planning, and comprises the following steps: based on a multi-modal large model, each first sub-problem under a basic problem category and a second sub-problem causing the first sub-problem are traversed, and based on the first sub-problem and the second sub-problem under each basic problem category, a level is gradually expanded until all typical problems are covered, and a tree branch is obtained; each basic problem category is fused based on the tree branch, and a typical problem structure tree is obtained; a typical problem knowledge graph is constructed by taking the problems represented by each node in the typical problem structure tree as an entity and taking the existing association relationship between the problems as an entity edge; and a typical problem library is constructed based on the bidirectional relationship mapping between the nodes in the typical problem structure tree and the entities in the typical problem knowledge graph. The application improves the processing efficiency of power distribution network planning operation problems.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution network planning, and in particular to a power distribution network planning operation typical problem library construction method, system, device and medium. BACKGROUND

[0002] In the field of power distribution network, the existing planning operation typical problem library construction method mainly depends on the traditional relational database. According to the established classification rules, various problems generated in the planning operation process of the power distribution network, such as device failure, abnormal power flow, and network structure defects, are simply classified and stored in different data tables. However, due to the complexity and dynamic nature of the power distribution network system, there are complex coupling and hierarchical relationships between problems, for example, device failure may be caused by multiple factors, and at the same time, it will have a chain effect on system power flow and power supply reliability. However, the traditional method can only realize the static storage and simple retrieval of the problem, and cannot effectively show the cause-and-effect relationship, subordinate relationship and hierarchical structure between the problems, which makes it difficult for operation and maintenance personnel to quickly sort out the overall situation and internal relations of the problem when facing actual problems, and cannot efficiently locate the associated problems and develop solutions, which seriously hinders the efficiency of solving the planning operation problems of the power distribution network. SUMMARY

[0003] In view of the above existing problems, the present application is proposed.

[0004] Therefore, the present application provides a power distribution network planning operation typical problem library construction method, system, device and medium to solve the problem of low processing efficiency of the existing planning operation typical problem library construction method.

[0005] To solve the above technical problems, the present application provides the following technical solutions:

[0006] In a first aspect, the present application provides a power distribution network planning operation typical problem library construction method, comprising:

[0007] Obtaining power distribution device data;

[0008] Based on the obtained data, classifying the power distribution network planning operation typical problems to obtain a plurality of basic problem categories;

[0009] Using a multi-modal large model to traverse each first sub-problem and the second sub-problem causing the first sub-problem under each basic problem category in the historical experience library, and gradually expanding the hierarchy based on the first sub-problem and the second sub-problem under each basic problem category until all typical problems are covered, to obtain a tree branch of each basic problem category;

[0010] Based on the tree branch of each basic problem category, a typical problem structure tree is obtained through index logical association analysis and node fusion;

[0011] Based on the typical problem structure tree, a typical problem knowledge graph is constructed by taking the problem represented by each node as an entity and the association relationship between problems as an entity edge.

[0012] Based on the nodes in the typical problem structure tree and the entities in the typical problem knowledge graph, a typical problem library is constructed through bidirectional relationship mapping.

[0013] As a preferred scheme of the power distribution network planning and operation typical problem library construction method, the typical problem structure tree is obtained by index logical association analysis and node fusion, including:

[0014] For any two first basic problem categories and second basic problem categories, if the semantic distance between the first semantic vector of the first basic problem category and the second semantic vector of the second basic problem category is less than the first preset distance threshold, the first keyword of the first basic problem category and the second keyword of the second basic problem category are extracted;

[0015] If the similarity coefficient between the first keyword and the second keyword is greater than the preset coefficient threshold, the co-occurrence frequency index is determined based on the number of common occurrence cases of the two basic problem categories and the number of individual occurrence cases of the first basic problem category and the second basic problem category;

[0016] If the co-occurrence frequency index is greater than the preset frequency threshold, a first logical derivation path from the first basic problem category to the second basic problem category is constructed, and a logical derivation association coefficient is determined based on the path complexity and the path existence probability of the first logical derivation path;

[0017] If the logical derivation association coefficient is greater than the preset association threshold, a connection between the first root node of the first basic problem category and the second root node of the second basic problem category is established;

[0018] The node fusion is performed based on the association relationship between the child nodes of each level in the tree branch of the basic problem category connected by the root node, and a typical problem structure tree is obtained.

[0019] As a preferred scheme of the power distribution network planning and operation typical problem library construction method, the node fusion is performed based on the association relationship between the child nodes of each level in the tree branch of the basic problem category connected by the root node, and a typical problem structure tree is obtained, including:

[0020] For a third basic problem category and a fourth basic problem category with established root node connection, based on the hierarchical depth of the tree branches of the third basic problem category and the fourth basic problem category, the two tree branches are adjusted to the same hierarchical structure to obtain a first adjusted tree branch and a second adjusted tree branch.

[0021] For any two first and second child nodes of the same level in the first and second adjusted tree branches, determine a Manhattan distance based on a first fusion feature vector of the first child node and a second fusion feature vector of the second child node; the fusion feature vector comprises a text feature vector, an image feature vector and an audio feature vector;

[0022] If the Manhattan distance is less than a second preset distance threshold, determine a context feature matching degree based on pre-node features and post-node features of the first and second child nodes;

[0023] If the context feature matching degree is greater than a preset matching degree, perform node fusion based on associated paths of the first and second child nodes to obtain a typical problem structure tree.

[0024] As a preferred scheme of the method for constructing a typical problem library for power distribution network planning and operation, the node fusion based on the associated paths of the first and second child nodes to obtain the typical problem structure tree comprises:

[0025] Search a child level downward from a current level of the first and second child nodes, and add child nodes with a Manhattan distance less than a second preset distance threshold and a context feature matching degree greater than a preset matching degree to a path set to obtain a first associated path and a second associated path respectively;

[0026] Determine a level number of a common child level in the first and second associated paths, and determine a feature similarity based on the first and second fusion feature vectors;

[0027] Determine a node connection strength between the first and second child nodes based on the context feature matching degree, the level number and the feature similarity;

[0028] Perform node fusion based on the node connection strength to obtain the typical problem structure tree.

[0029] As a preferred scheme of the method for constructing a typical problem library for power distribution network planning and operation, the node fusion based on the node connection strength to obtain the typical problem structure tree comprises:

[0030] If the node connection strength is greater than a preset strength threshold, construct a second logical deduction path based on problem descriptions of the first and second child nodes;

[0031] Perform logical contradiction detection based on a premise and a conclusion in a deduction process of the second logical deduction path to obtain a logical contradiction detection result;

[0032] If the logic contradiction detection result is no conflict, whether the knowledge domains involved in the second logic deduction path conflict with each other is determined to obtain a knowledge domain conflict detection result;

[0033] If the knowledge domain conflict detection result is no conflict, a connection between the first sub-node and the second sub-node is established to obtain a typical problem structure tree.

[0034] As a preferred scheme of the power distribution network planning and operation typical problem library construction method, the typical problem library is constructed through bidirectional relationship mapping, and the method comprises the following steps:

[0035] For a pair of nodes with a parent-child hierarchical relationship in the typical problem structure tree, a first entity and a second entity corresponding to the typical problem knowledge graph are matched, and a new first hierarchical association edge is created for the first entity and the second entity in the typical problem knowledge graph to obtain a first updated knowledge graph, and the hierarchical relationship of the structure tree in the structure tree is integrated into the knowledge graph.

[0036] Second hierarchical association edges corresponding to other types except the type corresponding to the first hierarchical association edge in the first updated knowledge graph are traversed, the first node and the second node corresponding to the first node and the second node in the typical problem structure tree are determined, and a new virtual node is created for the first node and the second node in the typical problem structure tree, an association relationship between the first node and the virtual node is established, and an association relationship between the second node and the virtual node is established to obtain a first updated structure tree, and the association relationship between the problems in the knowledge graph is embedded into the structure tree; the type of the association relationship is the same as that in the knowledge graph.

[0037] Bidirectional relationship mapping is performed between the nodes in the first updated structure tree and the entities in the first updated knowledge graph to construct a typical problem library.

[0038] As a preferred scheme of the power distribution network planning and operation typical problem library construction method, the typical problem library is constructed based on bidirectional relationship mapping between the nodes in the first updated structure tree and the entities in the first updated knowledge graph, and the method comprises the following steps:

[0039] In the first updated structure tree, if multiple nodes have the same type of association relationship with the same virtual node, the multiple nodes are directly associated, and the intermediate virtual node is deleted to obtain a second updated structure tree.

[0040] In the first updated knowledge graph, if there are at least two hierarchical association edges of the same type between the same two entities, only one hierarchical association edge is retained to obtain a second updated knowledge graph.

[0041] Traverse all nodes of the second updated structure tree and all entities of the second updated knowledge graph, and match the most similar target entity in the second updated knowledge graph for any target node in the second updated structure tree based on the question text similarity between the nodes and the entities;

[0042] A bidirectional relationship mapping between the target node and the target entity is established to obtain the typical question library.

[0043] In a second aspect, the present application provides a power distribution network planning and operation typical question library construction system, comprising:

[0044] An acquisition module is configured to acquire power distribution equipment data.

[0045] A question classification module is configured to classify power distribution network planning and operation typical questions based on the acquired data to obtain a plurality of basic question categories.

[0046] A tree branch construction module is configured to use a multi-modal large model to traverse a first sub-question under each basic question category and a second sub-question that triggers the first sub-question in a historical experience library, and gradually expand the level based on the first sub-question and the second sub-question under each basic question category until all typical questions are covered to obtain a tree branch of each basic question category.

[0047] A structure tree fusion module is configured to obtain a typical question structure tree through index logical association analysis and node fusion based on the tree branch of each basic question category.

[0048] A knowledge graph construction module is configured to construct a typical question knowledge graph by taking the question represented by each node as an entity and the association relationship between questions as an entity edge based on the typical question structure tree.

[0049] A question library construction module is configured to construct a typical question library through bidirectional relationship mapping based on the nodes in the typical question structure tree and the entities in the typical question knowledge graph.

[0050] In a third aspect, the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the power distribution network planning and operation typical question library construction method when executing the computer program.

[0051] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the power distribution network planning and operation typical question library construction method.

[0052] The application has the beneficial effects that: the application organizes power distribution network planning and operation problems in an orderly manner in a hierarchical manner by constructing a structure tree, clearly presents the classification and hierarchy of the problems from the root node to each level of the child nodes, and enables the operation and maintenance personnel to quickly grasp the overall architecture of the problems. Meanwhile, the knowledge graph is constructed and fused with the structure tree, and the correlation relationships such as causality and parallelism between the typical problems of the power distribution network planning and operation are intuitively presented, so that when searching and analyzing the problems, the operation and maintenance personnel can not only quickly locate the related problems according to the hierarchy, but also find other problems having the relationships such as causality and similarity according to the knowledge graph, thereby improving the processing efficiency of the power distribution network planning and operation problems. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0054] Figure 1 The basic flowchart of a power distribution network planning and operation typical problem library construction method provided by an embodiment of the application is shown. DETAILED DESCRIPTION

[0055] In order to make the above-mentioned purposes, features and advantages of the application more apparent and easy to understand, the specific embodiments of the application will be described in detail below with reference to the drawings of the specification. Obviously, the described embodiments are only a part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor should be within the protection scope of the application.

[0056] Embodiment 1, refer to Figure 1 For an embodiment of the application, a power distribution network planning and operation typical problem library construction method is provided, which comprises:

[0057] S100: acquiring power distribution equipment data;

[0058] S200: classifying the power distribution network planning and operation typical problems based on the acquired data, to obtain a plurality of basic problem categories;

[0059] S300: using a multi-modal large model to traverse each first sub-problem under each basic problem category and a second sub-problem causing the first sub-problem in the historical experience library, and gradually expanding the hierarchy based on the first sub-problem and the second sub-problem under each basic problem category, until all typical problems are covered, to obtain a tree branch of each basic problem category;

[0060] S400: Based on the tree branch of each basic problem category, the typical problem structure tree is obtained through index logical association analysis and node fusion;

[0061] S500: Based on the typical problem structure tree, the typical problem knowledge graph is constructed by taking the problems represented by each node as entities and the association relationship between problems as entity edges;

[0062] S600: Based on the nodes in the typical problem structure tree and the entities in the typical problem knowledge graph, the typical problem library is constructed through bidirectional relationship mapping.

[0063] It should be noted that the existing power distribution network planning and operation typical problem library construction method faces a series of significant challenges during operation, mainly including that the traditional method relies on a relational database for simple classification storage, which is difficult to effectively display the complex causal association, subordinate relationship and hierarchical structure between problems. This leads to the fact that the operation and maintenance personnel have difficulty in quickly sorting out the overall picture and internal relationship of the problem when facing actual problems, which reduces the efficiency of problem solving; the single mode processing method based on text data cannot fully utilize multi-source heterogeneous data (such as images, audio, etc.), which limits the ability to fully understand the problem; the existing method lacks dynamic adaptability and is difficult to flexibly cope with new problems brought by the changing power distribution network environment and technological progress; the traditional static storage and simple retrieval mechanism cannot support deep logical deduction and decision support, further restricting the efficiency and accuracy of operation and maintenance work. These problems jointly act on the existing typical problem library construction method, which faces many limitations in actual application.

[0064] Therefore, in order to solve the problem of low processing efficiency of the existing planning and operation typical problem library construction method, through the steps of S100-S600, the present application introduces a multi-modal large model, combines multiple types of data (such as text, image, audio) to classify and associate analyze the power distribution network planning and operation typical problems, and significantly improves the accuracy and comprehensiveness of the problem library construction. Specifically, the present application first classifies the power distribution equipment related problems into multiple basic problem categories, and gradually expands the hierarchy through a multi-modal large model to form detailed tree branches; based on semantic similarity, topic word similarity and other indicators, logical association analysis and node fusion are performed to construct a typical problem structure tree; the problems in the structure tree are used as entities, and the association relationship is used as entity edges to construct a typical problem knowledge graph; the structure tree and the knowledge graph are deeply integrated through bidirectional relationship mapping to form a typical problem library with rich content and clear relationship. This systematic method not only clearly shows the hierarchical structure and complex association of the problem, but also provides strong decision support capability, enabling operation and maintenance personnel to quickly locate related problems and develop effective solutions, thereby significantly improving the efficiency and accuracy of power distribution network planning and operation problem processing.

[0065] Embodiment 2, which is an embodiment of the present application, provides a power distribution network planning and operation typical problem library construction method based on the previous embodiment, comprising:

[0066] In the embodiment of the present application, the database construction system comprehensively analyzes the power distribution equipment in step S100, wherein the power distribution equipment includes but is not limited to substations, power distribution lines, power distribution transformers, switchgear, user terminals, etc.

[0067] In the embodiment of the present application, the database construction system classifies the power distribution network planning and operation typical problems according to the functions, characteristics and relevance of common problems of each power distribution equipment in the power distribution network planning and operation in step S200. In the classification process, the system will comprehensively consider factors such as the frequency of problem occurrence, the scope of influence, the difficulty of solution, etc., and finally form multiple basic problem categories.

[0068] In the embodiment of the present application, in the power distribution network in a certain area in step S200, the database construction system analyzes and finds that the problems related to substations are mainly concentrated in capacity shortage, power supply reliability decline caused by equipment aging, etc.; the power distribution line problems are mostly excessive line loss and frequent faults; the power distribution transformer problems are commonly seen in overload operation and substandard energy efficiency, etc. Based on these analyses, the system classifies the power distribution network planning and operation typical problems into three basic problem categories: “substation related problems”, “power distribution line related problems” and “power distribution transformer related problems”.

[0069] In the embodiment of the present application, the database construction system calls the multi-modal large model to conduct deep mining in the historical experience library for each basic problem category in step S300, wherein the historical experience library is a multi-modal data experience library of power grid guidelines, guidelines and standards, etc.

[0070] In the embodiment of the present application, the database construction system retrieves the first sub-problems under each basic problem category, i.e. the specific problems directly subordinate to the basic problem category, and then continues to find the second sub-problems that cause each first sub-problem.

[0071] In the embodiment of the present application, the database construction system takes each basic problem category as a root node, the first sub-problems and the second sub-problems as hierarchical nodes, and the first sub-problems and the second sub-problems as the basis to gradually expand the hierarchy downward according to the logical relationships between problems such as causality and subordination, constantly mining finer-grained problems, until all possible typical problems under the basic problem category are covered, obtaining a tree branch structure, i.e. the tree branch of each basic problem category.

[0072] Taking the basic problem category of "substation related problems" as an example, the database construction system retrieves in the historical experience library through the multi-modal large model to obtain the first sub-problem "substation capacity shortage". Further search finds that the second sub-problem causing "substation capacity shortage" is "regional power load growth too fast" and "insufficient capacity reservation in early planning". In view of "regional power load growth too fast", the sub-problems "new industrial park settlement leading to power demand surge" and "large-scale new residential area construction" are continuously excavated. Taking "substation related problems" as the root node, the sub-problems under it as the hierarchical node, and continuously expanding the hierarchy, a tree branch about "substation related problems" containing many sub-problems is finally formed. Similarly, the same operation is performed on the basic problem categories of "distribution line related problems" and "distribution transformer related problems" to obtain corresponding tree branches.

[0073] In the embodiment of the application, the database construction system fuses the tree branches of each basic problem category according to the root node and the hierarchical node in step S400 to obtain a typical problem structure tree, which is specifically described in steps S410 to S450.

[0074] In the embodiment of the application, the steps S410 to S450 are described as follows:

[0075] S410: For any two first basic problem categories and second basic problem categories, if the semantic distance between the first semantic vector of the first basic problem category and the second semantic vector of the second basic problem category is less than the first preset distance threshold, the first topic word of the first basic problem category and the second topic word of the second basic problem category are extracted.

[0076] In the embodiment of the application, for any two first basic problem categories and second basic problem categories, the database construction system converts the first basic problem category and the second basic problem category into corresponding first semantic vectors and second semantic vectors respectively by using the word vector model (such as BERT) in the natural language processing technology.

[0077] In the embodiment of the application, the database construction system calculates the semantic distance between the first semantic vector and the second semantic vector by the cosine similarity formula, and compares the semantic distance with the first preset distance threshold, wherein the first preset distance threshold is set according to actual conditions.

[0078] If it is determined that the semantic distance is greater than or equal to the first preset distance threshold, it means that there is a large difference in semantics between the first basic problem category and the second basic problem category, so the root node connection of the first basic problem category and the second basic problem category will not be established subsequently.

[0079] If it is determined that the semantic distance is less than the first preset distance threshold, it is indicated that the first basic question category and the second basic question category are relatively similar in semantics, and thus the first basic question category and the second basic question category are connected as the root node to be established. At this time, the database construction system uses a topic model (such as LDA) to analyze the two basic question categories, and extracts a first topic word of the first basic question category and a second topic word of the second basic question category.

[0080] In the embodiment of the application, the first basic question category is “substation related questions”, and the second basic question category is “distribution transformer related questions”. The database construction system converts the “substation related questions” into a first semantic vector, and converts the “distribution transformer related questions” into a second semantic vector. The calculated semantic distance is 0.2, and the first preset distance threshold is 0.3, which meets the condition. Then, the LDA model is used to analyze the two categories, and the first topic word “equipment aging” is extracted from the “substation related questions”, and the second topic word “component wear” is extracted from the “distribution transformer related questions”.

[0081] In the embodiment of the application, the first basic question category is “substation related questions”, and the second basic question category is “distribution transformer related questions”. The database construction system converts the “substation related questions” into a first semantic vector, and converts the “distribution transformer related questions” into a second semantic vector. The calculated semantic distance is 0.2, and the first preset distance threshold is 0.3, which meets the condition. Then, the LDA model is used to analyze the two categories, and the first topic word “equipment aging” is extracted from the “substation related questions”, and the second topic word “component wear” is extracted from the “distribution transformer related questions”.

[0082] In the embodiment of the application, the database construction system calculates the similarity coefficient between the first topic word and the second topic word by using the edit distance algorithm. Therefore, the database construction system calculates the number of edit operations (insertion, deletion, and replacement) required to convert a string of the first topic word into a corresponding string in the second topic word, wherein the number of edit operations represents the difference between the two topic words at the character level. The smaller the number of edit operations, the more similar the structures of the two topic words.

[0083] In the embodiment of the application, the database construction system calculates the length of the string of the first topic word, i.e., the number of characters contained in the first topic word, and calculates the length of the string of the second topic word.

[0084] In the embodiment of the application, the database construction system determines the length of the string of the first topic word and the length of the string of the second topic word, and calculates the quotient of the number of edit operations between the first topic word and the second topic word and the length of the string with the maximum length. The quotient obtained is the similarity coefficient between the first topic word and the second topic word.

[0085] In the embodiment of the present application, if the similarity coefficient is less than or equal to the preset coefficient threshold, it indicates that the first basic problem category and the second basic problem category have a large difference in the subject words, and therefore the root node connection between the first basic problem category and the second basic problem category will not be established subsequently, wherein the preset coefficient threshold is set according to actual conditions.

[0086] In the embodiment of the present application, if the similarity coefficient between the subject words is greater than the preset coefficient threshold, the database construction system counts the number of common occurrence cases of the two basic problem categories from the historical experience library , the number of individual occurrence cases of the first basic problem category , and the number of individual occurrence cases of the second basic problem category , and calculates the co-occurrence frequency using the following formula:

[0087] ,

[0088] wherein, represents the co-occurrence frequency, is the number of common occurrence cases of the two basic problem categories, is the number of individual occurrence cases of the first basic problem category, is the number of individual occurrence cases of the second basic problem category.

[0089] Taking the above-mentioned two subject words "equipment aging" and "component wear" as an example, the similarity coefficient calculated by the edit distance algorithm is 0.8, and the preset coefficient threshold is 0.7, which meets the condition. From the historical experience library, it is known that the number of common occurrence cases of "substation related problems" and "distribution transformer related problems" is , , . Substituting into the formula gives:

[0090] ,

[0091] Therefore, the co-occurrence frequency is 0.1.

[0092] S430: If the co-occurrence frequency index is greater than the preset frequency threshold, a first logical derivation path from the first basic problem category to the second basic problem category is constructed, and a logical derivation correlation coefficient is determined based on the path complexity and the path existence probability of the first logical derivation path.

[0093] In the embodiment of the present application, if the co-occurrence frequency index is less than or equal to the preset frequency threshold, it indicates that the first basic problem category and the second basic problem category have a large difference in the common occurrence cases, and therefore the root node connection between the first basic problem category and the second basic problem category will not be established subsequently.

[0094] In the embodiment of the present application, if the co-occurrence frequency indicator is greater than the preset frequency threshold, the database construction system constructs a first logical derivation path from the first basic problem category to the second basic problem category by analyzing the problem solving process and the causal relationship in the historical cases, wherein the path complexity is comprehensively evaluated by the number of nodes and the number of relationship types contained in the path, and the path existence probability is calculated based on the ratio of the number of times the path appears in the historical cases to the total number of cases. The logical derivation correlation coefficient is calculated using the following formula:

[0095] ,

[0096] wherein, is the logical derivation correlation coefficient, is the path existence probability, is the path complexity.

[0097] Continuing the above embodiment, the co-occurrence frequency indicator is 0.1, which is greater than the preset frequency threshold (0.08), so a logical derivation path from the "substation related problems" to the "distribution transformer related problems" is constructed. It is found that the path contains 5 nodes and 3 types of relationships, and the path complexity is evaluated as 7; in the 1000 historical related cases, this path appears 50 times, and the path existence probability is 0.05, so the logical derivation correlation coefficient is 0.001.

[0098] S440: If the logical derivation correlation coefficient is greater than the preset correlation threshold, a connection between the first root node of the first basic problem category and the second root node of the second basic problem category is established.

[0099] In the embodiment of the present application, if the logical derivation correlation coefficient is less than or equal to the preset correlation threshold, it indicates that there is a greater contradiction in the logical derivation between the first basic problem category and the second basic problem category, so the root node connection between the first basic problem category and the second basic problem category will not be established subsequently.

[0100] In the embodiment of the present application, if the logical derivation correlation coefficient is greater than the preset correlation threshold, the database construction system establishes a connection between the first root node of the first basic problem category and the second root node of the second basic problem category, wherein the root node connection represents a strong logical correlation between the two basic problem categories.

[0101] Continuing the above embodiment, the preset correlation threshold is 0.0008, and the calculated logical derivation correlation coefficient 0.001 is greater than the threshold, so a connection is established between the root node of the "substation related problems" and the root node of the "distribution transformer related problems", indicating that there is a close logical relationship between the two basic problem categories.

[0102] S450: Fuse nodes based on the association between the child nodes of each level in the tree branch of the root node connected basic problem category, to obtain a typical problem structure tree.

[0103] In the embodiment of the present application, for the root node connected basic problem category, the database construction system obtains the association between the child nodes of each level in the tree branch, and fuses the nodes based on the association between the child nodes of each level to obtain a typical problem structure tree, as described in steps S451 to S454.

[0104] In the embodiment of the present application, steps S451 to S454 are described as follows:

[0105] S451: For the third and fourth basic problem categories that have established root node connections, based on the level depth of the tree branches of the third and fourth basic problem categories, adjust the two tree branches to the same level structure to obtain a first adjusted tree branch and a second adjusted tree branch.

[0106] In the embodiment of the present application, for the third and fourth basic problem categories that have established root node connections, the database construction system obtains the level depth of each of the two tree branches. If the level depths of the two tree branches are different, the embodiment of the present application uses the method of node insertion or deletion to make them have the same level structure. In specific operation, for the tree branch with a shallower level, insert a virtual node (the virtual node is only used for level alignment and does not contain actual problem information) at a suitable position; for the tree branch with a deeper level, delete part of the redundant nodes without affecting the logical relationship of the problems. After adjustment, a first adjusted tree branch corresponding to the third basic problem category and a second adjusted tree branch corresponding to the fourth basic problem category are obtained.

[0107] In the embodiment of the present application, the third basic problem category is "power distribution line related problems", and the level depth of its tree branch is 4 layers; the fourth basic problem category is "power distribution transformer related problems", and the level depth of its tree branch is 3 layers. The database construction system adjusts the tree branch of "power distribution transformer related problems", inserts a virtual node under the third layer node to make the level depth also 4 layers, and obtains a first adjusted tree branch and a second adjusted tree branch with the same level structure as the tree branch of "power distribution line related problems".

[0108] S452: For any two first and second child nodes of the same level in the first and second adjusted tree branches, determine the Manhattan distance based on the first fusion feature vector of the first child node and the second fusion feature vector of the second child node. The fusion feature vector includes a text feature vector, an image feature vector, and an audio feature vector.

[0109] In the embodiment of the present application, for any two first and second child nodes at the same level in the first and second adjusted tree branches, the database construction system first extracts their fusion feature vectors, which are composed of text feature vectors, image feature vectors and audio feature vectors. The text feature vectors are obtained by encoding the node problem description text through a pre-trained language model (such as RoBERTa); the image feature vectors are obtained by feature extraction on the node-related schematic diagram through a convolutional neural network (such as a residual network ResNet); and the audio feature vectors are obtained by analyzing the node-related voice explanation through an audio processing model (such as WaveNet). Further, the database construction system calculates the distance between the fusion feature vectors of the two child nodes according to the Manhattan distance formula, which is:

[0110] ,

[0111] wherein, is the distance between the fusion feature vectors of the two child nodes, is the dimension of the fusion feature vector (the sum of the dimensions of the text feature vector, the image feature vector and the audio feature vector), is the first element of the fusion feature vector of the first child node, is the first element of the fusion feature vector of the second child node. Continuing the above embodiment, in the adjusted tree branch at the same level, the first child node is "insulation aging of power distribution line" and the second child node is "insulation layer damage of power distribution transformer". The fusion feature vectors of the two child nodes are obtained, for example, the dimension of the fusion feature vector is 100 (the dimension of the text feature vector is 30, the dimension of the image feature vector is 40, and the dimension of the audio feature vector is 30). After calculation, = 0.2,

[0112] = 0.3,..., = 0.4, = 0.5, the Manhattan distance is 15 calculated by substituting into the Manhattan distance formula. S453: If the Manhattan distance is less than the second preset distance threshold, determine the context feature matching degree based on the pre-node features and post-node features of the first and second child nodes.

[0113] In the embodiment of the present application, if it is determined that the Manhattan distance is greater than or equal to the second preset distance threshold, it means that there is a large difference between the first and second child nodes in the text feature, image feature or / and audio feature, so the child node connection between the first and second child nodes will not be established subsequently.

[0114] In the embodiment of the present application, if it is determined that the Manhattan distance is greater than or equal to the second preset distance threshold, it means that there is a large difference between the first and second child nodes in the text feature, image feature or / and audio feature, so the child node connection between the first and second child nodes will not be established subsequently.

[0115] ​In the embodiment of the present application, if the Manhattan distance of the two child nodes is less than the second preset distance threshold, the database construction system analyzes the preceding child node features and the following child node features of the first child node and the second child node, and determines the context feature matching degree according to the preceding child node features and the following child node features of the first child node and the second child node, wherein the preceding child node features and the following child node features are also represented in the form of feature vectors and are obtained in the same way as the child node features are extracted. The embodiment of the present application calculates the context feature matching degree by using the cosine similarity formula, and the formula is as follows:

[0116]

[0117] is the context feature matching degree, is the preceding child node feature vector and the following child node feature vector of the first child node, is the preceding child node feature vector and the following child node feature vector of the second child node.

[0118] Continuing with the above embodiment, the second preset distance threshold is 20, and the Manhattan distance of the "power distribution line insulation aging" and "power distribution transformer insulation layer damage" child nodes is 15, which is less than the threshold. The database construction system obtains the preceding child node "long-term high-load operation of line" feature vector of the "power distribution line insulation aging", the following child node "increased risk of line short circuit" feature vector , the preceding child node "transformer overheating" feature vector of the "power distribution transformer insulation layer damage", and the following child node "increased probability of transformer failure" feature vector , and substitutes them into the formula to obtain the context feature matching degree of 0.8.

[0119] S454: If the context feature matching degree is greater than the preset matching degree, node fusion is performed based on the associated paths of the first child node and the second child node to obtain the typical problem structure tree.

[0120] In the embodiment of the present application, if it is determined that the context feature matching degree is less than or equal to the preset matching degree, it indicates that there is a large difference in the context features between the first child node and the second child node, and therefore the child node connection of the first child node and the second child node will not be established subsequently.

[0121] In the embodiment of the present application, if it is determined that the context feature matching degree is greater than the preset matching degree, the database construction system performs node fusion based on the associated paths of the first child node and the second child node to obtain the typical problem structure tree, and the details are as described in steps S4541 to S4544.

[0122] ​​​​The embodiment of the present application accurately identifies the child nodes with similarity and correlation, integrates under the premise of retaining the logical relationship of the problem, so that the finally constructed typical problem structure tree can comprehensively present various problems and their internal relations in power distribution network planning and operation, clearly presents the classification and hierarchy of the problems, so that the operation and maintenance personnel can quickly grasp the overall architecture of the problems, and the processing efficiency of the power distribution network planning and operation problems is improved.

[0123] In the embodiment of the present application, the steps S4541 to S4544 are described as follows:

[0124] S4541: search the child level based on the current level of the first child node and the second child node, add the child node with a Manhattan distance less than a second preset distance threshold and a context feature matching degree greater than a preset matching degree to a path set, and respectively obtain a first associated path and a second associated path.

[0125] In the embodiment of the present application, the database construction system takes the current level where the first child node and the second child node are located as a starting point, and uses a depth-first search algorithm to search and traverse the child level downward. In the search process, for each encountered child node, the Manhattan distance between the child node and the corresponding node (the first child node corresponds to the subsequent child level node of the first child node, and the second child node corresponds to the subsequent child level node of the second child node) is calculated, and it is determined whether the distance is less than a second preset distance threshold. At the same time, the context feature matching degree between the child node and the corresponding node is calculated, and it is checked whether the matching degree is greater than a preset matching degree. If both conditions are met, the child node is added to the corresponding path set. After the search, the first associated path and the second associated path containing the child nodes meeting the conditions are obtained.

[0126] Continuing the above embodiment, the first child node is “distribution line insulation aging”, the second child node is “distribution transformer insulation layer damage”, the second preset distance threshold is 20, and the preset matching degree is 0.7. The database construction system starts searching downward from the level where the two child nodes are located. In the child level of the “distribution line insulation aging” child node, the “line insulation resistance decrease” child node is found, the Manhattan distance between the “line insulation resistance decrease” child node and the “distribution transformer insulation layer damage” child node in the child level of the “distribution transformer insulation layer damage” child node is calculated as 18 (less than 20), and the context feature matching degree between the two is 0.8 (greater than 0.7), so the “line insulation resistance decrease” and the “transformer insulation resistance decrease” are added to the first associated path and the second associated path respectively. Continue to search, and add other child nodes meeting the conditions to the corresponding path set, and complete the construction of the first associated path and the second associated path.

[0127] S4542: determine the number of levels of the common child level in the first associated path and the second associated path, and determine the feature similarity based on the first fusion feature vector and the second fusion feature vector.

[0128] In the embodiment of the present application, the database construction system determines the number of sub-levels of the common sub-levels in the first associated path and the second associated path. Continuing with the example of the two sub-nodes of "insulation aging of distribution line" and "insulation layer damage of distribution transformer" described above, the database construction system finds through analysis that the sub-levels passed by the first associated path are 2, 3, 4, 6, 7 and 9, the sub-levels passed by the second associated path are 2, 4, 6 and 10, and the common sub-levels are 2, 4 and 6, so the number of levels of the first associated path and the second associated path is 3 levels.

[0129] In the embodiment of the present application, the database construction system calculates the feature similarity based on the first fusion feature vector of the first sub-node and the second fusion feature vector of the second sub-node, and the embodiment of the present application uses the cosine similarity formula to calculate the feature similarity.

[0130] S4543: Based on the context feature matching degree, the number of levels and the feature similarity, the node contact strength between the first sub-node and the second sub-node is determined.

[0131] In the embodiment of the present application, the database construction system determines the node contact strength between the first sub-node and the second sub-node according to the context feature matching degree, the number of levels and the feature similarity, and the specific formula is:

[0132] ,

[0133] Wherein, is the node contact strength, is the number of levels of the common sub-levels in the first associated path and the second associated path, is the feature similarity calculated based on the first fusion feature vector and the second fusion feature vector.

[0134] In the embodiment of the present application, the feature similarity is calculated using the cosine similarity, that is,

[0135] ,

[0136] Wherein, , are the fusion feature vectors of the first sub-node and the second sub-node respectively.

[0137] S4544: Node fusion is performed based on the node contact strength to obtain a typical problem structure tree.

[0138] In the embodiment of the present application, the database construction system performs node fusion according to the node contact strength to obtain a typical problem structure tree, which is specifically described in steps S45441 to S45444.

[0139] The embodiment of the application fully considers the feature similarity, context association and hierarchical structure relationship of nodes, avoids information redundancy or loss caused by simple merging, and makes the finally constructed typical problem structure tree more accurately reflect the internal logical relationship of various problems in power distribution network planning and operation, clearly present the classification and hierarchy of the problems, and enable the operation and maintenance personnel to quickly grasp the overall architecture of the problems and improve the processing efficiency of power distribution network planning and operation problems.

[0140] In the embodiment of the application, the steps S45441 to S45444 are described as follows:

[0141] S45441: If the node contact strength is greater than the preset strength threshold, a second logical derivation path is constructed based on the problem description of the first child node and the second child node.

[0142] In the embodiment of the application, the database construction system judges whether the node contact strength is greater than the preset strength threshold, wherein the preset strength threshold is set according to actual conditions.

[0143] If the node contact strength is less than or equal to the preset strength threshold, it indicates that the first child node and the second child node are weak in node contact, and therefore the child node connection of the first child node and the second child node is not established.

[0144] In the embodiment of the application, if the node contact strength is greater than the preset strength threshold, the database construction system performs natural language processing on the problem description of the first child node and the second child node, and mines the potential logical relationship between the two problems through semantic analysis, causal relationship extraction and other technologies. Further, the database construction system constructs a second logical derivation path from the first child node to the second child node according to the logical relationship, which clearly shows the process and basis of how to derive from one problem to another problem.

[0145] Continuing the above embodiment, the first child node is "distribution line insulation aging", the second child node is "distribution transformer insulation layer damage", and the calculated node contact strength is 0.71, and the preset strength threshold is 0.7. Since the node contact strength is greater than the preset strength threshold, the database construction system analyzes the problem description of the two child nodes. Through semantic analysis, it is found that the insulation problems of the distribution line and the distribution transformer are related to long-term operation, environmental factors and other factors, and there is a potential causal relationship. The system constructs a second logical derivation path accordingly: "long-term operation and environmental factors cause the distribution line insulation to age, and similar operating conditions and environmental influences may also cause the distribution transformer insulation layer to be damaged".

[0146] S45442: Logical contradiction detection is performed on the premise condition and the conclusion in the derivation process based on the second logical derivation path, and a logical contradiction detection result is obtained.

[0147] In the embodiment of the present application, for the constructed second logical deduction path, the database construction system conducts a detailed review on the premise conditions and the derived conclusions in the deduction process, and the embodiment of the present application uses the logical inference rules and the logical relations in the knowledge base to check whether the premise conditions can reasonably deduce the conclusions and whether the conclusions are in conflict with the premise conditions. For example, whether there are conflicting factors in the premise conditions, whether the conclusions are contrary to the known logical rules, and the like, to obtain a logical conflict detection result.

[0148] Continuing the second logical deduction path from "insulation aging of distribution line" to "insulation layer damage of distribution transformer" constructed above, the database construction system reviews the premise condition "long-term operation and environmental factors", and finds that the premise condition is reasonable and has no internal conflict. Further checking the conclusion "similar operation conditions and environmental impact may also cause insulation layer damage of distribution transformer", by comparison with the logical rules about the relationship between equipment aging and environment and operation conditions in the knowledge base, it is found that the conclusion is consistent with the premise condition and there is no logical conflict, so the logical conflict detection result is no conflict.

[0149] S45443: If the logical conflict detection result is no conflict, based on whether the knowledge domains involved in the second logical deduction path conflict with each other, a knowledge domain conflict detection result is obtained.

[0150] In the embodiment of the present application, if the logical conflict detection result is conflict, it indicates that the first sub-node and the second sub-node are in conflict in logical conflict, so the sub-node connection of the first sub-node and the second sub-node is not established.

[0151] In the embodiment of the present application, if the logical conflict detection result is no conflict, the database construction system analyzes the knowledge domains involved in the second logical deduction path. Optionally, the embodiment of the present application matches the problems, factors and the like involved in the path with the preset knowledge domain classification, to determine whether there is conflict between different knowledge domains. For example, whether mutually contradictory professional theories, technical specifications and the like appear in the same deduction path, so as to obtain a knowledge domain conflict detection result. Continuing the second logical deduction path from "insulation aging of distribution line" to "insulation layer damage of distribution transformer", the knowledge domains involved are mainly the aging mechanism of power equipment, the influence of power system operation environment and the like. The database construction system compares these knowledge domains with the preset knowledge domain classification, and finds that they all belong to the knowledge category of power equipment maintenance and operation, and there is no mutually conflicting professional theory or technical specification. Therefore, the knowledge domain conflict detection result is no conflict.

[0152] S45444: If the knowledge domain conflict detection result is no conflict, the connection between the first sub-node and the second sub-node is established, and a typical problem structure tree is obtained.

[0153] In the embodiment of the present application, if the knowledge domain conflict detection result is that there is a conflict, it indicates that there is a conflict between the first sub-node and the second sub-node in the knowledge domain, and the connection of the first sub-node and the second sub-node is not established.

[0154] In the embodiment of the present application, if the knowledge domain conflict detection result is that there is no conflict, the database construction system establishes the connection between the first sub-node and the second sub-node, and obtains the typical problem structure tree.

[0155] Continuing the above embodiment, since the "power distribution line insulation aging" and "power distribution transformer insulation layer damage" sub-nodes meet the conditions through the above detection, the database construction system establishes the connection between the two sub-nodes, and integrates the problem description, the second logical deduction path and other information of the two sub-nodes. The same operation is performed on other sub-nodes that meet the conditions, and finally a complete typical problem structure tree is constructed, for example, in the tree structure, the "power distribution line insulation aging" and the "power distribution transformer insulation layer damage" are converged to the "power distribution network equipment insulation problem" node at a higher level through the connection.

[0156] The embodiment of the present application ensures the rationality and accuracy of the node connection by constructing the logical deduction path, detecting the logical contradiction and the knowledge domain conflict, so that the finally constructed typical problem structure tree not only reflects the association relationship between the problems, but also guarantees the logical rigor and the consistency of the knowledge, provides a reliable structured knowledge framework for the analysis, diagnosis and solution of the power distribution network planning and operation problems, and enables the operation and maintenance personnel to quickly grasp the overall architecture of the problems, thereby improving the processing efficiency of the power distribution network planning and operation problems.

[0157] The embodiment of the present application can accurately identify the potential association between problem categories, and construct a typical problem structure tree with rigorous logical structure, which not only clearly presents the hierarchical relationship of each problem category and its sub-problems, but also excavates the deep logical relationship between the problems, so that the operation and maintenance personnel can quickly grasp the overall architecture of the problems, thereby improving the processing efficiency of the power distribution network planning and operation problems.

[0158] In the embodiment of the present application, in step S500, the database construction system takes the problem represented by each node in the typical problem structure tree as an entity in the knowledge graph, for example, "substation capacity shortage", "regional power load growth too fast" and the like are taken as independent entities, and according to the hierarchical relationship, the causal relationship and other associations between the problems in the typical problem structure tree, the edges between the entities are determined, and the edges represent the association relationship between the problems.

[0159] In the embodiment of the present application, the database construction system adopts the knowledge graph construction technology to structurally store and visually display these entities and entity edges, and constructs a typical problem knowledge graph to clearly present the complex relationship network between the typical problems of power distribution network planning and operation.

[0160] Continuing the above in the typical problem structure tree, the "substation capacity shortage" node and the "regional electricity load growth too fast" node have a causal relationship, the database construction system takes "substation capacity shortage" and "regional electricity load growth too fast" as entities, creates two corresponding nodes in the knowledge graph, and then adds an edge from the "regional electricity load growth too fast" node to the "substation capacity shortage" node, and marks the relationship as "causes". In this way, all nodes and associated relationships in the typical problem structure tree are converted into entities and entity edges in the knowledge graph, and a complete typical problem knowledge graph is finally constructed, and through the visual interface, the association between problems can be intuitively seen.

[0161] In the embodiment of the present application, the database construction system in step S600 maps the nodes in the typical problem structure tree and the entities in the typical problem knowledge graph in both directions to embed the association relationship between the problems in the typical problem knowledge graph into the typical problem structure tree, and at the same time, the hierarchical architecture of the typical problem structure tree is reflected in the typical problem knowledge graph, and the typical problem library is constructed, as described in steps S610 to S630.

[0162] The embodiment of the present application organizes the problems of power distribution network planning and operation in a hierarchical manner by constructing a structure tree, clearly presents the classification and hierarchy of the problems from the root node to the child nodes, and enables the operation and maintenance personnel to quickly grasp the overall architecture of the problems. At the same time, the knowledge graph is constructed and integrated with the structure tree, and the causal, parallel and other association relationships between the typical problems of power distribution network planning and operation are intuitively presented, so that when searching and analyzing the problems, the operation and maintenance personnel can not only quickly locate the related problems according to the hierarchy, but also find other problems that have causal, similar and other relationships according to the knowledge graph, thereby improving the processing efficiency of the problems of power distribution network planning and operation.

[0163] In the embodiment of the present application, steps S610 to S630 are described as follows:

[0164] S610: For the node pair having a parent-child hierarchical relationship in the typical problem structure tree, match the first entity and the second entity in the typical problem knowledge graph. Create a new first-level association edge for the first entity and the second entity in the typical problem knowledge graph, and obtain a first updated knowledge graph to integrate the structure tree hierarchical relationship in the structure tree into the knowledge graph.

[0165] In the embodiment of the present application, the database construction system traverses the typical problem structure tree, and identifies the node pairs having parent-child hierarchical relationship in the typical problem structure tree. Further, for each pair of nodes having parent-child hierarchical relationship, the database construction system finds the first entity and the second entity corresponding to the pair of nodes in the typical problem knowledge graph, and creates a new first-level association edge for the first entity and the second entity in the typical problem knowledge graph after finding the corresponding entities, wherein the first-level association edge is specially used to represent the hierarchical relationship in the structure tree, and the type of the association edge has a clear identification. Therefore, by means of the above embodiment, the hierarchical relationship information in the typical problem structure tree is integrated into the typical problem knowledge graph, and the first updated knowledge graph is obtained.

[0166] Continuing the above embodiment, in the typical problem structure tree, the "substation related problem" is the parent node, and the "substation capacity shortage" is the child node, which is a pair of nodes having parent-child hierarchical relationship. The database construction system finds the entity A corresponding to the "substation related problem" and the entity B corresponding to the "substation capacity shortage" in the typical problem knowledge graph. Then, a new first-level association edge is created for the entity A and the entity B in the knowledge graph, and the type of the edge is marked as "parent-child hierarchical relationship", and the first updated knowledge graph is obtained, so that the hierarchical relationship can also be reflected in the knowledge graph.

[0167] S620: Traverse the second-level association edges of other types except the type corresponding to the first-level association edge in the first updated knowledge graph. Determine the first node and the second node corresponding to the two ends of the second-level association edge in the typical problem structure tree, and create a new virtual node for the first node and the second node in the typical problem structure tree. An association relationship between the first node and the virtual node is established, and an association relationship between the second node and the virtual node is established, and the first updated structure tree is obtained, so as to embed the association relationship between the problems in the knowledge graph into the structure tree. The type of the association relationship is the same as that in the knowledge graph.

[0168] In the embodiment of the present application, the database construction system traverses the first updated knowledge graph, and screens out the second-level association edges of other types except the type corresponding to the first-level association edge.

[0169] In the embodiment of the present application, for each second-level association edge, the database construction system determines the first node and the second node corresponding to the two end entities of the second-level association edge in the typical problem structure tree. Then, a new virtual node is created for the first node and the second node in the typical problem structure tree, where the virtual node acts as an “intermediary” or “carrier” to simulate and present the association relationship between the first node and the second node in the knowledge graph in the typical problem structure tree. At the same time, the database construction system establishes the association relationship between the first node and the virtual node and the association relationship between the second node and the virtual node, respectively, and ensures that the types of these association relationships are completely consistent with the types in the knowledge graph, to obtain the first updated structure tree, thereby embedding the association relationship between problems in the knowledge graph into the typical problem structure tree.

[0170] Continuing in the first updated knowledge graph, there is a second-level association edge connecting the entity “regional electricity load growth too fast” and the entity “substation capacity insufficient”, and the association edge type is “cause”. The database construction system determines that “regional electricity load growth too fast” corresponds to node C in the typical problem structure tree, and “substation capacity insufficient” corresponds to node D. A virtual node E is created for node C and node D in the typical problem structure tree. The “cause” relationship between node C and virtual node E and the “caused” relationship between node D and virtual node E are established, and finally the first updated structure tree is obtained, at this time the structure tree also has the same association relationship as the knowledge graph.

[0171] S630: Based on the nodes in the first updated structure tree and the entities in the first updated knowledge graph, bidirectional relationship mapping is performed to construct a typical problem library.

[0172] In the embodiment of the present application, the database construction system performs bidirectional relationship mapping between the nodes in the first updated structure tree and the entities in the first updated knowledge graph to construct a typical problem library, which is specifically described in steps S631 to S634.

[0173] The embodiment of the present application integrates the knowledge graph and the structure tree, and directly displays the causal and parallel association relationships between the typical problems of power distribution network planning and operation, so that when searching and analyzing problems, the operation and maintenance personnel can not only quickly locate related problems according to the hierarchy, but also find other problems with causal and similar relationships according to the knowledge graph, thereby improving the processing efficiency of power distribution network planning and operation problems.

[0174] In the embodiment of the present application, steps S631 to S634 are described as follows:

[0175] S631: In the first updated structure tree, if multiple nodes have the same type of association relationship with the same virtual node, the multiple nodes are directly associated, and the intermediate virtual node is deleted, to obtain a second updated structure tree.

[0176] In the embodiment of the application, the database construction system traverses the first updated structure tree to find multiple nodes having the same type of association relationship with the same virtual node. If such a combination of nodes is found, the nodes are directly associated, and the association type remains unchanged, and the intermediate virtual node is deleted. This operation eliminates the redundant connection in the structure tree, simplifies the tree structure, and obtains a second updated structure tree.

[0177] Continuing in the first updated structure tree, the virtual association node E has a “cause” relationship with node C (the regional power load grows too fast) and a “caused” relationship with node D (the substation capacity is insufficient). The virtual association node E also has a “cause” relationship with node F (the line is overloaded). It is detected that node C and node F are connected to the same virtual node E through the “cause” relationship, and a “cause” association relationship is directly established between node C and node F, and the virtual association node E is deleted, to obtain a second updated structure tree.

[0178] S632: In the first updated knowledge graph, if there are at least two association edges of the same type between the same two entities, one association edge is retained, to obtain a second updated knowledge graph.

[0179] In the embodiment of the application, the database construction system checks the first updated knowledge graph to find multiple association edges of the same type between the same two entities. For such redundant edges, the system retains only one association edge and deletes the other redundant edges, to obtain a second updated knowledge graph. This operation eliminates the redundant connection in the knowledge graph and optimizes the structure of the graph.

[0180] Continuing in the first updated knowledge graph, there are two association edges of “parent-child hierarchical relationship” between entity A (the substation related problem) and entity B (the substation capacity is insufficient). After detecting this redundancy, one of the association edges is retained and the other is deleted, to obtain a second updated knowledge graph.

[0181] S633: All nodes in the second updated structure tree and all entities in the second updated knowledge graph are traversed, and the most similar target entity in the second updated knowledge graph is matched for any target node in the second updated structure tree based on the problem text similarity between the nodes and the entities.

[0182] In the embodiment of the present application, the database construction system traverses all nodes of the second updated structure tree and all entities of the second updated knowledge graph, and for any target node in the second updated structure tree, the Jaccard similarity coefficient is combined with the TF-IDF (term frequency-inverse document frequency) algorithm to calculate the question text similarity between the target node and each entity in the second updated knowledge graph. The specific calculation formula is as follows:

[0183] ,

[0184] wherein, is the question text similarity, denotes the question text segmentation set of the target node, denotes the question text segmentation set of the entity in the knowledge graph, denotes the number of intersection elements of the two sets, denotes the number of union elements, is the segmented word, is the word in the target node question text, is the word in the entity question text.

[0185] In the embodiment of the present application, the database construction system determines the entity with the largest question text similarity to the target node in the second updated knowledge graph as the most similar target entity, thereby matching the most similar target entity in the second updated knowledge graph for any target node in the second updated structure tree.

[0186] Continuing the above embodiment, the question corresponding to the target node C in the second updated structure tree is "high power loss of distribution circuit", and the word set obtained after the question text is segmented is called "target node question segmentation set"; the question corresponding to the entity Z in the second updated knowledge graph is "line resistance causes loss problem", and the segmentation result is called "target entity question segmentation set".

[0187]

[0188] The actual segmentation content is as follows:

[0189] The "target node question segmentation set" contains the words: "distribution circuit", "loss", "too high", "current", "resistance";

[0190] The "target entity question segmentation set" contains the words: "line", "resistance", "loss", "electric energy loss".

[0191] Let the "target node question segmentation set" be , and the "target entity question segmentation set" be​ Then we have:

[0192] = 3, = 6,

[0193] The words in the intersection are "loss", "resistance", and "line". For each word , the TF-IDF value in the node and entity problem text is calculated respectively and substituted into the following formula:

[0194] ,

[0195] After calculation, the problem text similarity = 0.75. After traversing all entities in the knowledge graph, it is determined that the similarity between entity Z and target node C is the highest, so entity Z is taken as the matching target entity of target node C.

[0196] S634: Establish a bidirectional relationship mapping between the target node and the target entity to obtain a typical problem library.

[0197] In the embodiment of the application, the database construction system establishes a bidirectional relationship mapping between each target node in the second updated structure tree and the target entity matched in the second updated knowledge graph. When establishing the mapping, not only the attribute information (such as problem description, related parameters, etc.) of the target node and the target entity is recorded, but also the association path information from the target node to the target entity and from the target entity to the target node is recorded in detail. Through this comprehensive bidirectional relationship mapping, the information in the structure tree and the knowledge graph is deeply integrated, and finally a typical problem library with rich content and clear relationship is constructed.

[0198] The database construction system continues to establish a bidirectional relationship mapping for the above matched target node C (overly high loss of distribution line) and target entity Z (line resistance causes loss problem). The attribute information of target node C is recorded as "description: distribution line causes overly high loss due to current passing through; influence: reduces power supply efficiency", and the attribute information of target entity Z is recorded as "description: line resistance causes power loss; solution: optimize line material". At the same time, the association path from node C to entity Z is recorded as "structure tree path: distribution network line problem → overly high loss of distribution line, knowledge graph path: line problem → line resistance problem → loss problem", and the reverse association path from entity Z to node C. Through similar bidirectional relationship mapping operations for all nodes and entities, a complete typical problem library is finally constructed, realizing systematic and structured management of typical problems of distribution network planning and operation.

[0199] The embodiment of the application combines the hierarchical structure of the structure tree and the associated network of the knowledge graph, and the constructed typical problem library can comprehensively present the hierarchical relationship, the associated relationship and detailed problem information between the typical problems of power distribution network planning and operation, so that the operation and maintenance personnel can not only quickly locate the related problems according to the level, but also can find other problems related to the knowledge graph, thereby improving the processing efficiency of the power distribution network planning and operation problems.

[0200] Embodiment 3, which is an embodiment of the application, differs from the first embodiment in that a power distribution network planning and operation typical problem library construction system is provided.

[0201] It should be noted that the technical scheme of the power distribution network planning and operation typical problem library construction system belongs to the same concept as the technical scheme of the power distribution network planning and operation typical problem library construction method described above. The technical scheme of the power distribution network planning and operation typical problem library construction system in this embodiment is not described in detail, and the description of the technical scheme of the power distribution network planning and operation typical problem library construction method described above can be referred to.

[0202] The power distribution network planning and operation typical problem library construction system in this embodiment comprises:

[0203] The acquisition module is configured to acquire power distribution equipment data.

[0204] The problem classification module is configured to classify the power distribution network planning and operation typical problems based on the acquired data to obtain a plurality of basic problem categories.

[0205] The tree branch construction module is configured to use a multi-modal large model to traverse each first sub-problem and a second sub-problem causing the first sub-problem under each basic problem category in the historical experience library, and gradually expand the level based on the first sub-problem and the second sub-problem under each basic problem category until all typical problems are covered, to obtain a tree branch of each basic problem category.

[0206] The structure tree fusion module is configured to obtain a typical problem structure tree based on the tree branch of each basic problem category through index logical association analysis and node fusion.

[0207] The knowledge graph construction module is configured to construct a typical problem knowledge graph by taking the problems represented by each node as entities and the associated relationships between the problems as entity edges based on the typical problem structure tree.

[0208] The problem library construction module is configured to construct a typical problem library by bidirectional relationship mapping based on the nodes in the typical problem structure tree and the entities in the typical problem knowledge graph.

[0209] The embodiment also provides an electronic device suitable for the power distribution network planning and operation typical problem library construction method, which comprises:

[0210] a memory and a processor; the memory is configured to store computer-executable instructions, and the processor is configured to execute the computer-executable instructions to implement the method for constructing a typical problem library of power distribution network planning and operation as described in the above embodiments.

[0211] The embodiment also provides a storage medium having a computer program stored thereon, where the computer program, when executed by a processor, implements the method for constructing a typical problem library of power distribution network planning and operation as described in the above embodiments.

[0212] The storage medium provided by the embodiment belongs to the same inventive concept as the method for constructing a typical problem library of power distribution network planning and operation provided by the above embodiments, and the technical details not described in the embodiment can be referred to the above embodiments, and the embodiment has the same beneficial effects as the above embodiments.

[0213] From the above description about the embodiments, those skilled in the art can clearly understand that the present application can be realized by means of software and necessary universal hardware, and of course can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a floppy disk, a ROM, a RAM, a FLASH, a hard disk, or an optical disc, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of various embodiments of the present application.

[0214] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for constructing a power distribution network planning operation typical problem library, characterized in that, The method comprises the following steps: acquiring power distribution equipment data; based on the acquired data, classifying power distribution network planning and operation typical problems to obtain multiple basic problem categories; using a multi-modal large model to traverse a first sub-problem and a second sub-problem causing the first sub-problem under each basic problem category in a historical experience library, and gradually expanding the level based on the first sub-problem and the second sub-problem under each basic problem category until all typical problems are covered, to obtain a tree branch of each basic problem category; based on the tree branch of each basic problem category, obtaining a typical problem structure tree through index logical association analysis and node fusion; based on the nodes in the typical problem structure tree and the entities in the typical problem knowledge graph, constructing a typical problem library through bidirectional relationship mapping; the bidirectional relationship mapping to construct the typical problem library comprises: for a pair of nodes with a parent-child hierarchical relationship in the typical problem structure tree, matching a first entity and a second entity in the typical problem knowledge graph; creating a new first-level association edge for the first entity and the second entity in the typical problem knowledge graph to obtain a first updated knowledge graph, and integrating the structure tree hierarchical relationship in the structure tree into the knowledge graph; traversing a second-level association edge corresponding to a type other than the first-level association edge in the first updated knowledge graph; determining a first node and a second node corresponding to the first node and the second node in the typical problem structure tree, and creating a new virtual node for the first node and the second node in the typical problem structure tree; establishing an association relationship between the first node and the virtual node, and establishing an association relationship between the second node and the virtual node, to obtain a first updated structure tree, and embedding the association relationship between problems in the knowledge graph into the structure tree; the type of the association relationship is the same as that in the knowledge graph; based on the nodes in the first updated structure tree and the entities in the first updated knowledge graph, constructing a typical problem library through bidirectional relationship mapping; the bidirectional relationship mapping to construct the typical problem library based on the nodes in the first updated structure tree and the entities in the first updated knowledge graph comprises: in the first updated structure tree, if multiple nodes have the same type of association relationship with the same virtual node, the multiple nodes are directly associated, and the intermediate virtual node is deleted to obtain a second updated structure tree; in the first updated knowledge graph, if there are at least two hierarchical association edges of the same type between the same two entities, only one hierarchical association edge is retained to obtain a second updated knowledge graph; traversing all nodes of the second updated structure tree and all entities of the second updated knowledge graph, and matching the most similar target entity in the second updated knowledge graph for any target node in the second updated structure tree based on the problem text similarity between the nodes and the entities; establishing bidirectional relationship mapping between the target node and the target entity to obtain a typical problem library. the typical problem structure tree is obtained through index logical association analysis and node fusion, which comprises:

2. The method of claim 1, wherein the method comprises: determining a plurality of typical problems of the power distribution network; and constructing a library of the typical problems of the power distribution network. ​ For any two first basic question categories and second basic question categories, if a semantic distance between a first semantic vector of the first basic question category and a second semantic vector of the second basic question category is less than a first preset distance threshold, first keywords of the first basic question category and second keywords of the second basic question category are extracted; If a similarity coefficient between the first keywords and the second keywords is greater than a preset coefficient threshold, a co-occurrence frequency indicator is determined based on a number of co-occurrence cases of the two basic question categories and a number of individual occurrence cases of the first basic question category and the second basic question category; If the co-occurrence frequency indicator is greater than a preset frequency threshold, a first logical derivation path from the first basic question category to the second basic question category is constructed, and a logical derivation correlation coefficient is determined based on a path complexity and a path existence probability of the first logical derivation path; If the logical derivation correlation coefficient is greater than a preset correlation threshold, a connection between a first root node of the first basic question category and a second root node of the second basic question category is established; Node fusion is performed based on an association relationship between child nodes at each level in a tree branch of the basic question category connected by the root node, to obtain a typical question structure tree.

3. The method of claim 2, wherein the method further comprises: determining the operation mode of the power distribution network; and determining the operation mode of the power distribution network based on the operation mode of the power distribution network. The node fusion based on the association relationship between the child nodes at each level in the tree branch of the basic question category connected by the root node to obtain the typical question structure tree includes: For a third basic question category and a fourth basic question category for which the root node connection has been established, the tree branches of the third basic question category and the fourth basic question category are adjusted to have the same level structure based on a level depth of the tree branches, to obtain a first adjusted tree branch and a second adjusted tree branch; For any two first child nodes and second child nodes at the same level in the first adjusted tree branch and the second adjusted tree branch, a Manhattan distance is determined based on a first fusion feature vector of the first child node and a second fusion feature vector of the second child node; the fusion feature vector includes a text feature vector, an image feature vector, and an audio feature vector; If the Manhattan distance is less than a second preset distance threshold, a context feature matching degree is determined based on a preceding child node feature and a following child node feature of the first child node and the second child node; If the context feature matching degree is greater than a preset matching degree, node fusion is performed based on an association path of the first child node and the second child node, to obtain the typical question structure tree.

4. The method of claim 3, wherein the method further comprises: determining the operation mode of the power distribution network; and determining the operation mode of the power distribution network based on the operation mode of the power distribution network. The node fusion based on the association path of the first child node and the second child node to obtain the typical question structure tree includes: A child level below the current level of the first child node and the second child node is searched, and child nodes having a Manhattan distance less than the second preset distance threshold and a context feature matching degree greater than the preset matching degree are added to a path set, to obtain a first association path and a second association path, respectively; A number of levels of a common child level in the first association path and the second association path is determined, and a feature similarity is determined based on the first fusion feature vector and the second fusion feature vector; A node contact strength between the first child node and the second child node is determined based on the context feature matching degree, the number of levels, and the feature similarity; Fuse nodes based on node contact strength to obtain a typical problem structure tree.

5. The method of claim 4, wherein the method further comprises: determining the power grid planning operation typical problem library according to the power grid planning operation typical problem library construction method. The node fusion based on the node contact strength to obtain the typical problem structure tree comprises: If the node contact strength is greater than a preset strength threshold, a second logical derivation path is constructed based on the problem description of the first and second child nodes; Logical contradiction detection is performed on the premise condition and conclusion in the derivation process based on the second logical derivation path to obtain a logical contradiction detection result; If the logical contradiction detection result is no conflict, a knowledge domain conflict detection result is obtained based on whether the knowledge domains involved in the second logical derivation path conflict with each other; If the knowledge domain conflict detection result is no conflict, a connection between the first and second child nodes is established to obtain the typical problem structure tree.

6. A system for constructing a typical problem library for power distribution network planning and operation, applying the method for constructing a typical problem library for power distribution network planning and operation according to any one of claims 1-5, characterized in that, Comprise: An acquisition module is configured to acquire power distribution equipment data; A problem classification module is configured to classify power distribution network planning and operation typical problems based on the acquired data to obtain a plurality of basic problem categories; A tree branch construction module is configured to use a multi-modal large model to traverse a first sub-problem and a second sub-problem that triggers the first sub-problem under each basic problem category in a historical experience library, and gradually expand the level based on the first sub-problem and the second sub-problem under each basic problem category until all typical problems are covered to obtain a tree branch of each basic problem category; A structure tree fusion module is configured to obtain a typical problem structure tree through index logical correlation analysis and node fusion based on the tree branch of each basic problem category; A knowledge graph construction module is configured to construct a typical problem knowledge graph by taking the problems represented by each node as entities and the association relationships between the problems as entity edges based on the typical problem structure tree; A problem library construction module is configured to construct a typical problem library by bidirectional relationship mapping based on the nodes in the typical problem structure tree and the entities in the typical problem knowledge graph. 7.An electronic device comprising a memory and a processor, the memory storing a computer program, wherein, The processor executes the computer program to realize the steps of the power distribution network planning and operation typical problem library construction method of any one of claims 1 to 5.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the power distribution network planning and operation typical problem library construction method of any one of claims 1 to 5.

Images