A power equipment knowledge graph ontology construction method, system, device and medium

Abstract

The application discloses a power equipment knowledge graph ontology construction method, system, device and medium, and the power equipment knowledge graph ontology construction method comprises the following steps: acquiring power equipment data; wherein the power equipment data comprises maintenance account data, maintenance guidelines and technical standards of the power equipment; based on the power equipment data, a power transmission ontology framework, a power transformation ontology framework and a power distribution ontology framework of the power equipment are constructed; the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework are fused to obtain a power equipment knowledge graph ontology. The method provided by the application can solve the technical problems that a heterogeneous knowledge graph system is not conducive to the optimization of a business process and the design of characteristics, and can improve the data access efficiency and convenience of an upper-layer application based on a power operation and inspection field knowledge graph.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of artificial intelligence, relates to the field of electric power operation and inspection, and particularly relates to an electric power equipment knowledge graph ontology construction method and system, equipment and a medium. BACKGROUND

[0002] With the rise of knowledge graph technology, experts and scholars have actively explored the field of electric power operation and inspection in recent years, and have researched knowledge graph construction for power transmission, power distribution and power transformation and other subdivided electric power businesses and key technologies for auxiliary decision-making based on the knowledge graph in the field of electric power operation and inspection. Among them, as the core part of the knowledge graph, the ontology introduces new problems while solving the sharing of field knowledge.

[0003] In the field of electric power operation and inspection, the ontology of each subdivided business field such as power transmission, power distribution and power transformation is usually constructed and maintained by business experts in the field. Different subjective understandings of knowledge by each business expert in the field will cause certain differences. At the same time, since there is an intersection between the practice generation and operation environment of power transmission, power distribution and power transformation, the content described by the corresponding ontology usually has overlapping or associated relationships in semantics, and therefore there are differences in the ontology representation language and representation model used, resulting in ontology heterogeneity. SUMMARY

[0004] The purpose of the present application is to provide an electric power equipment knowledge graph ontology construction method, system, equipment and medium to solve the technical problems of the above-mentioned existing technology that the heterogeneous knowledge graph is not conducive to the optimization of business processes and the design of characteristics. The method provided by the present application can solve the technical problems that the heterogeneous knowledge graph system is not conducive to the optimization of business processes and the design of characteristics, and can improve the data access efficiency and convenience of upper-layer applications based on the knowledge graph in the field of electric power operation and inspection.

[0005] To achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] The electric power equipment knowledge graph ontology construction method provided in the first aspect of the present application comprises the following steps:

[0007] Obtaining electric power equipment data; wherein the electric power equipment data comprises maintenance records data, maintenance guidelines and technical standards of electric power equipment;

[0008] Based on the electric power equipment data, a power transmission ontology framework, a power transformation ontology framework and a power distribution ontology framework of electric power equipment are constructed and obtained;

[0009] The power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework constructed and obtained are fused to obtain an electric power equipment knowledge graph ontology.

[0010] The further improvement of the present application is that the step of constructing the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework of the power equipment based on the power equipment data comprises:

[0011] Based on the maintenance account data and combined with the maintenance guide and the technical standard, the logical relationship between the nodes and the nodes is obtained;

[0012] Based on the obtained logical relationship, the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework of the power equipment are respectively constructed; wherein the maintenance account data comprises structured data, semi-structured data and unstructured data; the structured data constructs the ontology by mapping; the semi-structured data and the unstructured data form the ontology by identifying the logical relationship between the nodes and the nodes by information extraction.

[0013] The further improvement of the present application is that the step of constructing the ontology by the structured data by mapping comprises:

[0014] The structured data is preprocessed; after preprocessing, the column representing the power equipment is taken as the core node, the remaining other columns are taken as the graph nodes or attributes, the values are taken as the node names or node attributes, and the relationship names between the nodes are reused as the column names;

[0015] Among them, the preprocessing includes: performing case normalization processing on the English field; and uniformly processing the string encoding into UTF-8.

[0016] The further improvement of the present application is that the step of forming the ontology by identifying the logical relationship between the nodes and the nodes by information extraction for the semi-structured data and the unstructured data comprises:

[0017] For the semi-structured data and the unstructured data, an end-to-end information extraction model DYGIEPP is used for event node extraction to form the ontology;

[0018] Among them, the end-to-end information extraction model DYGIEPP comprises:

[0019] An input layer is configured to convert the semi-structured data or the unstructured data into a character vector sequence;

[0020] A character fragment coding layer is configured to obtain a character fragment sequence and a fragment code of each character fragment based on the character vector sequence;

[0021] A fragment graph propagation is configured to take the obtained fragment code as input and perform fragment graph propagation through a multi-layer neural network calculation; wherein in each iteration t, the result of the last iteration t-1 is taken as input;

[0022] ​a classification layer for classification and output; wherein for the subtask of abstracting into segment classification, the following is adopted classification, the following is adopted classification; FFNN represents a Softmax-based classifier, and T represents the last output of segment graph propagation;

[0023] In the input layer, the character vector sequence is represented as In the formula, x i represents the BERT encoding vector corresponding to the i th character, n represents the length of the text, d bert represents the dimension of the BERT-Chinese output layer;

[0024] In the character segment encoding layer, the character segment sequence is represented as [s1, s2, …, s i , …, s m ]; each segment s i is represented as Each e i is obtained by splicing the BERT encoding vector corresponding to the first character, the BERT encoding vector corresponding to the last character, and a vector related to the length of the segment, and is represented as In the formula, s i represents the i th character segment, e i represents the vector representation of the i th character segment obtained by traversal; represents the e i first character corresponding to the position sequence number in the text, represents the BERT encoding vector corresponding to the first character; represents the e i last character corresponding to the position sequence number, represents the BERT encoding vector corresponding to the last character; represents the length of the segment vector;

[0025] In the segment graph propagation, the calculation process of different tasks x propagating information in different segments is as follows:

[0026]

[0027]

[0028]

[0029] In the formula, Layer represents the output of the t th layer corresponding to the i th segment; represents the information encoding obtained by the i th segment from the t th layer input corresponding to other segments for task x; denotes a segment vector according to the t-th layer input and the weight W between the two segments obtained x is a randomly initialized weight.

[0030] The further improvement of the present application is that the step of fusing the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework to obtain the final power equipment knowledge graph ontology comprises:

[0031] Based on the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework, the same nodes are merged, and different nodes are merged into one ontology.

[0032] The merging of the same nodes comprises: based on the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework, co-reference resolution is performed on the nodes in all triple data by using a synonym table retrieval method, node disambiguation is performed based on a graph, and disambiguation is performed based on a probabilistic generative model.

[0033] The further improvement of the present application is that the step of performing node disambiguation based on a graph comprises:

[0034] A reference graph is established based on the obtained power equipment data and a preset weight; the reference graph represents the node-entity and entity-entity relationships; the nodes and the weighted edges of the candidate in the reference graph represent local dependencies, and the weighted edges between entities represent the semantic correlation between entities.

[0035] The initial importance measure of the node is taken as an initial basis for entity disambiguation, and is transmitted on the reference graph; based on the transmission result of the entity disambiguation basis, a disambiguation target function combining local consistency and global dependency is calculated, a target entity that can maximize the target function is determined for each node, and a disambiguation result is obtained.

[0036] The further improvement of the present application is that the step of performing disambiguation based on a probabilistic generative model comprises:

[0037] A model is established based on the joint probability of the entity mention and the entity, and the disambiguation problem is solved by reasoning of the established model to obtain a disambiguation result.

[0038] The generation process of the entity mention comprises: selecting the entity corresponding to the entity mention according to the probability distribution of the entity; selecting the name of the entity mention according to the conditional probability of the entity name of the given entity; and outputting the context of the entity mention according to the conditional probability of the context of the given entity.

[0039] The joint probability of an entity and an entity mention is represented as P(m, e) = P(s, c, e) = P(e)P(s|e)P(c|e);

[0040] In the formula, P(m, e) represents the joint probability that an entity description m in the text is linked to an entity e in the database; p(s, c, e) represents the joint probability that an entity description m=(s, c) with an entity name s and other context c corresponds to an entity e in the database; p(e) represents the prior probability that an entity e in the database appears in the text; p(s|e) represents the conditional probability that s is the name of the entity e given that the entity e appears in the text; and p(c|e) represents the conditional probability that the entity e appears in the text with the context c;

[0041] P(e) corresponds to the popularity of the entity, P(s|e) corresponds to the entity name knowledge, and P(c|e) corresponds to the context knowledge; when the entity mention m is given, the calculation expression of the semantic calculation between the candidate entities is,

[0042]

[0043] Further improvements of the present application are that when the same nodes are merged, the method further comprises:

[0044] The entity disambiguation based on deep learning comprises: adopting word2vec to perform vector representation on concepts; input information corresponding to each concept comprises a concept E, a relationship R owned by the concept, a concept type ET and a concept description D; based on bag-of-words and one-hot representation input, dimension reduction is performed through a word hashing layer, and then nonlinear transformation of a multilayer neural network is performed to obtain a semantic layer representation vector; the cosine similarity of the semantic layer representation vectors of two concept nodes is calculated as their correlation; and entity disambiguation is completed based on comparison of the similarity with a preset threshold.

[0045] The second aspect of the present application provides an electric power equipment knowledge graph ontology construction system, comprising:

[0046] An original data acquisition module is configured to acquire electric power equipment data; wherein the electric power equipment data comprises maintenance account data, maintenance guidelines and technical standards of the electric power equipment;

[0047] An ontology framework construction module is configured to construct a power transmission ontology framework, a power transformation ontology framework and a power distribution ontology framework of the electric power equipment based on the electric power equipment data;

[0048] An ontology fusion module is configured to fuse the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework to obtain an electric power equipment knowledge graph ontology.

[0049] The further improvement of the present application is that in the ontology framework construction module, the step of constructing the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework of the power equipment based on the power equipment data comprises:

[0050] Based on the maintenance account data and combined with the maintenance guide and the technical standard, the logical relationship between the nodes and the nodes is obtained;

[0051] Based on the obtained logical relationship, the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework of the power equipment are constructed respectively; wherein the maintenance account data comprises structured data, semi-structured data and unstructured data; the structured data constructs the ontology by mapping; the semi-structured data and the unstructured data identify the logical relationship between the nodes and the nodes by information extraction to form the ontology.

[0052] The further improvement of the present application is that in the ontology framework construction module, the step of constructing the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework of the power equipment based on the power equipment data comprises:

[0053] The structured data is preprocessed; after preprocessing, the column representing the power equipment is taken as the core node, the remaining other columns are taken as the atlas node or the attribute, the value is taken as the node name or the node attribute, and the relationship name between the nodes is reused as the column name;

[0054] Among them, the preprocessing includes: performing case normalization processing on the English field; and uniformly processing the string encoding into UTF-8.

[0055] The further improvement of the present application is that in the ontology framework construction module, the step of constructing the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework of the power equipment based on the power equipment data comprises:

[0056] For semi-structured data and unstructured data, an end-to-end information extraction model DYGIEPP is used for event node extraction to form an ontology;

[0057] Among them, the end-to-end information extraction model DYGIEPP comprises:

[0058] An input layer is configured to convert the semi-structured data or the unstructured data into a character vector sequence;

[0059] A character fragment coding layer is configured to obtain a character fragment sequence and a fragment code of each character fragment based on the character vector sequence;

[0060] A fragment graph propagation is configured to take the obtained fragment code as input and perform fragment graph propagation through a multi-layer neural network calculation; wherein in each iteration t, the result of the last iteration t-1 is taken as input; ​

[0061] A classification layer is used for classification and output; specifically, for subtasks abstracted as fragment classification, a classification layer is employed. Classification is performed, and for subtasks abstracted into relational classification, the following methods are used: Classification is performed; FFNN represents a softmax-based classifier, and T represents the final output of the fragment graph propagation;

[0062] In the input layer, the character vector sequence is represented as follows: In the formula, x i This represents the BERT encoding vector corresponding to the i-th character, where n represents the text length and d represents the BERT encoding vector. bert Indicates the dimension of the BERT-Chinese output layer;

[0063] In the character segment encoding layer, the character segment sequence is represented as [s1, s2, ..., s i ,…,s m ]; Each fragment s i Represented as Each e i It is obtained by concatenating the BERT encoding vector corresponding to the first character, the BERT encoding vector corresponding to the last character, and a vector related to the segment length, and is represented as follows: In the formula, s i Represents the i-th character segment, e i This represents the vector representation of the i-th character segment obtained through traversal; e i The position number of the first character in the text. This represents the BERT encoding vector corresponding to the first character; e i The position number corresponding to the last character. This represents the BERT encoding vector corresponding to the last character; This represents the length vector of the segment;

[0064] In the fragment graph propagation, the calculation process for different tasks x to propagate information in different fragments is as follows:

[0065]

[0066]

[0067]

[0068] In the formula, The layer represents the output of the t-th layer corresponding to the i-th segment; represents the information encoding obtained by the i-th segment from the t-th layer input corresponding to other segments for task x; represents the segment vector according to the t-th layer input and the weight W between the two segments obtained x is a randomly initialized weight.

[0069] Further improvement of the present application is that in the ontology fusion module, the step of implementing ontology fusion of the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework to obtain the final power equipment knowledge graph ontology includes:

[0070] Based on the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework, the same nodes are merged, and different nodes are merged into one ontology.

[0071] The merging of the same nodes includes: based on the power transmission ontology framework, the power transformation ontology framework and the power distribution ontology framework, the nodes in all triple data are co-reference resolved by adopting a synonym table retrieval method, the nodes are disambiguated based on a graph, and disambiguation is performed based on a probabilistic generative model.

[0072] Further improvement of the present application is that in the ontology fusion module, the step of implementing disambiguation of nodes based on a graph includes:

[0073] Based on the obtained power equipment data and the preset weight, a reference graph is established; wherein the reference graph represents the node-entity and entity-entity relationship; the nodes and the weighted edges of the candidate in the reference graph represent local dependency, and the weighted edges between entities represent semantic correlation between entities;

[0074] The initial importance measure of the node is taken as an initial basis for entity disambiguation and is transmitted on the reference graph; based on the transmission result of the entity disambiguation basis, a disambiguation target function combining local compatibility and global dependency is calculated, a target entity that can maximize the target function is determined for each node, and a disambiguation result is obtained.

[0075] Further improvement of the present application is that in the ontology fusion module, the step of implementing disambiguation based on a probabilistic generative model includes:

[0076] Based on the probabilistic generative model, a model is established for the joint probability of the entity mention and the entity, the disambiguation problem is solved through reasoning of the established model, and a disambiguation result is obtained.

[0077] The generation process of the entity mention includes: selecting an entity corresponding to the entity mention according to the probability distribution of the entity; selecting the name of the entity mention according to the conditional probability of the entity name of the given entity; and outputting the context of the entity mention according to the conditional probability of the context of the given entity.

[0078] The joint probability of entity and entity mention is represented as P(m, e) = P(s, c, e) = P(e)P(s|e)P(c|e);

[0079] In the formula, P(m, e) represents the joint probability that a piece of entity description m in the text is linked to entity e in the database; p(s, c, e) represents the joint probability that the entity description m=(s, c) with the entity name s and other context c corresponds to the entity e in the database; p(e) represents the prior probability that the entity e in the database appears in the text; p(s|e) represents the conditional probability that s is the name of the entity e given that the entity e appears in the text; and p(c|e) represents the conditional probability that the entity e appears in the text with the context c;

[0080] P(e) corresponds to the popularity of the entity, P(s|e) corresponds to the entity name knowledge, and P(c|e) corresponds to the context knowledge; when the entity mention m is given, the calculation expression of the semantic calculation between the candidate entities is,

[0081]

[0082] Further improvement of the present application is that in the ontology fusion module, when the same nodes are merged, the ontology fusion module further comprises:

[0083] The entity disambiguation based on deep learning comprises: adopting word2vec to perform vector representation on concepts; input information corresponding to each concept comprises a concept E, a relationship R owned by the concept, a concept type ET and a concept description D; based on bag-of-words and one-hot representation input, dimension reduction is performed through a word hashing layer, and then nonlinear transformation of a multilayer neural network is performed to obtain a semantic layer representation vector; the cosine similarity of the semantic layer representation vectors of two concept nodes is calculated as their correlation; and entity disambiguation is completed based on comparison of the similarity and a preset threshold.

[0084] The third aspect of the present application provides an electronic device, comprising:

[0085] at least one processor; and

[0086] a memory in communication connection with the at least one processor; wherein

[0087] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the power equipment knowledge graph ontology construction method as described in any one of the above aspects of the present application.

[0088] The fourth aspect of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the power equipment knowledge graph ontology construction method of any of the above aspects of the present application.

[0089] Compared with the prior art, the present application has the following beneficial effects:

[0090] The power equipment knowledge graph ontology construction method provided by the present application provides a technical route for constructing a power operation and inspection field knowledge graph ontology based on transmission, transformation and distribution of account data, and the final power equipment knowledge graph ontology framework is obtained through unified standard maintenance and management of data and ontology fusion, which can solve the technical problem that heterogeneous knowledge graph systems are not conducive to optimization and characteristic design of business processes, and can improve the data access efficiency and convenience of upper-layer applications based on the power operation and inspection field knowledge graph. BRIEF DESCRIPTION OF DRAWINGS

[0091] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following briefly introduces the drawings needed to be used in the embodiments or prior art description; obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0092] Figure 1 is a flowchart of a power equipment knowledge graph ontology construction method provided by an embodiment of the present application;

[0093] Figure 2 is a flowchart of a power equipment knowledge graph ontology construction method provided by another embodiment of the present application;

[0094] Figure 3 is a reference graph in the node disambiguation process based on a graph in an embodiment of the present application;

[0095] Figure 4 is a disambiguation diagram based on a probability generation model in an embodiment of the present application;

[0096] Figure 5 is a fused ontology framework diagram in an embodiment of the present application;

[0097] Figure 6 is a diagram of a power equipment knowledge graph ontology construction system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0098] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort should fall within the protection scope of the present application.

[0099] It should be noted that the terms "first", "second", and the like in the specification and claims of the present application and the above-described accompanying drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or device including a series of steps or units does not have to be limited to only those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product, or device.

[0100] The present application will be described in further detail below in conjunction with the accompanying drawings:

[0101] Embodiment One

[0102] Please refer to Figure 1 The embodiment of the present application provides a power equipment knowledge graph ontology construction method for a power operation and inspection field, which comprises the following steps:

[0103] Step 1, obtaining the maintenance ledger data (specifically, for example, a circuit breaker, a transformer, a lightning rod, etc.), maintenance guidelines, and preset technical standards (specifically, for example, one or more of national standards, industry standards, team standards, and enterprise standards) of the preset power equipment (specifically, for example, a circuit breaker, a transformer, a lightning rod, etc.);

[0104] Step 2, based on the obtained data of the preset power equipment, constructing a power transmission ontology framework, a power transformation ontology framework, and a power distribution ontology framework of the power equipment;

[0105] Step 3, performing ontology fusion on the constructed power transmission ontology framework, power transformation ontology framework, and power distribution ontology framework of the power equipment to obtain a fused power equipment knowledge graph ontology.

[0106] The embodiment of the application explains the problems existing in the current power service field. Firstly, different experts in the same power service field construct subgraph topologies with similar structures, but there are usually a large number of heterogeneous topologies, and knowledge sharing between power software systems based on different ontologies is difficult. Secondly, due to the differences in the knowledge systems of power field experts and the differences in the understanding of the business, it is not realistic to completely manually establish and maintain a complete and unified power equipment knowledge graph ontology. Thirdly, with the more and more extensive application of ontology, the knowledge sharing bottleneck caused by ontology heterogeneity is becoming more and more prominent. Therefore, solving the ontology heterogeneity of the power equipment knowledge graph and eliminating the knowledge sharing obstacles are an urgent problem faced by the application based on ontology, and ontology fusion is an effective way to solve the ontology heterogeneity problem. In the device knowledge graph ontology construction method for the power operation and inspection field provided by the embodiment of the application, a technical route for constructing the knowledge graph ontology of the power operation and inspection field based on the accounting data of the power transmission, power transformation and power distribution is provided, and through the way of unified and standardized maintenance and management of data, the technical problem that the heterogeneous knowledge graph system is not conducive to the optimization and characteristic design of the business process can be solved.

[0107] Embodiment two

[0108] Please refer to Figures 2 to 5 The device knowledge graph ontology construction method for the power operation and inspection field provided by the embodiment of the application comprises the following steps:

[0109] The power transmission, power transformation and power distribution ontology framework is constructed, and the steps comprise the following: based on the device maintenance accounting data and combined with the maintenance guide and technical standards, the logical relationship between the nodes such as power equipment, the city to which the power equipment belongs, the maintenance team and the like and the logical relationship between the nodes are sorted out, and then the power transmission, power transformation and power distribution operation and inspection device knowledge graph ontology architecture is constructed. It is further specifically preferred in the embodiment of the application that the device maintenance accounting data comprises structured data, semi-structured data and unstructured data; wherein the structured data can be constructed into an ontology by using a mapping method, the semi-structured and unstructured data identify the nodes and the logical relationship between the nodes by using an information extraction technology, and a small amount of expert auditing work is assisted, and finally the ontology is formed.

[0110] As a further preferred technical scheme of the embodiment of the application, the above steps specifically comprise the following steps:

[0111] The step of constructing the ontology by using the mapping method for the structured data comprises the following: for the structured data from the relational database, the column name field is preprocessed, the English field is normalized in uppercase and lowercase, the string coding is unified as UTF-8, the column representing the power equipment is the core node, other columns are used as graph nodes or attributes, the specific values are used as node names or node attributes, and the relationship name between the nodes is reused for the column name;

[0112] For semi-structured and unstructured data, an end-to-end information extraction model DYGIEPP is adopted for event node extraction, and the specific steps are as follows:

[0113] (1) Input layer: BERT-Chinese is adopted to convert the input text into a character vector sequence, which is represented as:

[0114]

[0115] In the formula, x i represents the bert encoding vector corresponding to the i th character, n represents the length of the text, d bert represents the dimension of the BERT-Chinese output layer.

[0116] (2) Character fragment encoding layer: according to the characteristics of the power inspection field corpus, all character fragments with a length not greater than 10 characters are traversed to obtain a fragment sequence [s1, s2, …, s i ,…,s m ];

[0117] For each fragment s i is represented as follows:

[0118]

[0119]

[0120] In the formula, s i represents the i th character fragment, e i represents the vector representation of the i th character fragment obtained by traversal; represents the e i first character corresponding to the position number in the text, represents the bert encoding vector corresponding to the first character; represents the e i corresponding to the position number corresponding to the last character, represents the bert encoding vector corresponding to the last character; represents the length of the fragment vector. Each e i is obtained by splicing the bert encoding vector corresponding to the first character, the bert encoding vector corresponding to the last character, and a vector related to the length of the fragment;

[0121] (3) Fragment graph propagation: taking the obtained fragment encoding E=[e1, e2,...] as input, the fragment graph propagation is calculated through a multi-layer neural network, and the output of the t-1 th layer is taken as the input of the t th layer, that is, in each iteration t, the result of the last iteration t-1 is taken as the input of the next iteration t As input, information is propagated in different segments according to different tasks x, and the calculation process is as follows:

[0122]

[0123]

[0124]

[0125] In the formula, The layer represents the output of the t-th layer corresponding to the i-th segment; This represents the information encoding obtained by the i-th segment from the t-th layer input corresponding to other segments for task x; This represents the fragment vector based on the input from the t-th layer. and The weights W between the two segments are obtained. x Initialize the weights randomly;

[0126] (4) Classification Layer: A Softmax-based classifier is represented using FFNN. For subtasks that can be abstracted as segment classification, including entity recognition and trigger word detection, a classification layer is adopted. Classification is performed; for subtasks that can be abstracted into relation classification, including relation extraction and role classification, the following methods are adopted: Where T represents the final output of the fragment graph propagation.

[0127] The obtained power equipment transmission, substation, and distribution ontology frameworks are integrated. This mainly involves merging the transmission, distribution, and substation frameworks constructed in the above steps. Specifically, identical nodes are merged, including mapping information such as categories, attributes, and relationships, and different nodes are merged into one ontology.

[0128] As a further preferred embodiment of the present invention, the above steps specifically include the following steps:

[0129] Based on the transmission ontology framework, substation ontology framework, and distribution ontology framework, a thesaurus retrieval method is used to resolve coreferences of nodes in all triplet data. Preferably, after the coreference resolution, the following steps may be included: using surface name expansion technology, other possible expansion variants (e.g., full names) are identified from the relevant documents where the node appears; then, these expansion forms are used to form a candidate entity set for the node. Specifically, surface name expansion can be implemented using a heuristic pattern matching method.

[0130] The technical solution of this invention also includes graph-based concept node disambiguation. The graph-based method represents nodes E and their relationships in the form of a graph, and then performs collaborative reasoning on the relationships between nodes E and between candidate entities on the graph.

[0131] Reference Graph-based Collaborative Entity Linking Method Figure 3 As shown: First, based on the acquired device data and preset weights (obtained based on expert experience or by statistical methods), a reference diagram is established as shown in the figure. The relationship between concept nodes and entities, and between entities, is represented in the figure. The concept nodes and candidate weighted edges in the figure represent their local dependencies. The weighted edges between entities represent the semantic relevance between entities, providing a basis for global collaborative entity disambiguation.

[0132] After calculating the initial importance metric for each node, it is used as the initial basis for entity disambiguation and passed on to the reference map. Finally, based on the result of the entity disambiguation metric transfer, a disambiguation objective function combining local compatibility and global dependency is calculated. For each concept node, a target entity that maximizes the objective function is determined, thus obtaining the disambiguation result.

[0133] The technical solution of this invention also includes disambiguation based on a probabilistic generation model. The joint probability of concept nodes (which can be understood as entity mentions) and entities is modeled based on the probabilistic generation model, and the disambiguation problem is solved through model inference. In the entity disambiguation method based on the probabilistic generation model, entity mentions are modeled as generated samples, and their generation process is as follows: Figure 4 The final integrated framework is as follows Figure 5 As shown.

[0134] First, the model selects the entity corresponding to the entity mention based on the entity's probability distribution P(e), such as "500kV Kangjia Second Line" and "Wanquan 220kV #4 and #5 Busbars" in the example. Then, the model selects the entity mention's name based on the conditional probability P(s|e) of the given entity name e, such as "Line" and "Busbar" in the example. Finally, the model outputs the entity mention's context based on the conditional probability P(c|e) of the given entity context e. According to the above entity mention generation process, the joint probability of an entity and its mention can be defined as:

[0135] P(m,e)=P(s,c,e)=P(e)P(s|e)P(c|e)(7)

[0136] In the formula, P(m,e) represents the joint probability that an entity description m in the text is linked to an entity e in the database; p(s,c,e) represents the joint probability that an entity description m = (s,c) with entity name s and other context c corresponds to entity e in the database; p(e) represents the prior probability that entity e in the database appears in the text; p(s|e) represents the conditional probability that, given entity e, it appears in the text with the name s; p(c|e) represents the conditional probability that, given entity e, it appears in the text with context c.

[0137] In this method, P(e) corresponds to the popularity of the entity, P(s|e) corresponds to the entity name knowledge, and P(c|e) corresponds to the context knowledge. When a given entity mentions m, the semantic computation between it and the candidate entity is achieved through the following formula:

[0138]

[0139] A further preferred embodiment of the present invention includes a deep learning-based entity disambiguation method during ontology fusion. This primarily employs word2vec to represent concepts as vectors. Specifically, at the input layer, the input information for each concept includes concept E, the relation R possessed by the concept, concept type ET, and concept description D. The input, based on bag-of-words and one-hot representations, undergoes dimensionality reduction through a word hash layer, followed by a nonlinear transformation through a multi-layer neural network to obtain the concept representation at the semantic layer. The cosine similarity between the semantic layer representation vectors of two concepts is calculated as their relevance. Based on a pre-set threshold th1, if Sim(x1,x2)>th1, the two concept nodes can be fused, and the result is output; for example, the threshold th1 can be set to 0.5.

[0140] To further facilitate querying, a power knowledge reference table retrieval system was constructed. This includes: building a query power knowledge reference table based on maintenance log data, establishing a correspondence between concepts and candidates; this table is a key-value mapping, where one key can correspond to one or more values. After the reference table is constructed, candidates can be directly obtained from the table by concept. A query entity reference table can also be constructed, extracting the correspondence between concepts and other candidates based on term pages, knowledge graphs, and knowledge maps from the power encyclopedia website.

[0141] In summary, power operation and maintenance (POM) covers the subdivided businesses of transmission, substation, and distribution. Currently, the information sharing and data interaction methods between the various business systems lack unified standards, resulting in high implementation costs and hindering maintenance and development. To help break down knowledge barriers and achieve data collaboration and integration across subdivided business areas, this invention proposes a technical approach and application cases for constructing a power MPM domain knowledge graph ontology based on ledger data for transmission, substation, and distribution. This aims to address the problems of heterogeneous knowledge graph systems hindering business process optimization and feature design. By unifying and standardizing data maintenance and management methods, it improves the efficiency and convenience of data access for upper-layer applications based on the power MPM domain knowledge graph. This invention provides a method for constructing an equipment knowledge graph ontology for the power operation and maintenance field: First, based on existing equipment ledger data in the power operation and maintenance field, a knowledge graph ontology for transmission, distribution, and substation subdivisions is constructed; then, a rule-based method is used to perform coreference disambiguation using a domain-accumulated and solidified thesaurus; finally, for semantically similar concepts, a rule-based and machine learning method is used to calculate similarity, and concept nodes with similarity greater than a threshold are fused; for concepts that do not need to be fused, the original information is directly retained and added to a unified ontology.

[0142] The following are embodiments of the apparatus of the present invention, which can be used to execute embodiments of the method of the present invention. For details not omitted in the apparatus embodiments, please refer to the embodiments of the method of the present invention.

[0143] Please see Figure 6 Another embodiment of the present invention provides a power equipment knowledge graph ontology construction system, comprising:

[0144] The raw data acquisition module is used to acquire power equipment data; wherein, the power equipment data includes power equipment maintenance log data, maintenance guidelines and technical standards;

[0145] The ontology framework construction module is used to construct and obtain the power transmission ontology framework, substation ontology framework, and distribution ontology framework of the power equipment based on the power equipment data.

[0146] The ontology fusion module is used to fuse the constructed transmission ontology framework, the substation ontology framework, and the distribution ontology framework to obtain a power equipment knowledge graph ontology.

[0147] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions from the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used in the operation of a power equipment knowledge graph ontology construction method.

[0148] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the power equipment knowledge graph ontology construction method in the above embodiments.

[0149] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure One One or more processes and / or boxes Figure One A device that provides the functions specified in one or more boxes.

[0151] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure One One or more processes and / or boxes Figure One The function specified in one or more boxes.

[0152] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure One One or more processes and / or boxes Figure One The steps of the function specified in one or more boxes.

[0153] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for constructing a knowledge graph ontology for power equipment, characterized in that, Includes the following steps: Acquire power equipment data; wherein, the power equipment data includes power equipment maintenance log data, maintenance guidelines and technical standards; Based on the power equipment data, construct and obtain the power transmission body framework, substation body framework and distribution body framework of the power equipment. The constructed power transmission ontology framework, power substation ontology framework, and power distribution ontology framework are fused to obtain a power equipment knowledge graph ontology. in, The steps of constructing the power equipment transmission body framework, substation body framework, and distribution body framework based on the power equipment data include: obtaining nodes and the logical relationships between nodes based on the maintenance log data and in conjunction with maintenance guidelines and technical standards; and constructing the power equipment transmission body framework, substation body framework, and distribution body framework based on the obtained logical relationships, respectively. The maintenance log data includes structured data, semi-structured data, and unstructured data; the structured data is constructed using a mapping method; and the semi-structured and unstructured data are used to identify nodes and the logical relationships between nodes to form the body. The steps for identifying nodes and logical relationships between nodes to form an ontology using information extraction methods for semi-structured and unstructured data include: For semi-structured and unstructured data, the end-to-end information extraction model DYGIEPP is used to extract event nodes to form an ontology; The end-to-end information extraction model DYGIEPP includes: The input layer is used to convert semi-structured or unstructured data into a sequence of character vectors. A character fragment encoding layer is used to obtain a character fragment sequence and a fragment encoding for each character fragment based on the character vector sequence; Fragment graph propagation is used to propagate the obtained fragments as input through a multi-layer neural network; in each iteration... In the previous iteration The result For input; A classification layer is used for classification and output; specifically, for subtasks abstracted as fragment classification, a classification layer is employed. Classification is performed, and for subtasks abstracted into relational classification, the following methods are used: To perform classification; FFNN represents a softmax-based classifier. This represents the final output of the fragment graph propagation; In the input layer, the character vector sequence is represented as follows: In the formula, This represents the BERT encoding vector corresponding to the i-th character, where n represents the text length. Indicates the dimension of the BERT-Chinese output layer; In the character segment encoding layer, the character segment sequence is represented as follows: Each segment Represented as Each It is obtained by concatenating the BERT encoding vector corresponding to the first character, the BERT encoding vector corresponding to the last character, and a vector related to the segment length, and is represented as follows: In the formula, This represents the i-th character segment. This represents the vector representation of the i-th character segment obtained through traversal; express The position number of the first character in the text. This represents the BERT encoding vector corresponding to the first character; express The position number corresponding to the last character. This represents the BERT encoding vector corresponding to the last character; This represents the length vector of the segment; In the propagation of the fragment graph, different tasks The calculation process for propagating information across different segments is as follows: ； ； ； In the formula, The layer represents the output of the t-th layer corresponding to the i-th segment; This represents the information encoding obtained by the i-th segment from the t-th layer input corresponding to other segments for task x; This represents the fragment vector based on the input from the t-th layer. and The weights between the two segments are obtained. Initialize the weights randomly; The step of fusing the power transmission body framework, the substation body framework, and the distribution body framework to obtain the final power equipment knowledge graph ontology includes: Based on the power transmission body framework, the substation body framework, and the power distribution body framework, identical nodes are merged, and different nodes are merged into one body. The process of merging identical nodes includes: using a thesaurus search method based on the transmission, substation, and distribution frameworks to resolve common references among nodes in all triplet data; performing node disambiguation based on the graph; and performing disambiguation based on a probabilistic generation model.

2. The method for constructing a knowledge graph ontology for power equipment according to claim 1, characterized in that, The steps for constructing an ontology from the structured data using a mapping method include: The structured data is preprocessed; the columns representing power equipment after preprocessing are the core nodes, and the remaining columns are used as graph nodes or attributes. The values ​​are used as node names or node attributes, and the column names are reused for the relationships between nodes. The preprocessing includes: normalizing the case of English fields; and unifying the string encoding to UTF-8.

3. The method for constructing a knowledge graph ontology for power equipment according to claim 1, characterized in that, The steps for graph-based node disambiguation include: A reference image is created based on the acquired power equipment data and preset weights; wherein, the reference image represents the node-entity and entity-entity relationships; the weighted edges of nodes and candidates in the reference image represent local dependencies, and the weighted edges between entities represent the semantic relevance between entities; The initial importance metric of the node is used as the initial basis for entity disambiguation and is passed on the reference. Based on the passing result of the entity disambiguation basis, the disambiguation objective function combining local compatibility and global dependency is calculated, and the target entity that maximizes the objective function is determined for each node to obtain the disambiguation result.

4. The method for constructing a knowledge graph ontology for power equipment according to claim 1, characterized in that, The steps for disambiguation based on the probability generation model include: A model is established based on the joint probability of entity mentions and entities using a probabilistic generative model. The disambiguation problem is solved by reasoning through the established model to obtain the disambiguation result. The entity mention generation process includes: selecting the entity corresponding to the entity mention based on the entity's probability distribution; selecting the name of the entity mention based on the conditional probability of the entity name of a given entity; and outputting the context of the entity mention based on the conditional probability of the given entity context. The joint probability of an entity and its mentions is represented as follows: ; In the formula, P(m,e) represents the joint probability that an entity description m in the text is linked to an entity e in the database; p(s,c,e) represents the joint probability that an entity description m = (s,c) with entity name s and other context c corresponds to entity e in the database; p(e) represents the prior probability that entity e in the database appears in the text; p(s|e) represents the conditional probability that a given entity e appears in the text with the name s; p(c|e) represents the conditional probability that a given entity e appears in the text with context c. The popularity of the corresponding entity, Corresponding entity name knowledge, Corresponding contextual knowledge; when a given entity is mentioned When the semantic computation between the candidate entity and the candidate entity is, the computational expression is, 。 5. The method for constructing a knowledge graph ontology for power equipment according to claim 1, characterized in that, When merging identical nodes, the process also includes: Entity disambiguation based on deep learning includes: using word2vec to represent concepts as vectors; the input information for each concept includes concept E, the relation R that the concept possesses, concept type ET, and concept description D; the input based on bag-of-words and one-hot representations is dimensionality-reduced through a word hash layer, and then subjected to a nonlinear transformation through a multi-layer neural network to obtain a semantic layer representation vector; the cosine similarity between the semantic layer representation vectors of two concept nodes is calculated as their relevance; entity disambiguation is completed based on the comparison of the similarity with a preset threshold.

6. A knowledge graph ontology construction system for power equipment, characterized in that, include: The raw data acquisition module is used to acquire power equipment data; wherein, the power equipment data includes power equipment maintenance log data, maintenance guidelines and technical standards; The ontology framework construction module is used to construct and obtain the power transmission ontology framework, substation ontology framework, and distribution ontology framework of the power equipment based on the power equipment data. The ontology fusion module is used to fuse the constructed power transmission ontology framework, the substation ontology framework, and the distribution ontology framework to obtain a power equipment knowledge graph ontology. in, The ontology framework construction module includes the following steps for constructing the power equipment's transmission ontology framework, substation ontology framework, and distribution ontology framework based on the power equipment data: obtaining nodes and their logical relationships based on the maintenance log data and in conjunction with maintenance guidelines and technical standards; and constructing the power equipment's transmission ontology framework, substation ontology framework, and distribution ontology framework based on the obtained logical relationships. The maintenance log data includes structured data, semi-structured data, and unstructured data. The structured data is constructed using a mapping method. The semi-structured and unstructured data are extracted to identify nodes and their logical relationships to form the ontology. In the ontology framework construction module, the steps for identifying nodes and logical relationships between nodes to form an ontology using information extraction methods for semi-structured and unstructured data include: For semi-structured and unstructured data, the end-to-end information extraction model DYGIEPP is used to extract event nodes to form an ontology; The end-to-end information extraction model DYGIEPP includes: The input layer is used to convert semi-structured or unstructured data into a sequence of character vectors. A character fragment encoding layer is used to obtain a character fragment sequence and a fragment encoding for each character fragment based on the character vector sequence; Fragment graph propagation is used to propagate the obtained fragments as input through a multi-layer neural network; in each iteration... In the previous iteration The result For input; A classification layer is used for classification and output; specifically, for subtasks abstracted as fragment classification, a classification layer is employed. Classification is performed, and for subtasks abstracted into relational classification, the following methods are used: To perform classification; FFNN represents a softmax-based classifier. This represents the final output of the fragment graph propagation; In the input layer, the character vector sequence is represented as follows: In the formula, This represents the BERT encoding vector corresponding to the i-th character, where n represents the text length. Indicates the dimension of the BERT-Chinese output layer; In the character segment encoding layer, the character segment sequence is represented as follows: Each segment Represented as Each It is obtained by concatenating the BERT encoding vector corresponding to the first character, the BERT encoding vector corresponding to the last character, and a vector related to the segment length, and is represented as follows: In the formula, This represents the i-th character segment. This represents the vector representation of the i-th character segment obtained through traversal; express The position number of the first character in the text. This represents the BERT encoding vector corresponding to the first character; express The position number corresponding to the last character. This represents the BERT encoding vector corresponding to the last character; This represents the length vector of the segment; In the propagation of the fragment graph, different tasks The calculation process for propagating information across different segments is as follows: ； ； ； In the formula, The layer represents the output of the t-th layer corresponding to the i-th segment; This represents the information encoding obtained by the i-th segment from the t-th layer input corresponding to other segments for task x; This represents the fragment vector based on the input from the t-th layer. and The weights between the two segments are obtained. Initialize the weights randomly; In the ontology fusion module, the steps for fusing the power transmission ontology framework, the substation ontology framework, and the distribution ontology framework to obtain the final power equipment knowledge graph ontology include: merging identical nodes and merging different nodes into one ontology based on the power transmission ontology framework, the substation ontology framework, and the distribution ontology framework; wherein merging identical nodes includes: performing coreference resolution on nodes in all triplet data using a thesaurus search method based on the power transmission ontology framework, the substation ontology framework, and the distribution ontology framework; performing node disambiguation based on the graph; and performing disambiguation based on a probabilistic generation model.

7. A power equipment knowledge graph ontology construction system according to claim 6, characterized in that, In the ontology framework construction module, the steps for constructing an ontology from structured data using a mapping method include: The structured data is preprocessed; the columns representing power equipment after preprocessing are the core nodes, and the remaining columns are used as graph nodes or attributes. The values ​​are used as node names or node attributes, and the column names are reused for the relationships between nodes. The preprocessing includes: normalizing the case of English fields; and unifying the string encoding to UTF-8.

8. A power equipment knowledge graph ontology construction system according to claim 6, characterized in that, In the ontology fusion module, the steps for implementing graph-based node disambiguation include: A reference image is created based on the acquired power equipment data and preset weights; wherein, the reference image represents the node-entity and entity-entity relationships; the weighted edges of nodes and candidates in the reference image represent local dependencies, and the weighted edges between entities represent the semantic relevance between entities; The initial importance metric of the node is used as the initial basis for entity disambiguation and is passed on the reference. Based on the passing result of the entity disambiguation basis, the disambiguation objective function combining local compatibility and global dependency is calculated, and the target entity that maximizes the objective function is determined for each node to obtain the disambiguation result.

9. A power equipment knowledge graph ontology construction system according to claim 6, characterized in that, The ontology fusion module includes the following steps for disambiguation based on a probabilistic generation model: A model is established based on the joint probability of entity mentions and entities using a probabilistic generative model. The disambiguation problem is solved by reasoning through the established model to obtain the disambiguation result. The entity mention generation process includes: selecting the entity corresponding to the entity mention based on the entity's probability distribution; selecting the name of the entity mention based on the conditional probability of the entity name of a given entity; and outputting the context of the entity mention based on the conditional probability of the given entity context. The joint probability of an entity and its mentions is represented as follows: ; In the formula, P(m,e) represents the joint probability that an entity description m in the text is linked to an entity e in the database; p(s,c,e) represents the joint probability that an entity description m = (s,c) with entity name s and other context c corresponds to entity e in the database; p(e) represents the prior probability that entity e in the database appears in the text; p(s|e) represents the conditional probability that a given entity e appears in the text with the name s; p(c|e) represents the conditional probability that a given entity e appears in the text with context c. The popularity of the corresponding entity, Corresponding entity name knowledge, Corresponding contextual knowledge; when a given entity is mentioned When the semantic computation between the candidate entity and the candidate entity is, the computational expression is, 。 10. A power equipment knowledge graph ontology construction system according to claim 6, characterized in that, The ontology fusion module, when merging identical nodes, also includes: Entity disambiguation based on deep learning includes: using word2vec to represent concepts as vectors; the input information for each concept includes concept E, the relation R that the concept possesses, concept type ET, and concept description D; the input based on bag-of-words and one-hot representations is dimensionality-reduced through a word hash layer, and then subjected to a nonlinear transformation through a multi-layer neural network to obtain a semantic layer representation vector; the cosine similarity between the semantic layer representation vectors of two concept nodes is calculated as their relevance; entity disambiguation is completed based on the comparison of the similarity with a preset threshold.

11. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the power equipment knowledge graph ontology construction method as described in any one of claims 1 to 5.

12. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the power equipment knowledge graph ontology construction method as described in any one of claims 1 to 5.

Images