A knowledge graph construction method for substation project site selection

Abstract

The present invention discloses a method for constructing a knowledge graph for substation project site selection, which comprises the following steps: collecting relevant specifications and standards for substation project site selection, obtaining a concept set and a relationship set through term extraction and screening; establishing "concept-relationship-concept" triples to construct an ontology layer knowledge graph; collecting operating data of existing substations, obtaining an entity set, a relationship set and an attribute set through knowledge extraction and screening; establishing "instance-relationship-instance" and "instance-relationship-attribute" triples to construct an instance layer knowledge graph; constructing retrieval learning and auxiliary decision-making for substation project site selection knowledge based on the ontology layer knowledge graph and the instance layer knowledge graph and the mapping relationship between them; the present invention constructs a relatively complete knowledge graph system for substation site selection; and provides new project retrieval and auxiliary decision-making by forming a rich case knowledge base.

Description

Technical Field

[0001] The present invention relates to the field of power engineering and knowledge engineering technology, and in particular to a method for constructing a knowledge graph for substation project site selection. Background Art

[0002] Substation planning is an important part of power grid planning, which directly affects the structure, investment, economy and reliability of the power grid in the planning area. Traditional substation planning mainly relies on manual experience and field surveys. It is not only labor-intensive and time-consuming, but also affected by many subjective factors. It is difficult to fully consider the factors affecting the site selection of substations, making the site selection of substations very difficult. Against the background of the continuous expansion of the scale of power grid construction and the rapid development of power grid informatization, considering the development direction of power grid planning business comprehensively, it is necessary to study and realize the intelligent site selection of substations based on power big data, apply cutting-edge information technology, integrate various professional information resources, and deeply explore the potential value of data. Summary of the invention

[0003] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0004] In view of the above problems existing in the site selection of existing substation projects, the present invention is proposed.

[0005] Therefore, the purpose of the present invention is to provide a method for constructing a knowledge graph for substation project site selection, which is suitable for structured decomposition and knowledge graph construction of substation project site selection knowledge, improving the management efficiency of existing substation site selection knowledge, assisting engineering decision-making, and further improving the intelligence of substation project site selection.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for constructing a knowledge graph for substation project site selection, which comprises the following steps: collecting relevant specifications and standards for substation project site selection, and obtaining key terms of the knowledge graph for substation project site selection through term extraction; referring to the relevant specifications and standards for existing site selection, combining key terms, and obtaining the concept set and relationship set of substation project site selection through screening; establishing "concept-relationship-concept" triples according to the concept set and relationship set, and constructing an ontology layer knowledge graph; collecting the operating data of existing substations, and obtaining the entity set, relationship set and attribute set of substation project site selection knowledge after knowledge extraction and screening; establishing "instance-relationship-instance" and "instance-relationship-attribute" triples through the entity set, relationship set and attribute set, and constructing an instance layer knowledge graph; constructing a knowledge system for substation project site selection based on the ontology layer knowledge graph and the instance layer knowledge graph and the mapping relationship between them.

[0007] As a preferred scheme of the method for constructing a knowledge graph for substation project site selection described in the present invention, the relevant specifications and standards for substation project site selection are collected, and key terms of the knowledge graph for substation project site selection are obtained through term extraction, including the following steps: term extraction, first using a threshold classifier to extract two-word candidate words in the corpus, and then expanding the candidate words to a certain extent on the left and right sides, and screening multi-word candidate words that meet the requirements; using the TF-IDF algorithm to filter the obtained candidate words to obtain the final result, that is, the terms of the knowledge graph for substation project site selection.

[0008] As a preferred solution of the method for constructing the knowledge graph of substation project site selection of the present invention, wherein: the filtering of the obtained candidate words using the TF-IDF algorithm to obtain the terms of the knowledge graph of substation project site selection includes the following steps:

[0009] First, calculate the word frequency tf ij ,Right now:

[0010]

[0011] Among them, n i，j Is the word in file d j The number of occurrences in n k，j The denominator is file d j The sum of the number of occurrences of all words in ;

[0012] Then, calculate the inverse document frequency IDF, that is:

[0013]

[0014] Where, |D| is the total number of documents in the collected substation engineering site selection knowledge, |{j:t i ∈d j}| indicates the total number of documents containing the word (i.e. n i，j ≠0 total number of files).

[0015] Finally, calculate TF-IDF, that is: TF-IDF = TF × IDF

[0016] Among them, TF (tf ij ) represents word frequency and IDF represents inverse document frequency.

[0017] As a preferred scheme of the method for constructing a knowledge graph for substation project site selection described in the present invention, wherein: referring to the relevant specifications and standards for existing site selection, combined with key terms, and screened to obtain a concept set and a relationship set for substation project site selection, it includes the following steps: based on key terms and collected substation site selection related specifications and standards, a concept set of the substation project site selection knowledge graph ontology layer is obtained through screening; based on the obtained concept set, a remote supervision algorithm is used to extract relationships from the collected substation project site selection related specifications and standards, and then a relationship set of the substation project site selection knowledge graph ontology layer is obtained through screening.

[0018] As a preferred solution of the method for constructing the substation project site selection knowledge graph described in the present invention, the following steps are also included in the process of obtaining the ontology layer relationship set of the substation project site selection knowledge graph: performing concept comparison based on the obtained concept set in combination with the knowledge base, and performing relationship annotation on the collected substation site selection related specifications, standards and other materials; performing remote supervision relationship extraction model training based on the acquired knowledge, and using the obtained model to extract relationships to obtain the ontology layer relationship set of the substation project site selection knowledge graph.

[0019] As a preferred solution of the substation project site selection knowledge graph construction method described in the present invention, the concept set and relationship set of the knowledge graph ontology layer are checked and evaluated, including two aspects: one is whether the structure of the core concept is reasonable, and the other is the rationality, necessity and distinction between similar concepts of each concept.

[0020] As a preferred solution of the method for constructing a knowledge graph for substation project site selection according to the present invention, the operation data of existing substations are collected, and after knowledge extraction and screening, an entity set, a relationship set and an attribute set of substation project site selection knowledge are obtained, including the following steps: the collected operation data of existing substations are classified to obtain structured knowledge and unstructured knowledge of substation project site selection; wherein the operation data includes the ledger, operation history and overload of the substation; for the knowledge extraction of structured knowledge, the R2RML language is used to map the knowledge into RDF format triples to obtain corresponding entities, relationships and attributes; for the knowledge extraction of unstructured knowledge, the LSTM-CRF (machine learning-deep learning) model, remote supervision algorithm and MetaPAD (Meta Pattern Discovery from Massive Text) are used in combination with the constructed ontology layer knowledge graph. Corpora, discovers meta-patterns from massive text corpora) to perform entity extraction, relationship extraction and attribute extraction respectively; the entities, relationships and attributes obtained in the first two steps are combined, and after verification and filling in the gaps, the entity set, relationship set and attribute set of the initial substation project site selection knowledge graph are obtained.

[0021] As a preferred scheme of the method for constructing the substation project site selection knowledge graph described in the present invention, the entity set, relationship set and attribute set of the initial substation project site selection knowledge graph obtained are clustered with synonyms and near-synonyms to determine a unique standard expression, and obtain a streamlined initial substation project site selection knowledge graph entity set, relationship set and attribute set.

[0022] As a preferred solution of the method for constructing the substation project site selection knowledge graph described in the present invention, the following steps are included for entity extraction, relationship extraction and attribute extraction: according to the acquired ontology, the collected substation operation data is subjected to entity extraction using the LSTM-CRF model to obtain a substation project site selection knowledge graph entity set; according to the acquired entity, the collected substation operation data is subjected to relationship extraction using the remote supervision model to obtain a substation project site selection knowledge graph instance layer relationship set; according to the acquired entity, the collected substation operation data is subjected to attribute extraction using MetaPAD to obtain a substation project site selection knowledge graph attribute set.

[0023] Beneficial effects of the present invention:

[0024] The present invention constructs a relatively complete knowledge graph system for substation site selection; it can be used as a site selection knowledge guide to provide practical and effective guidance for site selection; by forming a rich case knowledge base, it can be used for new project retrieval and auxiliary decision-making; and it can provide data support and scientific basis for power grid planning decisions, achieve accurate planning, assist relevant technical personnel in making decisions, and achieve the purpose of improving work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0026] Figure 1 A schematic diagram of the knowledge graph construction process of the substation project site selection knowledge graph construction method of the present invention.

[0027] Figure 2 A partial schematic diagram of the ontology layer knowledge graph of the substation project site selection knowledge graph construction method of the present invention.

[0028] Figure 3 It is a partial schematic diagram of the instance layer knowledge graph of the substation project site selection knowledge graph construction method of the present invention.

[0029] Figure 4 This is an example diagram of the attributes of an example of the method for constructing a knowledge graph for substation project site selection according to the present invention.

[0030] Figure 5 This is a schematic diagram of the mapping relationship between the ontology layer and the instance layer of the substation project site selection knowledge graph construction method of the present invention. DETAILED DESCRIPTION

[0031] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0032] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0033] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0034] Secondly, the present invention is described in detail with reference to the schematic diagram. When describing the embodiments of the present invention in detail, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0035] Through in-depth analysis of the source, content, characteristics and application requirements of knowledge in the field of substation project site selection, it is proposed to construct concept sets and relationship sets, establish an ontology-level knowledge graph, and store static knowledge such as technical regulations. Based on the ontology, the knowledge related to substation project site selection is extracted to obtain the entity set, attribute set and relationship set of substation project site selection knowledge, establish an instance-level knowledge graph, and store dynamic knowledge such as cases and expert experience.

[0036] Example 1

[0037] Reference Figures 1 to 5 , which is the first embodiment of the present invention, provides a method for constructing a knowledge graph for substation project site selection, and the graph construction method comprises the following steps:

[0038] S1: Collect relevant specifications and standards for substation project site selection, and obtain key terms of the substation project site selection knowledge graph through term extraction.

[0039] Specifically, the existing vocabulary and terms related to the specifications and standards of substation project site selection are collected to form a corpus and knowledge base, and the key terms directly related to the substation project site selection knowledge graph are extracted from the corpus through term extraction.

[0040] Furthermore, key terms are obtained by combining term extraction with the TF-IDF algorithm, as shown in the following steps:

[0041] S11: Term extraction, first use the threshold classifier to extract the two-word candidate words in the corpus, and then expand the candidate words to the left and right to a certain extent to select the multi-word candidate words that meet the requirements.

[0042] S12: The obtained candidate words are then filtered through the TF-IDF algorithm to obtain the final result, i.e., the key terms of the substation project site selection knowledge graph.

[0043] Furthermore, when the TF-IDF algorithm is used, first, the word frequency TF (tf ij ),Right now:

[0044]

[0045] Term frequency (TF) is equal to the number of times a word appears in a text / the total number of words in the text, where ni，j is the number of times the word appears in document d j , and n k，j The denominator is the total number of times all words appear in document d j .

[0046] Then, calculate the inverse document frequency IDF, that is:

[0047]

[0048] The inverse document frequency (IDF) is equal to log(total number of documents in the corpus / number of documents containing the word + 1). The most common words ("of", "is", "in", "China", etc.) are given the smallest weights, and the less common words ("substation", "a certain power company") are given larger weights. Among them, |D| is the total number of documents in the collected knowledge of substation project site selection, and |{j: t i ∈d j}| represents the total number of documents containing the word (i.e., the total number of documents where n i，j ≠0). If the word is not in the corpus, it will cause the denominator to be zero. Therefore, generally, 1 + |{j: t i ∈d j}| is used.

[0049] Finally, calculate TF-IDF, that is:

[0050] TF-IDF = TF x IDF

[0051] Among them, TF (tf ij ) represents the term frequency, and IDF represents the inverse document frequency.

[0052] The two values of "term frequency (TF)" and "inverse document frequency (IDF)" are multiplied to obtain the TF-IDF value of a certain word. The higher the importance of a certain word, the larger its TF-IDF value. That is, generally speaking, the importance of this word in this article will be higher. Therefore, by calculating the TF-IDF of each word in the document and sorting them from largest to smallest, the first few words are the keywords of the article, thus achieving the purpose of filtering the vocabulary. Using the TF-IDF algorithm to filter the key terms is simple, fast, and easy to understand, and the accuracy rate of term extraction has increased by nearly 78%.

[0053] S2: Refer to the relevant specifications and standards of the existing site selection, and combine with the key terms. After screening, obtain the concept set and relationship set of the substation project site selection.

[0054] Specifically, according to the existing relevant specifications and standards for substation project site selection, combined with the extracted key terms, through screening by professionals (experts) in the field of substation site selection technology, a concept set and relationship set of site selection related terms are obtained.

[0055] Specifically, the following steps are included:

[0056] S21: Based on the key terms and the collected substation site selection related specifications and standards, the concept set of the substation project site selection knowledge graph ontology layer is obtained through screening;

[0057] S22: Based on the acquired concept set, a remote supervision algorithm is used to extract relationships from the collected substation project site selection related specifications and standards, and then the substation project site selection knowledge graph ontology layer relationship set is obtained through screening.

[0058] Furthermore, in the process of obtaining the ontology layer relationship set of the substation project site selection knowledge graph, the following processing was performed:

[0059] S221: Perform concept comparison based on the obtained concept set and the knowledge base, and perform relationship annotation on the collected substation site selection related specifications, standards and other information;

[0060] S222: Perform remote supervision relationship extraction model training based on the knowledge acquired in the previous step, use the obtained model to extract relationships, and obtain a set of relationships at the ontology level of the knowledge graph for substation project site selection;

[0061] S223: Check and evaluate the concept set and relationship set at the ontology level of the knowledge graph, including two aspects: one is whether the structure of the core concepts is reasonable, and the other is the rationality and necessity of each concept and the distinction between similar concepts.

[0062] S3: Figure 2 As shown in , according to the concept set and relationship set obtained in step S2, a "concept-relationship-concept" triple is established to construct an ontology layer knowledge graph. That is, two or more concept terms are matched through a mapping relationship to form an ontology layer knowledge graph.

[0063] Further, S4: collect the operation data of existing substations, and obtain the entity set, relationship set and attribute set of substation project site selection knowledge after knowledge extraction and screening; wherein, it is necessary to collect a large amount of existing site selection and operation data of substations in operation, and construct the knowledge graph instance layer set of the instance layer after knowledge extraction, specifically, including the following steps:

[0064] S41: Classify the collected operation data of existing substations to obtain structured knowledge and unstructured knowledge for substation project site selection. The operation data includes the substation ledger, operation history, and overload conditions;

[0065] S42: For knowledge extraction of structured knowledge, the R2RML language is used to map the knowledge into RDF format triples to obtain the corresponding entities, relationships and attributes;

[0066] S43: For knowledge extraction of unstructured knowledge, combined with the ontology layer knowledge graph constructed in step S3, the LSTM-CRF model, remote supervision algorithm and MetaPAD are used to perform entity extraction, relationship extraction and attribute extraction respectively.

[0067] For entity extraction, relationship extraction, and attribute extraction, the following processing steps are taken:

[0068] S431: Using the LSTM-CRF model to extract entities from the collected substation operation data according to the acquired ontology, and obtaining a substation project site selection knowledge graph entity set;

[0069] S432: extracting relationships from the collected substation operation data using a remote supervision model based on the acquired entities, and obtaining a substation project site selection knowledge graph instance layer relationship set;

[0070] S433: MetaPAD is used to extract attributes of the collected substation operation data according to the acquired entities to obtain a substation project site selection knowledge graph attribute set.

[0071] It should be noted that MetaPAD is a context-aware segmentation method that carefully determines the boundaries of patterns by learning a pattern quality assessment function; identifies and groups synonymous meta-patterns from multiple aspects, including their types, contexts, and extractions; examines the type distribution of entities in the instances extracted by each pattern group, and seeks the appropriate type level to make the discovered patterns precise.

[0072] In this embodiment, the type of each word is marked according to the obtained entity set, and MetaPAD is used to extract patterns (samples) from massive data, and the pattern frequency is counted. Through TrueIE, high-quality sentence patterns that describe entity attributes are screened out, and the words in them are extracted as candidate attribute words. Then, a dictionary is generated for all candidate attribute words, and the corpus just now is segmented again, and all attribute word tags are classified as "Attribute". A second pattern extraction is performed, and only patterns related to the "Attribute" type are taken, and the text in them is screened out as the attribute value. For attributes whose attribute values ​​are sentences, the relevant sentences are directly intercepted as the answer.

[0073] Further, S44: the entities, relationships, and attributes obtained in step S42 and step S43 are combined, and after verification and supplementation, the entity set, relationship set, and attribute set of the initial substation project site selection knowledge graph are obtained.

[0074] S45: Synonym and near-synonym clustering is performed on the entity set, relationship set and attribute set of the obtained initial substation project site selection knowledge graph, and a unique standard expression is determined to obtain a streamlined initial substation project site selection knowledge graph entity set, relationship set and attribute set.

[0075] S5: Figure 3 and Figure 4 As shown in , "instance-relationship-instance" and "instance-relationship-attribute" triplets are established through the acquired entity set, relationship set, and attribute set to build an instance-level knowledge graph; it is used to store dynamic knowledge such as cases and expert experience.

[0076] S6: Figure 5 As shown in , based on the ontology layer knowledge graph and the instance layer knowledge graph and the mapping relationship between them, a knowledge system for substation project site selection is constructed to help retrieval learning and auxiliary decision-making in substation project site selection.

[0077] Through the above steps, a complete substation site selection knowledge graph system can be constructed, which can be used as a site selection knowledge guide to provide practical and effective guidance for site selection; by forming a rich case knowledge base, it can be used for new project retrieval and auxiliary decision-making; and it can provide data support and scientific basis for power grid planning decisions, realize accurate planning, assist relevant technical personnel in making decisions, and achieve the purpose of improving work efficiency.

[0078] Example 2

[0079] In order to verify and illustrate the technical effect of this method, this embodiment is verified and illustrated by a specific example. Figures 2 to 4 as shown in .

[0080] 1. By sorting and screening the collected substation site selection related specifications and standards, a corpus is formed, and the following Figure 2 A local example of the ontology layer knowledge graph in . Then perform term extraction on the corpus.

[0081] Regarding term extraction: Initially, term extraction was mainly done manually, which was a huge workload and slow progress. Secondly, some automatic term extractors have been implemented based on statistical methods. Among them, the most representative ones are based on statistical corpora and are language-independent, that is, they can extract corpora in different languages ​​(such as Chinese and English). However, they rely solely on statistical methods to process corpora without considering or less considering the use of grammatical rules. The accuracy of terms obtained by this method is relatively low.

[0082] In this scheme, candidate words are extracted through a threshold classifier, and bidirectional expansion and screening of the words are performed to obtain multi-word candidate words, and then the obtained candidate words are filtered through the TF-IDF algorithm to obtain the final key terms.

[0083] According to the algorithm steps in the above embodiment, a term extraction experiment was conducted on the corpus in the field of substation engineering site selection, and then compared with the manual extraction results, the accuracy rate was 64.49%. After analyzing the extraction results, it was found that there were a large number of "correct" terms in the extraction results, but they were not the terms in the field of substation engineering site selection that we wanted, which had the greatest impact on the accuracy rate. Then we used the TF-IDF algorithm to filter the terms, and then compared with the manual extraction results, the accuracy rate was further improved to 78%.

[0084] 2. Classify and correspond the large amount of substation records, operation history, overload conditions and other information collected to form a knowledge base, and obtain Figure 3 and 4 A local example of the instance-level knowledge graph in . Extract knowledge from instances in the knowledge base.

[0085] For knowledge extraction: a method combining machine learning model and deep learning (LSTM-CRF) is used. In order to measure the recognition effect of the algorithm on entities and the generalization ability of the model, this method uses the commonly used indicators of information classification: recall rate (Recall, denoted as Rec); precision (precision, denoted as Pre); F1-Measure (comprehensive evaluation indicator, denoted as F1). Among them, the larger the Pre and Rec are at the same time, the better the effect of the recognition model. F1 is a comprehensive indicator of Pre and REC. Using Pre, Rec and F1 at the same time can evaluate the model more objectively and reasonably.

[0086] In order to test the LSTM-CRF method's ability to identify entities in the field of substation project site selection, the following comparative experiments were conducted: LSTM-CRF, LSTM, CRF, HMM four methods of entity recognition comparative experiments, respectively comparing the Pre, Rec and F1 values ​​of the four methods. The experimental results show that HMM has the worst recognition effect, and the three indicators of HMM are all about 80%. The recognition effects of CRF and LSTM are better than HMM. LSTM-CRF has the best recognition effect, and the three indicators are all 93%, which is better than the other three methods. This experiment proves that LSTM-CRF, a combination of LSTM and CRF, is more suitable for the task of entity recognition in the field of substation project site selection.

[0087] 3. Further, we extract terms from the existing substation site selection related specifications and standards, and extract knowledge from a large number of substation records, operation history, overload conditions and other data. The extracted data results obtained at this stage are shown in the following table:

[0088] type Total number concept 21 Examples 32234 property 194232 Concepts and the relationships between them 21 Relationship between concepts and instances 32234 Relationships between instances and attributes 194232

[0089] This extracted data contains 21 concepts, derived from 32,234 instances, which have a total of 194,232 attributes. The above data constitutes a complete knowledge graph for substation site selection. Furthermore, the above data can be added or deleted on the existing basis; and the more instances the knowledge graph stores, the higher the accuracy of its guidance on site selection decisions (that is, the greater the value of the knowledge graph).

[0090] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention 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 invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for constructing a knowledge graph for substation project site selection, characterized in that: The following steps are included: Collect relevant specifications and standards for substation project site selection, and extract key terms from the knowledge graph of substation project site selection through term extraction; By referring to the existing relevant specifications and standards for site selection and combining key terms, the concept set and relationship set for substation project site selection are obtained through screening; It includes the following steps: Based on the key terms and the collected substation site selection related specifications and standards, the concept set of the substation project site selection knowledge graph ontology layer is obtained through screening; According to the acquired concept set, the remote supervision algorithm is used to extract the relationship between the collected substation project site selection related specifications and standards, and then the substation project site selection knowledge graph ontology layer relationship set is obtained through screening; In the process of obtaining the ontology layer relationship set of the knowledge graph for substation project site selection, the following steps are also included: Based on the obtained concept set and combined with the knowledge base, the concepts are compared and the relationship between the collected substation site selection related specifications and standard data is marked; The remote supervision relationship extraction model is trained based on the acquired knowledge, and the obtained model is used to extract relationships to obtain the ontology layer relationship set of the substation project site selection knowledge graph; Establish "concept-relationship-concept" triples based on concept sets and relationship sets, and construct the ontology layer knowledge graph; Collect the operation data of existing substations, and obtain the entity set, relationship set and attribute set of substation project site selection knowledge after knowledge extraction and screening; it includes the following steps: Classify the collected operating data of existing substations to obtain structured and unstructured knowledge for substation project site selection; the operating data includes the substation ledger, operating history, and overload conditions; For knowledge extraction of structured knowledge, R2RML language is used to map knowledge into RDF format triples to obtain corresponding entities, relations and attributes; For knowledge extraction of unstructured knowledge, the LSTM-CRF model, remote supervision algorithm and MetaPAD are used to extract entities, relations and attributes respectively, in combination with the constructed ontology-level knowledge graph; The entities, relationships, and attributes obtained in the first two steps are combined, and after verification and supplementation, the entity set, relationship set, and attribute set of the initial substation project site selection knowledge graph are obtained; For entity extraction, relationship extraction, and attribute extraction, take the following steps: According to the acquired ontology, the LSTM-CRF model is used to extract entities from the collected substation operation data to obtain the entity set of the substation project site selection knowledge graph; According to the acquired entities, the remote supervision model is used to extract the relationship of the collected substation operation data, and the instance layer relationship set of the substation project site selection knowledge graph is obtained; According to the acquired entities, MetaPAD is used to extract attributes from the collected substation operation data to obtain the attribute set of the substation project site selection knowledge graph; Establish "instance-relationship-instance" and "instance-relationship-attribute" triples through entity sets, relationship sets, and attribute sets to build instance-level knowledge graphs; Based on the ontology layer knowledge graph and the instance layer knowledge graph and the mapping relationship between them, a knowledge system for substation project site selection is constructed.

2. The method for constructing a knowledge graph for substation project site selection according to claim 1, characterized in that: The collecting of relevant specifications and standards for substation project site selection and obtaining key terms of the substation project site selection knowledge graph through term extraction include the following steps: Term extraction: First, use the threshold classifier to extract two-word candidate words in the corpus, and then expand the candidate words to the left and right sides to select multi-word candidate words that meet the requirements; The obtained candidate words are filtered using the TF-IDF algorithm to obtain the final result, which is the key terms of the knowledge graph of substation project site selection.

3. The method for constructing a knowledge graph for substation project site selection according to claim 2 is characterized in that: The method of filtering the obtained candidate words using the TF-IDF algorithm to obtain key terms of the substation project site selection knowledge graph includes the following steps: First, calculate the word frequency ,Right now: ;in, Is the word in the file The number of times it appears in The denominator is the file The sum of the number of occurrences of all words in ; Then, calculate the inverse document frequency IDF, that is: ;in, is the total number of documents in the collected knowledge of substation project site selection, Represents the total number of documents containing the term, i.e. Total number of files; Finally, calculate TF-IDF, which is: ; Among them, TF ( ) is represented by word frequency and IDF is represented by inverse document frequency.

4. The method for constructing a knowledge graph for substation project site selection according to claim 3 is characterized in that: The concept set and relationship set at the ontology level of the knowledge graph are checked and evaluated, including two aspects: one is the rationality of the core concept structure, and the other is the rationality, necessity and distinction between similar concepts of each concept.

5. The method for constructing a knowledge graph for substation project site selection according to claim 4 is characterized in that: The entity set, relationship set and attribute set of the obtained initial substation project site selection knowledge graph are clustered by synonyms and near-synonyms to determine the unique standard expression, and the streamlined initial substation project site selection knowledge graph entity set, relationship set and attribute set are obtained.

Images