Power grid enterprise supply chain procurement standard digitization system based on knowledge graph

Abstract

The invention discloses a power grid enterprise supply chain purchase standard digitization system based on a knowledge graph, and relates to the technical field of digitization. A data processing layer; a knowledge extraction layer; a knowledge model layer; a knowledge management layer; a knowledge application layer; according to the invention, based on related technologies in the field of knowledge graph construction, the supply chain purchase standard knowledge graph is constructed to realize the linkage relationship of parameters between the purchase standard and the supply chain standard, so that an offline supply chain purchase standard document which cannot be directly understood by a machine is efficiently converted into a digital knowledge system; technical support is provided for supply chain purchase standard digital application, the supply chain purchase standard query time is saved, and the accuracy of supply chain purchase standard application is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field related to digital systems, and specifically to a knowledge graph-based digital system for supply chain procurement standards of power grid enterprises. Background Art

[0002] Faced with the complex and ever-changing big data environment, the knowledge graph, an emerging cognitive method developed from artificial intelligence technology, has been initially applied in the power field. More and more researchers are applying knowledge graphs to power grid dispatching, power equipment operation and maintenance, power marketing and other power fields. They build power knowledge graphs based on graph databases, merge scattered data sources, and describe the complex relationships between concepts and entities in the power system in a structured form.

[0003] The current power grid has the following main problems in standard application:

[0004] Data comes from many sources, with varying quality, making it difficult to update;

[0005] There are conflicts between different standard clauses and indicators for the same equipment, which are difficult to detect;

[0006] The feedback cycle of enterprise standard application is long and cannot be revised in time;

[0007] The means of utilizing standards are single and inefficient, and they fail to play their due value. Summary of the invention

[0008] In order to solve the defects of the prior art, the present invention provides a digital system for power grid enterprise supply chain procurement standards based on knowledge graph.

[0009] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0010] The present invention provides a knowledge graph-based digital system for procurement standards of power grid enterprises' supply chain, including:

[0011] Data layer: used to collect structured, semi-structured and unstructured data involved in procurement standards;

[0012] Data processing layer: used to pre-process structured data and unstructured data using different methods, and to label entities and relationships;

[0013] Knowledge extraction layer: The knowledge extraction layer is the core of building the knowledge graph. It uses database-based, rule-based, and deep learning methods to extract entities and relationships for different types of input labeled data.

[0014] Knowledge model layer: This layer uses the k-means clustering algorithm and cosine similarity algorithm to integrate the identified entities and relationships;

[0015] Knowledge management layer: manages the constructed control knowledge graph. Professionals review and verify the constructed knowledge graph according to the dispatching regulations, work instructions, and operating procedures to ensure that it meets the actual application requirements;

[0016] Knowledge application layer: The knowledge graph constructed by the application supports application scenarios such as procurement material catalog retrieval, standard material recommendation, structured preparation of technical specifications, comparison and recommendation of technical standard indicators, and intelligent preparation of procurement documents.

[0017] As a preferred technical solution of the present invention, the goal of the k-means clustering algorithm is to minimize the sum of the distances from all data points to their cluster centers;

[0018] Let x i is the ith data point, μ c(i) is x i The center of the cluster c(i), the objective function

[0019]

[0020] Where n is the total number of data points.

[0021] As a preferred technical solution of the present invention, the data layer is based on the digital application requirements of supply chain procurement standards, with national standards, industry standards, group standards, company enterprise standards, and general design, general equipment, and anti-accident measures standardization construction results, and the procurement standards include general and special parts.

[0022] As a preferred technical solution of the present invention, the general part can specify the general technical requirements of the product, product technical interfaces, quality inspection and testing methods, packaging and transportation; the special part specifies the product-specific technical requirements, necessary technical parameters and engineering environment usage conditions. Structured data can be used directly, and semi-structured and unstructured data need to be sorted in the data processing layer to complete the construction of the knowledge graph data layer.

[0023] As a preferred technical solution of the present invention, the data processing layer uses different methods to preprocess the supply chain procurement standard structured data and unstructured data, and divides the procurement standard data including the general part and the special part according to characteristic indicators, and distinguishes them into training samples and test samples.

[0024] As a preferred technical solution of the present invention, the entity and relationship marking can be carried out by manual marking in accordance with the current standard BIOES specification and data enhancement tool marking. The enhancement tool can be implemented by engineering-based synonym conversion and K-nearest neighbor classification, and the data processing layer of the procurement standard knowledge graph is constructed by preprocessing the data and marking the entities and relationships.

[0025] As a preferred technical solution of the present invention, based on the database knowledge extraction, the structured data is converted into entities and relationships through extraction-translation-loading;

[0026] Knowledge extraction based on the rules is implemented using Chinese word segmentation tools and regular expressions;

[0027] The knowledge extraction based on the deep learning uses Chinese word segmentation and Word2Vec to convert text into word vectors, takes the word vectors as input, and uses a bidirectional long short-term memory network + conditional random field to identify relationship labels.

[0028] As a preferred technical solution of the present invention, typical scenarios are determined based on the types and characteristics of technical parameters in the procurement standards, such as general technical requirements, product technical interfaces, quality inspection and testing methods, packaging and transportation, special technical requirements, necessary technical parameters and engineering environment use conditions;

[0029] The types of technical parameters and characteristic values ​​in the knowledge graph data layer are organized into a strategy set. The Jieba word segmentation tool is used to segment the technical parameter text. The accuracy of word segmentation is increased by establishing a professional dictionary of technical parameters. The dictionary words all have the default maximum word frequency, and regular expressions are used to identify the entities and relationships of the technical parameter knowledge graph.

[0030] As a preferred technical solution of the present invention, the identified entities and relationships are aligned, linked, and disambiguated, and entity links are performed through manual maintenance, same character rule matching, and k-means clustering to form a complete technical parameter knowledge graph;

[0031] Store the knowledge graph and form a procurement standard technical parameter association library to provide a data basis for the next step of multi-level label generation and application.

[0032] The beneficial effects of the present invention are:

[0033] This knowledge graph-based digital system for supply chain procurement standards of power grid enterprises builds field-related technologies based on knowledge graphs. By building a knowledge graph of supply chain procurement standards to realize the linkage relationship between procurement standards and supply chain standards, it efficiently converts offline supply chain procurement standard documents that cannot be directly understood by machines into a digital knowledge system, provides technical support for the digital application of supply chain procurement standards, saves supply chain procurement standard query time, greatly improves the accuracy of supply chain procurement standard application, overcomes the limitations of performance indicators such as recognition accuracy and running time, and improves the depth and breadth of digital application of supply chain procurement standards in business systems; by building a knowledge graph and application of supply chain procurement standards, it realizes the comprehensive digitalization of procurement standard query, compilation, management and application, carries out the practice of digital transformation of supply chain procurement standards, and significantly improves the efficiency and intelligence level of supply chain management. These innovative practices have promoted the transformation and upgrading of power grid enterprises. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0035] Figure 1 It is a diagram of a supply chain procurement standard knowledge graph construction method of a power grid enterprise supply chain procurement standard digital system based on a knowledge graph of the present invention;

[0036] Figure 2 It is a flow chart for constructing a supply chain procurement standard knowledge graph of a power grid enterprise supply chain procurement standard digitization system based on a knowledge graph in the present invention. DETAILED DESCRIPTION

[0037] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0038] Example: Figure 1 and Figure 2 As shown, the present invention provides a knowledge graph-based digital system for supply chain procurement standards of power grid enterprises, including:

[0039] Data layer: used to collect structured, semi-structured and unstructured data involved in procurement standards;

[0040] Data processing layer: used to pre-process structured data and unstructured data using different methods, and to label entities and relationships;

[0041] Knowledge extraction layer: The knowledge extraction layer is the core of building the knowledge graph. It uses database-based, rule-based, and deep learning methods to extract entities and relationships for different types of input labeled data.

[0042] Knowledge model layer: This layer uses the k-means clustering algorithm and cosine similarity algorithm to integrate the identified entities and relationships;

[0043] Knowledge management layer: manages the constructed control knowledge graph. Professionals review and verify the constructed knowledge graph according to the dispatching regulations, work instructions, and operating procedures to ensure that it meets the actual application requirements;

[0044] Knowledge application layer: The knowledge graph constructed by the application supports application scenarios such as procurement material catalog retrieval, standard material recommendation, structured preparation of technical specifications, comparison and recommendation of technical standard indicators, and intelligent preparation of procurement documents.

[0045] Among them, the goal of the k-means clustering algorithm is to minimize the sum of the distances from all data points to the center of their clusters;

[0046] Let x i is the ith data point, μ c(i) is x i The center of the cluster c(i), the objective function

[0047]

[0048] Where n is the total number of data points;

[0049] For example, suppose there are 4 data points x1, x2, x3, and x4. After clustering, x1 and x2 belong to cluster C1, whose center is μ1, and x3 and x4 belong to cluster C2, whose center is μ2.

[0050] J=d(x1,μ1) 2 +d(x2,μ1) 2 +d(x3,μ2) 2 +d(x3,μ2) 2

[0051] The algorithm will continue to iterate, making the value of continue to decrease until the stopping condition is met (such as the cluster center no longer changes or the maximum number of iterations is reached).

[0052] The data layer is based on the digital application requirements of supply chain procurement standards, and includes general and special parts of procurement standards based on national standards, industry standards, group standards, company standards, and the results of standardization construction of general design, general equipment, and anti-accident measures;

[0053] The supply chain procurement process involves supplier evaluation, risk assessment, procurement cost calculation, procurement volume decision, and procurement volume based on demand forecast;

[0054] The comprehensive scoring formula for supplier evaluation is:

[0055]

[0056] Where S is the supplier's comprehensive score, ω i is the weight of the i-th evaluation indicator (e.g., product quality weight ω1, delivery punctuality weight ω2, price rationality weight ω3), and

[0057]

[0058] x i is the supplier's score on the i-th evaluation indicator;

[0059] The risk assessment formula is:

[0060]

[0061] Among them, R f is the financial risk assessment value, L is the supplier's liabilities, A is the supplier's assets, D is the supplier's current liabilities, E is the supplier's owner's equity, k1 and k2 are weight coefficients set based on industry experience, which are used to adjust the impact of liabilities and equity related indicators on risk assessment;

[0062] The formula for calculating the total purchase cost is:

[0063] C t =C p +C s +C h +C o

[0064] Among them, C t is the total purchase cost, C p is the product purchase price (calculated based on the purchase contract unit price and purchase quantity), C s is the transportation cost (the freight is determined by factors such as transportation method, transportation distance, cargo weight and volume, etc.), C h is the storage cost (calculated based on storage time, storage space occupied, etc.), C o are other costs (such as procurement fees and insurance premiums);

[0065] Economic purchase quantity decision formula:

[0066]

[0067] Among them, D is the total annual demand, S is the fixed cost of each purchase (such as purchase order processing fees), and H is the annual holding cost per unit of goods (including storage costs and capital occupation costs). By calculating EOQ, the purchase quantity for each purchase can be determined under the optimal cost condition.

[0068] Procurement volume formula based on demand forecast:

[0069]

[0070] Among them, Q t is the purchase quantity at time, D t is the demand forecast value of time (obtained through demand forecasting models such as time series analysis and regression analysis), Z is the safety factor related to the service level (determined according to the acceptable out-of-stock risk of the enterprise), and σt is the standard deviation of the demand forecast (reflecting the uncertainty of demand)

[0071]

[0072] It is the total amount of inventory that has been purchased but not yet consumed.

[0073] Furthermore, the general part may specify the general technical requirements of the product, product technical interfaces, quality inspection and testing methods, packaging and transportation; the special part may specify the product-specific technical requirements, necessary technical parameters and engineering environment usage conditions. Structured data can be used directly, and semi-structured and unstructured data need to be sorted in the data processing layer to complete the construction of the knowledge graph data layer.

[0074] The data processing layer uses different methods to pre-process the structured data and unstructured data of the supply chain procurement standards, and divides the procurement standards, including the general part and the special part data, into training samples and test samples according to characteristic indicators;

[0075] The most common data standard structured processing is Z-score standardization. For the data column χ = [χ1, χ2, ..., χ n ], standardized χ 1 The calculation formula is:

[0076]

[0077] Where u is the mean of the x data column, that is

[0078]

[0079] σ is the standard deviation

[0080]

[0081] There is also minimum to maximum normalization, the formula is

[0082]

[0083] where χ min and χ max are the minimum and maximum values ​​in the data column respectively;

[0084] Specifically, unstructured data preprocessing includes:

[0085] Text cleaning, remove special characters, punctuation marks and extra spaces in the text. For example, for text strings, you can use regular expressions to remove non-alphanumeric characters. For example, in Python, you can use the re module: import re; t = re.sub('[^a-zA-Z0-9]', ", t), replace non-alphanumeric characters with spaces, and then use the split function to remove extra spaces;

[0086] Lexical analysis: for Chinese text, word segmentation is required. For example, use the jieba word segmentation tool, import jieba; words = jieba.cut (text), where text is the Chinese text to be segmented, and words is the result after word segmentation (an iterable object).

[0087] Among them, the entity and relationship marking can be carried out by manual marking according to the current standard BIOES specifications and data enhancement tool marking. The enhancement tool can be implemented by engineering-based synonym conversion and K-nearest Neighbors (KNN) classification, which is relatively easy to implement. By preprocessing the data and marking the entities and relationships, the data processing layer of the procurement standard knowledge graph is constructed.

[0088] Wherein, based on the database knowledge extraction, the structured data is converted into entities and relationships through extraction-translation-loading (ETL);

[0089] Knowledge extraction based on the rules is implemented using Chinese word segmentation tools and regular expressions;

[0090] The knowledge extraction based on the deep learning uses Chinese word segmentation and Word2Vec to convert text into word vectors, takes the word vectors as input, uses Bilateral Long-Short Term Memory-Conditional Random Field (BiLSTM-CRF) to identify entity labels, and uses Convolutional Neural Network (CNN) to identify relationship labels;

[0091] Chinese word segmentation converts text into word vectors using the One-Hot encoding algorithm formula

[0092] First, we need to build a vocabulary. Assume that the vocabulary size is N, which contains all the words ω1, ω2, …, ω that appear in the text. N , for any word ω in the text i , the dimension of its One-Hot vector v is N, and the calculation formula of the vector is as follows:

[0093]

[0094] Furthermore, typical scenarios are determined based on the types and characteristics of technical parameters in the procurement standards, such as general technical requirements, product technical interfaces, quality inspection and testing methods, packaging and transportation, special technical requirements, necessary technical parameters and engineering environment use conditions;

[0095] The types and characteristic values ​​of technical parameters in the knowledge graph data layer are organized into a strategy set. The Jieba word segmentation tool is used to segment the technical parameter text. The accuracy of word segmentation is increased by establishing a professional dictionary of technical parameters. The dictionary vocabulary is the default maximum word frequency. Then, regular expressions are used to identify entities and relationships in the technical parameter knowledge graph. The formula for regular expressions to identify product name entities is:

[0096] p product =r i [α-zA-Z0-9 _ ]3,20 1

[0097] [a-zA-Z0-9_] means matching any letter (uppercase or lowercase), number, underscore, or hyphen. {3, 20} means the preceding character combination must appear at least 3 times and at most 20 times.

[0098] Specifically, the identified entities and relationships are aligned, linked, and disambiguated, and entity links are performed through manual maintenance, same character rule matching, and k-means clustering to form a complete technical parameter knowledge graph;

[0099] Store the knowledge graph and form a procurement standard technical parameter association library to provide a data basis for the next step of multi-level label generation and application.

[0100] It should also be noted that the knowledge graph of the present invention is constructed in two ways: top-down and bottom-up. The top-down method extracts information from structured data sources and is suitable for building a basic knowledge base. The bottom-up method extracts information from public data sets through technical means to achieve knowledge extraction, construction, optimization and enhancement, forming a comprehensive and multi-dimensional knowledge graph. The knowledge graph construction flow chart is as follows: Figure 2 As shown;

[0101] In the initial stage of building the knowledge graph of supply chain procurement standards, the first task is to clearly define the data source, covering national standards, industry standards, group standards, company standards, as well as standardization results such as universal design, universal equipment, and anti-accident measures. It can be connected to the structured supply chain professional digital standard library, and can also import semi-structured and unstructured data. Through the system database knowledge extraction model, term extraction and standardization are carried out, and the linkage relationship between the parameters of procurement standards and supply chain standards is established. Through knowledge reasoning and fusion, the comprehensive digitalization of procurement standard query, compilation, management and application is realized.

[0102] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A knowledge graph-based digital system for power grid enterprise supply chain procurement standards, characterized by: Included are: Data layer: used to collect structured, semi-structured and unstructured data involved in procurement standards; Data processing layer: used to pre-process structured data and unstructured data using different methods and mark entities and relationships; Knowledge extraction layer: The knowledge extraction layer is the core of building the knowledge graph. It uses database-based, rule-based, and deep learning methods to extract entities and relationships for different types of input labeled data. Knowledge model layer: This layer uses the k-means clustering algorithm and cosine similarity algorithm to integrate the knowledge of the identified entities and relationships; Knowledge management layer: manages the constructed control knowledge graph. Professionals review and verify the constructed knowledge graph according to the dispatching regulations, work instructions, and operating procedures to ensure that it meets the actual application requirements; Knowledge application layer: The knowledge graph constructed by the application supports application scenarios such as procurement material catalog retrieval, standard material recommendation, structured preparation of technical specifications, comparison and recommendation of technical standard indicators, and intelligent preparation of procurement documents.

2. According to the knowledge graph-based digital system for supply chain procurement standards of power grid enterprises according to claim 1, it is characterized in that: The goal of the k-means clustering algorithm is to minimize the sum of the distances from all data points to the centers of their clusters. Let x i is the ith data point, μ c(i) is x i The center of the cluster c(i), the objective function Where n is the total number of data points.

3. According to the knowledge graph-based digital system for supply chain procurement standards of power grid enterprises according to claim 1, it is characterized in that: The data layer is based on the digital application requirements of supply chain procurement standards, and uses national standards, industry standards, group standards, company enterprise standards, and the results of standardization construction of general design, general equipment, and anti-accident measures to include general and special parts in the procurement standards.

4. According to claim 3, a knowledge graph-based digital system for supply chain procurement standards of power grid enterprises is characterized in that: The general part may specify the general technical requirements of the product, product technical interfaces, quality inspection and testing methods, packaging and transportation; the special part may specify the special technical requirements of the product, necessary technical parameters and engineering environment usage conditions. Structured data can be used directly, and semi-structured and unstructured data need to be sorted in the data processing layer to complete the construction of the knowledge graph data layer.

5. According to the knowledge graph-based digital system for supply chain procurement standards of power grid enterprises according to claim 1, it is characterized in that: The data processing layer uses different methods to pre-process the structured data and unstructured data of the supply chain procurement standards, and divides the procurement standards including the general part and the special part data according to feature indicators, and distinguishes them into training samples and test samples.

6. According to claim 5, a knowledge graph-based digital system for supply chain procurement standards of power grid enterprises is characterized in that: For entity and relationship labeling, the current standard BIOES specification manual labeling and data enhancement tool labeling can be adopted. The enhancement tool can be implemented by engineering-based synonym conversion and K-nearest neighbor classification, which are relatively easy to implement. By preprocessing the data and labeling the entities and relationships, the data processing layer of the procurement standard knowledge graph is constructed.

7. According to claim 1, a knowledge graph-based digital system for supply chain procurement standards of power grid enterprises is characterized in that: Extracting knowledge from the database and converting structured data into entities and relationships through extraction-translation-loading; Knowledge extraction based on the rules is implemented using Chinese word segmentation tools and regular expressions; The knowledge extraction based on the deep learning uses Chinese word segmentation and Word2Vec to convert text into word vectors, takes the word vectors as input, uses a bidirectional long short-term memory network + conditional random field to identify entity labels, and uses a convolutional neural network to identify relationship labels.

8. According to the knowledge graph-based digital system for power grid enterprise supply chain procurement standards in claim 1, it is characterized by: Determine typical scenarios based on the types and characteristics of technical parameters in the procurement standards, such as general technical requirements, product technical interfaces, quality inspection and testing methods, packaging and transportation, special technical requirements, necessary technical parameters and engineering environment use conditions; The types of technical parameters and characteristic values ​​in the knowledge graph data layer are organized into a strategy set. The Jieba word segmentation tool is used to segment the technical parameter text. The accuracy of word segmentation is increased by establishing a professional dictionary of technical parameters. The dictionary words all have the default maximum word frequency, and regular expressions are used to identify the entities and relationships of the technical parameter knowledge graph.

9. According to claim 8, a knowledge graph-based digital system for supply chain procurement standards of power grid enterprises, characterized in that: Align, link, and disambiguate the identified entities and relationships, and link entities through manual maintenance, same character rule matching, and k-means clustering to form a complete technical parameter knowledge graph; Store the knowledge graph and form a procurement standard technical parameter association library to provide a data basis for the next step of multi-level label generation and application.

Images