A Precise Retrieval and Intelligent Question Answering System and Method for Electrical Knowledge Base Based on AI Large Model and Knowledge Graph

Abstract

This invention discloses a precise retrieval and intelligent question-answering system and method for electrical knowledge base based on AI large-scale models and knowledge graphs, belonging to the field of data processing technology. This invention collects and processes historical electrical knowledge data, constructs an electrical knowledge recognition model using natural language processing, and identifies the processed historical electrical knowledge data. Simultaneously, it uses data association analysis to associate the entity information of the identified historical electrical knowledge data and outputs the data association results. After obtaining the data association results, it analyzes and matches the entity information of the identified historical electrical knowledge data with the corresponding output data association results. Furthermore, it fuses the analyzed and matched historical electrical knowledge data using knowledge fusion methods. Finally, it constructs a knowledge graph based on the fused historical electrical knowledge data and performs intelligent retrieval and question answering through semantic recognition, thereby improving the accuracy of electrical knowledge base retrieval.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a precise retrieval and intelligent question-answering system and method for electrical knowledge base based on AI large models and knowledge graphs. Background Technology

[0002] Currently, the management and retrieval of power equipment information mainly rely on manual or simple keyword search methods, which are difficult to meet the query needs of large-scale, highly complex datasets and suffer from problems such as inaccurate queries and low efficiency.

[0003] Existing technologies, such as the invention patent application with publication number CN119597929A, disclose a novel intelligent retrieval method and system for data-driven architecture knowledge in power system construction. The method includes: collecting various data types of assets related to the enterprise's digital architecture; tagging the collected data; using RAG models and graph databases to perform data fusion and knowledge extraction to construct a knowledge graph; performing semantic modeling based on deep learning models to construct a precise retrieval based on semantic understanding; combining RAG and GPT to construct a generative retrieval to meet the needs of enterprise architecture knowledge retrieval; and modularizing the precise retrieval and generative retrieval technologies.

[0004] As can be seen from the above solutions, the current management and retrieval of power equipment information still relies on tagging to determine search keywords, lacking the processing of data identification, fusion, analysis, and matching, which has certain limitations. Summary of the Invention

[0005] The purpose of this invention is to provide a precise retrieval and intelligent question-answering system and method for electrical knowledge base based on AI large model and knowledge graph, which solves the problems existing in the background technology.

[0006] To address the aforementioned technical problems, this invention adopts the following technical solution: This invention provides a method for accurate retrieval and intelligent question answering of electrical knowledge bases based on AI large models and knowledge graphs, specifically including the following steps:

[0007] S1. Collect historical electrical knowledge data, process the collected historical electrical knowledge data to obtain processed historical electrical knowledge data, and construct a sample set based on the obtained processed historical electrical knowledge data;

[0008] S2. Construct an electrical knowledge recognition model using natural language processing, and then use the constructed electrical knowledge recognition model to identify the processed historical electrical knowledge data to obtain the entity information of the identified historical electrical knowledge data.

[0009] S3. Use data association analysis to perform data association on the entity information of the identified historical electrical knowledge data, and output the data association results. The data association results include: performance parameter association of the corresponding entity information, working principle association of the corresponding entity information, and type association of the corresponding entity information.

[0010] S4. Based on the entity information of the identified historical electrical knowledge data and the corresponding output data association results, deep learning is used to analyze and match the data to obtain the analyzed and matched historical electrical knowledge data.

[0011] S5. Based on the analyzed and matched historical electrical knowledge data, the data is fused using a knowledge fusion method to obtain the fused historical electrical knowledge data.

[0012] S6. Construct a knowledge graph based on the fused historical electrical knowledge data, and perform intelligent retrieval and question answering through semantic recognition.

[0013] Preferably, the steps of collecting historical electrical knowledge data, processing the collected historical electrical knowledge data to obtain processed historical electrical knowledge data, and constructing a sample set based on the obtained processed historical electrical knowledge data include the following steps:

[0014] S11. Traverse the collected historical electrical knowledge data, locate the missing data in the historical electrical knowledge data, and fill the missing data in the historical electrical knowledge data through data filling.

[0015] S12. Summarize the historical electrical knowledge data after it has been filled in, and expand the historical electrical knowledge data after it has been filled in by data expansion and replacement to obtain the expanded historical electrical knowledge data.

[0016] Based on the electrical equipment models in the collected historical electrical knowledge data, the electrical equipment models are classified. After the classification is completed, each group of data in the historical electrical knowledge data is segmented and expanded. An entity noun is randomly selected from the corresponding electrical equipment model classification to replace the electrical equipment entity noun in a group of historical electrical knowledge data, forming a new group of historical electrical knowledge data.

[0017] S13. Summarize and expand the historical electrical knowledge data to construct a sample set and remove duplicates;

[0018] The deduplication process is as follows:

[0019] K = {d|d, id are unique and in the sample set};

[0020] Where K represents the deduplication result of the sample set, and d represents the constructed sample set.

[0021] Preferably, the process of traversing and collecting historical electrical knowledge data, locating missing data in the historical electrical knowledge data, and filling in the missing data in the historical electrical knowledge data through data filling includes the following steps:

[0022] The collected historical electrical knowledge data includes: electrical equipment model, electrical equipment ID, electrical equipment operating status, equipment location, and equipment performance parameters;

[0023] A set of standard historical electrical knowledge data must include: electrical equipment model, electrical equipment ID, electrical equipment operating status, equipment location, and equipment performance parameters;

[0024] When missing data is detected in the collected historical electrical knowledge data during the data traversal process, it is filled in using data filling methods;

[0025] The data filling formula is shown below:

[0026]

[0027] Where k represents the filler value for missing data, x1 represents the value of the historical electrical knowledge data corresponding to the missing data, x represents the average value of the historical electrical knowledge data corresponding to the missing data, y represents the number of historical electrical knowledge data, and y1 represents the number of missing data.

[0028] Preferably, the step of constructing an electrical knowledge recognition model using natural language processing, and then identifying the processed historical electrical knowledge data based on the constructed electrical knowledge recognition model to obtain the entity information of the identified historical electrical knowledge data includes the following steps:

[0029] S21. Randomly select a set of processed historical electrical knowledge data from the sample set, extract all entity nouns, and construct entity noun set B;

[0030] S22. Based on the constructed entity noun set B, calculate the entity noun weights using natural language processing.

[0031] S23. Construct an electrical knowledge recognition model based on the calculated entity noun weights, and summarize the recognized entity nouns to obtain the entity information of the historical electrical knowledge data after recognition;

[0032] The electrical knowledge recognition model is shown below:

[0033]

[0034] Where d is the identification adjustment coefficient, B i W(B) represents the i-th entity noun in the set of entity nouns. iB represents the recognition weight of the i-th entity noun in the entity noun set. j It represents the j-th entity noun that has the same type as the i-th entity noun in the entity noun set.

[0035] Preferably, the step of performing data association analysis on the entity information of the identified historical electrical knowledge data and outputting the data association results includes: performance parameter association of the corresponding entity information, working principle association of the corresponding entity information, and type association of the corresponding entity information, comprising the following steps:

[0036] S31. Construct association relationships based on entity information from the identified historical electrical knowledge data;

[0037] S32. Based on the established association relationships, perform data association on the entity information of the identified historical electrical knowledge data.

[0038] Preferably, the process of constructing associations based on entity information from identified historical electrical knowledge data includes the following steps:

[0039] Two sets of entity information from the identified historical electrical knowledge data are randomly selected, and the entity information from the two sets of identified historical electrical knowledge data is classified according to the data type in the historical electrical knowledge data, specifically divided into: electrical equipment model type, electrical equipment ID type, electrical equipment operating status type, equipment location type, and equipment performance parameter type;

[0040] Based on the entity information of the two sets of identified historical electrical knowledge data after classification, the relationship is associated by calculating the correlation degree between the two sets of entity information.

[0041] The formula for calculating the relevance of entity information is as follows:

[0042]

[0043] Where S(I,J) represents the relevance of historical electrical knowledge data entity information, and I and J represent different historical electrical knowledge data entity information, respectively. W(I) Len represents the breadth relevance value of the identified entity group I. W(J) Dep represents the breadth relevance value of the J-th group of entity information after identification. W(I,J) This represents the depth correlation value between the entity information of group I and group J after identification.

[0044] Preferably, the step of analyzing and matching the entity information and corresponding output data association results based on the identified historical electrical knowledge data using deep learning to obtain the analyzed and matched historical electrical knowledge data includes the following steps:

[0045] S41. Construct the LSTM network structure;

[0046] LSTM networks include input gates, output gates, and forget gates;

[0047] S42. Input the entity information of the identified historical electrical knowledge data as the input value of the input gate into the LSTM network, and input the output data association result as the candidate value into the LSTM network.

[0048] S43. Process the entity information of the current input and the historical electrical knowledge data identified in the previous time step through the forget gate;

[0049] S44. The entity information processed by the forget gate is output through the output gate to obtain the data association result predicted at the next time step;

[0050] S45. Iteratively use an LSTM network to determine the historical electrical knowledge data after analysis and matching;

[0051] Based on the updated data association results output by the LSTM network after iteration, the entity information of the continuously input historical electrical knowledge data after recognition is analyzed and matched to obtain the analyzed and matched historical electrical knowledge data.

[0052] Preferably, the process of fusing historical electrical knowledge data based on the analyzed and matched data through a knowledge fusion method to obtain the fused historical electrical knowledge data includes the following steps:

[0053] The knowledge fusion formula is as follows:

[0054] e = e z +e c +e l ;

[0055] Where e represents the merged historical electrical knowledge data, e z e represents the word vector in the fused historical electrical knowledge data. c e represents the word vector in the fused historical electrical knowledge data. l This represents the category vector of the word.

[0056] Preferably, the process of constructing a knowledge graph based on fused historical electrical knowledge data and performing intelligent retrieval and question answering through semantic recognition includes the following steps:

[0057] A knowledge graph is constructed by combining the merged historical electrical knowledge data with the updated data association results.

[0058] Real-time collection of electrical knowledge data input by users, and analysis and matching of the electrical knowledge data input by users based on the constructed knowledge graph;

[0059] The formula for analyzing and matching is:

[0060]

[0061] Wherein, KT represents the electrical knowledge data input by the user, KN represents the fused historical electrical knowledge data, and A(KN,KT) represents the similarity between data KT and data KN;

[0062] The fused historical electrical knowledge data with the highest similarity is selected as the analysis and matching result, and the data association results related to the fused historical electrical knowledge data with the highest similarity are fed back to the user as the search and answer results.

[0063] This invention discloses a precise retrieval and intelligent question-answering system for electrical knowledge base based on AI large model and knowledge graph, which is used to realize the precise retrieval and intelligent question-answering method for electrical knowledge base based on AI large model and knowledge graph. The system includes: a data acquisition module, a data processing module, a data recognition module, a data association analysis module, an analysis and matching module, a data fusion module, and an intelligent retrieval module.

[0064] The data acquisition module is used to collect historical electrical knowledge data and user-input electrical knowledge data;

[0065] The data processing module is used to process the collected historical electrical knowledge data;

[0066] The data recognition module is used to identify the processed historical electrical knowledge data;

[0067] The data association module is used to perform data association on the identified historical electrical knowledge data through data association analysis methods;

[0068] The analysis and matching module is used to analyze and match the identified historical electrical knowledge data and the corresponding output data association results;

[0069] The data fusion module is used to fuse historical electrical knowledge data after analysis and matching through knowledge fusion.

[0070] The intelligent retrieval module is used to perform intelligent retrieval and question answering through semantic recognition.

[0071] The beneficial effects of this invention are as follows:

[0072] (1) This invention collects and processes historical electrical knowledge data, constructs an electrical knowledge recognition model using natural language processing, and identifies the processed historical electrical knowledge data. Simultaneously, it uses data association analysis to associate the entity information of the identified historical electrical knowledge data and outputs the data association results. After obtaining the data association results, it analyzes and matches the entity information of the identified historical electrical knowledge data with the corresponding output data association results. Simultaneously, it fuses the historical electrical knowledge data based on the analysis and matching through knowledge fusion. Finally, it constructs a knowledge graph based on the fused historical electrical knowledge data and performs intelligent retrieval and question answering through semantic recognition, thereby improving the accuracy of electrical knowledge base retrieval.

[0073] (2) This invention improves the effectiveness of data processing by traversing the collected historical electrical knowledge data and filling in the missing data. After filling, the historical electrical knowledge data is expanded by data expansion and replacement. Finally, the expanded historical electrical knowledge data is summarized to construct a sample set and deduplicated, which improves the effectiveness of data processing.

[0074] (3) This invention randomly selects a set of processed historical electrical knowledge data and extracts entity names. After extraction, the entity name weights are calculated using natural language processing. At the same time, an electrical knowledge recognition model is constructed based on the calculated entity name weights. The entity names are then summarized to obtain the entity information of the identified historical electrical knowledge data, thus ensuring the reliability of electrical knowledge data recognition.

[0075] (4) The present invention constructs an association relationship based on the entity information of the identified historical electrical knowledge data, and at the same time performs data association on the entity information of the identified historical electrical knowledge data based on the constructed association relationship, thereby improving the reliability of data association.

[0076] (5) This invention uses deep learning to analyze and match the entity information of the identified historical electrical knowledge data with the corresponding output data association results, thereby improving the efficiency of electrical knowledge data analysis. Attached Figure Description

[0077] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0078] Figure 1 This is a schematic diagram of the electrical knowledge base accurate retrieval and intelligent question answering method of the present invention. Detailed Implementation

[0079] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0080] In a specific embodiment of the present invention,

[0081] Reference Figure 1 As shown, this invention provides a method for accurate retrieval and intelligent question answering of electrical knowledge bases based on AI large models and knowledge graphs, specifically including the following steps:

[0082] S1. Collect historical electrical knowledge data, process the collected historical electrical knowledge data to obtain processed historical electrical knowledge data, and construct a sample set based on the obtained processed historical electrical knowledge data;

[0083] S2. Construct an electrical knowledge recognition model using natural language processing, and then use the constructed electrical knowledge recognition model to identify the processed historical electrical knowledge data to obtain the entity information of the identified historical electrical knowledge data.

[0084] S3. Use data association analysis to perform data association on the entity information of the identified historical electrical knowledge data, and output the data association results. The data association results include: performance parameter association of the corresponding entity information, working principle association of the corresponding entity information, and type association of the corresponding entity information.

[0085] S4. Based on the entity information of the identified historical electrical knowledge data and the corresponding output data association results, deep learning is used to analyze and match the data to obtain the analyzed and matched historical electrical knowledge data.

[0086] S5. Based on the analyzed and matched historical electrical knowledge data, the data is fused using a knowledge fusion method to obtain the fused historical electrical knowledge data.

[0087] S6. Construct a knowledge graph based on the fused historical electrical knowledge data, and perform intelligent retrieval and question answering through semantic recognition;

[0088] Furthermore, referring to Figure 1 As shown, the process of collecting historical electrical knowledge data, processing the collected historical electrical knowledge data to obtain processed historical electrical knowledge data, and constructing a sample set based on the processed historical electrical knowledge data includes the following steps:

[0089] S11. Traverse the collected historical electrical knowledge data, locate the missing data in the historical electrical knowledge data, and fill the missing data in the historical electrical knowledge data through data filling.

[0090] The collected historical electrical knowledge data includes: electrical equipment model, electrical equipment ID, electrical equipment operating status, equipment location, and equipment performance parameters;

[0091] A set of standard historical electrical knowledge data must include: electrical equipment model, electrical equipment ID, electrical equipment operating status, equipment location, and equipment performance parameters;

[0092] When missing data is detected in the collected historical electrical knowledge data during the data traversal process, it is filled in using data filling methods;

[0093] The data filling formula is shown below:

[0094]

[0095] Where k represents the imputation value for missing data, x1 represents the value of the historical electrical knowledge data corresponding to the missing data, x represents the average value of the historical electrical knowledge data corresponding to the missing data, y represents the number of historical electrical knowledge data, and y1 represents the number of missing data.

[0096] S12. Summarize the historical electrical knowledge data after it has been filled in, and expand the historical electrical knowledge data after it has been filled in by data expansion and replacement to obtain the expanded historical electrical knowledge data.

[0097] Based on the electrical equipment models in the collected historical electrical knowledge data, the electrical equipment models are classified. After the classification is completed, each group of data in the historical electrical knowledge data is segmented and expanded. An entity noun is randomly selected from the corresponding electrical equipment model classification to replace the electrical equipment entity noun in a group of historical electrical knowledge data, forming a new group of historical electrical knowledge data.

[0098] S13. Summarize and expand the historical electrical knowledge data to construct a sample set and remove duplicates;

[0099] The deduplication process is as follows:

[0100] K = {d|d, id is unique and exists in the sample set};

[0101] Where K represents the deduplication result of the sample set, and d represents the constructed sample set;

[0102] Furthermore, referring to Figure 1As shown, an electrical knowledge recognition model is constructed using natural language processing, and the processed historical electrical knowledge data is then identified based on this model to obtain entity information from the identified historical electrical knowledge data. This process includes the following steps:

[0103] S21. Randomly select a set of processed historical electrical knowledge data from the sample set, extract all entity nouns, and construct entity noun set B;

[0104] S22. Based on the constructed entity noun set B, calculate the entity noun weights using natural language processing.

[0105]

[0106] Among them, w i f represents the weight value of the i-th entity noun. i Let n represent the number of times the i-th entity noun appears, and N represent the number of entity noun sets. j Indicates the number of entity nouns that need to be identified;

[0107] S23. Construct an electrical knowledge recognition model based on the calculated entity noun weights, and summarize the recognized entity nouns to obtain the entity information of the historical electrical knowledge data after recognition;

[0108] The electrical knowledge recognition model is shown below:

[0109]

[0110] Where d is the identification adjustment coefficient, B i W(B) represents the i-th entity noun in the set of entity nouns. i B represents the recognition weight of the i-th entity noun in the entity noun set. j This represents the j-th entity noun that has the same type as the i-th entity noun in the entity noun set;

[0111] Furthermore, referring to Figure 1 As shown, the entity information of the identified historical electrical knowledge data is correlated using a data association analysis method, and the data association results are output. The data association results include: performance parameter association of corresponding entity information, working principle association of corresponding entity information, and type association of corresponding entity information. The steps include:

[0112] S31. Construct association relationships based on entity information from the identified historical electrical knowledge data;

[0113] Two sets of entity information from the identified historical electrical knowledge data are randomly selected, and the entity information from the two sets of identified historical electrical knowledge data is classified according to the data type in the historical electrical knowledge data, specifically divided into: electrical equipment model type, electrical equipment ID type, electrical equipment operating status type, equipment location type, and equipment performance parameter type;

[0114] Furthermore, based on the entity information of the two sets of identified historical electrical knowledge data after classification, the relationship between the two sets of entity information is determined by calculating the correlation degree between the two sets of entity information;

[0115] The formula for calculating the relevance of entity information is as follows:

[0116]

[0117] Where S(I,J) represents the relevance of historical electrical knowledge data entity information, and I and J represent different historical electrical knowledge data entity information, respectively. W(I) Len represents the breadth relevance value of the identified entity group I. W(J) Dep represents the breadth relevance value of the J-th group of entity information after identification. W(I,J) This represents the depth correlation value between the entity information of group I and group J after identification;

[0118] S32. Based on the established association relationships, perform data association on the entity information of the identified historical electrical knowledge data;

[0119] The data association formula is shown below:

[0120]

[0121] Where H represents the data association result, θ represents the association parameter, b represents the attribute value of the historical electrical knowledge data entity information, and w represents the weight;

[0122] Furthermore, referring to Figure 1 As shown, based on the entity information of the identified historical electrical knowledge data and the corresponding output data association results, deep learning is used for analysis and matching to obtain the analyzed and matched historical electrical knowledge data, including the following steps:

[0123] S41. Construct the LSTM network structure;

[0124] LSTM networks include input gates, output gates, and forget gates;

[0125] S42. Input the entity information of the identified historical electrical knowledge data as the input value of the input gate into the LSTM network, and input the output data association result as the candidate value into the LSTM network.

[0126] The input gate's formula for processing input values ​​is as follows:

[0127] α t =σ(ω·[h t-1 ,x t ]+b α );

[0128] Where, α t The input values ​​of the processed input gate are represented by ω, and the weight matrix is ​​represented by b. α The input gate bias value is represented by σ, which represents the sigmoid function, and h is the input gate bias value. t-1 x represents the correlation result of the input data at time t-1. t This represents a set of entity information input at time t;

[0129] S43. Process the entity information of the current input and the historical electrical knowledge data identified in the previous time step through the forget gate;

[0130] The formula for the forgetting gate is as follows:

[0131] f t =σ(ω·[h t-1 ,x t ]+b f );

[0132] Among them, f t b represents the output value of the forget gate at time t. f Indicates the bias value of the forget gate;

[0133] The formula for updating the data association results is as follows:

[0134]

[0135] Among them, E t E represents the updated data association result. t-1 This indicates the data association result from the previous time step. This represents the candidate value at time t;

[0136] S44. The entity information processed by the forget gate is output through the output gate to obtain the data association result predicted at the next time step;

[0137] The output gate formula is as follows:

[0138] h t =o t ×tanh(E t );

[0139] o t =σ(W·[h t-1 ,x t ]+b o);

[0140] Among them, h t This represents the data association result predicted at time t, o t Let t represent the output value of the output gate at time t, tanh represent the hyperbolic tangent function, and b o This represents the bias value of the output gate;

[0141] S45. Iteratively use an LSTM network to determine the historical electrical knowledge data after analysis and matching;

[0142] Based on the updated data association results output by the LSTM network after iteration, the entity information of the continuously input historical electrical knowledge data after recognition is analyzed and matched to obtain the analyzed and matched historical electrical knowledge data.

[0143] Furthermore, referring to Figure 1 As shown, based on the analyzed and matched historical electrical knowledge data, the fused historical electrical knowledge data is obtained through knowledge fusion, including the following steps:

[0144] The knowledge fusion formula is as follows:

[0145] e = e z +e c +e l ;

[0146] Where e represents the merged historical electrical knowledge data, e z e represents the word vector in the fused historical electrical knowledge data. c e represents the word vector in the fused historical electrical knowledge data. l The category vector representing the word;

[0147] Furthermore, referring to Figure 1 As shown, the process of constructing a knowledge graph based on fused historical electrical knowledge data and performing intelligent retrieval and question answering through semantic recognition includes the following steps:

[0148] A knowledge graph is constructed by combining the merged historical electrical knowledge data with the updated data association results.

[0149] Furthermore, electrical knowledge data input by users is collected in real time, and the electrical knowledge data input by users is analyzed and matched based on the constructed knowledge graph;

[0150] The formula for analyzing and matching is:

[0151]

[0152] Wherein, KT represents the electrical knowledge data input by the user, KN represents the fused historical electrical knowledge data, and A(KN,KT) represents the similarity between data KT and data KN;

[0153] The fused historical electrical knowledge data with the highest similarity is selected as the analysis and matching result, and the data association results related to the fused historical electrical knowledge data with the highest similarity are fed back to the user as the search and answer results;

[0154] In one specific embodiment, the electrical knowledge base accurate retrieval and intelligent question answering system based on AI large model and knowledge graph is used to implement the electrical knowledge base accurate retrieval and intelligent question answering method based on AI large model and knowledge graph. The system includes: a data acquisition module, a data processing module, a data recognition module, a data association analysis module, an analysis and matching module, a data fusion module, and an intelligent retrieval module.

[0155] The data acquisition module is used to collect historical electrical knowledge data and user-input electrical knowledge data;

[0156] The data processing module is used to process the collected historical electrical knowledge data;

[0157] The data recognition module is used to identify the processed historical electrical knowledge data;

[0158] The data association module is used to perform data association on the identified historical electrical knowledge data through data association analysis methods;

[0159] The analysis and matching module is used to analyze and match the identified historical electrical knowledge data and the corresponding output data association results;

[0160] The data fusion module is used to fuse historical electrical knowledge data after analysis and matching through knowledge fusion.

[0161] The intelligent retrieval module is used to perform intelligent retrieval and question answering through semantic recognition.

[0162] It should be noted that

[0163] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. An AI large model and knowledge graph-based electrical knowledge base precise retrieval and intelligent question and answer method, characterized in that, Includes the following steps: S1. Collect historical electrical knowledge data, process the collected historical electrical knowledge data to obtain processed historical electrical knowledge data, and construct a sample set based on the obtained processed historical electrical knowledge data; S2. Construct an electrical knowledge recognition model using natural language processing, and then identify the processed historical electrical knowledge data based on the constructed electrical knowledge recognition model to obtain the identified historical electrical knowledge. S3. Use data association analysis to perform data association on the entity information of the identified historical electrical knowledge data, and output the data association results. The data association results include: performance parameter association of the corresponding entity information, working principle association of the corresponding entity information, and type association of the corresponding entity information. S4. Based on the entity information of the identified historical electrical knowledge data and the corresponding output data association results, deep learning is used to analyze and match the data to obtain the analyzed and matched historical electrical knowledge data. S41. Construct the LSTM network structure; LSTM networks include input gates, output gates, and forget gates; S42. Input the entity information of the identified historical electrical knowledge data as the input value of the input gate into the LSTM network, and input the output data association result as the candidate value into the LSTM network. S43. Process the entity information of the current input and the historical electrical knowledge data identified in the previous time step through the forget gate; S44. The entity information processed by the forget gate is output through the output gate to obtain the data association result predicted at the next time step; S45. Iteratively use an LSTM network to determine the historical electrical knowledge data after analysis and matching; Based on the updated data association results output by the LSTM network after iteration, the entity information of the continuously input historical electrical knowledge data after recognition is analyzed and matched to obtain the analyzed and matched historical electrical knowledge data. S5. Based on the analyzed and matched historical electrical knowledge data, the data is fused using a knowledge fusion method to obtain the fused historical electrical knowledge data. S6. Construct a knowledge graph based on the fused historical electrical knowledge data, and perform intelligent retrieval and question answering through semantic recognition.

2. The AI large model and knowledge graph-based electrical knowledge base precise retrieval and intelligent question and answer method according to claim 1, characterized in that, The process of collecting historical electrical knowledge data, processing the collected historical electrical knowledge data to obtain processed historical electrical knowledge data, and constructing a sample set based on the obtained processed historical electrical knowledge data includes the following steps: S11. Traverse the collected historical electrical knowledge data, locate the missing data in the historical electrical knowledge data, and fill the missing data in the historical electrical knowledge data through data filling. S12. Summarize the historical electrical knowledge data after it has been filled in, and expand the historical electrical knowledge data after it has been filled in by data expansion and replacement to obtain the expanded historical electrical knowledge data. Based on the electrical equipment models in the collected historical electrical knowledge data, the electrical equipment models are classified. After the classification is completed, each group of data in the historical electrical knowledge data is segmented and expanded. An entity noun is randomly selected from the corresponding electrical equipment model classification to replace the electrical equipment entity noun in a group of historical electrical knowledge data, forming a new group of historical electrical knowledge data. S13. Summarize and expand the historical electrical knowledge data to construct a sample set and remove duplicates; The deduplication process is as follows: ； wherein, represents the sample set deduplication result, represents the constructed sample set.

3. The method for accurate retrieval and intelligent question answering of electrical knowledge base based on AI large model and knowledge graph as described in claim 2, is characterized in that, The process of traversing and collecting historical electrical knowledge data, locating missing data within the historical electrical knowledge data, and filling in the missing data through data imputation includes the following steps: The collected historical electrical knowledge data includes: electrical equipment model, electrical equipment ID, electrical equipment operating status, equipment location, and equipment performance parameters; A set of standard historical electrical knowledge data must include: electrical equipment model, electrical equipment ID, electrical equipment operating status, equipment location, and equipment performance parameters; When missing data is detected in the collected historical electrical knowledge data during the data traversal process, it is filled in using data filling methods; The data filling formula is shown below: ； in, The filler value represents the missing data. This indicates the numerical value of the historical electrical knowledge data corresponding to the missing data. This indicates the average value of historical electrical knowledge data corresponding to missing data. This indicates the amount of historical electrical knowledge data. Indicates the number of missing data.

4. The method for accurate retrieval and intelligent question answering of electrical knowledge base based on AI large model and knowledge graph as described in claim 1, characterized in that, The process of constructing an electrical knowledge recognition model using natural language processing and then identifying the processed historical electrical knowledge data based on that model to obtain entity information from the identified historical electrical knowledge data includes the following steps: S21. Randomly select a set of processed historical electrical knowledge data from the sample set, extract all entity names, and construct an entity name set. ; S22, Based on the constructed entity noun set Entity noun weights are calculated using natural language processing. S23. Construct an electrical knowledge recognition model based on the calculated entity noun weights, and summarize the recognized entity nouns to obtain the entity information of the historical electrical knowledge data after recognition; The electrical knowledge recognition model is shown below: ； in, To identify the adjustment coefficient, This represents the i-th entity noun in the set of entity nouns. This represents the recognition weight of the i-th entity noun in the entity noun set. It represents the j-th entity noun that has the same type as the i-th entity noun in the entity noun set.

5. The method for accurate retrieval and intelligent question answering of electrical knowledge base based on AI large model and knowledge graph as described in claim 1, characterized in that, The method of associating entity information in the identified historical electrical knowledge data using data association analysis and outputting the data association results includes: performance parameter association, working principle association, and type association of the corresponding entity information. This includes the following steps: S31. Construct association relationships based on entity information from the identified historical electrical knowledge data; S32. Based on the established association relationships, perform data association on the entity information of the identified historical electrical knowledge data.

6. The method for accurate retrieval and intelligent question answering of electrical knowledge base based on AI large model and knowledge graph as described in claim 5, is characterized in that, The process of constructing associations based on entity information from identified historical electrical knowledge data includes the following steps: Two sets of entity information from the identified historical electrical knowledge data are randomly selected, and the entity information from the two sets of identified historical electrical knowledge data is classified according to the data type in the historical electrical knowledge data, specifically divided into: electrical equipment model type, electrical equipment ID type, electrical equipment operating status type, equipment location type, and equipment performance parameter type; Based on the entity information of the two sets of identified historical electrical knowledge data after classification, the relationship is associated by calculating the correlation degree between the two sets of entity information. The formula for calculating the relevance of entity information is as follows: ； in, This indicates the degree of correlation between historical electrical knowledge data entity information. These represent different historical electrical knowledge data entity information. This represents the breadth relevance value of the identified entity group I. This represents the breadth relevance value of the J-th group of entity information after identification. This represents the depth correlation value between the entity information of group I and group J after identification.

7. The method for accurate retrieval and intelligent question answering of electrical knowledge base based on AI large model and knowledge graph as described in claim 1, characterized in that, The process of fusing historical electrical knowledge data based on analysis and matching, and obtaining fused historical electrical knowledge data through knowledge fusion, includes the following steps: The knowledge fusion formula is shown below: ； in, This represents the merged historical electrical knowledge data. This represents the word vectors in the fused historical electrical knowledge data. This represents the word vectors in the fused historical electrical knowledge data. This represents the category vector of the word.

8. The method for accurate retrieval and intelligent question answering of electrical knowledge base based on AI large model and knowledge graph as described in claim 1, characterized in that, The process of constructing a knowledge graph based on fused historical electrical knowledge data and performing intelligent retrieval and question answering through semantic recognition includes the following steps: A knowledge graph is constructed by combining the merged historical electrical knowledge data with the updated data association results. Real-time collection of electrical knowledge data input by users, and analysis and matching of the electrical knowledge data input by users based on the constructed knowledge graph; The formula for analyzing and matching is: ； Where KT represents the electrical knowledge data input by the user, and KN represents the fused historical electrical knowledge data. This represents the similarity between data KT and data KN; The fused historical electrical knowledge data with the highest similarity is selected as the analysis and matching result, and the data association results related to the fused historical electrical knowledge data with the highest similarity are fed back to the user as the search and answer results.

9. A system for implementing the precise retrieval and intelligent question-answering method for electrical knowledge base based on AI large model and knowledge graph as described in any one of claims 1-8, characterized in that, include: The system includes a data acquisition module, a data processing module, a data recognition module, a data correlation analysis module, an analysis and matching module, a data fusion module, and an intelligent retrieval module. The data acquisition module is used to collect historical electrical knowledge data and user-input electrical knowledge data; The data processing module is used to process the collected historical electrical knowledge data; The data recognition module is used to identify the processed historical electrical knowledge data; The data association analysis module is used to perform data association on the identified historical electrical knowledge data through data association analysis methods. The analysis and matching module is used to analyze and match the identified historical electrical knowledge data and the corresponding output data association results; The data fusion module is used to fuse historical electrical knowledge data after analysis and matching through knowledge fusion. The intelligent retrieval module is used to perform intelligent retrieval and question answering through semantic recognition.

Images