Power supply station comprehensive analysis and evaluation method and system based on artificial intelligence

Abstract

The invention relates to a power supply station comprehensive analysis and evaluation method and system based on artificial intelligence, and the method comprises the following steps: S1, obtaining the data of all levels of a power supply station, and carrying out the preprocessing through automatic extraction, cleaning and automatic alignment; s2, removing redundant data and unifying a service aperture for the preprocessed data through feature engineering to obtain a key feature data set; s3, according to the key feature data set, constructing an AI-driven heart index library and an index relationship; s4, on the basis of the heart index library and the index relationship, setting a scientific weight, calculating the score of each index and forming a total score; and S5, according to a scoring result, performing transverse ranking and score assignment on the same kind of power supply stations in the whole region, and performing segmentation through clustering analysis. The comprehensive analysis and evaluation efficiency and reliability of the power supply station are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of power supply station management, and in particular to a comprehensive analysis and evaluation method and system for power supply stations based on artificial intelligence. Background Technology

[0002] Currently, the power supply station's operation, maintenance, and marketing data are scattered across multiple systems (such as Marketing 2.0, PMS 3.0, and the data acquisition system). The data structures, naming conventions, and data granularity differ across these systems, making data acquisition difficult to automate and standardize, resulting in frequent data redundancy and inconsistencies. Manual sorting and extraction are error-prone and time-consuming, and inconsistent data definitions hinder the accuracy and timeliness of subsequent analysis.

[0003] Most existing power supply station performance evaluation systems rely on manually assigned weights or single standards, lacking data-driven approaches and business logic connections, making it difficult to dynamically adapt to business changes. Redundancy or missing indicators are common, and there is a lack of systematic modeling of causal, impact, and business co-occurrence relationships between indicators, resulting in limited reference value and enforceability of evaluation results, making it difficult to accurately reflect the actual operational level and management shortcomings of power supply stations. Summary of the Invention

[0004] To address the aforementioned problems, the present invention aims to provide a method and system for comprehensive analysis and evaluation of power supply stations based on artificial intelligence, which effectively improves the efficiency and reliability of comprehensive analysis and evaluation of power supply stations.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A comprehensive analysis and evaluation method for power supply stations based on artificial intelligence includes the following steps: S1: Acquire data from all levels of the power supply station and preprocess it through automatic extraction, cleaning, and automatic alignment; S2: After preprocessing, feature engineering is used to remove redundant data and unify business definitions to obtain a key feature dataset. S3: Based on the key feature dataset, construct an AI-driven heart indicator library and indicator relationships; S4: Based on the heart indicator library and indicator relationships, set scientific weights, calculate the score of each indicator and form the total score; S5: Based on the scoring results, assign scores to the horizontal ranking of similar power supply stations across the entire region, and segment them using cluster analysis.

[0006] Furthermore, data from various levels of the power supply station is acquired and preprocessed through automatic extraction, cleaning, and alignment, as detailed below: Collect data from business systems at all levels, including Marketing 2.0, PMS 3.0, and the data collection system; and set data synchronization strategies to collect and store data at different granularities such as daily, hourly, and real-time according to system characteristics, and automatically transfer data from multiple systems to the data fusion platform. The data fusion platform performs deduplication, completion, anomaly handling, and format standardization on the data, and maps and associates the data of the same business entity in various systems to eliminate redundancy and discrepancies.

[0007] Furthermore, data from the same business entity across different systems is mapped and associated to eliminate redundancy and discrepancies, as detailed below: To achieve unified entity / field alignment and master data association for data items from different sources, with different names, and different granularities, the Jaccard coefficient is used to measure the similarity between two sets A and B.

[0008] If the similarity J(A,B)>θ, where θ is a preset threshold, then A and B are considered to be the same business entity. And by calculating the correlation matrix, highly redundant features r were discovered. xy: ; Where, x a ′、y a 'a' represents the a'th data point of features x and y, respectively; n' represents the amount of data. and Let x be the mean of feature x and feature y; When |r xy When |>0.8, the two features are considered highly correlated, and only one is retained; Multi-feature data can be dimensionality reduced and principal components extracted using PCA, thus reducing the analytical burden. The core data of all power supply stations, after extraction, cleaning, alignment, and dimensionality reduction, are transformed into multi-system fused data.

[0009] Furthermore, after preprocessing, feature engineering is used to remove redundant data and unify business definitions to obtain a key feature dataset, as follows: Based on the preprocessed multi-system fusion data, key features are selected through feature engineering. The IV value is used to measure the predictive power of an indicator on a target variable. ; Among them, IV j The information value of feature j; n j Number of bins for feature j; Good ja The number of positive samples in the a-th bin; Badja G represents the number of negative samples in the a-th bin of feature j; total B total The values ​​represent the total number of positive and negative samples, respectively; indicators with IV ≥ 0.1 are retained, while those with IV < 0.02 are removed. Based on the reduction of information entropy during decision tree splitting: ; Where: S is the sample set; Values(j) are the possible values ​​(or binning) of feature j; S v Let S be the subset of samples where feature j takes the value v; H(S) is the entropy of the current sample set S; Gain(S,j) is the information gain of feature j on the sample set S; H(S) v ) is a subset S v entropy; Calculate the information gain of all features, select the features with the largest Z% gain, and calculate the feature contribution using a tree model: ; Where, φ ij φ is the SHAP value of the j-th feature of the i-th sample; j The average absolute contribution of feature j; n is the number of samples; The rate of change of the index IV value is calculated using a time sliding window; if the decay exceeds a threshold, replacement is triggered. ; Among them, DR j Let be the IV decay rate of feature j; The IV value of the latest time window feature j; The IV value of feature j within the historical time window; By aligning business definitions with a rules engine, key features are unified to the same dimension and distribution, resulting in a key feature dataset.

[0010] Furthermore, based on the key feature dataset, an AI-driven mental indicator library and indicator relationships are constructed, as follows: S31: Transform key features into interpretable business metrics, obtain a list of business metrics, and establish a cross-domain semantic association matrix; S32: Combine the business indicator list and the cross-domain semantic association matrix to construct an indicator relationship knowledge graph, and use the graph structure to express the logical, causal, and business relationships between indicators.

[0011] Furthermore, key features are transformed into interpretable business metrics, and cross-domain semantic relationships are established, as follows: Establish a feature-indicator mapping dictionary, define composite indicators according to business rules, map each feature to a business-interpretable atomic or composite indicator, and supplement the generated indicators with name, business definition, domain, and unit of measurement metadata to obtain a list of business indicators. Using a pre-trained BERT model, the metric names and business definitions are converted into vector representations E': E'=BERT(metric name, business definition); And calculate the cosine similarity of each index: ; Where Sim(U,V) represents the semantic similarity between indices U and V, and E' represents the semantic similarity between them. U 、E' V These are vector representations of indices U and V, respectively; For vector E' U The model; For vector E' V The modulus is determined; and indicators whose Sim(U,V) is greater than the threshold are grouped into the same business domain or cross-domain group to construct a cross-domain semantic association matrix.

[0012] Furthermore, by combining the business indicator list and the cross-domain semantic association matrix, an indicator relationship knowledge graph is constructed, as follows: Generate graph node V based on the business indicator list, including atomic indicator nodes, composite indicator nodes and business entity nodes; Edge relationship E modeling includes three types of relationships: logical dependency, causal inference, and business co-occurrence. The logical dependency is obtained through explicit calculation between indicators; the causal relationship is obtained through time series causal inference; and the business co-occurrence is obtained through the co-occurrence frequency of indicators in the same business scenario. Based on the graph nodes V and edge relationships E, an index relationship knowledge graph G=(V,E) is constructed.

[0013] Furthermore, based on the indicator database and the relationships between indicators, scientific weights are set, scores for each indicator are calculated, and a total score is generated, as follows: For each index j' of each object i', standardize: ; Where, x i′j′ z is the j'th original index value of the i'th object; i′j′ The standard score of the j'th metric for the i'th object after standardization; max i′ x i′j′ The maximum value of the j'-th index among all objects; min i′ x i′j′The j'th index is the minimum value among all objects; Calculate the objective weights using the entropy weight method: ; ; ; Where, p i′j′ Ej′ represents the normalized weight of object i' on index j'; Ej′ represents the information entropy of the j'-th index. Let be the objective weight of the j'th indicator; M is the total number of objects, and N is the total number of indicators; Calculate subjective weights using AHP expert weights: Judgment Matrix ; The weights are obtained by normalizing the eigenvectors: ; The weights are adjusted based on the relationships in the knowledge graph to obtain the final weights. C is the relational influence coefficient matrix: ; Where I is the identity matrix; The result is either AHP or entropy weighting. Then the final weight of indicator j' is w. j' : ; in, α, β, γ The fusion ratio of expert subjective weight, objective weight, and relationship-adjusted weight; Then the score of a single indicator S i'j' : ; Further obtain the overall evaluation score S of the i'th evaluation object. i' : .

[0014] Furthermore, based on the scoring results, a horizontal ranking score is assigned to all similar power supply stations in the region, and cluster analysis is used for segmentation, as follows: Based on the comprehensive score obtained from S4, the scores of each power supply station are ranked from high to low; K-means clustering is used to scientifically segment the comprehensive score and key business indicator scores of all similar power supply stations into high-performance, medium-performance, and low-performance segments; finally, the final label for each power supply station is obtained, including its comprehensive score, horizontal ranking and ranking score in the region, and the segment it belongs to from which cluster it was segmented.

[0015] An artificial intelligence-based comprehensive analysis and evaluation system for power supply stations includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the artificial intelligence-based comprehensive analysis and evaluation method for power supply stations as described above.

[0016] The present invention has the following beneficial effects: 1. This invention ensures the fullness, integration, and consistency of data by collecting, extracting, deduplicating, aligning, and fusing data from multiple systems, sources, and granularities in power supply stations. It also utilizes feature engineering models and AI algorithms to automatically select the most critical and representative core features of the business, and combines information value, feature importance, and dynamic changes to achieve feature optimization. Based on business rules, it automatically constructs a clear indicator library and indicator relationship network, generating a traceable and interpretable multi-level business indicator system. This invention integrates three weighting sources: objective, subjective, and knowledge graph-driven correlation, dynamically and adaptively allocating the importance of various business indicators to ensure the comprehensiveness and authority of the evaluation system. Based on this, combined with a standardized data process, it achieves unbiased weighted scoring for each indicator. Through horizontal comparison of comprehensive scores, it achieves automatic ranking. Simultaneously, hierarchical aggregation and multi-dimensional interpretable output facilitate differentiated analysis and targeted improvements for power supply stations, effectively enhancing the management efficiency and quality of power supply stations. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] The following is in conjunction with the appendix Figure 1 The present invention will be further described in detail with reference to specific embodiments: A comprehensive analysis and evaluation method for power supply stations based on artificial intelligence includes the following steps: S1: Acquire data from all levels of the power supply station and preprocess it through automatic extraction, cleaning, and automatic alignment; S2: After preprocessing, feature engineering is used to remove redundant data and unify business definitions to obtain a key feature dataset. S3: Based on the key feature dataset, construct an AI-driven heart indicator library and indicator relationships; S4: Based on the heart indicator library and indicator relationships, set scientific weights, calculate the score of each indicator and form the total score; S5: Based on the scoring results, assign scores to the horizontal ranking of similar power supply stations across the entire region, and segment them using cluster analysis.

[0019] In this embodiment, data from various levels of the power supply station is acquired and preprocessed through automatic extraction, cleaning, and automatic alignment, as detailed below: Collect data from business systems at all levels, including Marketing 2.0, PMS 3.0, and the data collection system; and set data synchronization strategies to collect and store data at different granularities such as daily, hourly, and real-time according to system characteristics, and automatically transfer data from multiple systems to the data fusion platform. The data fusion platform performs deduplication, completion, anomaly handling, and format standardization on the data. It also maps and associates the data of the same business entity (such as the same power supply station or equipment) in various systems to eliminate redundancy and discrepancies.

[0020] In this embodiment, data of the same business entity (such as the same power supply station or equipment) in various systems are mapped and associated to eliminate redundancy and discrepancies, as detailed below: To achieve unified entity / field alignment and master data association for data items from different sources, with different names, and different granularities, the Jaccard coefficient is used to measure the similarity between two sets A and B (such as the word segmentation results of the equipment description field). ; If the similarity J(A,B)>θ, where θ is a preset threshold, then A and B are considered to be the same business entity. And by calculating the correlation matrix, highly redundant features r were discovered. xy: ; Where, x a ′、y a 'a' represents the a'th data point of features x and y, respectively; n' represents the amount of data. and Let x be the mean of feature x and feature y; When |r xy When |>0.8, the two features are considered highly correlated, and only one is retained; Multi-feature data can be dimensionality reduced and principal components extracted using PCA, thus reducing the analytical burden. The core data of all power supply stations, after extraction, cleaning, alignment, and dimensionality reduction, are transformed into multi-system fused data.

[0021] In this embodiment, the preprocessed data is subjected to feature engineering to remove redundant data and unify business definitions, resulting in a key feature dataset, as follows: Based on the preprocessed multi-system fusion data, key features are selected through feature engineering. The IV (Information Value) value is used to measure the predictive power of an indicator for a target variable (such as the level of a power supply station). ; Among them, IV j The information value of feature j; n jNumber of bins for feature j; Good ja The number of positive samples in the a-th bin; Bad ja G represents the number of negative samples in the a-th bin of feature j; total B total The values ​​represent the total number of positive and negative samples, respectively; indicators with IV ≥ 0.1 are retained, while those with IV < 0.02 are removed. Based on the reduction of information entropy during decision tree splitting: ; Where: S is the sample set; Values(j) are the possible values ​​(or binning) of feature j; S v Let S be the subset of samples where feature j takes the value v; H(S) is the entropy of the current sample set S; Gain(S,j) is the information gain of feature j on the sample set S; H(S) v ) is a subset S v entropy; Calculate the information gain of all features, select the features with the largest Z% gain, and calculate the feature contribution using a tree model (such as XGBoost): ; Where, φ ij φ is the SHAP value of the j-th feature of the i-th sample; j The average absolute contribution of feature j; n is the number of samples; The rate of change of the index IV value is calculated using a sliding window over time. If the decay exceeds a threshold (e.g., 30%), a replacement is triggered. ; Among them, DR j Let be the IV decay rate of feature j; The IV value of the latest time window feature j; The IV value of feature j within the historical time window; By aligning business definitions with a rules engine, key features are unified to the same dimension and distribution, resulting in a key feature dataset.

[0022] In this embodiment, an AI-driven mental indicator library and indicator relationships are constructed based on the key feature dataset, as detailed below: S31: Transform key features into interpretable business metrics, obtain a list of business metrics (including metadata, formulas, and semantic information), and establish a cross-domain semantic association matrix; S32: Combine the business indicator list and the cross-domain semantic association matrix to construct an indicator relationship knowledge graph, and use the graph structure to express the logical, causal, and business relationships between indicators.

[0023] In this embodiment, key features are transformed into interpretable business metrics, and cross-domain semantic associations are established, as detailed below: Establish a feature-indicator mapping dictionary, define composite indicators according to business rules, map each feature to a business-interpretable atomic or composite indicator, and supplement the generated indicators with name, business definition, domain, and unit of measurement metadata to obtain a list of business indicators. Using a pre-trained BERT model, the metric names and business definitions are converted into vector representations E': E'=BERT(metric name, business definition); And calculate the cosine similarity of each index: ; Where Sim(U,V) represents the semantic similarity between indices U and V, and E' represents the semantic similarity between them. U 、E' V These are vector representations of indices U and V, respectively; For vector E' U The model; For vector E' V The modulus is determined; and indicators whose Sim(U,V) is greater than the threshold are grouped into the same business domain or cross-domain group to construct a cross-domain semantic association matrix.

[0024] In this embodiment, a knowledge graph of indicator relationships is constructed by combining the business indicator list and the cross-domain semantic association matrix, as follows: Generate graph node V based on the business indicator list, including atomic indicator nodes, composite indicator nodes and business entity nodes; Edge relationship E modeling includes three types of relationships: logical dependency, causal inference, and business co-occurrence. The logical dependency is obtained through explicit calculation between indicators; the causal relationship is obtained through time series causal inference; and the business co-occurrence is obtained through the co-occurrence frequency of indicators in the same business scenario (work order / report). Based on the graph nodes V and edge relationships E, an index relationship knowledge graph G=(V,E) is constructed.

[0025] In this embodiment, based on the cardiac indicator database and indicator relationships, scientific weights are set, scores for each indicator are calculated, and a total score is formed, as detailed below: For each index j' of each object i', standardize: ; Where, x i′j′ z is the j'th original index value of the i'th object; i′j′ The standard score of the j'th metric for the i'th object after standardization; max i′ x i′j′The maximum value of the j'-th index among all objects; min i′ x i′j′ The j'th index is the minimum value among all objects; Calculate the objective weights using the entropy weight method: ; ; ; Where, p i′j′ Ej′ represents the normalized weight of object i' on index j'; Ej′ represents the information entropy of the j'-th index. Let be the objective weight of the j'th indicator; M is the total number of objects, and N is the total number of indicators; Calculate subjective weights using AHP expert weights: Judgment Matrix ; The weights are obtained by normalizing the eigenvectors: ; The weights are adjusted based on the relationships in the knowledge graph to obtain the final weights. C is the relational influence coefficient matrix: ; Where I is the identity matrix; The result is either AHP or entropy weighting. Then the final weight of indicator j' is w. j' : ; in, α, β, γ The fusion ratio of expert subjective weight, objective weight, and relationship-adjusted weight; Then the score of a single indicator S i'j' : ; Further obtain the overall evaluation score S of the i'th evaluation object. i' : .

[0026] In this embodiment, based on the scoring results, the horizontal ranking of similar power supply stations across the entire region is assigned scores, and cluster analysis is used for segmentation, as follows: Based on the comprehensive score obtained in S4, the scores of each power supply station are sorted from high to low; K-means clustering is used to scientifically segment the comprehensive scores and key business indicator scores of all similar power supply stations into high-performance, medium-performance, and low-performance segments; finally, the final label of each power supply station is obtained, including its comprehensive score, horizontal ranking and ranking score across the entire region, and the segment it belongs to (e.g., high-performance, medium-performance, low-performance, etc.).

[0027] A power supply substation comprehensive analysis and evaluation system based on artificial intelligence is characterized by comprising a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the aforementioned artificial intelligence-based power supply substation comprehensive analysis and evaluation method. Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0028] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0029] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0030] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0031] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A comprehensive analysis and evaluation method for power supply stations based on artificial intelligence, characterized in that, Includes the following steps: S1: Acquire data from all levels of the power supply station and preprocess it through automatic extraction, cleaning, and automatic alignment; S2: After preprocessing, feature engineering is used to remove redundant data and unify business definitions to obtain a key feature dataset. S3: Based on the key feature dataset, construct an AI-driven heart indicator library and indicator relationships; S4: Based on the core indicator library and indicator relationships, set scientific weights, calculate the score of each indicator and form the total score; S5: Based on the scoring results, assign scores to the horizontal ranking of similar power supply stations across the entire region, and segment them using cluster analysis.

2. The comprehensive analysis and evaluation method for power supply stations based on artificial intelligence according to claim 1, characterized in that, The process of acquiring data from various levels of the power supply station and preprocessing it through automatic extraction, cleaning, and automatic alignment is as follows: Collect data from business systems at all levels, including Marketing 2.0, PMS 3.0, and the data collection system; and set data synchronization strategies to collect and store data at different granularities such as daily, hourly, and real-time according to system characteristics, and automatically transfer data from multiple systems to the data fusion platform. The data fusion platform performs deduplication, completion, anomaly handling, and format standardization on the data, and maps and associates the data of the same business entity in various systems to eliminate redundancy and discrepancies.

3. The comprehensive analysis and evaluation method for power supply stations based on artificial intelligence according to claim 2, characterized in that, The data of the same business entity in various systems are mapped and associated to eliminate redundancy and discrepancies, as detailed below: To achieve unified entity / field alignment and master data association for data items from different sources, with different names, and different granularities, the Jaccard coefficient is used to measure the similarity between two sets A and B. ； If the similarity J(A,B)>θ, where θ is a preset threshold, then A and B are considered to be the same business entity. And by calculating the correlation matrix, highly redundant features r were discovered. xy: ; Where, x a ′、y a 'a' represents the a'th data point of features x and y, respectively; n' represents the amount of data. and Let x be the mean of feature x and feature y; When |r xy When |>0.8, the two features are considered highly correlated, and only one is retained; Multi-feature data can be dimensionality reduced and principal components extracted using PCA, thus reducing the analytical burden. The core data of all power supply stations, after extraction, cleaning, alignment, and dimensionality reduction, are transformed into multi-system fused data.

4. The comprehensive analysis and evaluation method for power supply stations based on artificial intelligence according to claim 1, characterized in that, The preprocessed data is then subjected to feature engineering to remove redundant data and unify business definitions, resulting in a key feature dataset, as detailed below: Based on the preprocessed multi-system fusion data, key features are selected through feature engineering. The IV value is used to measure the predictive power of an indicator on a target variable. ； Among them, IV j The information value of feature j; n j Number of bins for feature j; Good ja The number of positive samples in the a-th bin; Bad ja G represents the number of negative samples in the a-th bin of feature j; total B total These represent the total number of positive and negative samples, respectively; indicators with IV ≥ 0.1 are retained, while those with IV < 0.02 are removed. Based on the reduction of information entropy during decision tree splitting: ； Where: S is the sample set; Values(j) are the possible values ​​(or binning) of feature j; S v Let S be the subset of samples where feature j takes the value v; H(S) is the entropy of the current sample set S; Gain(S,j) is the information gain of feature j on the sample set S; H(S) v ) is a subset S v entropy; Calculate the information gain of all features, select the features with the largest Z% gain, and calculate the feature contribution using a tree model: ； Where, φ ij φ is the SHAP value of the j-th feature of the i-th sample; j The average absolute contribution of feature j; n is the number of samples; The rate of change of the index IV value is calculated using a time sliding window; if the decay exceeds a threshold, replacement is triggered. ； Among them, DR j Let be the IV decay rate of feature j; The IV value of the latest time window feature j; The IV value of feature j within the historical time window; By aligning business definitions with a rules engine, key features are unified to the same dimension and distribution, resulting in a key feature dataset.

5. The comprehensive analysis and evaluation method for power supply stations based on artificial intelligence according to claim 1, characterized in that, The AI-driven mental indicator library and indicator relationships are constructed based on the key feature dataset, as detailed below: S31: Transform key features into interpretable business metrics, obtain a list of business metrics, and establish a cross-domain semantic association matrix; S32: Combine the business indicator list and the cross-domain semantic association matrix to construct an indicator relationship knowledge graph, and use the graph structure to express the logical, causal, and business relationships between indicators.

6. The comprehensive analysis and evaluation method for power supply stations based on artificial intelligence according to claim 5, characterized in that, Key features are transformed into interpretable business metrics, and cross-domain semantic relationships are established, as follows: Establish a feature-indicator mapping dictionary, define composite indicators according to business rules, map each feature to a business-interpretable atomic or composite indicator, and supplement the generated indicators with name, business definition, domain, and unit of measurement metadata to obtain a list of business indicators. Using a pre-trained BERT model, the metric names and business definitions are converted into vector representations E': E'=BERT(metric name, business definition); And calculate the cosine similarity of each index: ； Where Sim(U,V) represents the semantic similarity between indices U and V, and E' represents the semantic similarity between them. U 、E' V These are vector representations of indices U and V, respectively; For vector E' U The model; For vector E' V The modulus is determined; and indicators whose Sim(U,V) is greater than the threshold are grouped into the same business domain or cross-domain group to construct a cross-domain semantic association matrix.

7. The comprehensive analysis and evaluation method for power supply stations based on artificial intelligence according to claim 5, characterized in that, The method of combining the business indicator list and the cross-domain semantic association matrix to construct an indicator relationship knowledge graph is as follows: Generate graph node V based on the business indicator list, including atomic indicator nodes, composite indicator nodes and business entity nodes; Edge relationship E modeling includes three types of relationships: logical dependency, causal inference, and business co-occurrence. The logical dependency is obtained through explicit calculation between indicators; the causal relationship is obtained through time series causal inference; and the business co-occurrence is obtained through the co-occurrence frequency of indicators in the same business scenario. Based on the graph nodes V and edge relationships E, an index relationship knowledge graph G=(V,E) is constructed.

8. The comprehensive analysis and evaluation method for power supply stations based on artificial intelligence according to claim 1, characterized in that, Based on the heart indicator database and indicator relationships, scientific weights are set, scores for each indicator are calculated, and a total score is formed, as detailed below: For each index j' of each object i', standardize: ； Where, x i′j′ z is the j'th original index value of the i'th object; i′j′ The standard score of the j'th metric for the i'th object after standardization; max i′ x i′j′ The maximum value of the j'-th index among all objects; min i′ x i′j′ Let j' be the minimum value of the j'-th index among all objects; Calculate the objective weights using the entropy weight method: ; ; ; Where, p i′j′ Ej′ represents the normalized weight of object i' on index j'; Ej′ represents the information entropy of the j'-th index. Let be the objective weight of the j'th indicator; M is the total number of objects, and N is the total number of indicators; Calculate subjective weights using AHP expert weights: Judgment Matrix ; The weights are obtained by normalizing the eigenvectors: ； The weights are adjusted based on the relationships in the knowledge graph to obtain the final weights. C is the relational influence coefficient matrix: ； Where I is the identity matrix; The result is either AHP or entropy weighting. Then the final weight of indicator j' is w. j' : ； in, α, β, γ The fusion ratio of expert subjective weight, objective weight, and relationship-adjusted weight; Then the score of a single indicator S i'j' : ； Further obtain the overall evaluation score S of the i'th evaluation object. i' : 。 9. The comprehensive analysis and evaluation method for power supply stations based on artificial intelligence according to claim 8, characterized in that, Based on the scoring results, the horizontal ranking of similar power supply stations across the entire region is assigned scores, and cluster analysis is used for segmentation, as follows: Based on the comprehensive score obtained from S4, the scores of each power supply station are ranked from high to low; K-means clustering is used to scientifically segment the comprehensive scores and key business indicator scores of all similar power supply stations into high-performance, medium-performance, and low-performance segments; finally, the final label of each power supply station is obtained, including its comprehensive score, horizontal ranking and ranking score across the entire region, and the segment it belongs to in which cluster.

10. A comprehensive analysis and evaluation system for power supply stations based on artificial intelligence, characterized in that, It includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it specifically performs the steps in the comprehensive analysis and evaluation method for power supply stations based on artificial intelligence as described in any one of claims 1-9.

Images