Digital management method of power system business environment based on big data analysis
Through big data analysis and optimization algorithms, combined with natural language processing and blockchain technology, the problem that traditional power system management models are difficult to adapt to regional needs is solved, and the optimization of the power system business environment and the flexibility and transparency of policy suggestions are achieved.
Patent Information
- Application Number
- CN202411400593.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-09
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-10-09
AI Technical Summary
The traditional power system management model is difficult to adapt to the specific needs and changing environments in different regions, which makes it difficult to ensure the efficiency of policy formulation and implementation, and lacks a systematic analysis and feedback mechanism, which affects the optimization of the business environment.
Using a method based on big data analysis, user portrait analysis and business environment scoring are carried out through natural language processing, machine learning, blockchain technology and optimization algorithms, dynamically adjust and iterate power policy plans, and generate policy suggestions for multi-objective comprehensive optimization .
It has achieved comprehensive optimization of the business environment of the power system, improved the flexibility and effectiveness of policies, ensured the adaptability and transparency of policies, and reduced the risks of artificial errors and data tampering.
Smart Images

Figure CN118898498B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of big data analysis, and in particular to a digital management method for the business environment of an electric power system based on big data analysis. Background Art
[0002] The management and optimization of power systems are directly related to social and economic development and the quality of life of residents. With the rapid development of economic globalization and information technology, the management methods of power systems are also constantly innovating. Especially in the current era of big data, the traditional power system management model can no longer meet the growing needs and increasingly complex management environment.
[0003] Moreover, in the management of power systems, the formulation and optimization of power policy solutions are crucial. However, traditional power policy formulation relies on manual experience, lacks systematic analysis and optimization methods, and is often difficult to adapt to the specific needs and changing environments of different regions. There is also a lack of systematic evaluation and feedback mechanisms for the implementation effects of policy solutions and their impact on the business environment. This makes it difficult to ensure the efficiency and accuracy of policy implementation, affecting the optimization of the power business environment. Summary of the invention
[0004] In order to solve the above problems, the purpose of the present invention is to provide a digital management method for the business environment of the power system based on big data analysis. Through a comprehensive analysis of user portraits and business environment scores, combined with the practical application of optimization algorithms, it is possible to generate multi-objective comprehensive optimization policy recommendations. Dynamic adjustment and iterative strategies ensure the flexibility and effectiveness of the policies, and gradually achieve comprehensive optimization of the business environment of the power system.
[0005] To achieve the above object, the present invention adopts the following technical solutions:
[0006] A digital management method for the business environment of a power system based on big data analysis includes the following steps:
[0007] S1: Obtain the electricity policy plan, use natural language processing technology to perform word segmentation, named entity recognition and relationship extraction on the policy content, and dynamically generate a form based on different policy types:
[0008] S2: Use machine learning models and pattern recognition to conduct intelligent form review, and use blockchain technology to record and trace the review process to ensure transparency and reliability;
[0009] S3: Obtain data including regional policy implementation, business environment data, economic data and customer feedback, and pre-process the collected data;
[0010] S4: Store the preprocessed data into the database and use blockchain technology to record the data source and operation history;
[0011] S5: Based on the preprocessed data, conduct user portrait analysis and power system business environment analysis to obtain regional user portraits and business environment scores;
[0012] S6: Based on the regional user profile and business environment score, optimize the electricity policy plan and generate an electricity policy plan suitable for the region.
[0013] Furthermore, S1 is specifically:
[0014] Obtain the latest text data on power policy plans related to the power system business environment, use natural language processing technology to analyze policy texts, and build a policy knowledge graph;
[0015] According to the extracted policy type P=f(E,R), P represents the identified policy type, f represents the policy type identification model, whose input is the extracted entity E and relationship R, selects the corresponding form template from the form template library, and automatically fills the extracted entity information into the corresponding fields of the form to generate the final form.
[0016] Furthermore, natural language processing technology is used to analyze the policy text and construct a policy knowledge graph, as follows:
[0017] Suppose the original text is T, and the original text is preprocessed, and the preprocessed text is T′:
[0018] ;
[0019] Among them, clean() is a text cleaning function; the function clean() includes removing HTML tags, removing special symbols, converting to standard format, removing extra spaces and data conversion;
[0020] Segment the cleaned text T′ to obtain a word set ,in w u Indicates the uth word or phrase in the word segmentation result;
[0021] Perform named entity recognition on the segmented text to obtain an entity set:
[0022] ;
[0023] Among them, e i is the i-th entity identified, label i is the label of the entity;
[0024] Extract the relationship between entities from the text and obtain a relationship set:
[0025]
[0026] in, For entity e i and e j the relationship between
[0027] Construct a policy knowledge graph G=(R,E) based on the identified entities and relations.
[0028] Furthermore, the policy type recognition model is constructed based on the multi-layer perceptron model, as follows:
[0029] By converting the information of entities and relations into vector representation, we get:
[0030] V E =[v E1 ,v E2 ,…,v Ek ];
[0031] V R =[v R1 ,v R2 ,…,v Rl ];
[0032] Among them, V E is the vector representation of the entity; V R is the vector representation of the relationship; v Rl represents the number of relations in the lth category, where v Ek represents the number of entities in the kth category;
[0033] Construct the input vector for the multilayer perceptron model:
[0034] V=[V E, V R ];
[0035] The hidden layer of the multilayer perceptron model calculates:
[0036] h = σ(WV + b);
[0037] Among them, σ is the activation function, W is the hidden layer weight, b is the hidden layer bias; h is the hidden layer output;
[0038] Output layer calculation:
[0039] ;
[0040] Among them, softmax is the activation function, and are the output layer weights and biases respectively
[0041] The final P is the probability distribution set of each policy type, from which the policy type corresponding to the maximum probability is selected.
[0042] Furthermore, blockchain technology is used to record data sources and operation history, specifically:
[0043] Based on the Hyperledger Fabric blockchain platform, operations include data insertion create, data update update, and data deletion delete;
[0044] For each operation, the following information is recorded: data ID; timestamp: operation time; operation type; operator information; data content: data content before, after, or after the operation;
[0045] Then, a blockchain transaction consists of:
[0046] Data Insertion Transaction:
[0047] Tinsert=(dataID,timestamp,operationType='create',operatorInfo,newDataContent);
[0048] Among them, dataID is the ID that uniquely identifies the data item, timestamp is the timestamp, that is, the exact time when the operation occurred; operationType is the operation type; operatorInfo is the operator information; newDataContent is the specific content of the new data inserted;
[0049] Data update transaction:
[0050] ;
[0051] Among them, oldDataContent is the data content before updating;
[0052] Data Deletion Transactions:
[0053] Tdelete=(dataID,timestamp,operationType='delete',operatorInfo,deletedDataContent)
[0054] Among them, deletedDataContent is the deleted data content;
[0055] Send this information as a transaction record to the blockchain;
[0056] And define smart contracts on the blockchain platform to manage data storage and traceability.
[0057] Further, the user portrait analysis is as follows:
[0058] Extract features from electricity consumption, service feedback, and economic data to form a user feature vector, including electricity consumption features, service feedback features, and economic data features;
[0059] The power consumption characteristics include average daily power consumption , Peak power consumption P peak、 Valley power consumption P valley And seasonal electricity consumption characteristics S season ;
[0060] The service feedback characteristics include average satisfaction , satisfaction standard deviation Service response time T response and the number of complaints per month C monthly ;
[0061] The economic data characteristics include the average income of users , income standard deviation σ I , Regional GDP growth rate G GDP and the proportion of electricity consumption by industry industry ;
[0062] Get the user feature vector:
[0063] X=[ ,P peak , P valley ,S season , , ,T response ,C monthly , , σ I ,G GDP ,R industry ];
[0064] Based on the user feature vector, a hierarchical clustering algorithm is used to classify users into different clusters;
[0065] Use clustering results for supervised learning and build a user portrait model.
[0066] Furthermore, based on the user feature vector, a hierarchical clustering algorithm is used to classify users into different clusters, as follows:
[0067] (1) Use Z-score to standardize user feature vectors;
[0068] (2) Calculate the Euclidean distance between user features;
[0069] (3) Adopting the agglomerative hierarchical clustering method, the minimum chain method is used to calculate the distance between clusters, and the two clusters with the smallest distance are selected to merge;
[0070] ;
[0071] in, Represent two clusters, is the Euclidean distance between the points in the cluster; They are Points in a cluster;
[0072] (4) Update the distance matrix and repeat step (3) until all users are merged into one cluster; generate a dendrogram;
[0073] (5) Observe the maximum vertical distance through the dendrogram to preliminarily determine the optimal clipping point, and compare the silhouette coefficients under different cluster numbers to determine the final optimal cluster number;
[0074] sample The silhouette coefficient for:
[0075] ;
[0076] in, It is a sample The average distance to other points in the same cluster; It is a sample The average distance to the nearest neighbor cluster;
[0077] (6) Use the determined optimal number of clusters for final clustering and add labels to the original data to obtain the final clustering result.
[0078] Further, the business environment analysis is as follows:
[0079] Integrate policy data, user feedback, and economic data to build a comprehensive feature vector dataset ;
[0080] Build a business environment scoring model based on random forest to predict the business environment score:
[0081] Choose the number of trees Q and hyperparameters;
[0082] For each decision tree, a subset of the comprehensive feature vector dataset is randomly selected for training;
[0083] For each node, a certain number of features are randomly selected and divided based on the optimal partition point until the stopping condition is met
[0084] Integrate all decision trees and make the final prediction by averaging the prediction results of each tree;
[0085] Assume that the random forest contains Q decision trees, and the tth tree is f t , then the random forest prediction value is:
[0086] .
[0087] Furthermore, based on the regional user profile and business environment score, the power policy plan is optimized to generate a power policy plan suitable for the region, as follows:
[0088] According to the obtained regional user portrait feature vector Business environment score ;
[0089] Set the optimization objective function and constraints, and apply genetic algorithms to generate the optimal strategy;
[0090] Among them, the objective function is:
[0091] ;
[0092] in, , , and is the weight parameter, Energy loss of the power system; is the maximum energy loss of the power system; The average electricity cost for users; Score personalized needs based on user portraits;
[0093] The constraints include:
[0094] Electricity supply and demand balance constraints:
[0095] ;
[0096] in, is the number of users, For the The electricity demand of each user; The total amount of electricity supplied for electricity;
[0097] User satisfaction constraints:
[0098] Satisfaction min ≤Satisfaction≤Satisfaction max ;
[0099] Among them, Satisfaction is user satisfaction; Satisfaction minand Satisfaction max They are the lowest and highest user satisfaction respectively;
[0100] Environmental impact constraints:
[0101] E emission ≤E emissionlimit;
[0102] Among them, E emission is the emission during electricity production, E emissionlimit for environmental emission limits;
[0103] Generate optimized power policy solutions based on the optimal strategy;
[0104] Apply the optimized electricity policy to actual scenarios, continuously monitor policy effects, and dynamically adjust and optimize based on feedback.
[0105] Furthermore, the genetic algorithm is used to generate the optimal strategy, as follows:
[0106] (1) Generate an initial population with a population size of , which consists of each individual electricity policy parameter:
[0107] ;
[0108] (2) For each individual, calculate the fitness and crowding degree based on the objective function and dominance relationship;
[0109] (3) According to Pareto superiority and crowding, the selection operation is performed, and the individual with higher Pareto superiority and greater crowding is selected. The selection probability is:
[0110] ;
[0111] in, To select individuals probability; For individuals The non-dominated sorting level of ; For individuals The degree of congestion; For individuals The degree of congestion;
[0112] (4) Performing crossover operations on selected individuals based on simulated binary crossover to generate offspring, and the new individuals undergo mutation operations using multiple nonlinear mutations;
[0113] (5) Use the new offspring and the previous generation mixed population to form a new population based on non-dominated sorting and crowding;
[0114] (6) Repeat steps (2)-(5) until the maximum number of generations is reached or the convergence condition is met.
[0115] The present invention has the following beneficial effects:
[0116] 1. The present invention can generate policy recommendations for multi-objective comprehensive optimization through comprehensive analysis of user portraits and business environment scores, combined with the practical application of optimization algorithms. Dynamic adjustment and iteration strategies ensure the flexibility and effectiveness of policies, and gradually achieve comprehensive optimization of the business environment of the power system.
[0117] 2. The present invention uses natural language processing technology to perform word segmentation, named entity recognition and relationship extraction on policy content, and can automatically generate structured forms to ensure that policy information is complete and standardized. It also uses machine learning models and pattern recognition technology to automatically conduct intelligent review of form content, improve review efficiency and accuracy, and reduce manual errors. The review process is recorded using blockchain technology, which can trace the review process, ensure transparency and reliability, and prevent data tampering;
[0118] 3. The present invention optimizes the electricity policy plan based on regional user portraits and business environment scores, generates policies suitable for specific regions, ensures the effective implementation of policies, and continuously optimizes the electricity policy plan through dynamic data updates and analysis, improves the flexibility and adaptability of policies, and is able to cope with rapidly changing market environments and user needs. BRIEF DESCRIPTION OF THE DRAWINGS
[0119] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0120] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:
[0121] refer to Figure 1 In this embodiment, a digital management method for the business environment of a power system based on big data analysis is provided, comprising the following steps:
[0122] S1: Obtain the electricity policy plan, use natural language processing technology to perform word segmentation, named entity recognition and relationship extraction on the policy content, and dynamically generate a form based on different policy types:
[0123] S2: Use machine learning models and pattern recognition to conduct intelligent form review, and use blockchain technology to record and trace the review process to ensure transparency and reliability;
[0124] S3: Obtain data including regional policy implementation status (such as policy implementation progress, social impact, economic impact, etc.), business environment data, economic data and customer feedback data, and pre-process the collected data;
[0125] S4: Store the preprocessed data in the database and use blockchain technology to record the data source and operation history; ensure data accuracy, completeness and traceability;
[0126] S5: Based on the preprocessed data, conduct user portrait analysis and power system business environment analysis to obtain regional user portraits and business environment scores;
[0127] S6: Based on the regional user profile and business environment score, optimize the electricity policy plan and generate an electricity policy plan suitable for the region.
[0128] In this embodiment, S1 is specifically:
[0129] Obtain the latest text data on power policy plans related to the power system business environment, use natural language processing (NLP) technology to analyze policy texts, and build a policy knowledge graph;
[0130] According to the extracted policy type P=f(E,R), P represents the identified policy type, f represents the policy type identification model, whose input is the extracted entity E and relationship R, selects the corresponding form template from the form template library, and automatically fills the extracted entity information into the corresponding fields of the form to generate the final form.
[0131] In this embodiment, natural language processing technology is used to analyze the policy text and construct a policy knowledge graph, as follows:
[0132] Suppose the original text is T, and the original text is preprocessed, and the preprocessed text is T′:
[0133] ;
[0134] Among them, clean() is a text cleaning function; (the function clean() includes removing HTML tags, removing special symbols, converting to standard format, removing extra spaces and data conversion);
[0135] Segment the cleaned text T′ to obtain a word set ,in w u Indicates the uth word or phrase in the word segmentation result;
[0136] Perform named entity recognition on the segmented text to obtain an entity set:
[0137] ;
[0138] Among them, e iis the i-th entity identified, label i is the label of the entity;
[0139] Extract the relationship between entities from the text and obtain a relationship set:
[0140]
[0141] in, For entity e i and e j the relationship between
[0142] Construct a policy knowledge graph G=(R,E) based on the identified entities and relations.
[0143] In this implementation, an example is given as follows:
[0144] Tokenization: Split text into individual words.
[0145] For example, use the Jieba word segmentation tool to segment the text "encourage enterprises to increase innovation in power technology" to get a word list ={"encourage", "enterprise", "increase", "electricity", "technology", "innovation"}, where Represents the word list after word segmentation.
[0146] Named Entity Recognition (NER): Identify key entities such as policy objectives, policy objects, and policy measures.
[0147] For example, to identify entities in the text "In order to promote the development of new energy (e1, goal), relevant departments will provide financial subsidies (e2, measures)", we get the named entity set: E = {(e1, goal), (e2, measures)}, where E represents the identified named entity set, e i represents the i-th entity, label i Indicates the type of this entity.
[0148] Relationship extraction: Analyze the relationships between entities, such as subject-predicate-object relationships, modification relationships, etc., to build a policy knowledge graph.
[0149] For example, extracting the text "department(e1)" will simplify(r 12 ) approval process (e2), reduce (r 13 ) Enterprise electricity cost (e3)" and obtain the relationship set:
[0150] R={(e1,r 12 ,e2),(e1,r 13 ,e3)}, where R represents the extracted relation set, ei and e j Represents two entities, r ij Indicates the type of relationship between them.
[0151] In this embodiment, the policy type identification model is constructed based on a multi-layer perceptron model, as follows:
[0152] By converting the information of entities and relations into vector representation, we get:
[0153] V E =[v E1 ,v E2 ,…,v Ek ];
[0154] V R =[v R1 ,v R2 ,…,v Rl ];
[0155] Among them, V E is the vector representation of the entity; V R is the vector representation of the relationship; v Rl represents the number of relations in the lth category, where v Ek represents the number of entities in the kth category;
[0156] Construct the input vector for the multilayer perceptron model:
[0157] V=[V E, V R ];
[0158] The hidden layer of the multilayer perceptron model calculates:
[0159] h = σ(WV + b);
[0160] Among them, σ is the activation function, W is the hidden layer weight, b is the hidden layer bias; h is the hidden layer output;
[0161] Output layer calculation:
[0162] ;
[0163] Among them, softmax is the activation function, and are the output layer weights and biases respectively
[0164] The final P is the probability distribution set of each policy type, from which the policy type corresponding to the maximum probability is selected.
[0165] In this embodiment, blockchain technology is used to record data sources and operation history, specifically:
[0166] Based on the Hyperledger Fabric blockchain platform, operations include data insertion create, data update update, and data deletion delete;
[0167] For each operation, the following information is recorded: data ID; timestamp: operation time; operation type; operator information; data content: data content before, after, or after the operation;
[0168] Then, a blockchain transaction consists of:
[0169] Data Insertion Transaction:
[0170] Tinsert=(dataID,timestamp,operationType='create',operatorInfo,newDataContent);
[0171] Among them, dataID is the ID that uniquely identifies the data item, timestamp is the timestamp, that is, the exact time when the operation occurred; operationType is the operation type; operatorInfo is the operator information; newDataContent is the specific content of the new data inserted;
[0172] Data update transaction:
[0173] ;
[0174] Among them, oldDataContent is the data content before updating;
[0175] Data Deletion Transactions:
[0176] Tdelete=(dataID,timestamp,operationType='delete',operatorInfo,deletedDataContent)
[0177] Among them, deletedDataContent is the deleted data content;
[0178] Send this information as a transaction record to the blockchain;
[0179] And define smart contracts on the blockchain platform to manage data storage and traceability.
[0180] In this embodiment, the user portrait analysis is as follows:
[0181] Extract features from electricity consumption, service feedback, and economic data to form a user feature vector, including electricity consumption features, service feedback features, and economic data features;
[0182] The power consumption characteristics include average daily power consumption , Peak power consumption P peak、 Valley power consumption P valley And seasonal electricity consumption characteristics S season ;
[0183] The service feedback characteristics include average satisfaction , satisfaction standard deviation Service response time T response and the number of complaints per month C monthly ;
[0184] The economic data characteristics include the average income of users , income standard deviation σ I , Regional GDP growth rate G GDP and the proportion of electricity consumption by industry industry ;
[0185] Get the user feature vector:
[0186] X=[ ,P peak , P valley ,S season , , ,T response ,C monthly , , σ I ,G GDP ,R industry ];
[0187] Based on the user feature vector, a hierarchical clustering algorithm is used to classify users into different clusters;
[0188] Use clustering results for supervised learning and build a user portrait model.
[0189] In this embodiment, based on the user feature vector, a hierarchical clustering algorithm is used to classify users into different clusters, as follows:
[0190] (1) Use Z-score to standardize user feature vectors;
[0191] (2) Calculate the Euclidean distance between user features;
[0192] (3) Adopting the agglomerative hierarchical clustering method, the minimum chain method is used to calculate the distance between clusters, and the two clusters with the smallest distance are selected to merge;
[0193] ;
[0194] in, Represent two clusters, is the Euclidean distance between the points in the cluster; They are Points in a cluster;
[0195] (4) Update the distance matrix and repeat step (3) until all users are merged into one cluster; generate a dendrogram;
[0196] (5) Observe the maximum vertical distance through the dendrogram to preliminarily determine the optimal clipping point, and compare the silhouette coefficients under different cluster numbers to determine the final optimal cluster number;
[0197] sample The silhouette coefficient for:
[0198] ;
[0199] in, It is a sample The average distance to other points in the same cluster; It is a sample The average distance to the nearest neighbor cluster;
[0200] (6) Use the determined optimal number of clusters for final clustering and add labels to the original data to obtain the final clustering result.
[0201] In this embodiment, the business environment analysis is as follows:
[0202] Integrate policy data, user feedback, and economic data to build a comprehensive feature vector dataset ;
[0203] Build a business environment scoring model based on random forest to predict the business environment score:
[0204] Choose the number of trees Q and hyperparameters;
[0205] For each decision tree, a subset of the comprehensive feature vector dataset is randomly selected for training;
[0206] For each node, a certain number of features are randomly selected and divided based on the optimal partition point until the stopping condition is met
[0207] Integrate all decision trees and make the final prediction by averaging the prediction results of each tree;
[0208] Assume that the random forest contains Q decision trees, and the tth tree is ft , then the random forest prediction value is:
[0209] .
[0210] In this embodiment, based on the regional user profile and business environment score, the power policy plan is optimized to generate a power policy plan suitable for the region, as follows:
[0211] According to the obtained regional user portrait feature vector Business environment score ;
[0212] Set the optimization objective function and constraints, and apply genetic algorithms to generate the optimal strategy;
[0213] Among them, the objective function is:
[0214] ;
[0215] in, , , and is the weight parameter, Energy loss of the power system; is the maximum energy loss of the power system; The average electricity cost for users; Score personalized needs based on user portraits;
[0216] The constraints include:
[0217] Electricity supply and demand balance constraints:
[0218] ;
[0219] in, is the number of users, For the The electricity demand of each user; The total amount of electricity supplied for electricity;
[0220] User satisfaction constraints:
[0221] Satisfaction min ≤Satisfaction≤Satisfaction max ;
[0222] Among them, Satisfaction is user satisfaction; Satisfaction min and Satisfaction max They are the lowest and highest user satisfaction respectively;
[0223] Environmental impact constraints:
[0224] E emission ≤E emissionlimit ;
[0225] Among them, E emission is the emission during electricity production, E emissionlimit for environmental emission limits;
[0226] Generate optimized power policy solutions based on the optimal strategy;
[0227] Apply the optimized electricity policy to actual scenarios, continuously monitor policy effects, and dynamically adjust and optimize based on feedback.
[0228] In this embodiment, a genetic algorithm is used to generate the optimal strategy, as follows:
[0229] (1) Generate an initial population with a population size of , which consists of each individual electricity policy parameter:
[0230] ;
[0231] (2) For each individual, calculate the fitness and crowding degree based on the objective function and dominance relationship;
[0232] (3) According to Pareto superiority and crowding, the selection operation is performed, and the individual with higher Pareto superiority and greater crowding is selected. The selection probability is:
[0233] ;
[0234] in, To select individuals probability; For individuals The non-dominated sorting level of ; For individuals The degree of congestion; For individuals The degree of congestion;
[0235] (4) Performing crossover operations on selected individuals based on simulated binary crossover to generate offspring, and the new individuals undergo mutation operations using multiple nonlinear mutations;
[0236] (5) Use the new offspring and the previous generation mixed population to form a new population based on non-dominated sorting and crowding;
[0237] (6) Repeat steps (2)-(5) until the maximum number of generations is reached or the convergence condition is met.
[0238] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented 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.
[0239] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0240] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0241] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0242] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.
Claims
1. Digital management method of power system business environment based on big data analysis, It is characterized in that The following steps are involved: S1: Obtain the electricity policy plan, perform word segmentation, named entity recognition and relationship extraction on the policy content through natural language processing technology, and dynamically generate forms according to different policy types; S2: Use machine learning models and pattern recognition to conduct intelligent audits of forms, and use blockchain technology to record and trace the audit process; S3: Obtain data including regional policy implementation, business environment data, economic data and customer feedback, and pre-process the collected data; S4: Store the preprocessed data into the database and use blockchain technology to record the data source and operation history; S5: Based on the preprocessed data, conduct user portrait analysis and power system business environment analysis to obtain regional user portraits and business environment scores; The user portrait analysis is as follows: Extract features from electricity consumption, service feedback, and economic data to form a user feature vector, including electricity consumption features, service feedback features, and economic data features; The power consumption characteristics include average daily power consumption , Peak power consumption P peak、 Valley power consumption P valley And seasonal electricity consumption characteristics S season ; The service feedback characteristics include average satisfaction , Satisfaction standard deviation , service response time T response and the number of complaints per month C monthly ; The economic data characteristics include the average income of users , income standard deviation σ I , Regional GDP growth rate G GDP and the proportion of electricity consumption by industry industry ; Get the user feature vector: X=[ ,P peak , P valley ,S season , , ,T response ,C monthly , , σ I ,G GDP ,R industry ]; Based on the user feature vector, a hierarchical clustering algorithm is used to classify users into different clusters; Use clustering results for supervised learning and build a user portrait model; Based on the user feature vector, a hierarchical clustering algorithm is used to classify users into different clusters, as follows: (1) Use Z-score to standardize user feature vectors; (2) Calculate the Euclidean distance between user features; (3) Adopting the agglomerative hierarchical clustering method, the minimum chain method is used to calculate the distance between clusters, and the two clusters with the smallest distance are selected to merge; ; in, Represent two clusters, is the Euclidean distance between the points in the cluster; They are Points in a cluster; (4) Update the distance matrix and repeat step (3) until all users are merged into one cluster; generate a dendrogram; (5) Observe the maximum vertical distance through the dendrogram to preliminarily determine the optimal clipping point, and compare the silhouette coefficients under different cluster numbers to determine the final optimal cluster number; sample The silhouette coefficient for: ; in, It is a sample The average distance to other points in the same cluster; It is a sample The average distance to the nearest neighbor cluster; (6) Use the determined optimal number of clusters for final clustering and add the labels to the original data to obtain the final clustering result; S6: Based on the regional user profile and business environment score, optimize the electricity policy plan and generate an electricity policy plan suitable for the region.
2. According to the digital management method of the power system business environment based on big data analysis according to claim 1, It is characterized in that The S1 is specifically: Obtain the latest text data on power policy plans related to the power system business environment, use natural language processing technology to analyze policy texts, and build a policy knowledge graph; According to the extracted policy type P=f(E,R), P represents the identified policy type, f represents the policy type identification model, whose input is the extracted entity E and relationship R, selects the corresponding form template from the form template library, and automatically fills the extracted entity information into the corresponding fields of the form to generate the final form.
3. According to the digital management method of the power system business environment based on big data analysis according to claim 2, It is characterized in that The use of natural language processing technology to analyze policy texts and construct a policy knowledge graph is as follows: Suppose the original text is T, and the original text is preprocessed, and the preprocessed text is T′: ; Among them, clean() is a text cleaning function; the function clean() includes removing HTML tags, removing special symbols, converting to standard format, removing extra spaces and data conversion; Segment the cleaned text T′ to obtain a word set ,in w u Indicates the uth word or phrase in the word segmentation result; Perform named entity recognition on the segmented text to obtain an entity set: ; Among them, e i is the i-th entity identified, label i is the label of the entity; Extract the relationship between entities from the text and obtain a relationship set: in, For entity e i and e j the relationship between Construct a policy knowledge graph G=(R,E) based on the identified entities and relations.
4. According to the digital management method of the power system business environment based on big data analysis according to claim 3, It is characterized in that The policy type identification model is constructed based on a multi-layer perceptron model, as follows: By converting the information of entities and relations into vector representation, we get: V E =[v E1 ,v E2 ,…,v Ek ]; V R =[v R1 ,v R2 ,…,v Rl ]; Among them, V E is the vector representation of the entity; V R is the vector representation of the relationship; v Rl represents the number of relations in the lth category, where v Ek represents the number of entities in the kth category; Construct the input vector for the multilayer perceptron model: V=[V E, V R ]; The hidden layer of the multilayer perceptron model calculates: h = σ(WV + b); Among them, σ is the activation function, W is the hidden layer weight, b is the hidden layer bias; h is the hidden layer output; Output layer calculation: ; Among them, softmax is the activation function, and are the output layer weights and biases respectively The final P is the probability distribution set of each policy type, from which the policy type corresponding to the maximum probability is selected.
5. According to the digital management method of the power system business environment based on big data analysis according to claim 1, It is characterized in that The use of blockchain technology to record data sources and operation history is specifically as follows: Based on the Hyperledger Fabric blockchain platform, operations include data insertion create, data update update, and data deletion delete; For each operation, the following information is recorded: Data ID; Timestamp: operation time; Operation type; Operator information; Data content: data content before, after, or after the operation; Then, a blockchain transaction consists of: Data Insertion Transaction: Tinsert=(dataID,timestamp,operationType='create',operatorInfo,newDataContent); Among them, dataID is the ID that uniquely identifies the data item, timestamp is the timestamp, that is, the exact time when the operation occurred; operationType is the operation type; operatorInfo is the operator information; newDataContent is the specific content of the new data inserted; Data update transaction: ; Among them, oldDataContent is the data content before updating; Data Deletion Transactions: Tdelete=(dataID,timestamp,operationType='delete',operatorInfo,deletedDataContent) Among them, deletedDataContent is the deleted data content; Send this information as a transaction record to the blockchain; And define smart contracts on the blockchain platform to manage data storage and traceability.
6. According to claim 1, the digital management method for the business environment of the power system based on big data analysis, It is characterized in that The business environment analysis is as follows: Integrate policy data, user feedback, and economic data to build a comprehensive feature vector dataset ; Build a business environment scoring model based on random forest to predict the business environment score: Choose the number of trees Q and hyperparameters; For each decision tree, a subset of the comprehensive feature vector dataset is randomly selected for training; For each node, a certain number of features are randomly selected and divided based on the optimal partition point until the stopping condition is met Integrate all decision trees and make the final prediction by averaging the prediction results of each tree; Assume that the random forest contains Q decision trees, and the tth tree is f t , then the random forest prediction value is: 。 7. According to claim 1, the digital management method for the business environment of the power system based on big data analysis, It is characterized in that The above-mentioned optimization of the power policy plan based on the regional user profile and business environment score, and the generation of a power policy plan suitable for the region are as follows: According to the obtained regional user portrait feature vector Business environment score ; Set the optimization objective function and constraints, and apply genetic algorithms to generate the optimal strategy; Among them, the objective function is: ; in, , , and is the weight parameter, Energy loss of the power system; is the maximum energy loss of the power system; The average electricity cost for users; Score personalized needs based on user portraits; The constraints include: Electricity supply and demand balance constraints: ; in, is the number of users, For the The electricity demand of each user; The total amount of electricity supplied for electricity; User satisfaction constraints: Satisfaction min ≤Satisfaction≤Satisfaction max ; Among them, Satisfaction is user satisfaction; Satisfaction min and Satisfaction max They are the lowest and highest user satisfaction respectively; Environmental impact constraints: AND emission ≤E emissionlimit; Among them, E emission is the emission during electricity production, E emissionlimit for environmental emission limits; Generate optimized power policy solutions based on the optimal strategy; Apply the optimized electricity policy to actual scenarios, continuously monitor policy effects, and make dynamic adjustments and optimizations based on feedback.
8. According to the digital management method of the power system business environment based on big data analysis according to claim 7, It is characterized in that The application of genetic algorithm to generate the optimal strategy is as follows: (1) Generate an initial population with a population size of , which consists of each individual electricity policy parameter: ; (2) For each individual, calculate the fitness and crowding degree based on the objective function and dominance relationship; (3) Based on Pareto superiority and crowding, the selection operation is performed, and the individual with higher Pareto superiority and greater crowding is selected. The selection probability is: ; in, To select individuals probability; For individuals The non-dominated sorting level of ; For individuals The degree of congestion; For individuals The degree of congestion; (4) Performing crossover operations on selected individuals based on simulated binary crossover to generate offspring, and the new individuals undergo mutation operations using multiple nonlinear mutations; (5) Use the new offspring and the previous generation mixed population to form a new population based on non-dominated sorting and crowding; (6) Repeat steps (2)-(5) until the maximum number of generations is reached or the convergence condition is met.
Citation Information
Cited By
Method and system for enhancing investment attraction environment assessment accuracy based on cross semantic verification
CN122472046A
