Intelligent management method and system for power grid regulation and control service

Abstract

The invention discloses a power grid regulation and control service intelligent management method and system, relates to the technical field of power grid regulation and control, and aims to solve the problems that the query efficiency is low and risk rules lag policy changes. The method comprises the following steps: extracting multivariate heterogeneous data, and constructing a differentiated data index and a dynamic supervision rule base so as to perform bidirectional mapping on risk points in the multivariate heterogeneous data and corresponding policy terms; a verification rule of the power grid regulation and control data is set in a user-defined mode, and an AI large model and a computer vision fusion algorithm are used for analyzing the unstructured data to complete structured conversion; automatically checking the analyzed structured data and multivariate heterogeneous data according to a checking rule, and if the checking is not passed, triggering early warning, generating a triple and then carrying out state circulation until the checking is passed; and automatically updating the risk point and the checking rule, and periodically predicting the recurrence probability of the rectified risk by adopting a linear regression algorithm. The method has the effect of improving the power grid supervision efficiency and accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid regulation, and in particular to a power grid regulation service intelligent management method and system. BACKGROUND

[0002] With the deepening of the construction of new power systems, the scale of power grids is growing exponentially, and power grid regulation services cover multiple core fields such as grid connection management, power market operation, and new energy consumption. The volume and complexity of the services are increasing simultaneously. At the same time, regulatory agencies are increasingly strict about the compliance of power grid dispatching behavior, and there is an urgent need for high-precision and high-efficiency compliance supervision technical solutions.

[0003] In the prior art, power grid regulation service compliance supervision mainly relies on manual verification and offline process flow, and there are three major technical defects: first, multi-source heterogeneous data (time series operation data, text policy files, and topological device data) lack unified fusion and indexing technology, resulting in incomplete supervision data acquisition and low query efficiency; second, risk rules and policy clauses lack a dynamic correlation mechanism, and rule updates lag behind policy changes, resulting in compliance supervision blind spots; third, risk responsibility is not digitized, and the rectification process lacks closed-loop control technical support, resulting in risk disposal lag and difficulty in responsibility tracing. SUMMARY

[0004] The purpose of the present application is to provide a power grid regulation service intelligent management method and system to solve the technical problems of high manual dependence, tedious data sorting, risk identification lag, low supervision response efficiency, and difficulty in responsibility tracing in existing power grid regulation service compliance management.

[0005] To achieve the above purpose, the present application provides the following technical solutions: A power grid regulation service intelligent management method, comprising: A pre-grid preparation stage: multi-element heterogeneous data is extracted from multiple heterogeneous source systems, and a differentiated data index and a dynamic supervision rule library are constructed for the multi-element heterogeneous data to bidirectionally map risk points in the multi-element heterogeneous data to corresponding policy clauses; A pre-grid intelligent verification stage: verification rules for power grid regulation data are defined according to the dynamic supervision rule library, and unstructured data is analyzed using an AI large model and a computer vision fusion algorithm to complete structured conversion; A pre-grid commissioning stage: the analyzed structured data and multi-element heterogeneous data are automatically checked according to the verification rules, and if the verification fails, a warning is triggered and a triple including a risk point ID, a responsibility subject ID, and a control action ID is generated for state transfer until the verification passes; During the formal grid connection and operation phase: By connecting to the standard knowledge base program interface and calling the TF-IDF algorithm and cosine similarity matching algorithm, risk points and verification rules are automatically updated; a linear regression algorithm is used to periodically predict the recurrence probability of rectified risks.

[0006] A further technical solution is that the multi-dimensional heterogeneous data includes text data, time-series data, and topological data.

[0007] A further technical solution is that: the construction of a differentiated data index for the diverse heterogeneous data specifically includes: For text-based data, a custom word segmentation dictionary is built based on a power grid professional dictionary. The word weights are calculated using the TF-IDF algorithm, and an inverted index is established. Time-series data is indexed in a segmented manner using a three-level granularity of minutes, hours, and days. Topological data is indexed using adjacency lists built based on graph theory.

[0008] A further technical solution is that the verification rules include grid connection data integrity verification rules, protocol clause standardization verification, and early warning thresholds.

[0009] A further technical solution is that the state transition includes early warning triggering, task decomposition, collaborative handling, effect verification, and closed-loop archiving.

[0010] Compared with existing technologies, the intelligent management method for power grid dispatching provided by this invention has the following advantages: 1. By analyzing multi-professional business processes, core risk points are identified, and a structured regulatory rule base containing fields such as risk ID, trigger threshold, and traceability policy ID is constructed. The TF-IDF algorithm and cosine similarity matching algorithm are used to achieve a two-way mapping between risk points and policy provisions. Through a dynamic update mechanism and version control technology that connects to the standard knowledge base API, the risk rules are kept in real time synchronized with the latest regulatory policies, fundamentally solving the problems of disconnect between risk rules and policies, delayed updates, and compliance blind spots in existing technologies.

[0011] 2. A mapping matrix is ​​constructed using triples (risk point ID, responsible entity ID, and control action ID) and stored in a distributed graph database. Responsibility is quantified and allocated through the Word2Vec model. A risk transition model and rule engine with five states are designed to achieve full-process monitoring and anomaly interception in risk handling. The linear regression algorithm is combined to predict the probability of risk recurrence and generate periodic review tasks, which completely solves the defects of ambiguous responsibility definition and lack of closed-loop rectification process in the existing technology.

[0012] 3. Through the visual modular configuration engine architecture, the self-defined configuration of unstructured data analysis rules, checking dimensions and other parameters is supported, and new business systems can be quickly connected through API; AI large models, computer vision and natural language processing technology are integrated to realize efficient analysis and structured conversion of unstructured data, risk grading and differentiated early warning are realized by combining 5*5 dynamic threshold algorithm, a full-process technical closed loop of "configuration-analysis-checking-early warning" is formed, and the supervision mode is transformed from "after-the-fact remediation" to "pre-event warning and in-process control".

[0013] 4. According to the differentiated characteristics of text, time sequence and topology data, inverted index based on power grid professional dictionary, segmented time sequence index with three-level time granularity and adjacency list index based on graph theory are constructed respectively; incremental updating algorithm and Redis distributed lock are used to realize index consistency maintenance, solving the problems of difficult multi-source data fusion and low query efficiency in the prior art.

[0014] The application also provides a power grid regulation and control business intelligent management system, comprising: A data index construction module is configured to extract multi-element heterogeneous data from multiple heterogeneous source systems, and construct differentiated data indexes and a dynamic supervision rule library for the multi-element heterogeneous data, so as to bidirectionally map risk points in the multi-element heterogeneous data to corresponding policy provisions; A risk checking module is configured to set checking rules for power grid regulation and control data according to the dynamic supervision rule library, and analyze unstructured data by using AI large models and computer vision fusion algorithm, complete structured conversion, and automatically check the structured data and multi-element heterogeneous data after analysis according to the checking rules; A risk control module is configured to trigger early warning and generate a triple including a risk point ID, a subject ID and a control action ID when the checking fails, and then perform state flow transfer until the checking passes; A state monitoring module is configured to automatically update risk points and checking rules by interfacing with a standard knowledge base program interface, calling TF-IDF algorithm and cosine similarity matching algorithm, and periodically predict the recurrence probability of the rectified risk by using linear regression algorithm.

[0015] Further technical solutions are as follows: the multi-element heterogeneous data comprises text data, time sequence data and topology data.

[0016] Further technical solutions are as follows: the construction of differentiated data indexes for the multi-element heterogeneous data specifically comprises: A self-defined word segmentation dictionary is constructed for text data based on a power grid professional dictionary, inverted index is established by calculating word weights by using TF-IDF algorithm; segmented time sequence index is constructed for time sequence data according to minute / hour / day three-level granularity; adjacency list index is constructed for topology data based on graph theory.

[0017] Further technical solutions are that the check rules include grid connection data integrity check rules, protocol clause normative check and early warning threshold.

[0018] Further technical solutions are that the state transition includes early warning triggering, task decomposition, cooperative disposal, effect verification and closed-loop archiving.

[0019] Compared with the prior art, the power grid regulation business intelligent management system provided by the application has the same beneficial effects as the power grid regulation business intelligent management method described above, and will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0020] The accompanying drawings, which are included to provide a further understanding of the application, form a part of the application and help to explain the application together with the specific embodiments described herein and are not intended to limit the application. In the drawings: Figure 1 A method flowchart is provided for the embodiments of the application. DETAILED DESCRIPTION

[0021] This specific embodiment is only an explanation of the application and is not a limitation of the application. Those skilled in the art can make modifications to the embodiment without creative contribution after reading the specification, as long as the modifications are within the scope of the claims of the application.

[0022] It should be noted that in the application, the words "exemplary" or "for example" are used to mean serving as an example, instance, or illustration. Any embodiment or design described as "exemplary" or "for example" in the application should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of the words "exemplary" or "for example" is intended to present concepts in a concrete manner.

[0023] The application takes "data access-fusion processing-core analysis-early warning anchoring-closed-loop management-application output" as the link process, sequentially accesses multi-source heterogeneous data and processes fusion index, dynamically constructs and intelligently checks the double-core module through the risk list, realizes risk early warning and responsibility anchoring, and finally outputs in multiple modes through full-process closed-loop management, forms a full-link technology closed loop, and achieves the intelligent compliance supervision goal of the whole life cycle of power grid regulation.

[0024] The embodiments of the application will be further described in detail below with reference to the accompanying drawings.

[0025] Please refer to Figure 1 The embodiments of the application provide a power grid regulation business intelligent management method, which includes the following steps: Step S1: In the pre-grid connection preparation stage, multi-element heterogeneous data is extracted from multiple heterogeneous source systems, and a differentiated data index and a dynamic supervision rule library are constructed for the multi-element heterogeneous data to map risk points in the multi-element heterogeneous data to corresponding policy clauses in both directions.

[0026] Among them, the multi-element heterogeneous data includes text data, time series data and topology data, and the multi-element heterogeneous data can be obtained by adopting three ways of interface calling, message bus subscription and file timing import. For example: obtaining power station time series operation data (output, voltage, current, etc.) from an EMS system; obtaining text data such as grid connection application and protocol contract from a new energy grid connection management system; obtaining power station equipment topology correlation data from a power grid topology management system.

[0027] Specifically, the text data constructs a self-defined segmentation dictionary based on a power grid professional dictionary (such as device name, operation term, policy keyword), calculates the word weight by a TF-IDF algorithm and establishes an inverted index, and the index fields include word hash value, associated data source ID and policy clause ID. The time series data adopts a segmented time series index structure, constructs a segmented time series index in minute / hour / day three-level granularity, and each data segment contains a timestamp index and a data value range index, supporting fast query based on a time window. The topology data constructs a device correlation index model based on graph theory, taking devices as nodes and connection relationships as edges, constructs an adjacency table index by a breadth-first search (BFS) algorithm, and supports millisecond-level positioning of associated device data.

[0028] A structured storage model (fields include risk ID, risk description, trigger threshold, traceability policy ID and applicable business scenario) is adopted to associate and bind risk points in multiple professional fields (regulatory operation, power market, new energy consumption, etc.) with policy clauses, forming a dynamic supervision rule library. Then, a TF-IDF algorithm is adopted to convert policy text into a feature vector, a cosine similarity algorithm is adopted to calculate the matching degree (matching degree threshold ≥ 85%) of risk description and policy clause, a positive mapping of "risk point→policy clause" and a reverse mapping of "policy clause→associated risk point" are established, and association and binding with responsible departments and checking rules are completed.

[0029] The positive mapping and the reverse mapping are illustrated below.

[0030] Positive mapping: risk point→policy clause (from risk positioning to compliance basis).

[0031] Example Scenario: Risk of deviation in output curve of new energy power plant; Risk point information (sourced from structured regulatory rule base): Risk ID: RISK-2023-008, Risk description: The deviation between the actual output curve of the new energy power plant and the declared curve of the day-ahead exceeds ±10%, and the duration is ≥1 hour; Trigger threshold: Deviation rate >10% and duration ≥60 minutes; Applicable business scenario: Compliance supervision of power market dispatch during the formal operation phase of new energy power plants; Policy text data (sourced from policy documents in the standard knowledge base): Policy name: "Administrative Measures for Dispatch and Operation of New Energy Power Plants", Policy clause ID: POLICY-2022-15, Clause content: "New energy power plants shall strictly execute dispatch instructions in accordance with the output curve declared of the day-ahead. The deviation between the actual output and the declared curve shall not exceed ±10%, and the duration of continuous deviation shall not exceed 1 hour; if it exceeds the prescribed range, the dispatching agency has the right to conduct an assessment in accordance with regulations."

[0032] The forward mapping process is as follows: (1) Using the TF-IDF algorithm, keywords for risk description and policy clauses were extracted and weights were calculated: the keywords and weights for risk description were “output curve (0.35), declaration curve (0.30), deviation (0.25), ±10% (0.05), 1 hour (0.05)”; the keywords and weights for policy clauses were “output curve (0.30), declaration curve (0.28), deviation (0.22), ±10% (0.10), 1 hour (0.10)”.

[0033] (2) After converting the two into feature vectors, the matching degree is calculated by cosine similarity algorithm: The risk description matches the above policy clauses by 92.3%, which meets the threshold requirement of ≥85%.

[0034] (3) The system automatically establishes a positive mapping relationship: RISK-2023-008→POLICY-2022-15, that is, the risk of deviation of the output curve directly corresponds to Article 15 of the "Management Measures for Dispatch and Operation of New Energy Power Stations", providing a clear policy basis for compliance judgment after risk identification.

[0035] Reverse mapping: Policy provisions → Related risk points (risk management is synchronized with policy updates).

[0036] Example scenario: Policy clause update triggers risk point related update; Policy update information (source: Standard Knowledge Base API push): Original policy clause ID: POLICY-2022-15 (i.e., the aforementioned clause); Updated policy clause ID: POLICY-2023-18; Clause content revision: "The deviation between the actual output of new energy power plants and the day-ahead declared curve shall not exceed ±8%, and the duration of continuous deviation shall not exceed 45 minutes; if the deviation exceeds the prescribed range, a written explanation shall be submitted within 15 minutes." (Core revision: The deviation threshold is adjusted from ±10% to ±8%, and the duration is shortened from 1 hour to 45 minutes).

[0037] The reverse mapping implementation process is as follows: (1) The system captures the revision notice of POLICY-2022-15 in real time by connecting to the standard knowledge base API, automatically extracts the keywords of the updated clause (POLICY-2023-18) and converts them into feature vectors. The keywords and weights are "output curve (0.30), declaration curve (0.28), deviation (0.22), ±8% (0.10), 45 minutes (0.10)".

[0038] (2) Call the cosine similarity algorithm to compare the feature vector of the updated clause with the description vector of all risk points in the regulatory rule base in batches, and filter out risk points with a matching degree of ≥85%.

[0039] (3) After calculation, the matching degree between the original risk point RISK-2023-008 and the updated clause POLICY-2023-18 is 89.7%, which meets the threshold requirement. The system automatically establishes a reverse mapping relationship: POLICY-2023-18→RISK-2023-008.

[0040] (4) Generate a “Risk Point Update Recommendation List”, prompting that the trigger threshold of RISK-2023-008 be revised to “deviation rate > 8% and duration ≥ 45 minutes”, and record the rule update trajectory through version control technology (updated from POLICY-2022-15 to POLICY-2023-18), supporting subsequent rule backtracking.

[0041] The dynamic regulatory rule base includes a dynamic update mechanism, connects to the standard knowledge base to publish APIs, and obtains policy update notifications in real time. When policies change, the system automatically calls the aforementioned cosine similarity algorithm to generate a "risk point update suggestion list" and records the rule update trajectory through version control technology, supports rule backtracking, and ensures that risk management is synchronized with the latest regulatory requirements.

[0042] To clearly explain the relationship between association binding and two-way mapping, the following definition is provided: Association binding is a multi-dimensional association system centered on risk points. The binding objects are policy provisions, responsible departments, and verification rules, as shown in the table below: Binding relationship type Binding subject Binding core purpose Core base binding Risk point ↔ policy clause Explicit risk control compliance basis, realizing "rules to follow" Responsibility landing binding Risk point ↔ responsibility department (such as dispatch department, water and new energy department, etc.) Digitally anchor risk responsibility subject, eliminate shirking Execution logic binding Risk point ↔ review rules (including verification dimension, judgment standard, early warning threshold, etc.) Provide execution logic for automatic risk verification, realize "rules must be checked" The three types of binding form an integrated chain of "risk points - policy provisions - responsible departments - verification rules", ensuring that compliance requirements are connected throughout the entire chain from "basis" to "control" and then to "responsibility".

[0043] Association binding and two-way mapping are related as containment and being contained, and as foundation and extension. They have clear differences in core definition, coverage, technical support, and functional goals, as shown in the table below: Comparison dimension Two-way mapping Association binding Core definition Specifically, the two-way association between "risk points" and "policy clauses" is a single-dimensional technical mapping Based on two-way mapping, the multi-dimensional association system includes "risk point-policy clause", "risk point-responsibility subject", "risk point-review rule" and other multiple bindings Coverage Only involves "risk points" and "policy clauses", single range Covering "risk points, policy clauses, responsibility departments, review rules", the range is more extensive Technical support Core relies on TF-IDF algorithm (feature vector transformation), cosine similarity algorithm (matching degree ≥ 85%), which is a pure technical level of accurate matching Based on the algorithm technology of two-way mapping, superimposed with business process division logic (responsibility department binding), verification execution logic (review rule binding), it is a "technology + business" composite association Functional target Solve the problem of "risk points without clear compliance basis" and "policy updates unable to synchronize risk rules", ensure that risk control is consistent with policy requirements In addition to achieving the core goal of two-way mapping, it also solves the problems of "unclear responsibility subject" and "no standard for verification logic", providing full-process support for subsequent intelligent verification, responsibility tracing, and closed-loop management Hierarchical relationship It is the core basis and prerequisite of association binding, without two-way mapping, it is impossible to achieve accurate binding of compliance basis It is an extension and expansion of two-way mapping, and builds a complete regulatory execution system on the basis of two-way mapping By constructing differentiated data indexes and dynamic regulatory rule bases, the unified integration of multi-source data from multiple heterogeneous systems has been successfully achieved, with a data coverage rate of 100%. Through incremental index update technology, data retrieval latency is controlled within 50ms, improving retrieval efficiency by 80% compared to traditional databases. Multiple regulatory points are precisely bound to policy provisions, and the latest policy updates can be automatically synchronized after the rule base is initialized, eliminating blind spots in compliance supervision. Regulatory responsibilities are solidified to specific departments through the rule base, preventing the phenomenon of shirking responsibility.

[0044] Step S2: In the pre-grid connection intelligent verification stage, the verification rules for power grid control data are set according to the dynamic monitoring rule library, and the unstructured data is parsed using the AI ​​big model and computer vision fusion algorithm to complete the structured transformation.

[0045] The verification rules can be configured through the visualization module, specifically including rules for verifying the integrity of grid connection data, rules for parsing unstructured data, risk verification, verification of the standardization of agreement terms, verification cycle, and early warning thresholds. The system has a built-in standardized data interaction protocol, which can be quickly connected to new business systems through API interfaces, adapting to the differentiated regulatory needs of different regions and professions without modifying the underlying code.

[0046] Unstructured data includes PDFs, Word documents, and scanned copies. Natural Language Processing (NLP) technology is used to transform the key information in unstructured data into structured data.

[0047] Through a dynamic regulatory rule base and the fusion of AI large models and computer vision algorithms, the accuracy rate of unstructured data parsing reached 96.8%, higher than the preset standard of 95%. The efficiency of grid connection document compliance review was improved by 65% ​​compared with traditional manual review, and the review time for a single project was reduced from 12 hours to 4.2 hours. Through practical statistics, a total of 32 new energy power plants completed the pre-grid connection document verification, and 47 problems such as missing documents and non-standard clauses were found. Early warnings were given for all of them, eliminating the risk of non-compliant grid connection from the source.

[0048] Step S3: During the grid connection trial operation phase: Automatically verify the parsed structured data and multi-variable heterogeneous data according to the verification rules. If the verification fails, trigger an early warning and generate a triplet including risk point ID, responsible entity ID, and control action ID, and then proceed with the status transition until the verification passes.

[0049] Among these measures, when verifying the parsed structured data, a consistency comparison is made with the filing information or required rules, the standardization of key agreement clauses in the structured data is verified, and missing information or non-standard clauses are automatically marked and trigger warnings.

[0050] When verifying diverse and heterogeneous data, the system automatically checks the data's completeness, the test items' integrity, and the performance indicators' compliance based on verification rules. If time-series data is not reported on time or indicators fail to meet standards, resulting in verification failure, an alert is automatically triggered. A triplet is generated based on a "risk-position-rectification" three-dimensional mapping model, including the risk point ID, the responsible entity ID, and the control action ID, clearly identifying the responsible department and person, and automatically pushing an alert work order. The alert processing flow is monitored through a rules engine, following a state flow of "alert triggering → task decomposition → collaborative handling → effect verification → closed-loop archiving," intercepting abnormal closed-loop behaviors that skip effect verification.

[0051] In the triplet, the risk point includes the risk level (levels 1-5), trigger condition parameters, and quantitative value of the scope of impact; the subject of authority and responsibility adopts a hierarchical structure, and the business process nodes and job responsibility texts are converted into vectors through the Word2Vec model, and the cosine similarity is calculated to obtain the authority and responsibility boundary coefficient (0-1) to realize the quantitative allocation of responsibility; the control action includes the measure type (preventive / process / emergency), execution threshold, and related business system interface.

[0052] Accurate risk warnings were provided during the grid-connected trial operation phase, improving the timeliness of problem detection by 55% compared to traditional manual methods. Responsibility for multiple issues during the trial operation phase was precisely assigned, and the entire handling process was traceable. Abnormal closed-loop behaviors were intercepted multiple times through the rule engine, ensuring verifiable handling results. All issues were rectified within the stipulated timeframe, achieving a 100% rectification rate and guaranteeing the stable operation capability of the power station before its commercial conversion.

[0053] Step S4: During the formal grid connection and operation phase: By connecting to the standard knowledge base API and calling the TF-IDF algorithm and cosine similarity matching algorithm, the risk points and verification rules are automatically updated; the linear regression algorithm is used to periodically predict the recurrence probability of rectified risks.

[0054] During this phase, the configuration engine is set to automatically trigger verification tasks at fixed times every day. The verification scope includes core regulatory points such as the consistency between the unit's daily reported curve and the actual output curve, and the completeness of the exemption application approval process.

[0055] When the probability of risk recurrence is ≥30%, two risks (output curve deviation and untimely information disclosure) are automatically generated to form a periodic review task, thereby preventing risk recurrence.

[0056] During the formal grid connection and operation phase, the core business of fully automated and routine supervision can complete daily verification tasks in ≤30 minutes, improving efficiency by 90% compared to manual verification. Through practical statistics, the accuracy rate of assessment calculations reaches 100%, and the assessment disputes caused by human error have decreased by 35% compared to before. The rule base can automatically synchronize policy updates to ensure that regulatory requirements are adapted in real time. Periodic review tasks effectively reduce the probability of risk recurrence, reducing the probability of risk recurrence from 20% to below 5%, significantly improving the credibility of dispatching.

[0057] This invention also discloses an intelligent management system for power grid dispatching, comprising: The data indexing construction module is used to extract diverse heterogeneous data from multiple heterogeneous source systems and construct differentiated data indexes and dynamic regulatory rule bases for the diverse heterogeneous data, so as to bidirectionally map the risk points in the diverse heterogeneous data with the corresponding policy provisions; wherein, the diverse heterogeneous data includes text data, time series data and topological data; The risk verification module sets verification rules for power grid control data according to the dynamic regulatory rule base, and uses AI big data model and computer vision fusion algorithm to parse unstructured data and complete the structure transformation. It automatically verifies the parsed structured data and multi-variable heterogeneous data according to the verification rules. The verification rules include grid connection data integrity verification rules, agreement clause standardization verification, and early warning thresholds. The risk management module is used to trigger an early warning and generate a triplet when the verification fails, and then proceed with the state transition until the verification passes; wherein, the triplet includes risk point ID, responsible entity ID, and management action ID; the state transition includes early warning triggering - task decomposition - collaborative handling - effect verification - closed-loop archiving; The status monitoring module automatically updates risk points and verification rules by connecting to the standard knowledge base program interface and calling the TF-IDF algorithm and cosine similarity matching algorithm; it uses a linear regression algorithm to periodically predict the recurrence probability of rectified risks.

[0058] More specifically, text-based data is constructed using a custom word segmentation dictionary built on a power grid professional dictionary (including equipment names, operational terms, and policy keywords). The TF-IDF algorithm is used to calculate the weights of professional terms and to build an inverted index. The index fields include the word hash value, the associated data source ID, and the policy clause ID.

[0059] Time-series data uses a segmented time-series index structure, which includes data segments divided into three time granularities: minutes, hours, and days. Each data segment contains a timestamp index and a data value range index, supporting fast queries based on time windows.

[0060] Topological data uses graph theory to build a device association index model, with devices as nodes and connections as edges. It constructs an adjacency list index using the breadth-first search (BFS) algorithm, supporting millisecond-level location of associated device data.

[0061] The beneficial effects of the intelligent management system for power grid dispatching provided by this invention are the same as those of the intelligent management method for power grid dispatching described in the above technical solution, and will not be repeated here.

[0062] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely exemplary descriptions of the invention as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A method for intelligent management of power grid dispatching operations, characterized in that, include: Preparatory stage before grid connection: Extract diverse and heterogeneous data from multiple heterogeneous source systems, and construct a differentiated data index and dynamic regulatory rule base for the diverse and heterogeneous data to bidirectionally map the risk points in the diverse and heterogeneous data with the corresponding policy provisions; Intelligent verification stage before grid connection: The verification rules for power grid control data are set according to the dynamic monitoring rule base, and the unstructured data is parsed and structured by using the fusion algorithm of AI big model and computer vision. During the grid connection trial operation phase: The parsed structured data and multi-variate heterogeneous data are automatically checked according to the verification rules. If the verification fails, an early warning is triggered and a triplet including risk point ID, responsible entity ID and control action ID is generated, and the status transition continues until the verification passes. During the formal grid connection and operation phase: By connecting to the standard knowledge base program interface and calling the TF-IDF algorithm and cosine similarity matching algorithm, risk points and verification rules are automatically updated; a linear regression algorithm is used to periodically predict the recurrence probability of rectified risks.

2. The intelligent management method for power grid dispatching operations according to claim 1, characterized in that, The diverse heterogeneous data includes textual data, time-series data, and topological data.

3. The intelligent management method for power grid dispatching operations according to claim 2, characterized in that, The construction of a differentiated data index for the diverse heterogeneous data specifically includes: For text-based data, a custom word segmentation dictionary is built based on a power grid professional dictionary. The word weights are calculated using the TF-IDF algorithm, and an inverted index is established. Time-series data is indexed in a segmented manner using a three-level granularity of minutes, hours, and days. Topological data is indexed using adjacency lists built based on graph theory.

4. The intelligent management method for power grid dispatching operations according to claim 1, characterized in that, The verification rules include rules for verifying the integrity of grid connection data, checking the standardization of agreement terms, and setting early warning thresholds.

5. The intelligent management method for power grid dispatching operations according to claim 1, characterized in that, The state transition includes early warning triggering, task decomposition, collaborative handling, effect verification, and closed-loop archiving.

6. An intelligent management system for power grid dispatching operations, characterized in that, include: The data index building module is used to extract diverse heterogeneous data from multiple heterogeneous source systems and build a differentiated data index and dynamic regulatory rule base for the diverse heterogeneous data, so as to bidirectionally map the risk points in the diverse heterogeneous data with the corresponding policy provisions. The risk verification module sets verification rules for power grid control data according to the dynamic regulatory rule base, and uses AI big data model and computer vision fusion algorithm to parse unstructured data, complete the structure transformation, and automatically verify the parsed structured data and multi-variable heterogeneous data according to the verification rules. The risk control module is used to trigger an alert and generate a triplet including the risk point ID, the responsible entity ID, and the control action ID when the verification fails, and then proceed with the status transition until the verification passes. The status monitoring module automatically updates risk points and verification rules by connecting to the standard knowledge base program interface and calling the TF-IDF algorithm and cosine similarity matching algorithm; it uses a linear regression algorithm to periodically predict the recurrence probability of rectified risks.

7. The intelligent management system for power grid dispatching and control according to claim 6, characterized in that, The diverse heterogeneous data includes textual data, time-series data, and topological data.

8. The intelligent management system for power grid dispatching and control according to claim 7, characterized in that, The construction of a differentiated data index for the diverse heterogeneous data specifically includes: For text-based data, a custom word segmentation dictionary is built based on a power grid professional dictionary. The word weights are calculated using the TF-IDF algorithm, and an inverted index is established. Time-series data is indexed in a segmented manner using a three-level granularity of minutes, hours, and days. Topological data is indexed using adjacency lists built based on graph theory.

9. The intelligent management system for power grid dispatching and control according to claim 6, characterized in that, The verification rules include rules for verifying the integrity of grid connection data, checking the standardization of agreement terms, and setting early warning thresholds.

10. The intelligent management system for power grid dispatching and control according to claim 6, characterized in that, The state transition includes early warning triggering, task decomposition, collaborative handling, effect verification, and closed-loop archiving.

Images