A power signal intelligent processing and data knowledge construction system and method

Abstract

This invention provides an intelligent processing and knowledge-based data construction system and method for power bidding information, belonging to the field of data processing technology. Specifically, it includes: a data acquisition module that obtains bidding information page fields, screenshots, and attachments, and stores them with a unique index; a file processing module that outputs standardized files for subsequent parsing modules; a platform segmentation module that determines the verification processing targets and verification processing schemes for updated business extraction rules in the update control platform; an intelligent parsing module that performs three-level header matching and field semantic recognition; and a standardization processing module that performs unified encoding, format conversion, and data cleaning on the parsing results, and automatically updates the knowledge base when new business extraction rules exist, thereby improving the stability and reliability of power bidding information data extraction.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a system and method for intelligent processing of power information and the construction of data knowledge. Background Technology

[0002] With the deepening of market-oriented reforms in the power industry, bidding activities have become increasingly frequent, and major power generation groups have established their own e-commerce or supply chain platforms to publish bidding information. These bidding information are key data sources reflecting industry dynamics, project opportunities, and market trends. However, due to platform heterogeneity, inconsistent formats, and diverse publishing formats (HTML, PDF, Word, Excel, etc.), there is a large amount of non-standardized expression.

[0003] Currently, the industry mainly relies on manual methods to collect and process bidding information, which is inefficient, costly, and prone to errors. Some solutions using general web crawlers are insufficient in parsing complex unstructured documents (such as scanned PDFs and multi-page tables), resulting in low data collection coverage and poor accuracy in structured data extraction. While web crawlers and semantic models can automate the extraction and processing of power bidding information data and create a knowledge base, they suffer from the following drawbacks:

[0004] The formats of different platforms may vary to some extent, which leads to differences in business extraction rules (i.e., semantic extraction rules). If a unified knowledge base is used to store business extraction rules, once the business extraction rules in one platform are updated, the accuracy of the extraction results of power information data in other platforms may not meet the requirements. Therefore, how to implement segmented management of the platform's knowledge base and determine the verification strategy before promoting its application on multiple platforms based on the extracted data, thereby improving the reliability of power information data extraction, has become an urgent technical problem to be solved.

[0005] Therefore, there is an urgent need for an intelligent processing system and method for power information and a data knowledge-based construction system. Summary of the Invention

[0006] To achieve the objectives of this invention, the following technical solution is adopted:

[0007] Specifically, this application provides an intelligent processing and data knowledge construction system for the entire process of power bidding, which includes:

[0008] The data acquisition module constructs a targeted crawling system for power system platforms, covering multiple platforms, acquiring fields, screenshots, and attachments from bidding information pages, and storing them with a unique index.

[0009] The file processing module is responsible for unifying the formats of the collected multi-format files, performing file structure and content legality checks, and outputting standardized files for use by the subsequent parsing module.

[0010] The platform segmentation module determines the update control platform in the platform based on the update data of the knowledge base of the platform, and determines the verification processing target and verification processing scheme of the business extraction rules of the update control platform based on the update processing data of the independent knowledge base of the update control platform.

[0011] The intelligent parsing module is responsible for building a parsing module based on the energy model, integrating business extraction rules with deep learning capabilities, and realizing three-level header matching and field semantic recognition;

[0012] The standardization processing module is responsible for uniformly encoding, format conversion, and data cleaning of the parsing results, and automatically updating the knowledge base when new business extraction rules are available.

[0013] The beneficial effects of this invention are as follows:

[0014] The update control platform is determined based on the updated data in the knowledge base of the platform. This avoids the technical problem of mutual interference between business extraction rules of different platforms after the business extraction rules of many platforms are updated in the knowledge base. Platforms with high update frequency or a large number of platforms with updated business extraction rules are designated as update control platforms, and independent knowledge bases are built for them to ensure the reliability of extraction from multiple platforms.

[0015] In terms of data collection, this invention improves the stability and coverage of data collection by constructing a targeted crawling system, combined with the Quartz timed scheduling framework and dynamic IP pool technology, reducing the cost of manual intervention and possessing high automation and scalability.

[0016] In the file processing stage, this invention introduces a format unification and legality verification mechanism, which performs normalization conversion and validity screening for various file types (such as PDF, Word, Excel, etc.), significantly improving the data quality before parsing, simplifying the subsequent processing flow, reducing the parsing failure rate caused by format differences, and enhancing the system's compatibility and robustness.

[0017] This invention employs an intelligent parsing approach driven by a large energy model, integrating power industry business rules with deep learning capabilities to achieve accurate identification and field matching of complex table structures, cross-page content, and diverse table headers. Based on the Transformer architecture and fine-tuned for industry applications, this module possesses strong semantic understanding and generalization capabilities, overcoming the limitations of traditional rule matching and significantly improving parsing accuracy and processing efficiency.

[0018] This invention designs algorithms for monetary unit conversion, date format calibration, and dual primary key deduplication to ensure data consistency and standardization when used across platforms. These algorithms are characterized by simple structure and high computational efficiency, enabling them to maintain high performance in massive data processing scenarios and effectively improving data availability and service quality.

[0019] Based on the update processing data of the independent knowledge base of the update management platform, the verification processing targets of the updated business extraction rules in the update management platform are determined. This ensures that the update management platform with a large number of updated business extraction rules can be jointly verified and processed on multiple platforms. By verifying and processing the updated business extraction rules in batches, the timeliness of the update of business extraction rules on multiple platforms is guaranteed, while also further improving the stability of the extraction and processing of power bidding data models.

[0020] Specifically, obtaining the fields, screenshots, and attachments of the bidding information page includes:

[0021] Set differentiated data collection strategies for the update frequency of different platforms;

[0022] It integrates a dynamic IP proxy pool to obtain and process tag page fields, page screenshots, and attachments by rotating IP addresses and simulating browser behavior.

[0023] Specifically, standardized output files are provided for use by subsequent parsing modules, including:

[0024] The collected multi-format files are converted to obtain structured data objects through format unification.

[0025] A two-layer validity verification method is used to determine the integrity of the detected documents and whether they fall within the valid standard range.

[0026] The preprocessed files are uniformly converted into standard JSON format, i.e., standardized files, and metadata is attached to provide high-quality input for subsequent parsing.

[0027] Secondly, this application provides a method for intelligent processing and data knowledge construction of power information, applied to the aforementioned intelligent processing and data knowledge construction system for power information, specifically including:

[0028] S1 acquires and processes power bidding data from multiple platforms based on an energy model, updates the knowledge base based on the acquisition and processing results, determines the update control platform in the platform based on the update data of different platforms in the knowledge base, constructs an independent knowledge base for the update control platform, and uses it as an independent knowledge base.

[0029] S2 uses the update data of the business extraction rules of the update management platform and the independent knowledge base of the update management platform in the platform to determine the method of obtaining the power bidding information data of the update management platform, performs the acquisition processing of the power bidding information data of the update management platform based on the acquisition method, and performs the update processing of the business extraction rules of the independent knowledge base of different update management platforms.

[0030] S3 When the consistency of the business extraction rules of the independent knowledge base cannot meet the requirements, a verification processing scheme for the business extraction rules is determined based on the extraction data of the business extraction rules of the update management platform and the update data of the business extraction rules of different update management platforms.

[0031] Furthermore, the power bidding data consists of parsed bidding documents from different platforms.

[0032] Furthermore, the knowledge base will be updated, specifically including:

[0033] Based on the analysis results of power bidding data from different platforms, bids with keywords that are inconsistent with the keywords in the original knowledge base are identified.

[0034] The knowledge base is updated based on the keywords of the tender documents that are inconsistent with the keywords of the original knowledge base and the extraction strategy of power tender information data.

[0035] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0036] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0037] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0038] Figure 1 This is a framework diagram of a power information intelligent processing and data knowledge construction system;

[0039] Figure 2 This is a flowchart illustrating the methods for obtaining fields, screenshots, and attachments from the bidding information page.

[0040] Figure 3 This is a flowchart illustrating the method for outputting standardized files for use by subsequent parsing modules;

[0041] Figure 4 This is a flowchart of a method for intelligent processing of power information and knowledge-based data construction. Detailed Implementation

[0042] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0043] Example 1

[0044] like Figure 1 As shown, this application provides an intelligent processing and data knowledge construction system for the entire process of power bidding, specifically including:

[0045] The data acquisition module constructs a targeted crawling system for power system platforms, covering multiple platforms, acquiring fields, screenshots, and attachments from bidding information pages, and storing them with a unique index.

[0046] The file processing module is responsible for unifying the formats of the collected multi-format files, performing file structure and content legality checks, and outputting standardized files for use by the subsequent parsing module.

[0047] The platform segmentation module determines the update control platform in the platform based on the update data of the knowledge base of the platform, and determines the verification processing target and verification processing scheme of the business extraction rules of the update control platform based on the update processing data of the independent knowledge base of the update control platform.

[0048] The intelligent parsing module is responsible for building a parsing module based on the energy model, integrating business extraction rules with deep learning capabilities, and realizing three-level header matching and field semantic recognition;

[0049] The standardization processing module is responsible for uniformly encoding, format conversion, and data cleaning of the parsing results, and automatically updating the knowledge base when new business extraction rules are available.

[0050] Specifically, such as Figure 2 As shown, the acquisition of the bidding information page fields, page screenshots, and attachments specifically includes:

[0051] Set differentiated data collection strategies for the update frequency of different platforms;

[0052] It integrates a dynamic IP proxy pool to obtain and process tag page fields, page screenshots, and attachments by rotating IP addresses and simulating browser behavior.

[0053] The system employs the Quartz timed scheduling framework, setting differentiated collection strategies based on the update frequency of different platforms or the acquisition method of power bidding data. For example, the State Grid platform collects data every 30 minutes, while the Southern Power Grid platform collects data every hour. The system integrates a dynamic IP proxy pool, collecting data by rotating IP addresses and simulating browser behavior (such as User-Agent randomization, cookie management, and JavaScript rendering support). During collection, the system not only extracts structured fields from the bidding page (such as project name, bidding unit, amount, and time), but also automatically downloads associated PDF, Word, and Excel attachments. It uses a unique indexing mechanism (such as generating a hash value based on "platform + release time + bidding ID") to associate and store page data with attachments, ensuring data chain integrity and traceability.

[0054] Specifically, such as Figure 3 As shown, the output standardized file is used by the subsequent parsing module, specifically including:

[0055] The collected multi-format files are converted to obtain structured data objects through format unification.

[0056] A two-layer validity verification method is used to determine the integrity of the detected documents and whether they fall within the valid standard range.

[0057] The preprocessed files are uniformly converted into standard JSON format, i.e., standardized files, and metadata is attached to provide high-quality input for subsequent parsing.

[0058] In the file preprocessing stage, the system first performs a standardized format conversion on the collected multi-format files. For example, PDF files are converted into editable text using OCR technology while preserving the original layout structure; Word and Excel files are parsed into structured data objects. Next, the system performs a two-layer validity check: at the structural level, it checks whether the file is complete and whether it is encrypted or corrupted; at the content level, it uses a rule base (such as keyword filtering and regular expression matching) to determine whether the file falls within the valid target range, removing interfering files such as advertisements, notices, and invalid announcements. The preprocessed files are uniformly converted to standard JSON format and include metadata (such as file type, size, and source platform) to provide high-quality input for the subsequent parsing module.

[0059] Specifically, the intelligent parsing module includes:

[0060] The parser scheduling center automatically allocates the optimal parser based on the file type. The content extraction engine uses a fusion of computer vision and natural language processing technology. It first identifies table areas through layout analysis, and then uses row and column detection algorithms to reconstruct the table structure.

[0061] Based on the Transformer architecture, a pre-trained model is used, combined with the matching of tabular data and business extraction rules, to perform three-level header matching mechanism: the first level matches standard headers, the second level matches synonym headers, and the third level performs semantic inference, thus performing three-level header matching and field semantic recognition.

[0062] The intelligent parsing module employs a hybrid parsing strategy driven by a large energy model. The parser scheduling center automatically allocates the optimal parser based on file type (e.g., PDF tables, Word documents, Excel files). For complex tables (e.g., multi-page tables, merged cells), the content extraction engine uses a fusion of computer vision and natural language processing technologies. It first identifies table regions through layout analysis, and then reconstructs the table structure using row and column detection algorithms. The core parsing unit, based on a pre-trained model of the Transformer architecture and the matching of data in the table with business extraction rules, utilizes the matched business extraction rules and the pre-trained model to perform field semantic recognition and three-level header matching: the first level matches standard headers (e.g., "project name"), the second level matches synonym headers (e.g., "project name" and "project engineering name"), and the third level performs semantic inference (e.g., inferring "tender amount" from "budget amount"), thereby achieving high-precision field extraction and mapping.

[0063] This invention designs an intelligent processing mechanism driven by a large energy model and possessing self-evolution capabilities. The core of this mechanism lies in its ability to automatically analyze and generate new extraction rules when the system cannot find a match in the existing rule base, leveraging the domain semantic understanding capabilities of the large model. When a rule match is successful, it is not simply applied; instead, a collaborative verification and joint extraction process between the rule and the large model is initiated, thereby achieving highly robust and accurate information extraction.

[0064] I. Adaptive rule generation and learning process when there are no matching rules:

[0065] When a header in the bidding form cannot be matched using the existing standard header library or thesaurus, the system determines it to be in a "no rule" state and immediately triggers the rule self-generation process of the energy big data model. First, the unidentified header and its local context (such as adjacent headers in the same row and the table title) are fed into the big data model, which has been pre-trained and fine-tuned using massive amounts of text in the power sector. Based on its deep domain knowledge, the model performs a deep understanding and reasoning of the header's semantics, outputting one or more possible standardized field mapping suggestions, and assigning a confidence score to each suggestion. For example, for the unseen header "Total Estimated Project Cost (Ten Thousand Yuan)," the model might infer a confidence score of 0.94 for mapping it to the "Bidding Amount" field and 0.88 for mapping it to the "Budget Amount" field.

[0066] Subsequently, the system enters a rigorous rule generation and validation phase. The highest-confidence mapping suggestion output by the model will be adopted as the core of a "candidate rule." The system will automatically analyze the text features of the table header, such as extracting key patterns (e.g., words like "price," "amount," and "yuan"), and generate corresponding regular expression patterns or keyword lists. Simultaneously, the system will retrieve existing synonyms related to the candidate field and attempt to establish associations between the current table header text and these synonyms. This information is integrated and encapsulated to form a new, structured business extraction rule.

[0067] II. Model and rule joint extraction and validation process when matching rules exist:

[0068] When the table header successfully matches an existing business rule, this invention employs a more cautious and precise "dual-path collaborative" extraction strategy, rather than relying solely on a rule. One path is the traditional "rule extraction path," which quickly locates and extracts target information from the document based on the patterns, locations, or logic defined in the rules. The other path is the "model semantic extraction path," where the energy big model is not bypassed but given a new task: guided by the mapping clues provided by known rules (e.g., knowing that the current table header corresponds to "project name"), it performs deeper semantic analysis and integrity verification on the target cell and its context.

[0069] After the two paths are executed in parallel, their results converge at a "consistency arbitration center." Ideally, the rule extraction result is completely consistent with the model's semantic inference result. In this case, the system outputs the result with extremely high overall confidence and marks it as "double verification passed." However, when the results of the two paths diverge, a refined conflict resolution phase begins. The system evaluates multiple pieces of evidence, including the accuracy statistics of the historical rule in past use, the confidence score of the model's output result in the current instance, and the logical consistency between the extracted result and other related fields in the document (e.g., whether the extracted "bidding unit" name appears frequently in other parts of the document). Based on these dynamic weights, the system calculates the final adopted result through a weighted decision algorithm. Simultaneously, the complete context of this conflict, the arbitration basis, and the final decision are recorded in detail, forming a traceable audit log. This joint mechanism greatly reduces the reliance on potentially outdated, rigid, or poorly handled edge cases of a single rule, providing crucial error correction and reinforcement capabilities for rule extraction through the model's semantic understanding.

[0070] III. Mechanism Integration and System Advantages:

[0071] The above two processes are integrated within a unified intelligent parsing and scheduling framework. The system makes real-time judgments on each header to be processed, seamlessly switching to the "rule generation mode" or "joint extraction mode", thus constructing a stable and flexible information processing pipeline. The core advantage of this mechanism lies in its organic combination of "automated learning" and "intelligent verification". It can not only actively adapt to the diversity and variability of the power industry's standard information formats, automatically discover and learn new information expression patterns, but also, based on existing knowledge, ensure the extreme accuracy of information extraction through double verification. This significantly enhances the practicality, reliability, and long-term sustainable evolution ability of the entire system in real and complex business scenarios, laying a solid technical foundation for the in-depth value mining of power standard information data.

[0072] Specifically, the standardization processing module specifically includes:

[0073] The data standardization module performs unified encoding and cleaning on the parsed fields to form a standard information data knowledge base for power standard information data. In specific implementation, the system has a built-in intelligent conversion algorithm for monetary units: first, it identifies the unit keywords in the monetary string (such as "yuan", "ten thousand yuan", "hundred million yuan"), and then uniformly converts them to the "yuan" unit, supporting the processing of thousand-separator and digital format verification. The date field adopts a multi-format parsing and anomaly detection mechanism, supporting the automatic recognition and conversion of multiple formats such as "YYYY-MM-DD", "YYYY / MM / DD", "YYYY year MM month DD day", and automatically correcting or marking obviously abnormal dates (such as future dates, unreasonable years and months). The deduplication engine generates hash fingerprints based on the double primary keys of "standard information ID + tender invitation number", and achieves efficient deduplication through a Bloom filter, supporting two modes of real-time deduplication and batch deduplication to ensure data uniqueness.

[0074] Second, as Figure 4 shown, this application provides a method for intelligent processing of power standard information and knowledge-based construction of data, which is applied to the above-mentioned system for intelligent processing of power standard information and knowledge-based construction of data, and specifically includes:

[0075] S1 Based on the energy model, obtain and process the power standard information data of multiple platforms, update the knowledge base according to the obtained and processed results, determine the update control platform in the platform according to the updated data in the knowledge base for different platforms, construct an independent knowledge base for the update control platform, and use it as an independent knowledge base;

[0076] Furthermore, the power standard information data is the parsed data of tenders in different platforms.

[0077] Furthermore, the update process of the knowledge base specifically includes:

[0078] Based on the analysis results of power bidding data from different platforms, bids with keywords that are inconsistent with the keywords in the original knowledge base are identified.

[0079] The knowledge base is updated based on the keywords of the tender documents that are inconsistent with the keywords of the original knowledge base and the extraction strategy of power tender information data.

[0080] Specifically, the method for determining the update management platform within the aforementioned platform is as follows:

[0081] To ensure the stable operation of the power bidding information intelligent processing system and the effective management of the knowledge base in a multi-platform environment, this invention proposes a mechanism for dynamically identifying and isolating "update control platforms." This mechanism aims to identify specific bidding platforms that require extensive and frequent adjustments to business extraction rules due to frequent updates or drastic format changes, and to independently control these platforms to reduce their impact on the overall system stability.

[0082] S11 determines the updated data of the business extraction rules of the knowledge base for the platform based on the updated data of different platforms in the knowledge base;

[0083] It should be noted that if the number of updates to the business extraction rules of the knowledge base on the platform does not meet the requirements within the most recent preset time period in the above steps, the platform is determined to be an update control platform; otherwise, the process proceeds to step S12.

[0084] In the above steps, the preliminary screening based on the updated data of the knowledge base refers to the following: Platform: refers to the power industry procurement information release website or system connected to the system, such as the State Grid e-commerce platform; Knowledge base: a structured database that stores business extraction rules, thesaurus, and standard field mapping relationships.

[0085] Business extraction rules: A set of matching patterns or logical conditions used to identify and extract specific fields (such as amount and date) from bidding documents.

[0086] Preset time period: The statistical time window set by the system (such as the last 7 days or 30 days). Update quantity requirement: The threshold number of times the rules for determining whether the platform is active or changing (such as less than 2 times).

[0087] This step is a pre-filtering process. It uses quantitative statistics to quickly eliminate platforms with stable rules, avoiding complex calculations on all platforms, improving judgment efficiency and reducing computational overhead. It only performs in-depth analysis on platforms with active rule updates, and quickly responds to stable platforms. For platforms with mature formats, it maintains the original processing flow to ensure the continuity of core business.

[0088] S12 uses the platform data and the update data of the business extraction rules of different platforms in the knowledge base to determine the update control platform in the platform.

[0089] In the above steps, a multi-dimensional judgment and update management platform is established, with a preset platform number threshold: a critical value for dividing the "platform cluster size" (e.g., 20) to distinguish processing strategies; a rule change platform ratio: the proportion of platforms that have rule updates within the statistical period to the total number of platforms; an update number threshold: a threshold for the number of rule updates for high-variability platforms among the rule change platforms (e.g., 10); and an update management platform: a specific platform that is isolated by the system and managed with an independent knowledge base due to frequent or large rule changes.

[0090] This step is the core of dynamic decision-making. Based on the scale and distribution of changes in the platform cluster, it adaptively selects the most suitable control strategy to achieve a balance between precise isolation and system stability. The strategy is adaptive: it adopts differentiated control logic for platform clusters of different sizes to improve the universality of the method. Precise risk isolation: it focuses on the platforms that have a high impact on the system to avoid the waste of resources caused by "one-size-fits-all" approaches. Maintaining the overall stability of the system: through an independent knowledge base mechanism, it prevents the high-frequency changes of local platforms from spreading to the whole system.

[0091] A power data service provider's intelligent processing system is connected to 30 major power bidding platforms nationwide for the automatic collection and analysis of daily bidding information. The system performs an "update management platform" check every seven days to maintain the stability of the knowledge base.

[0092] Step S11: Preliminary screening. The system scans the knowledge base update logs of all 30 platforms within the last 7 days.

[0093] Statistics show that 23 platforms updated their rules 0 or 1 time, which is lower than the preset update requirement (2 times), while 7 platforms updated their rules 3 to 15 times, which is higher than the requirement.

[0094] The system marked these 7 platforms as "platforms to be analyzed in depth", while the remaining 23 platforms were determined to be stable platforms and were not included in the management process.

[0095] It should be noted that the above steps include the following situations:

[0096] Case 1: If the number of platforms is greater than the preset platform number threshold, then all platforms that have updated business extraction rules in the knowledge base within the most recent preset time period are considered as update control platforms.

[0097] The system determines that the total number of access platforms (30) is greater than the preset platform number threshold (20), and enters "Scenario 1: Large-scale platform cluster strategy". According to this strategy, the system identifies all 7 platforms that have updated their rules in the last 7 days as "update control platforms". These 7 platforms include: the bidding platform of a provincial power company of State Grid (updated 12 rules) and the electronic procurement system of a branch of China Southern Power Grid (updated 8 rules).

[0098] Case 2: If the number of platforms is not greater than the preset platform number threshold, then the updated data of the business extraction rules of the platform in the knowledge base is determined, and the proportion of the number of update control platforms among the platforms with updated business extraction rules is used as the extraction matching proportion. If the extraction matching proportion is greater than the preset number threshold, then the remaining platforms are determined not to belong to the update control platform.

[0099] Case 3: If the proportion of extracted matching numbers is not greater than the preset number threshold, then all platforms whose number of business extraction rule updates in the most recent preset time period is greater than the preset update number threshold will be regarded as update control platforms.

[0100] Summary of beneficial effects:

[0101] Intelligent decision-making: By quantifying rules and making multi-condition judgments, it automatically identifies highly variable platforms and reduces human intervention.

[0102] Elastic architecture support: The independent knowledge base mechanism provides a "sandbox environment" for the system, supporting trial and error and iteration of high-risk rules.

[0103] Business continuity assurance: Isolation strategies ensure that core services are not affected by local changes, meeting enterprises' needs for highly stable data services.

[0104] Scalability: The method is suitable for scenarios where the number of platforms increases or decreases dynamically, and supports smooth scaling.

[0105] This method achieves an organic balance between system stability, adaptability, and operational efficiency in a complex and ever-changing power information data environment, and is a key technological guarantee for supporting large-scale, highly reliable intelligent power data processing services.

[0106] S2 uses the update data of the business extraction rules of the update management platform and the independent knowledge base of the update management platform in the platform to determine the method of obtaining the power bidding information data of the update management platform, performs the acquisition processing of the power bidding information data of the update management platform based on the acquisition method, and performs the update processing of the business extraction rules of the independent knowledge base of different update management platforms.

[0107] This technical solution proposes an intelligent and adaptive data acquisition strategy decision-making mechanism specifically for data collection and management of the "update control platform" in the power bidding information processing system. The update control platform refers to a platform that requires an independent knowledge base for rule isolation due to frequent changes in platform format. This solution achieves efficient data acquisition and rule verification for highly volatile platforms while ensuring overall system stability through multi-dimensional status monitoring, real-time impact assessment, and tiered response strategies. Its core innovation lies in dynamically binding the data acquisition method to the system's risk status.

[0108] Specifically, the method for determining the acquisition method of the power bidding information data of the updated management and control platform is as follows:

[0109] S21 uses the update management platform data in the platform to determine the composition ratio of the update management platform in the platform, and uses it as the composition ratio of the management platform;

[0110] It should be noted that if the proportion of the control platform is greater than the preset proportion threshold in the above steps, the update efficiency of the business extraction rules of some platforms will be slow. Therefore, the method for obtaining the power bidding data of the updated control platform is real-time acquisition and processing, which can quickly determine the update status of the business extraction rules of different updated control platforms, and thus lay the foundation for further verification and promotion of business extraction rules.

[0111] The calculation and preliminary decision-making for the composition ratio of the management and control platform: Updating the management and control platform refers to power procurement platforms that require an independent knowledge base for business extraction rule isolation due to frequent changes in platform format. The composition ratio of the management and control platform refers to the percentage of platforms currently marked as "updating management and control platforms" out of the total number of connected platforms. The calculation formula is: Number of updated management and control platforms ÷ Total number of platforms × 100%. The preset composition ratio threshold is a system-preset judgment standard, typically set to 60%, used to distinguish between "partial changes" and "overall changes" in the system.

[0112] Real-time acquisition and processing: This refers to increasing the data collection frequency to the highest level (e.g., once every 5 minutes) to achieve near real-time data updates and rule verification.

[0113] This step is a crucial part of system-level situational awareness. By quantitatively assessing the proportion of control platforms, the system can determine the current level of environmental stability from a macro perspective. When most platforms are in a highly volatile state, it means that the entire bidding environment is undergoing significant adjustments, requiring aggressive data acquisition strategies to accelerate rule verification.

[0114] Rapid response to systemic changes: When industry policies are adjusted or platforms undergo collective upgrades, the system can automatically identify and adjust its strategies.

[0115] Avoiding response lag: Traditional fixed-period data collection can lead to rule verification lag when the environment changes drastically. This method solves this problem. Intelligent resource allocation: High-cost real-time data collection is only enabled when necessary, optimizing the use of computing and bandwidth resources.

[0116] Example: Handling scenarios with concentrated platform changes:

[0117] A power data service provider's system connects to 30 power procurement platforms. After rule change analysis, it was determined that 7 platforms required isolation by establishing independent knowledge bases due to frequent format changes. These 7 update control platforms include: a bidding platform of a provincial power company under the State Grid, and an electronic procurement system of a branch of the Southern Power Grid. The system calculated the proportion of control platforms to be 7 / 30 = 23.3%, which is less than the preset threshold of 60%. The system determined that the current situation was a localized change scenario and did not trigger the global real-time acquisition strategy, proceeding to step S22 for in-depth analysis. This decision avoided unnecessarily enabling high-cost real-time data acquisition for all control platforms, saving system resources.

[0118] S22 takes the platform excluding the update management platform as other platforms, obtains the update data of the independent knowledge base of different update management platforms, determines the independent knowledge base with updated business extraction rules based on the update data, and uses it as the update knowledge base;

[0119] Furthermore, in the above steps, it is also necessary to determine the number of other platforms and whether the number of other platforms is greater than the preset threshold for the number of general platforms. If so, then the number of other platforms affected by the update of the business extraction rules is relatively large, and it is determined that the method for obtaining the power information data of the update control platform is real-time acquisition and processing. This allows for a quick determination of the update status of the business extraction rules of different update control platforms, thereby laying the foundation for further verification and promotion of the business extraction rules. If not, proceed to the next step.

[0120] Update knowledge base identification and impact assessment; other platforms: refers to stable operating platforms that continue to use public knowledge bases but are not included in the control scope. These platforms have relatively stable formats and mature business extraction rules.

[0121] Update knowledge base: Specifically refers to knowledge base instances in independent knowledge bases that have undergone operations such as adding, modifying or deleting business extraction rules. Preset general platform quantity threshold: Usually set to 10, used to determine whether the scale of "other platforms" is large enough and whether the business is important enough. The specific setting can be customized according to user needs.

[0122] This step enables a precise assessment of the risk impact. By identifying which independent knowledge bases have been actually updated and evaluating these updates, and considering the stable platform scale that cannot be effectively updated due to the construction of independent knowledge bases, the system can determine whether it is necessary to accelerate the rule verification of the control platform in order to protect the timeliness of updates to the business rules of most platforms.

[0123] Example: Impact Assessment of Management Platform Rule Updates on Platform Stability

[0124] Continuing with the previous embodiment, among the 30 total platforms, 7 are update control platforms, and 23 are other stable platforms. The system scanned the 7 independent knowledge bases and found that 3 of them had rule updates in the last 24 hours: Platform A added 5 rules, Platform B modified 3 rules, and Platform C added 8 rules. These 3 were marked as updated knowledge bases. The system assessed the number of stable platforms at 23, which is greater than the preset threshold of 10 for general platforms. The system determined that although the proportion of control platforms is not high (23.3%), a large number of stable platforms (23) may not be able to effectively update business extraction rules due to the construction of independent knowledge bases. Therefore, a real-time acquisition strategy needs to be adopted for these control platforms to accelerate rule verification. The system decided to initiate real-time acquisition processing for the control platforms corresponding to these 3 updated knowledge bases to ensure that new rules can be quickly verified, laying the foundation for subsequent promotion to the 23 stable platforms.

[0125] S23 determines the method for obtaining power bidding information data of the updated management and control platform based on the composition ratio of the management and control platform, other platforms, and the updated knowledge base data.

[0126] Based on comprehensive and refined decision-making based on multiple factors, data is collected according to a preset time period: this refers to adopting the system's default, low-frequency collection strategy, such as collecting data once per hour or every 4 hours.

[0127] Condition-triggered real-time acquisition: This means that real-time acquisition is only enabled when specific conditions are met, and regular acquisition is maintained at other times.

[0128] This step is the final stage of refining the strategy execution. Based on the macro-level judgments made in the first two steps—that is, when neither of the aforementioned macro-level judgments holds true—the system selects the data acquisition frequency that best matches the current system state, taking into account the number of knowledge bases undergoing rule updates, thus achieving an optimal balance between cost and benefit. By assessing the scale of knowledge base updates, the system can distinguish between minor changes on individual platforms and widespread changes across multiple platforms.

[0129] Ultimate resource optimization: High-cost real-time data collection is only enabled when absolutely necessary; progressive verification is supported: The system can verify new rules in a small-scale, incremental manner; multiple safeguards are provided: Multi-level judgments ensure the robustness and reliability of decisions.

[0130] Specifically, the above steps include:

[0131] S231 determines whether there is an updated knowledge base. If yes, proceed to the next step. If no, determine that the method for obtaining the power bidding data of the updated management platform is to obtain and process the power bidding data according to a preset time period.

[0132] S232 determines whether the number of updated knowledge bases exceeds a preset threshold. If so, it determines that the method for acquiring power bidding data of the update management platform is real-time acquisition and processing, which enables faster verification of updated business extraction rules and lays the foundation for further verification and promotion of business extraction rules. If not, it determines that the method for acquiring power bidding data of the update management platform is real-time acquisition and processing only when there is an update to the business extraction rules, which enables rapid verification of the reliability of the extraction of business extraction rules and lays the foundation for further promotion.

[0133] In a possible specific embodiment, continuing from the previous embodiment, the system has identified 3 updated knowledge bases. After entering step S23, the system first determines that there are updated knowledge bases (3), and proceeds to the next step. The system compares the number of updated knowledge bases (3) with the preset threshold for the number of updated knowledge bases (2), and finds that they are equal. According to the preset logic, when the number of updated knowledge bases reaches the threshold, real-time acquisition processing should be adopted to quickly verify the updated business extraction rules. The system decides to enable real-time acquisition processing for the management platforms corresponding to these 3 updated knowledge bases, adjusting the collection frequency from once per hour to once every 5 minutes. For the other 4 management platforms without rule updates, the system maintains the original collection frequency to avoid unnecessary resource consumption, and real-time acquisition processing is only required after the business extraction rules are updated. This refined strategy ensures that system resources are concentrated on the platforms that most need accelerated verification, improving overall efficiency.

[0134] S3 When the consistency of the business extraction rules of the independent knowledge base cannot meet the requirements, a verification processing scheme for the business extraction rules is determined based on the extraction data of the business extraction rules of the update management platform and the update data of the business extraction rules of different update management platforms.

[0135] This technical solution proposes a knowledge base quality assessment method based on rule consistency, specifically designed to determine whether the consistency of business extraction rules in independent knowledge bases established due to frequent platform format changes meets system requirements. This solution achieves an objective and quantitative assessment of rule consistency across multiple independent knowledge bases through three key steps: identifying biased knowledge bases, calculating their proportion, and setting quality thresholds. Its core value lies in establishing a standardized consistency evaluation system, providing a scientific basis for determining whether cross-knowledge base joint rule verification is necessary, and ensuring the consistency of business extraction rules across multiple platforms based on reliable verification of the business extraction rules.

[0136] Specifically, the consistency of business extraction rules for independent knowledge bases fails to meet the requirements, including:

[0137] S31 uses the business extraction rule data of different independent knowledge bases as a basis to identify independent knowledge bases whose business extraction rules are inconsistent with those of other independent knowledge bases, and uses them as deviation knowledge bases.

[0138] Deviation knowledge base identification; independent knowledge base: refers to a business rule storage repository, independently of the public knowledge base, specifically established for the update management platform that requires isolated processing due to frequent format changes. Business extraction rule data: refers to the set of structured rules stored in the independent knowledge base, including header matching rules, field extraction patterns, data cleaning logic, etc.

[0139] Deviation knowledge base: refers to a specific knowledge base whose business extraction rules are inconsistent with the rules of all other independent knowledge bases.

[0140] Criteria for determining rule inconsistency: This refers to a substantial difference between the business extraction rules adopted by the two knowledge bases.

[0141] This step is fundamental to anomaly detection and isolation. By identifying knowledge bases with completely unique rule sets among the full set of independent knowledge bases, joint verification is often required when the degree of inconsistency is high, thereby ensuring the consistency of rules invoked by the business.

[0142] Based on the preset criterion of "rules inconsistent with other independent knowledge bases", the system identified knowledge bases E and G as deviation knowledge bases. These two knowledge bases contain business extraction rules that cannot be found in any of the seven knowledge bases, exhibiting significant rule isolation.

[0143] S32 determines whether the consistency of the business extraction rules of the independent knowledge base meets the requirements based on the composition data of the deviation knowledge base.

[0144] Specifically, when the proportion of the deviation knowledge base in the independent knowledge base is greater than a preset proportion threshold, it is determined that the consistency of the business extraction rules of the independent knowledge base does not meet the requirements.

[0145] It should be noted that when the consistency of the business extraction rules of the independent knowledge base meets the requirements, there is no need to perform joint verification of the updated business extraction rules in different independent knowledge bases for the time being.

[0146] Consistency assessment and threshold determination, deviation knowledge base composition data: refers to the number, proportion, and specific details of rule deviations of deviation knowledge bases. Composition proportion: refers to the percentage of deviation knowledge bases out of the total number of independent knowledge bases. The calculation formula is: number of deviation knowledge bases ÷ total number of independent knowledge bases × 100%.

[0147] Preset composition percentage threshold: The system's preset quality control standard, usually set at 20%-30%, is used to determine whether the overall rule consistency of the independent knowledge base system is acceptable.

[0148] Consistency meets requirements: This means that the consistency of rules in the independent knowledge base system meets the system's preset standards, and there is no need to immediately conduct joint rule verification across knowledge bases.

[0149] This step is a core component of quality quantification and decision support. By calculating the proportion of the biased knowledge base and comparing it with a preset threshold, the system can objectively assess the overall rule consistency of the independent knowledge base system. This quantitative assessment provides data support for deciding whether to initiate costly joint rule verification.

[0150] Following the previous embodiments, the system has identified two out of seven independent knowledge bases as biased knowledge bases in its rule consistency assessment and decision-making. The system calculates the proportion of biased knowledge bases as 2 / 7 = 28.6%. The preset threshold for the proportion is 25%.

[0151] The system performs a threshold comparison: actual composition percentage: 28.6%, preset threshold: 25%, comparison result: 28.6%>25%, the system determines: the proportion of the deviation knowledge base composition exceeds the preset threshold, and the consistency of the business extraction rules of the independent knowledge base system does not meet the requirements.

[0152] It should be noted that the data extracted by the business extraction rules of the update management platform is determined based on the number of times the update management platform extracts power bidding information data using the business extraction rules.

[0153] Specifically, the method for determining the verification processing scheme of the business extraction rules is as follows:

[0154] This technical solution proposes an intelligent verification decision-making system based on rule update frequency and application effectiveness, specifically designed to determine how to process business extraction rules for verification across multiple update and management platforms. Through core steps such as quantifying the scale of rule updates, statistically analyzing rule application effects, and calculating verification requirement values, this solution evaluates the verification priority of different platforms from multiple dimensions. Its core innovation lies in establishing a multi-dimensional verification requirement assessment model, which dynamically determines the optimal verification scope and processing scheme while ensuring the stability of power bidding data extraction and processing across various platforms.

[0155] S41 takes the updated business extraction rules in the independent knowledge base of the update management platform as the update extraction rules, and determines the number of update extraction rules of the update management platform based on the update data of the business extraction rules of the update management platform;

[0156] Update extraction rule identification and quantity statistics. Update extraction rules: refers to the business extraction rules added or modified in the independent knowledge base of the update management platform. Update data of business extraction rules: refers to the detailed record of rule changes in the independent knowledge base, including change time, change type, rule ID, change content, etc.

[0157] Number of updated extraction rules: refers to the total number of business extraction rules that have been updated within a specific time window (such as the last 24 hours or the last 7 days).

[0158] This step is fundamental to quantifying rule changes. By accurately counting the number of rule updates across each management platform, the system can grasp the magnitude and frequency of changes on each platform, providing a data foundation for subsequent verification prioritization and real-time monitoring of the change trend. It quantifies the rule changes across platforms, promptly identifies platforms with high-frequency changes, and establishes a change baseline. This provides objective quantitative indicators for assessing platform stability and supports trend analysis. Long-term statistical analysis identifies platform change patterns and predicts future change trends.

[0159] Example: Multi-platform rule update statistics

[0160] A power data processing system has established independent knowledge bases for seven update and management platforms. The system monitors the rule updates of each platform over the past week.

[0161] Platform A (a provincial platform of the State Grid): 8 new rules, 3 modified rules, totaling 11 updates; Platform B (a branch platform of the Southern Power Grid): 5 new rules, 2 modified rules, totaling 7 updates; Platform C (a local power trading platform): 12 new rules, 5 modified rules, totaling 17 updates; Platform D (a new energy bidding platform): 3 new rules, 1 modified rule, totaling 4 updates; Platform E (a distribution network bidding platform): 6 new rules, 4 modified rules, totaling 10 updates; Platform F (a power equipment procurement platform): 2 new rules, 0 modified rules, totaling 2 updates; Platform G (a power service procurement platform): 9 new rules, 3 modified rules, totaling 12 updates. The system identified a total of 63 rule updates across these 7 platforms, with Platform C having the most updates (17) and Platform F having the fewest (2). This data provides the basic input for subsequent verification decisions.

[0162] S42 uses the extracted data from the business extraction rules of the update management platform to determine the number of times the power bidding data is extracted using the update extraction rules;

[0163] Update rule application effect statistics, business extraction rule extraction data: refers to the application record of each business extraction rule in the actual data extraction process, extraction number: refers to the number of times a specific business extraction rule is successfully applied to power bidding data extraction within the statistical period.

[0164] Preset extraction threshold: The system's preset judgment standard is used to distinguish between high-frequency and low-frequency usage rules, and is usually set to 50-100 times.

[0165] This step is crucial for rule value assessment. By statistically analyzing the actual application effects of rules, the system can identify those rules that have high reliability in real-world business applications. These rules, with their high stability, should be prioritized for verification.

[0166] Example: Statistical Analysis of Rule Application Frequency Continuing from the previous example, the system statistically analyzes the application of update rules on each platform in the past week: The number of times the 11 update rules of platform A were applied: R_A001: 85 times, R_A002: 92 times, R_A003: 78 times, and the number of times the remaining 8 rules were applied was between 15 and 45 times.

[0167] The number of times the 17 update rules of platform C were applied: R_C001: applied 120 times, R_C002: applied 115 times, R_C003: applied 98 times, and the remaining 14 rules were applied between 5 and 35 times.

[0168] The preset threshold for extraction times is 50 times. The system identified that: Platform A has 3 rules that are extraction adaptation rules (application times > 50), Platform C has 3 rules that are extraction adaptation rules, and the number of extraction adaptation rules for other platforms varies from 0 to 2.

[0169] S43 determines the verification and processing scheme for the business extraction rules based on the number of update extraction rules of different update management and control platforms and the number of times the power bidding data is extracted using the update extraction rules.

[0170] Furthermore, based on the number of update extraction rules on different update management platforms and the number of times power bidding data is extracted using the aforementioned update extraction rules, a verification and processing scheme for the business extraction rules is determined, specifically including:

[0171] Based on the number of extractions of power bidding data using the aforementioned update extraction rules, update extraction rules with extraction counts greater than a preset extraction count threshold are determined and used as extraction adaptation rules.

[0172] Based on the number of extraction adaptation rules and the sum of the number of update extraction rules in the update management platform, the verification requirement value of the update management platform is determined. The update management platform with the largest verification requirement value is selected as the verification requirement platform. Based on the verification requirement value of the verification requirement platform, the verification processing scheme for the business extraction rules is determined.

[0173] Verification processing plan comprehensive decision-making, verification demand value: a quantitative indicator comprehensively reflecting the platform's verification priority, calculated as: number of extracted adaptation rules + number of updated extracted rules. Verification demand platform: the platform with the highest verification demand value is the target platform that needs to be prioritized for rule verification.

[0174] This step is the core of multi-dimensional intelligent decision-making. By comprehensively analyzing the number of rule updates and their application effects, the system can dynamically formulate the optimal verification strategy, balancing verification comprehensiveness and resource efficiency.

[0175] Prioritize data collection: Intelligently identify the most critical targets for verification from multiple platforms; formulate differentiated strategies: Develop different verification scope plans based on different situations; balance cost-effectiveness: Optimize resource usage while ensuring verification effectiveness.

[0176] Verification requirement platform identification

[0177] The system calculates the verification requirement values ​​for each platform: Platform A: 3 extraction adaptation rules + 11 update rules = 14, Platform B: 1 extraction adaptation rule + 7 update rules = 8, Platform C: 3 extraction adaptation rules + 17 update rules = 20, Platform D: 0 extraction adaptation rules + 4 update rules = 4, Platform E: 2 extraction adaptation rules + 10 update rules = 12, Platform F: 0 extraction adaptation rules + 2 update rules = 2, Platform G: 2 extraction adaptation rules + 12 update rules = 14. Platform C has the highest verification requirement value (20) and is therefore identified as the platform with the verification requirement.

[0178] Furthermore, based on the verification requirement value of the verification requirement platform, a verification processing scheme for the business extraction rules is determined, specifically including:

[0179] S431 determines whether the verification requirement value of the verification requirement platform is greater than the preset verification requirement threshold. If yes, then the verification processing of the update extraction rule of the verification requirement platform is carried out in all platforms. If no, then proceed to the next step.

[0180] The platform-wide verification process has a preset verification threshold of 18. Platform C's verification threshold of 20 > 18, which should trigger platform-wide verification. However, if the preset verification threshold is 21, the process proceeds to the next step.

[0181] Full-platform verification judgment, preset verification requirement threshold: The system presets a critical value for determining whether to initiate full-platform verification, usually set to 15-20. This threshold is determined based on a combination of historical verification data and resource costs.

[0182] All platforms: refers to all power information processing platforms in the system, including update and management platforms and stability platforms, which are 30 platforms in this embodiment.

[0183] Full-platform verification: refers to the processing mode of testing and verifying the update extraction rules of the verification requirement platform on all 30 platforms.

[0184] This step is the first-level decision-making stage for the verification scope, aiming to identify rule update scenarios with extremely high verification requirements that may have a significant impact on the entire system. When the verification requirement value exceeds a preset threshold, it indicates that the scale and frequency of rule updates on the platform have reached a level that requires the attention of the entire system.

[0185] Comprehensive risk control: For major rule changes that may affect the entire system, the most comprehensive verification strategy is adopted to test the new rules across the entire platform, so as to achieve a comprehensive update of the business extraction rules and establish a reliable verification foundation for subsequent rule standardization and promotion.

[0186] S432 determines whether the verification requirement value of the verification requirement platform is less than the preset requirement threshold. If so, the verification process of the update extraction rule of the verification requirement platform is only performed on the update control platform where there is no update business rule. If not, proceed to the next step.

[0187] The preset requirement threshold is 6. Platform C's verification requirement value 20 > 6, which does not meet the minimum range verification condition. At this time, there are a large number of update business rules that need to be verified, and the initial verification reliability is high. Therefore, verification can be carried out in multiple update management platforms.

[0188] Preset requirement threshold: This is a pre-defined threshold value used by the system to determine whether only minimal verification is required, typically set between 0 and 8. This threshold represents the minimum acceptable standard for verification requirements. Minimum verification refers to a processing mode where rule verification is only performed on "update management platforms where there are no updated business rules." These platforms have stable rules and a relatively clean verification environment.

[0189] Update control platforms that do not update business rules: These are platforms that are listed as control platforms (due to high historical change frequency) but have not had any rule updates in the current statistical period.

[0190] This step is a secondary decision on the scope of verification, targeting scenarios with low verification requirements. When the verification requirement is below the preset threshold, it indicates that the impact of the rule update is limited, and basic verification only needs to be performed in the most stable environment.

[0191] Save verification resources: For rule updates with limited impact, avoid unnecessary platform-wide verification overhead and utilize a stable environment: Perform verification on a platform with stable rules to reduce interference from environmental variables and achieve rapid verification feedback: Simplify the verification process and shorten the verification cycle.

[0192] S433 determines whether all the update extraction rules of the verification requirement platform belong to the extraction adaptation rules. If so, the update extraction rules of the verification requirement platform are verified in all update management platforms. If not, proceed to the next step.

[0193] Full control platform verification and judgment, extraction of adaptation rules: refers to those update rules that have been extracted more than the preset extraction number threshold (such as 50 times) in actual applications. These represent core rules that are frequently used and have been verified to a certain extent, and their reliability is relatively high. Full control platform verification: refers to the processing mode of verifying rules in all update control platforms (not the stable platform).

[0194] Rule type consistency: This refers to whether all updated rules belong to the core rule types that are frequently used. This is a key criterion for determining whether a full management platform verification is required.

[0195] This step is a three-level decision-making process for the scope of validation, specifically designed for scenarios where the update rules are of high quality. When all update rules fall under the category of extracting adaptation rules, it indicates that these rules have undergone a certain degree of practical application validation, possess high initial quality, and are suitable for further validation in a broader but still controlled environment.

[0196] Rapid Promotion of High-Quality Rules: Adopt a more proactive verification strategy for high-quality updated rules to accelerate their application and promotion; Verification in Control Environments: Verify rules in control platform environments with a certain degree of change to assess their adaptability; Balancing Verification Efficiency and Quality: Improve verification efficiency while ensuring verification quality.

[0197] Example: Applicable scenarios for full control platform verification:

[0198] Assume platform A has a verification requirement of 20, with 10 update rules all being extraction and adaptation rules (each rule applied more than 50 times). System checks reveal that these 10 rules are frequently used and meet the verification criteria of the full-control platform. The system decides to verify these 10 rules across all 7 update control platforms.

[0199] S434 Based on the number of update extraction rules of the update management platform, determine whether the number of update extraction rules of the update management platform is greater than the preset update rule number threshold. If yes, determine that the update management platform does not belong to the verification processing platform of the update extraction rules of the verification requirement platform. If no, proceed to the next step.

[0200] Rule quantity filtering, preset update rule quantity threshold: The maximum number of existing update rules to be extracted by the system for a single platform, usually set to 5. Exceeding this threshold is considered excessively complex. Verification processing platform: Refers to the target platform that actually performs the rule verification operation. These platforms will apply the rules to be verified to process their own data. Verification complexity: Refers to the comprehensive evaluation index of the time, computing resources, and manual input required to perform the rule verification task.

[0201] This step verifies the adaptability of the verification platform's capabilities, aiming to ensure that the verification platform can effectively complete the verification task. When a platform has too many updated rules, it may not have sufficient processing capacity to verify the rules of other platforms, and therefore will be excluded from the verification platform list.

[0202] Ensure verification quality: Avoid the verification platform being overloaded and affecting the verification effect. Reasonably allocate verification load: Dynamically allocate verification tasks according to the actual capacity of the platform to prevent technical problems such as poor operation stability caused by a large number of business rules being updated on a single platform.

[0203] Platform C has 7 update rules to process, exceeding the preset threshold of 6. Platform C is marked as "unavailable" and will not participate in this verification task allocation.

[0204] S435 designates the update management platform that is not part of the verification processing platform for update extraction rules of the verification requirement platform as an unavailable platform. It then determines whether the proportion of unavailable platforms in the update management platform is greater than a preset proportion threshold. If so, all update management platforms excluding unavailable platforms are considered to be verification processing platforms for update extraction rules of the verification requirement platform. If not, the update management platform is determined to be a verification processing platform for update extraction rules of the verification requirement platform based on the number of update extraction rules excluding adaptation rules in the update management platform.

[0205] Furthermore, when the number of update extraction rules in the update management platform, excluding the extraction adaptation rules, is less than a preset threshold for the number of extraction rules, the update management platform is determined to be a verification processing platform for the update extraction rules of the verification requirement platform.

[0206] The final verification platform determines that unusable platforms refer to update control platforms that are excluded from the verification platform scope due to excessive number of rules, high load, or other reasons.

[0207] Preset quantity percentage threshold: refers to the maximum allowed percentage of unavailable platforms out of the total number of update control platforms, usually set to 30%.

[0208] Remaining rule count judgment: This refers to whether the number of remaining update rules in the platform is lower than a preset threshold (usually 4) after excluding the extracted adaptation rules.

[0209] This step is the final decision-making process for platform selection, determining the most suitable set of platforms for performing the verification tasks through multi-dimensional evaluation. This step ensures the reasonable selection of verification platforms, considering both platform availability and the suitability for the verification tasks. Optimizing platform selection: A comprehensive evaluation from multiple dimensions selects the most suitable verification platform, dynamically adapting to system status: Verification strategies are adjusted based on the real-time status of the platforms to reduce the impact on the verification processing platform.

[0210] The system currently has 7 update management platforms, among which platform C is marked as unavailable due to an excessive number of rules. The unavailable platform ratio is calculated to be 14.3%, which is lower than the preset threshold of 30%. The system proceeds to the next step of judgment, analyzing the number of update rules in each platform excluding the extraction and adaptation rules (the preset threshold is 4): Ultimately, platform F is determined as the verification platform, responsible for verifying and processing the update business rules of the verification request platform.

[0211] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0212] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0213] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A power signal full-process intelligent processing and data knowledge construction system, characterized in that, Specifically, it includes: The data acquisition module constructs a targeted crawling system for power system platforms, covering multiple platforms, and obtains bidding information page fields, page screenshots and attachments, which are then stored with a unique index. The file processing module is responsible for unifying the formats of the collected multi-format files, performing file structure and content legality checks, and outputting standardized files for use by the subsequent parsing module. The platform segmentation module determines the update control platform in the platform based on the update data of the knowledge base of the platform, and determines the verification processing target and verification processing scheme of the business extraction rules of the update control platform based on the update processing data of the independent knowledge base of the update control platform. The intelligent parsing module is responsible for building a parsing module based on the energy model, integrating business extraction rules with deep learning capabilities, and realizing three-level header matching and field semantic recognition; The standardization processing module is responsible for uniformly encoding, format conversion and data cleaning of the parsing results, and automatically updating the knowledge base when new business extraction rules are available. The updated management platform was isolated from the system due to frequent or significant rule changes, and an independent knowledge base was established for management. The method for determining the verification and processing scheme for business extraction rules is as follows: The update extraction rule is the business extraction rule that is added or modified in the independent knowledge base of the update management platform. The extraction adaptation rule is the update extraction rule whose extraction count is greater than the preset extraction count threshold. The verification requirement value of the update management platform is the sum of the number of extraction adaptation rules and the number of update extraction rules in the update management platform. The verification requirement platform is the update management platform with the largest verification requirement value. S431 determines whether the verification requirement value of the verification requirement platform is greater than the preset verification requirement threshold. If so, all platforms perform verification processing of the verification requirement platform's update extraction rules. If not, proceed to S432. S432 determines whether the verification requirement value of the verification requirement platform is less than the preset requirement threshold. If so, the verification process of the update extraction rule of the verification requirement platform is carried out on the update control platform where there is no update business rule. If not, proceed to S433. S433 Determine whether all the update extraction rules of the verification requirement platform belong to the extraction adaptation rules. If so, perform verification processing of the update extraction rules of the verification requirement platform in all update control platforms. If not, proceed to S434. S434 Determine whether the number of update extraction rules of the update management platform is greater than the preset update rule number threshold. If yes, the update management platform is not a verification processing platform for the update extraction rules of the verification requirement platform. If no, proceed to S435. S435 designates the update management platform that is not part of the verification processing platform for update extraction rules of the verification requirement platform as an unavailable platform. It then determines whether the proportion of unavailable platforms in the update management platform is greater than a preset proportion threshold. If so, the update management platform excluding unavailable platforms becomes the verification processing platform for update extraction rules of the verification requirement platform. If not, it determines whether the platform belongs to the verification processing platform for update extraction rules of the verification requirement platform based on the number of update extraction rules excluding adaptation rules in the update management platform.

2. The power signal full-process intelligent processing and data knowledge construction system of claim 1, wherein, The acquisition of the bidding information page fields, page screenshots, and attachments specifically includes: Set differentiated data collection strategies for the update frequency of different platforms; It integrates a dynamic IP proxy pool to obtain and process tag page fields, page screenshots, and attachments by rotating IP addresses and simulating browser behavior. 3.The power signal full-process intelligent processing and data knowledge construction system of claim 1, wherein, Output standardized files for use by subsequent parsing modules, specifically including: The collected multi-format files are converted to obtain structured data objects through format unification. A two-layer validity verification method is used to determine the integrity of the detected documents and whether they fall within the valid standard range. The preprocessed files are uniformly converted into standard JSON format, i.e., standardized files, and metadata is attached to provide high-quality input for subsequent parsing.

4. The intelligent processing and data knowledge construction system for the entire power bidding process as described in claim 1, characterized in that, The intelligent parsing module specifically includes: The parser scheduling center automatically allocates the optimal parser based on the file type. The content extraction engine uses a fusion of computer vision and natural language processing technology. It first identifies table areas through layout analysis, and then uses row and column detection algorithms to reconstruct the table structure. Based on the Transformer architecture, a pre-trained model is used, combined with the matching of tabular data and business extraction rules, to perform three-level header matching mechanism: the first level matches standard headers, the second level matches synonym headers, and the third level performs semantic inference, thus performing three-level header matching and field semantic recognition.

5. A method for intelligent processing and data knowledge construction of power information tags, applied to the intelligent processing and data knowledge construction system for the entire power information tag process as described in any one of claims 1-4, characterized in that, Specifically, it includes: Based on the energy model, power bidding data from multiple platforms are acquired and processed, and the knowledge base is updated according to the acquisition and processing results. The update control platform in the platform is determined according to the update data of different platforms in the knowledge base, and an independent knowledge base is constructed for the update control platform and used as an independent knowledge base. Based on the update data of the business extraction rules of the update management platform and the independent knowledge base of the update management platform in the platform, the method for obtaining the power bidding information data of the update management platform is determined. Based on the acquisition method, the power bidding information data of the update management platform is processed, and the business extraction rules of the independent knowledge base of different update management platforms are updated. When the consistency of business extraction rules in an independent knowledge base cannot meet the requirements, a verification processing scheme for business extraction rules is determined based on the extraction data of the business extraction rules of the update management platform and the update data of the business extraction rules of different update management platforms.

6. The method for intelligent processing and data knowledge construction of power information as described in claim 5, characterized in that, The power bidding data refers to the parsed data of bidding documents from different platforms.

7. The method for intelligent processing and data knowledge construction of power information as described in claim 5, characterized in that, The knowledge base is updated, specifically including: Based on the analysis results of power bidding data from different platforms, bids with keywords that are inconsistent with the keywords in the original knowledge base are identified. The knowledge base is updated based on the keywords of the tender documents that are inconsistent with the keywords of the original knowledge base and the extraction strategy of power tender information data.

8. The method for intelligent processing and data knowledge construction of power information as described in claim 5, characterized in that, The method for determining the update control platform in the aforementioned platform is as follows: Based on the updated data of different platforms in the knowledge base, determine the updated data of the business extraction rules of the platform in the knowledge base; By utilizing the platform data and the update data of the business extraction rules of different platforms in the knowledge base, the update control platform in the platform is determined.

9. The method for intelligent processing and data knowledge construction of power information as described in claim 5, characterized in that, The consistency of business extraction rules for independent knowledge bases fails to meet the requirements, specifically including: Based on the business extraction rule data of different independent knowledge bases, we identify the independent knowledge base whose business extraction rules are inconsistent with those of other independent knowledge bases and use it as the deviation knowledge base. Based on the composition data of the deviation knowledge base, determine whether the consistency of the business extraction rules of the independent knowledge base meets the requirements.

10. The method for intelligent processing and data knowledge construction of power information as described in claim 9, characterized in that, When the consistency of the business extraction rules of the independent knowledge base meets the requirements, there is no need to temporarily verify the updated business extraction rules in different independent knowledge bases.

Images