A multi-dimensional lookboard analysis method and system for grid-connected review content

Abstract

The application relates to a multi-dimensional board analysis method and system for grid-connected review content, comprising the following steps: acquiring grid-connected data records and generating a grid-connected review data set; constructing a progress time sequence feature vector and further constructing a risk prediction model for tolerance and overage, calculating the tolerance and overage risk probability of a target grid-connected project and generating an overage risk prediction list; generating multi-dimensional index board data based on the grid-connected review data set and the overage risk prediction list; and establishing a correlation mapping between risk items and the progress time sequence feature vector to generate a multi-dimensional analysis result. The application integrates multi-heterogeneous platform data to form a complete data set with a project as a unit, extracts progress time sequence features to construct a risk prediction model, realizes a forward-looking prediction of tolerance and overage risk, and establishes a multi-level index correlation multi-dimensional index board to support layer-by-layer drilling positioning. The application can solve the problems of cross-platform data integration difficulty, risk control lag and insufficient multi-dimensional analysis capability in the prior art.

Description

Technical Field

[0001] This application relates to the technical field of power system grid connection review, and in particular to a multi-dimensional dashboard analysis method and system for grid connection review content. Background Technology

[0002] With the continued large-scale development of new energy power generation projects, the volume of grid connection review work has increased significantly, and the complexity of the review process has also increased accordingly. The grid connection review process typically covers multiple stages, including document review, on-site verification, and rectification tracking. The business data generated at each stage is stored in heterogeneous platforms such as the grid connection process platform, ITSCADA system, and architecture management system. Due to the different construction periods of these systems, independent data identification rules, and the lack of unified storage architecture standards, cross-system data integration faces significant obstacles. Existing technologies are insufficient to form a complete data view with grid connection projects as the basic unit, and reviewers cannot obtain coherent and comprehensive project information.

[0003] Regarding progress control, existing management methods primarily rely on manual ledger recording and periodic report summaries, lacking automated support for the collection and analysis of project progress timeline data. While the conditional review mechanism allows grid-connected projects to proceed with the review process even with incomplete documentation, the resulting issue of overdue and unclosed conditional work orders is also significant. Limited by the data processing capabilities of the existing system, managers cannot proactively identify the risk of overdue conditional work orders based on historical progress data; they can only initiate reactive measures after overdue events occur, easily leading to delays in the review process and wasted review resources.

[0004] In terms of visualization analysis, existing grid connection review dashboards are mostly presented in the form of static reports or single-dimensional charts. There is no index relationship between the data of each indicator, making it difficult to support drill-down positioning operations from the overall project dimension to the review stage dimension and then to the work order dimension. Summary of the Invention

[0005] To address the aforementioned shortcomings, this application provides a multi-dimensional dashboard analysis method and system for network-connected review content.

[0006] The above-mentioned objective of this application is achieved through the following technical solution: A multi-dimensional dashboard analysis method for content reviewed online includes the following steps: Grid connection data records are obtained from the grid connection process platform, ITSCADA system and architecture control system, and integrated to generate a grid connection review dataset with grid connection projects as the basic unit. The grid connection review dataset is time-series aligned, and the timestamp sequences of each grid connection project are extracted to construct the progress time-series feature vector of each grid connection project. Based on the progress time-series feature vector, statistical fitting is performed on the time-series feature distribution corresponding to the tolerance delay events in historical grid-connected projects, the preceding feature interval of the tolerance delay events in the time-series feature vector is extracted, and a tolerance delay risk prediction model is constructed. Identify the target grid-connected project, input the progress time sequence feature vector of the target grid-connected project into the tolerance for overdue risk prediction model, calculate the tolerance for overdue risk probability of the target grid-connected project, and generate an overdue risk prediction list; Multidimensional indicator dashboard data is generated based on the grid connection review dataset and the overdue risk prediction list. Multi-level index relationships are established for the multidimensional indicator dashboard data to support drill-down positioning of indicator data. Establish a correlation mapping between the risk items in the overdue risk prediction list and the progress time sequence feature vectors of the corresponding grid-connected projects, and generate multi-dimensional analysis results of the grid connection review process.

[0007] The second objective of this invention is achieved through the following technical solution: A multi-dimensional dashboard analysis system for online content review includes: The data integration module is used to obtain grid connection data records from the grid connection process platform, ITSCADA system and architecture control system, and integrate the grid connection data records to generate a grid connection review dataset with grid connection projects as the basic unit. The time series processing module is used to perform time series alignment processing on the grid connection review dataset, extract the timestamp sequence of each grid connection project, and construct the progress time series feature vector of each grid connection project. The model building module is used to statistically fit the time-series feature distribution corresponding to the defective and overdue events in historical grid-connected projects based on the progress time-series feature vector, extract the preceding feature interval of the defective and overdue events in the time-series feature vector, and build a defective and overdue risk prediction model. The risk prediction module is used to identify target grid-connected projects. The progress time-series feature vector of the target grid-connected project is input into the tolerance for overdue risk prediction model to calculate the tolerance for overdue risk probability of the target grid-connected project and generate an overdue risk prediction list. The index association module is used to generate multi-dimensional indicator dashboard data based on the grid connection review dataset and the overdue risk prediction list, and to establish multi-level index association relationships for the multi-dimensional indicator dashboard data to support the drilling down of indicator data layer by layer. The results generation module is used to establish a correlation mapping between the risk items in the overdue risk prediction list and the progress time sequence feature vectors of the corresponding grid-connected projects, and generate multi-dimensional analysis results of the grid connection review process.

[0008] This application also relates to a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described multi-dimensional dashboard analysis method for network-connected review content.

[0009] This application also relates to a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method for multi-dimensional dashboard analysis of network-connected review content.

[0010] This application provides a multi-dimensional dashboard analysis method and system for grid connection review content. By integrating grid connection data from multiple heterogeneous platforms to form a complete dataset based on projects, it extracts progress time-series features to construct a risk prediction model for overdue tolerances, thereby achieving risk prediction. Furthermore, it establishes a multi-dimensional indicator dashboard with multi-level index associations to support drill-down positioning. This method can solve the problems of difficulty in cross-platform data integration, lagging risk control, and insufficient multi-dimensional analysis capabilities in existing technologies. It can effectively improve the efficiency of grid connection review, achieve forward-looking control of overdue tolerance risks, and enhance the level of refinement in grid connection review management. Attached Figure Description

[0011] Figure 1 This is a flowchart of an embodiment of a multi-dimensional dashboard analysis method for network-connected review content according to this application; Figure 2 This is a flowchart of step S10 in an embodiment of a multi-dimensional dashboard analysis method for network-connected review content in this application; Figure 3 This is a flowchart of the implementation after step S13 in an embodiment of the multi-dimensional dashboard analysis method for grid-connected review content of this application. Detailed Implementation

[0012] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0013] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0014] In one embodiment, such as Figure 1 As shown, this application discloses a multi-dimensional dashboard analysis method for grid connection review content, which specifically includes the following steps: S10: Obtain grid connection data records from the grid connection process platform, ITSCADA system, and architecture control system, and integrate the grid connection data records to generate a grid connection review dataset with grid connection projects as the basic unit; In this embodiment, the grid connection process platform refers to a business system used to manage the application, approval, and document submission processes of grid connection projects. It typically records the project's progress and related document information at different review stages. The ITSCADA system refers to a real-time data acquisition and monitoring system in the power system, which typically acquires real-time operating data, equipment status, and performance indicators of grid connection projects. The architecture control system refers to a system used to manage the physical architecture, equipment configuration, and asset information of grid connection projects. It typically stores detailed technical specifications and asset instance data for grid connection projects. Grid connection data records refer to raw data entries obtained from different data sources such as the aforementioned grid connection process platform, ITSCADA system, and architecture control system. These records may include various information such as project name, time, status, and equipment model. A grid connection project refers to an independent new energy power generation or energy storage project that requires grid connection review. In this embodiment, the grid connection project serves as the basic unit for data integration and analysis. The grid connection review dataset refers to a collection of review-related data from different data sources, centered on the grid connection project, after integration and processing.

[0015] S20: Perform time-series alignment processing on the grid connection review dataset, extract the timestamp sequence of each grid connection project, and construct the progress time-series feature vector of each grid connection project; In this embodiment, time-series alignment refers to the standardization of time-series data from different grid-connected projects or different events within the same grid-connected project, making them comparable in the time dimension. As an example, time-series alignment can be achieved through the following steps: First, convert the raw time data of all grid-connected projects into a unified standard format, eliminating time zone or unit differences and establishing a unified time benchmark. Then, addressing the issue of varying event densities across different projects, a unified time grid method is used. Fixed time intervals are set, and linear interpolation or forward padding is employed to transform irregular timestamp sequences into regular sequences with fixed frequencies, ensuring a one-to-one correspondence between data points for each project in the time dimension. Finally, based on the aligned time series, the state values ​​of each time slice are extracted, and statistical indicators such as mean and slope are calculated using sliding window techniques. These are then combined sequentially to form a progress time-series feature vector representing the evolution of project progress. A timestamp sequence is a set of time stamps recording key events or state changes occurring at different points in time within a grid-connected project. A progress time-series feature vector is a set of numerical features that characterizes the evolution of project progress, constructed by analyzing and extracting the timestamp sequences of grid-connected projects.

[0016] S30: Based on the progress time sequence feature vector, statistically fit the time sequence feature distribution corresponding to the defective overdue events in historical grid-connected projects, extract the preceding feature interval of the defective overdue events in the time sequence feature vector, and construct a defective overdue risk prediction model. In this embodiment, a "tolerance for incomplete documentation exceeding the deadline" event refers to an event during the grid connection review process where a work order allowed to proceed due to incomplete documentation fails to be closed within the stipulated time, resulting in an overdue event. The time-series characteristic distribution refers to the numerical change pattern or probability distribution of the grid connection project's progress time-series characteristic vector over time before the occurrence of the tolerance for incomplete documentation exceeding the deadline event. Statistical fitting refers to modeling the time-series characteristic distribution in historical data using mathematical statistical methods to identify its inherent patterns and regularities. The preceding characteristic interval refers to a specific numerical range or trend in the progress time-series characteristic vector that is strongly correlated with the occurrence of the tolerance for incomplete documentation exceeding the deadline event before its occurrence. The tolerance for incomplete documentation exceeding the deadline risk prediction model is a mathematical model or algorithm constructed based on the time-series characteristic distribution patterns of tolerance for incomplete documentation exceeding the deadline event in historical data, capable of predicting the probability of such a risk occurring in a target grid connection project.

[0017] S40: Identify the target grid-connected project, input the progress time sequence feature vector of the target grid-connected project into the tolerance for overdue risk prediction model, calculate the tolerance for overdue risk probability of the target grid-connected project, and generate an overdue risk prediction list; In this embodiment, the target grid-connected project refers to a specific grid-connected project that currently requires risk assessment and analysis. The probability of a tolerance delay exceeding the expiration date refers to the likelihood, expressed numerically, that a tolerance delay event will occur at a future point in time for the target grid-connected project. The expiration risk prediction list is a list containing the tolerance delay risk probabilities of all target grid-connected projects, which may be sorted according to risk level.

[0018] Table 1 shows an example of a list of risks of overdue payment.

[0019]

[0020] Table 1: List of Risks of Overdue Payments S50: Generate multi-dimensional indicator dashboard data based on the grid connection review dataset and the overdue risk prediction list, and establish multi-level index association relationships for the multi-dimensional indicator dashboard data to support the drilling down of indicator data layer by layer. In this embodiment, the multidimensional indicator dashboard data refers to the key indicators from the grid connection review dataset and the overdue risk prediction list, organized in an interactive manner for visual display. Multi-level index relationships refer to the logical connections established between indicators at different levels within the multidimensional indicator dashboard data, allowing users to gradually delve from a macro to a micro level to view detailed data. Drill-down positioning refers to the operation where users click or select higher-level indicators on the dashboard to progressively expand and view more detailed indicator data under them, until they pinpoint a specific problem or risk item.

[0021] S60: Establish a correlation mapping between the risk items in the overdue risk prediction list and the progress time sequence feature vector of the corresponding grid-connected project, and generate multi-dimensional analysis results of the grid connection review process.

[0022] In this embodiment, the association mapping refers to the logical connection established between the risk items in the overdue risk prediction list and the corresponding progress time-series feature vectors of the grid-connected projects, in order to trace the source of risks and analyze their evolution. The multi-dimensional analysis results refer to the insights and conclusions obtained after conducting in-depth analysis of the grid connection review process from multiple perspectives using the above methods, including risk warnings, progress assessments, and problem identification.

[0023] For example, as a specific implementation, suppose a power company manages a large number of renewable energy grid-connected projects and is facing problems such as data fragmentation, delayed risk warnings, and insufficient visualization and analysis capabilities. Specifically, the company's internal grid connection process platform records project approval and document submission status, the ITSCADA system provides real-time operational data, and the architecture management system maintains project equipment configuration information. However, these systems are independent of each other, and project identification is inconsistent, making it difficult to fully grasp the complete review status of a grid-connected project. For example, taking a photovoltaic power generation grid-connected project as Project A, its documents show as submitted on the grid connection process platform, but the ITSCADA system may not have corresponding real-time data, or the equipment information in the architecture management system may not completely match the project name in the process platform.

[0024] To address the aforementioned issues, grid connection data records are first obtained from the grid connection process platform, the ITSCADA system, and the architecture management system. For example, all raw data related to the grid connection project can be periodically extracted from these three heterogeneous data sources. The grid connection data records are then integrated, for instance, using a rule-based matching mechanism. If data records from different systems have the same project name and approximately the same application date, they are preliminarily identified as belonging to the same grid connection project. Specifically, for project A, the raw data recorded in the grid connection process platform may include fields such as project number, application date, and document submission status; the corresponding data records in the ITSCADA system may include equipment online status and collection point number; and the architecture management system may include configuration information such as inverter model and access voltage level. Data that cannot be automatically matched using rules can be marked and await manual verification and confirmation. Through this method, a grid connection review dataset can ultimately be generated, with each grid connection project as the basic unit. This dataset contains various review information, operational data, and architecture information for each grid connection project.

[0025] Furthermore, the grid connection review dataset undergoes time-series alignment processing to extract the timestamp sequences of each grid connection project, constructing a progress time-series feature vector for each project. For example, for project A, timestamps of all key events can be extracted from its review dataset, such as project initiation time, initial document review completion time, on-site verification start time, work order creation time, and work order closure time. Arranging the extracted key event timestamps in chronological order forms the timestamp sequence for project A. Then, based on the timestamp sequence, a progress time-series feature vector reflecting the progress evolution of project A is constructed. This progress time-series feature vector may include basic time-series features such as the time interval between events and the cumulative number of events.

[0026] Furthermore, based on the schedule time-series feature vector, statistical fitting is performed on the time-series feature distribution corresponding to the tolerable delay events in historical grid-connected projects. The preceding feature intervals of the tolerable delay events in this time-series feature vector are extracted, and a tolerable delay risk prediction model is constructed. For example, all grid-connected projects that have experienced tolerable delay events in the past can be analyzed to observe the common patterns presented by the schedule time-series feature vectors in the period before the tolerable work order expires. Through statistical analysis, it can be identified that the value range or trend of certain time-series features is significantly correlated with the occurrence of the delay event before it occurs. For example, the time interval between specific events is abnormally extended, or certain key events are delayed. The identified value ranges of these strongly correlated features can be defined as the preceding feature intervals of the tolerable delay events. Furthermore, a risk prediction model constructed based on these preceding feature intervals can identify whether the current project is entering a risky range.

[0027] Furthermore, the target grid-connected projects are identified, and their progress time-series feature vectors are input into the tolerance-based delay risk prediction model to calculate the probability of delay risk and generate a delay risk prediction list. For example, when project A is underway, its latest progress time-series feature vector can be extracted in real time and input into the constructed tolerance-based delay risk prediction model. The model will calculate the probability of a delay event occurring in project A based on the degree of matching between project A's current progress characteristics and historical risk patterns. If project A has a high risk probability, it will be included in the delay risk prediction list and may be marked as a high-risk project.

[0028] Furthermore, multi-dimensional indicator dashboard data is generated based on the grid connection review dataset and the overdue risk prediction list. A multi-level index relationship is established for this dashboard data to support drill-down positioning of the indicator data. For example, an interactive dashboard can be generated, with its top layer displaying an overall risk overview of all grid connection projects, such as the number of high-risk projects and the average review period. When a manager clicks on the "Number of High-Risk Projects" indicator, the dashboard drills down to display a list of all high-risk projects. At this point, the manager can click on a specific project, such as Project A, and the dashboard will further drill down to display detailed review progress, work order status, and corresponding risk probabilities for Project A, enabling managers to quickly locate and understand the risks.

[0029] Finally, a correlation mapping is established between the risk items in the overdue risk prediction list and the corresponding progress time sequence feature vectors of the grid-connected projects, generating multi-dimensional analysis results for the grid connection review process. For example, in the aforementioned dashboard, when managers view the detailed information of project A, the risk items of project A in the overdue risk prediction list are correlated with the progress time sequence feature vector of project A. Through this correlation, managers can not only know that project A has risks, but also intuitively understand the specific progress anomalies that cause these risks. For example, the creation time of a tolerance work order X is too early, but the closing time is delayed, or the time taken for a certain key review stage is far beyond the average level. Based on this, multi-dimensional analysis results including risk warning, progress analysis, and problem localization are generated.

[0030] Based on the above examples, in terms of data integration, existing technologies often face the problem of unifying multi-source heterogeneous data. This often relies on manual comparison or simple ID matching, which is inefficient and prone to errors. The solution in this embodiment, however, obtains and integrates grid connection data records from the grid connection process platform, ITSCADA system, and architecture management system to generate a grid connection review dataset with grid connection projects as the basic unit. This effectively solves the problems of data dispersion and inconsistent identification. Regarding the prediction of overdue tolerance risks, existing technologies mainly rely on post-event response strategies and cannot predict risks in advance. The solution in this embodiment, however, performs time-series alignment processing on the grid connection review dataset to construct the progress time-series feature vectors of each grid connection project. Based on the time-series feature distribution corresponding to overdue tolerance events in historical data, it performs statistical fitting to extract the preceding feature intervals, thereby constructing a overdue tolerance risk prediction model. At this point, the progress time-series feature vector of the target grid connection project can be input into the model to calculate its overdue tolerance risk probability and generate an overdue risk prediction list, realizing a shift from passive response to proactive early warning and improving the management efficiency of the review process. In terms of visualization and analysis, existing review dashboard systems typically present data in the form of static reports or single-dimensional charts, lacking multi-level index association and drill-down positioning capabilities. The solution in this embodiment, however, generates multi-dimensional indicator dashboard data based on the grid connection review dataset and the overdue risk prediction list, and establishes multi-level index associations for this data to support layer-by-layer drill-down positioning. Furthermore, it establishes a correlation mapping between risk items in the overdue risk prediction list and the corresponding grid connection project's progress time-series feature vectors, generating multi-dimensional analysis results of the grid connection review process. This enables managers to quickly locate high-risk projects and their specific abnormal nodes, thereby achieving in-depth analysis of the review process.

[0031] In summary, the technical solution of this embodiment provides a comprehensive and efficient grid connection review management method through the above process, which can effectively improve the level of refined management and risk prevention and control capabilities of grid connection review.

[0032] In one embodiment, such as Figure 2 As shown, step S10 includes: S11: Extract the project identifier field and project attribute field from the grid connection process platform, ITSCADA system and architecture control system respectively; In this embodiment, the project identifier field refers to key information used to uniquely identify a grid-connected project, such as project number and project name. The project attribute field refers to auxiliary information describing the characteristics of the grid-connected project, such as project type, region, construction unit, and equipment model. By extracting the project identifier field and project attribute field separately, data support can be provided for subsequent data cleaning and correlation. For example, data extraction tools or customized scripts can be used to access the databases or API interfaces of various systems to obtain the required fields. Another approach is to configure a data acquisition agent to periodically obtain raw data streams from the data interfaces of various systems and parse the project identifier field and project attribute field from them.

[0033] S12: Identify grid connection data records with missing or inconsistent project identification fields, and generate corresponding project fingerprint keys based on project attribute fields; In this embodiment, the absence or inconsistency of the project identifier field may lead to the same grid-connected project being identified as different records in different systems, or different projects being incorrectly associated. A project fingerprint key is a unique or near-unique identifier generated by combining or hashing project attribute fields. For example, multiple highly distinctive project attribute fields (such as project name, construction location, equipment serial number, etc.) can be selected, concatenated into strings, and then hashed to generate a fixed-length project fingerprint key. Another approach is to use machine learning algorithms to extract features and reduce the dimensionality of project attribute fields, generating a vector that represents the project's characteristics, and using this vector as the project fingerprint key.

[0034] S13: Calculate the attribution confidence between grid-connected data records based on the project fingerprint key, and associate grid-connected data records whose attribution confidence meets the preset conditions with the same grid-connected project to generate a grid-connected review dataset with a unified project primary key.

[0035] In this embodiment, the attribution confidence score measures the probability or similarity that two or more grid-connected data records belong to the same grid-connected project. For example, the similarity of the project fingerprint keys of different data records can be compared, such as Jaccard similarity, cosine similarity, etc. The higher the similarity, the higher the attribution confidence score. When the attribution confidence score meets a preset condition, such as reaching a preset threshold, the grid-connected data record corresponding to the attribution confidence score can be considered to belong to the same grid-connected project and assigned a unified project primary key. Another approach is to use a clustering algorithm to cluster grid-connected data records with similar project fingerprint keys together, with each cluster representing a grid-connected project and each cluster assigned a unique project primary key.

[0036] For example, as a specific implementation method, to generate a grid connection review dataset with grid connection projects as the basic unit, data integration can be performed as follows: Extract fields such as project number, project name, and applicant unit from the grid connection process platform as project identification fields and project attribute fields; extract fields such as power plant name, equipment serial number, and grid connection time from the ITSCADA system; and extract fields such as architecture application ID, asset instance name, and region from the architecture management system. During the extraction process, a pre-configured data integration platform can be used to configure data extraction tasks. Subsequently, it is possible to identify grid connection data records where, for example, a project record in the grid connection process platform lacks an equipment serial number from the ITSCADA system, or where the power plant name in the ITSCADA system differs slightly from the project name in the grid connection process platform. For these identified abnormal grid connection data records, a project fingerprint key can be generated by concatenating strings based on project attribute fields, such as "project name," "applicant unit," and "grid connection time," and performing an MD5 hash operation on the concatenated result. Furthermore, the project fingerprint keys of all grid-connected data records are compared. If two records have identical fingerprint keys, or if the similarity calculated by a fuzzy matching algorithm (such as Levenshtein distance) exceeds a preset threshold, such as 0.95, they can be considered to belong to the same grid-connected project. For example, if the record "Project A of a certain photovoltaic power station" in the grid connection process platform and the record "Project A of a certain photovoltaic power station" in the ITSCADA system are determined to have an attribution confidence score higher than 0.95 after fingerprint key calculation and similarity comparison, these two records can be associated with the same logical grid-connected project and assigned a unified project primary key, such as "Project_2023001". Through the above method, a grid connection review dataset with a clear structure and consistent data can be generated, in which each grid-connected project has a unified primary key and integrates data from different systems.

[0037] Through the above technical solution, this application can effectively solve the data integration problem caused by the missing or inconsistent project identifier fields in multi-source heterogeneous environments of grid-connected data records. Specifically, by introducing a project fingerprint key and combining it with attribution confidence calculation, intelligent matching and association of scattered data can be achieved, improving the accuracy and automation level of data integration. This ensures that the generated grid-connected review dataset is based on a unified project primary key, guaranteeing data integrity and consistency. Compared to simple data acquisition and integration, the solution in this embodiment, through the aforementioned data cleaning and association mechanisms, has the effect of improving the accuracy and efficiency of grid-connected review data analysis.

[0038] In one embodiment, such as Figure 3 As shown, after step S13, the following steps are also included: S131: Obtain the architecture application information corresponding to each grid-connected project from the architecture management and control system, and obtain the corresponding asset instance information and operation monitoring information respectively; In this embodiment, the architecture management system is typically used to centrally manage and maintain various architectural information within an enterprise or organization, including IT architecture, business architecture, and application architecture. Its function is to provide a standardized architectural view and record the architectural specifications and declarations that projects should follow during the planning, design, and implementation phases. The architecture management system can be a dedicated configuration management database or a platform integrating architecture design tools and approval processes. Architecture declaration information refers to declarative documents or data submitted to the architecture management department by the grid-connected project at its initiation or a specific stage, regarding its system architecture, component composition, technology selection, interface specifications, etc. This information reflects the expected state of the project during the design phase, and may include, for example, the number of servers planned for deployment, database types, network topology, and security policies. Asset instance information refers to the specific details of the physical or virtual assets actually deployed and running in the grid-connected project. This information typically originates from ITSCADA systems, asset management systems, or automated deployment tools, reflecting detailed data on the actual hardware configuration, software versions, network device connections, storage allocation, and other aspects of the actual operating environment. Operational monitoring information refers to various performance indicators, log data, alarm events, etc. generated during the operation of a grid-connected project. It usually comes from the monitoring system and may include, for example, server CPU utilization, memory utilization, network traffic, service response time, error logs, security event records, etc., reflecting the actual operating status and health status of the project.

[0039] S132: Perform a consistency comparison of the architecture declaration information, asset instance information and operation monitoring information to generate difference information, which includes missing differences, redundancy differences, attribution differences and connectivity differences. In this embodiment, consistency comparison refers to cross-validating and comparing data from different sources and of different types using preset rules, algorithms, or models to identify any mismatches, conflicts, or omissions. The comparison process can be implemented using various technologies, such as field value matching based on rule engines, connectivity analysis based on topology graphs, or content comparison based on semantic analysis. Difference information refers to inconsistencies discovered during the consistency comparison process. Missing differences refer to situations where an entity or state is declared in the architecture declaration information but cannot be found in the asset instance information or operational monitoring information. For example, a critical service may be declared, but no instance of that service is found in actual deployment. Redundant differences refer to entities or states that exist in the asset instance information or operational monitoring information but are not declared or authorized in the architecture declaration information. For example, additional servers may be deployed but not reflected in the architecture declaration. Attribution differences refer to situations where the actual attribution of an asset instance or operational status does not match the attribution declared in the architecture declaration information. For example, a database instance may belong to project A in the declaration but be marked as belonging to project B in the actual configuration. Connectivity discrepancies refer to discrepancies between the connectivity or dependencies declared in the architecture declaration information and the connectivity or dependencies reflected in the actual operational monitoring information. For example, the declaration states that two systems should be able to access each other, but monitoring data shows that their network connection is faulty. Architecture consistency discrepancy characteristics refer to the structured, quantified, or categorized discrepancy information generated above, which serves as a key indicator describing the compliance and operational status of the grid-connected project architecture. This information can be used for subsequent risk assessment, compliance audits, and decision support.

[0040] S133: Store the difference information as an architectural consistency difference feature in the grid connection review dataset of the corresponding grid connection project.

[0041] In this embodiment, storing the grid connection review dataset for the corresponding grid connection project means associating the architectural consistency difference characteristics with the previously integrated grid connection project data and persistently storing it as part of the project review dataset. This ensures that the grid connection review dataset not only contains the project's basic information and progress information, but also its compliance and consistency assessment results at the architectural and operational levels.

[0042] For example, as a specific implementation, suppose there is a grid connection project called "Smart Grid Dispatch System Upgrade". After completing the integration of grid connection data records and generating a preliminary grid connection review dataset, the following steps can be performed to further verify the compliance of its architecture and operation: First, obtain the architecture declaration information corresponding to the project from the architecture management system. For example, the declaration document clearly states that the project needs to deploy two highly available servers (ServerA, ServerB), an Oracle database instance, and a dedicated data exchange module, and ServerA and ServerB should communicate with DEM1 through a specific network link. Simultaneously, obtain the actual deployed asset instance information from the ITSCADA system. For example, ServerA, ServerB, ServerC (an undeclared server), DB1, and DEM1 are actually deployed. Furthermore, obtain operation monitoring information from the operation monitoring system. For example, monitoring data shows that the network connection between ServerA and DEM1 is intermittently interrupted, and the actual attribution label of DB1 is "test environment" rather than "production environment". Furthermore, a consistency comparison is performed on the information obtained above. The comparison results may generate the following discrepancies: ServerC is found, which is not mentioned in the architecture declaration information, constituting a redundancy discrepancy; DB1 is actually assigned to the "test environment," which is inconsistent with the declared "production environment," constituting an attribution discrepancy; the network connection between ServerA and DEM1 has intermittent interruptions, which does not meet the declared stable connectivity requirements, constituting a connectivity discrepancy. Finally, the specific discrepancy information obtained, such as "ServerC redundancy," "DB1 attribution error: test environment," and "ServerA and DEM1 connectivity anomaly," will be stored as architecture consistency discrepancy features in the grid connection review dataset of the "Smart Grid Dispatch System Upgrade" project.

[0043] Through the aforementioned technical solution, based on the integrated grid-connected data and confirmed project ownership, a mechanism is further introduced to compare the consistency of architecture declaration information, asset instance information, and operational monitoring information. This generates difference information including missing differences, redundancy differences, ownership differences, and connectivity differences as architecture consistency difference features. Finally, the generated architecture consistency difference features are stored in the corresponding grid-connected project's grid-connected review dataset, thereby increasing the dimensionality and depth of the grid-connected review dataset. This allows the grid-connected review to go beyond project progress and basic information, delving into the project's architecture compliance, the accuracy of asset deployment, and the consistency of actual operational status. By identifying and quantifying these differences, this application can effectively discover potential architecture non-compliance, resource waste, configuration errors, or operational risks in grid-connected projects. This provides reviewers with a more comprehensive basis for decision-making, not only improving the accuracy and reliability of grid-connected reviews but also helping to identify and correct problems in a timely manner before project launch, reducing potential risks caused by architecture inconsistencies or operational anomalies, and ultimately improving the quality and stability of grid-connected projects.

[0044] In one embodiment, step S20 includes: S21: Extract the planned time nodes, review and submission time nodes, and defective work order creation time nodes of each grid-connected project, align the time sequence with the planned time nodes, and generate a timestamp sequence for each grid-connected project. In this embodiment, the planned time node, the review and submission time node, and the work order creation time node are key milestones or event points in the grid connection project lifecycle, used to define the project's timeline. Specifically, the planned time node can be obtained by parsing the project plan table or task flow log in the grid connection process platform; the review and submission time node can be obtained from the submission records in the review system; and the work order creation time node can be obtained from the work order creation records in the work order management system. Furthermore, these key time node data can be periodically synchronized by interfaceing with the project management system, document management system, or manual data entry system. Time sequence alignment based on the planned time node can be achieved by normalizing the first planned time node of all projects (e.g., the project start date or the planned start date of the first review phase) to a zero point on the timeline, with other time nodes offset relative to this zero point; or, a planned time node with universal significance for all projects (e.g., the grid connection application submission date or the grid connection permit issuance date) can be selected as the alignment benchmark, adjusting the time series of all projects to a relative time referenced to this benchmark. The timestamp sequence for each grid-connected project can be generated by arranging the extracted planned time nodes, review submission time nodes, defective work order creation time nodes, and other important actual occurrence time nodes (such as review completion time, rectification completion time, etc.) in chronological order; or, based on the above key time nodes, sampling can be performed at fixed time intervals to generate a uniformly distributed timestamp sequence, and the project status can be recorded at each sampling point.

[0045] S22: Based on the timestamp sequence, calculate the progress deviation and node completion rate of each grid-connected project at all time nodes; In this embodiment, the schedule deviation value can be defined as the difference between the actual occurrence time and the planned occurrence time, for example, the actual submission date minus the planned submission date. The node completion rate can be defined as the proportion of completed nodes to the total planned nodes, or a weighted completion rate calculated based on the weight of each node. Alternatively, the schedule deviation value can also be calculated by comparing the actual duration with the planned duration, for example, dividing the difference between the actual duration and the planned duration by the planned duration. The node completion rate can be evaluated based on the completion status of preset milestones; for example, if a milestone is completed, its completion rate is 100%, otherwise it is 0%, and then a weighted average is calculated.

[0046] S23: Construct the progress time sequence feature vector of each grid-connected project based on timestamp sequence, progress deviation value and node completion rate.

[0047] In this embodiment, the progress time-series feature vector of each grid-connected project can be constructed by directly concatenating the timestamp sequence, the progress deviation value corresponding to each time point, and the node completion rate as feature dimensions into a high-dimensional vector; or, the original features can be further statistically analyzed or transformed, such as calculating the average, maximum, minimum, and rate of change of the progress deviation value over a period of time, as well as the trend of the node completion rate, and combining the statistical features into a progress time-series feature vector.

[0048] Specifically, the proposed solution extracts key time nodes from the grid-connected project lifecycle and aligns them with the planned time nodes as a unified benchmark to ensure comparability across different projects in terms of time. Based on this, it quantifies the difference between the actual progress and the plan, as well as the degree of completion at each stage, by calculating the progress deviation and completion rate at each time node. Finally, the above information is integrated to construct a progress time-series feature vector, providing standardized input for the tolerable delay risk prediction model. This method transforms raw project data into a unified format, accurately reflecting project dynamics and thus improving the accuracy and reliability of the tolerable delay risk prediction model.

[0049] Through the above technical solution, this application can accurately define key time nodes and align the timelines with the planned time nodes as a consistent benchmark, thereby effectively solving the data comparison problem caused by inconsistencies in the timelines of different projects. Simultaneously, by quantitatively calculating progress deviation values ​​and node completion rates, it can provide intuitive and objective project health indicators. The resulting processed and integrated time-series feature vector can provide high-quality input for the tolerance and delay risk prediction model, improving the accuracy and reliability of risk prediction, thereby enhancing the intelligence level and risk management capabilities of grid connection review.

[0050] In one embodiment, step S30 includes: S31: Select a set of positive samples from historical grid-connected projects that have experienced overdue tolerance events and a set of negative samples that have not experienced overdue tolerance events; In this embodiment, the positive sample set refers to project data from historical grid-connected projects that experienced overdue events, containing information about the causes and consequences of these events. The negative sample set refers to project data from historical grid-connected projects that did not experience overdue events, representing normal grid-connection processes. The screening process can be based on predefined business rules or event logs. For example, overdue events can be identified by checking the status of work orders or comparing completion times with planned times in the grid-connection process platform. Alternatively, expert experience or historical data analysis can be used to define triggering conditions for overdue events and classify historical projects accordingly.

[0051] S32: Extract the progress time-series feature vectors corresponding to the tolerance delay events of each grid-connected project in the positive sample set, and divide the progress time-series feature vectors into a pre-feature sequence and a post-feature sequence with the occurrence time of the tolerance delay event as the dividing point. In this embodiment, the progress time-series feature vector refers to the sequence data describing the progress status of a grid-connected project at different time nodes. During extraction, for each positive sample project experiencing a delay event, the progress time-series feature vector from project initiation to the event occurrence is obtained. This vector constitutes the preceding feature sequence, while data after the event occurrence constitutes the following feature sequence. The following feature sequence is typically used to verify or analyze the impact of the event. The dividing point can be precise to the specific timestamp of the event occurrence, or it can be the most recent fixed time interval before the event occurrence.

[0052] S33: Perform a sliding window scan on the preceding feature sequence, calculate the variance of the progress deviation value, the rate of decrease of the node completion rate, and the rate of change of the frequency of creating defective work orders within each time window, and construct the preceding feature distribution. In this embodiment, sliding window scanning refers to a time-series data analysis technique that performs statistical calculations on the data within a fixed-size window by moving it along the sequence. The variance of the schedule deviation value reflects the severity of project schedule fluctuations; the rate of decline in the node completion rate indicates whether project progress has stalled or regressed; and the rate of change in the frequency of defective work orders reflects changes in the pressure on the project in handling anomalies or defects. The calculation results of the variance, rate of decline, and rate of change together constitute the pre-feature distribution, used to characterize the dynamic evolution of the project state before an event occurs. The size and step size of the sliding window can be set based on historical data analysis or expert experience combined with the actual scenario; for example, it can be set as a certain percentage of the average project review cycle, which is clear to those skilled in the art.

[0053] S34: Perform kernel density estimation on the pre-feature distribution, and extract the feature value range where the distribution density exceeds the preset density threshold as the pre-feature range of the tolerance overdue event in the time series feature vector; In this embodiment, kernel density estimation refers to a non-parametric density estimation method used to estimate the probability density function of a random variable. It can smoothly depict the distribution of feature values. By performing kernel density estimation on the distribution of prior features, the probability density of different combinations of feature values ​​can be obtained. When the density of a certain feature value interval exceeds a preset density threshold, it indicates that the feature pattern within that interval has a high frequency or significance before the occurrence of the tolerance period expiration event, and therefore can be extracted as a prior feature interval. The preset density threshold can be determined based on historical data analysis, statistical methods, or business experience, combined with the actual scenario, which is clear to those skilled in the art.

[0054] S35: For each grid-connected project in the negative sample set, backtrack for a time window of the same length as the preceding feature sequence of the positive sample set, and extract the negative sample feature sequence; In this embodiment, since no tolerance overdue event occurred in the negative sample, the time of the event cannot be used as the dividing point. To maintain the consistency of the feature sequence length, a time window of equal length to the preceding feature sequence of the positive sample is extracted by tracing back from the grid connection completion time of the negative sample project. This window serves as the negative example feature sequence. Furthermore, by using the grid connection completion time as the anchor point for tracing back, it can be ensured that the negative example feature sequence reflects the state before the project was successfully completed, and thus has a certain degree of comparability with the preceding feature sequence of the positive sample in the time dimension.

[0055] S36: Construct a prediction model for the risk of exceeding the tolerance period based on the comparison and fitting of the preceding feature sequence of the positive sample set and the negative example feature sequence of the negative sample set.

[0056] In this embodiment, by comparing and analyzing the preceding feature sequences of positive samples and the negative example feature sequences of negative samples, the constructed tolerance delay risk prediction model can learn to distinguish the feature patterns that lead to tolerance delay events from the feature patterns of normal project progress. The comparison and fitting can employ various machine learning algorithms, such as support vector machines, logistic regression, decision trees, random forests, or neural networks. The goal of constructing the tolerance delay risk prediction model is to establish a mapping relationship that converts the input progress time-series feature vector into tolerance delay risk probabilities.

[0057] Furthermore, the tolerance for delay risk prediction model takes the progress time series feature vector as input, the matching degree of the preceding feature interval as the intermediate feature, and the probability of tolerance for delay risk as output. Among them, the matching degree of the preceding feature interval, as the intermediate feature, represents the similarity between the input vector and the identified preceding feature interval, which helps to improve the interpretability and prediction accuracy of the tolerance for delay risk prediction model.

[0058] Furthermore, the following provides an example of the construction and application of a tolerance for overdue risk prediction model.

[0059] Step 1: Sample preparation, including: 12 historical overdue items as positive samples and 24 normal items as negative samples.

[0060] Step 2, Feature Extraction, including: taking the positive sample "XX Photovoltaic Phase I" as an example: extract the progress features for the 90 days before the deadline, including the progress deviation variance (0.4), the decrease in node completion rate (-0.05 / day), and the growth rate of incomplete work orders (0.15 orders / day).

[0061] Step 3, model construction, includes: estimating kernel density for positive sample features, extracting high-risk intervals, specifically progress variance > 0.35, completion rate decrease < -0.04 / day, and work order growth rate > 0.02 / day; then, using the GBDT algorithm to fit the positive and negative sample features and train a binary classification model.

[0062] Step 4, Risk Prediction, includes: inputting the current 90-day characteristics of the target project "XX Photovoltaic Phase II"; outputting a risk probability of 0.87, which is greater than the threshold of 0.8, therefore it is judged as high risk.

[0063] Step 5: Output application, including: adding to the overdue risk list, warning level "red"; recommended measures: assign more data specialists and complete the missing documents within 3 days.

[0064] For example, as a specific implementation method, when constructing a tolerance delay risk prediction model, all projects that have experienced tolerance delay events can be selected from historical grid-connected projects over the past three years as a positive sample set. For instance, this can be done by querying projects in the grid connection process platform whose "Tolerance Work Order Status" field is "Overdue" or "Closed and Overdue". Simultaneously, all projects that have successfully completed grid connection without any tolerance delay events can be selected as a negative sample set. For each positive sample project, its progress time-series feature vector from project initiation to the time of tolerance work order delay can be extracted and divided into preceding feature sequences. For example, if a project delays on day 90, its progress time-series feature vector for the preceding 90 days is the preceding feature sequence. Furthermore, a sliding window scan can be performed on the preceding feature sequences, for example, setting the window size to 7 days and the step size to 1 day. Within each 7-day window, the variance of the progress deviation is calculated, and the square of the standard deviation can be used to quantify volatility. The rate of decline in node completion rate is calculated by calculating the slope of the linear regression of node completion rate within the window. The rate of change in the frequency of defective work order creation is calculated; for example, the number of newly created defective work orders within the window can be compared with the previous window. Furthermore, the above calculation results together constitute the pre-feature distribution. Subsequently, a Gaussian kernel function can be used to estimate the kernel density, and a density threshold, such as 0.05, can be set to identify the feature value intervals that frequently occur before the defective work order overdue event occurs. For negative sample projects, the grid connection completion time can be used as the anchor point to trace back to a time window with the same average length as the pre-feature sequence of positive samples, and the negative sample feature sequence can be extracted. Finally, a gradient boosting decision tree algorithm can be used, with the feature sequences of positive and negative samples as training data and the matching degree of the pre-feature interval as the intermediate layer feature, to train a defective work order overdue risk prediction model. When the progress time-series feature vector of a new target grid-connected project is input into this tolerance delay risk prediction model, it can calculate the probability of the project experiencing a tolerance delay event.

[0065] Through the aforementioned technical solution, this application can identify key pre-existing characteristic patterns leading to delayed events from complex grid connection review data. Furthermore, by processing historical data and building models, it can effectively capture abnormal changes in dynamic indicators such as project progress, milestone completion status, and anomaly handling frequency, quantifying these as predictable risk signals. This allows grid connection review managers to receive early warnings of predicted risks before delayed events actually occur, providing sufficient time to take intervention measures, avoid or mitigate delayed events, improve the efficiency and quality of grid connection reviews, and reduce project delay risks.

[0066] In one embodiment, step S33 includes: S331: Based on the review stage corresponding to the historical grid-connected projects, the pre-requisite feature sequence is divided into phased sub-sequences, which include the document review stage sub-sequence, the on-site verification stage sub-sequence, and the rectification tracking stage sub-sequence. In this embodiment, based on the review stages corresponding to historical grid-connected projects, the pre-construction feature sequence is divided into phased sub-sequences. This aims to identify the inherent phased characteristics in the grid-connected review process and decompose the continuous pre-construction feature sequence into more homogeneous phased sub-sequences, which facilitates corresponding analysis based on the characteristics of different stages. The division of the pre-construction feature sequence can be achieved through predefined business rules or configuration tables. For example, stage boundaries can be determined based on the average duration of each review stage or key node events in historical data; alternatively, machine learning algorithms can be used to perform cluster analysis on the progress data of historical grid-connected projects to identify review stages with similar characteristics.

[0067] S332: Set corresponding baseline window sizes for each staged subsequence, and perform multi-scale sliding window scanning on the staged subsequence with a preset ratio of the staged subsequence length as the step size to generate a window feature set; In this embodiment, multi-scale analysis can be achieved by setting a baseline window size and a preset step size to capture the characteristics of each stage. For example, a larger baseline window size can be set for stages with relatively flat data changes, while a smaller baseline window size can be set for stages with frequent data changes or a high density of key events. The step size can be set as a fixed percentage of the length of the subsequence in each stage, such as 5% or 10%, to ensure effective multi-scale scanning on subsequences of different lengths. Multi-scale sliding window scanning can be achieved by performing multiple scans using window sizes that are multiples of the baseline window size, or by adjusting the step size.

[0068] S333: Calculate the cross-scale mean of the variance of the progress deviation value, the rate of decline of the node completion rate, and the rate of change of the frequency of defective work orders in the window feature set, respectively, to form the cross-scale mean feature vector of the phased subsequence; In this embodiment, by calculating the cross-scale mean, noise or local fluctuations that may be introduced by a single-scale scan can be smoothed out, resulting in more stable stage features. For example, for the window feature set obtained after scanning windows at different scales for each stage subsequence, the arithmetic mean or weighted average of the variance of the progress deviation value, the rate of decline of the node completion rate, and the rate of change of the frequency of defective work orders can be calculated respectively. These values ​​are then combined to form a vector, namely the cross-scale mean feature vector of the stage subsequence, which can comprehensively reflect the characteristic performance of the stage at multiple scales.

[0069] S334: Extract overlapping feature windows at the boundaries of adjacent segmented sub-sequences, calculate the cosine similarity of the cross-scale mean feature vectors on both sides of the boundary, and mark the corresponding boundary as a stage jump point if the cosine similarity is lower than the preset similarity threshold. In this embodiment, by comparing the feature similarity at the boundary, it is possible to detect whether the stage transition is smooth and whether there are potential risk signals. For example, at the connection between two adjacent staged sub-sequences, an overlapping feature window can be defined, spanning the end and beginning of the two sub-sequences, for example, taking several data points from the end and beginning of each adjacent sub-sequence. Then, the cross-scale mean feature vectors of the overlapping feature window on both sub-sequence sides are calculated respectively, and the cosine similarity is used to quantify the similarity between the two vectors. If the similarity is lower than a preset similarity threshold, it can be considered that there is a significant feature change at the boundary, and it is marked as a stage transition point.

[0070] S335: The cross-scale mean feature vectors of the phased subsequences are concatenated sequentially according to the review stages, and the corresponding stage transition point markers are associated with the concatenated result to form the pre-feature distribution.

[0071] In this embodiment, the pre-feature distribution not only includes the internal features of each stage, but also highlights the key change points between each stage. For example, the cross-scale mean feature vectors of each sub-sequence can be concatenated sequentially according to the natural order of the grid connection review process. In the concatenated sequence, a marker bit, such as a Boolean value or a specific numerical value, can be inserted or associated at the position corresponding to the previously identified stage transition point to indicate that the position is a transition point.

[0072] For example, as a specific implementation, suppose a historical grid-connected project's preliminary characteristic sequence covers 30 days of data, and it is known that the project went through a 10-day document review phase, a 10-day on-site verification phase, and a 10-day rectification follow-up phase. First, this 30-day preliminary characteristic sequence can be divided into three sub-sequences: the document review phase sub-sequence (days 1-10), the on-site verification phase sub-sequence (days 11-20), and the rectification follow-up phase sub-sequence (days 21-30). For the document review phase sub-sequence, a baseline window size of 3 days can be set, and multi-scale sliding window scanning can be performed with a step size of 10% of its length. For example, in addition to the 3-day window, 1.5-day and 4.5-day windows can be scanned simultaneously. For the on-site verification phase sub-sequence, considering that it may require more detailed observation, a baseline window size of 2 days and a step size of 0.5 days can be set. The rectification follow-up phase sub-sequence can be set with a 4-day baseline window and a step size of 1 day. After completing the multi-scale scanning of each stage, the cross-scale mean of the variance of the progress deviation value, the rate of decline of the node completion rate, and the rate of change of the frequency of defective work order creation are calculated in all window feature sets within each stage, thus obtaining the cross-scale mean feature vector of the three sub-sequences. For example, the feature vector of the document review stage may be [0.4, -0.05, 0.01], the on-site verification stage may be [0.8, -0.15, 0.03], and the rectification tracking stage may be [0.2, -0.02, 0.005]. Furthermore, at the boundary between the document review stage sub-sequence and the on-site verification stage sub-sequence (i.e., between day 10 and day 11), an overlapping feature window is extracted, for example, containing data from day 9 to day 12. The cross-scale mean feature vector of this window is calculated on the document review stage side (days 9-10) and the on-site verification stage side (days 11-12), and the cosine similarity is calculated. If the cosine similarity is lower than a preset similarity threshold, the boundary is marked as a stage transition point. Similarly, the boundary between the on-site verification stage subsequence and the rectification tracking stage subsequence is processed. Finally, the cross-scale mean feature vectors of these three subsequences are concatenated according to the order of the review stages, and the identified stage transition point markers are associated with the concatenation result, thus forming the pre-feature distribution of this historical grid-connected project.

[0073] Through the above technical solution, this application can more accurately capture the dynamic characteristics of grid-connected projects at different review stages and the abnormal changes between stages, enabling the constructed pre-feature distribution to more comprehensively reflect potential risk signals in project progress, especially at critical moments during stage transitions. Therefore, when the pre-feature distribution generated in the above manner is input into the tolerance delay risk prediction model, it can more accurately identify target grid-connected projects with tolerance delay risks, thereby improving the accuracy and reliability of risk prediction.

[0074] In one embodiment, step S50 includes: S51: Extract project-dimensional indicators, review-stage-dimensional indicators, and tolerance-based work order-dimensional indicators from the grid connection review dataset, and extract the tolerance-based overdue risk probability of each target grid connection project from the overdue risk prediction list. Add the tolerance-based overdue risk probability as a risk weight to the corresponding project-dimensional indicators to form a multi-dimensional indicator dashboard data. In this embodiment, project-level indicators may include information such as project name, project number, region, and project type; review stage-level indicators may include important stages in the grid connection process, such as document review, on-site verification, and rectification tracking; and tolerance-based work order-level indicators may include detailed information such as the specific tolerance-based work order number, work order type, creation time, and processing status. These indicators can be extracted from relational databases using structured query language, or from unstructured or semi-structured data sources using data parsing tools. For example, field mapping rules for each dimension indicator can be predefined, and data extraction tools can automatically identify and extract the corresponding data fields from the grid connection review dataset; alternatively, a metadata-based management approach can be used to define the extraction logic for different dimension indicators through configuration, enabling flexible indicator extraction. The probability of tolerance-based overdue risk refers to a quantitative indicator that measures the likelihood of a tolerance-based overdue event occurring at a future point in time for the grid connection project. Its specific value is calculated by the tolerance-based overdue risk prediction model and recorded in the overdue risk prediction list. Extracting the probability of overdue tolerance risk typically involves querying a list of overdue risk predictions. The risk probability value is retrieved based on the unique identifier (e.g., project ID) of the target grid-connected project. For example, the probability of overdue tolerance risk can be directly obtained from a database table or file storing risk prediction results, using the project ID as the query condition. Alternatively, if the list is provided via an API, the corresponding risk probability data can be obtained by calling the API and passing in the project identifier. Risk weights are numerical values ​​used to quantify the risk level of various indicators, directly derived from the probability of overdue tolerance risk. Attaching risk probabilities to project-level indicators means that when constructing a multi-dimensional indicator dashboard, each project not only includes its basic attribute information but also an additional numerical value reflecting its overall risk level. This can be achieved by adding a risk weight field to the data structure of the project-level indicators and filling this field with the extracted probability of overdue tolerance risk. For example, in the generated multidimensional indicator dashboard data, an attribute named "risk_score" can be added to each project record, and the calculated probability of tolerance for overdue risk can be assigned to the "risk_score" field; or, the risk probability can be logically bound to the project dimension indicators and associated in the data display layer or index building layer, so that when querying project indicators, their risk weights can also be obtained together.

[0075] S52: Construct a three-level index tree based on multi-dimensional indicator dashboard data, sort the nodes of the same level in the three-level index tree in descending order of risk weight, generate a risk perception sorting index, and associate risk warning tags with nodes whose risk weight exceeds the preset risk threshold. In this embodiment, the construction process of the three-level index tree can be implemented in a top-down manner. First, a root node is created, then child nodes are created for each project node, and finally child nodes are created for each review stage node. For example, a tree database (such as Neo4j) or a graph database can be used to directly construct this hierarchical structure, using each dimension indicator as a node and establishing parent-child relationships; or, multiple tables can be created in a relational database and foreign key relationships can be defined to simulate the logical structure of the three-level index tree, and the tree can be traversed and manipulated by application logic. The risk-aware sorting index is used to prioritize the display of nodes with higher risks at the same level, thereby improving the efficiency of users in identifying high-risk items. Descending order means that nodes with higher risk weights are placed earlier in the list or display. Descending order can be achieved by sorting the list of child nodes of each node when constructing the index tree. For example, a list of child nodes can be maintained under each parent node, and sorted in descending order using algorithms such as quicksort and mergesort based on the risk weight values ​​stored in the child nodes. Alternatively, during database queries, the ORDER BY clause can be used to specify descending sorting by the risk weight field, thus achieving risk-aware sorting in the result set. Risk warning tags are visual or logical identifiers used to highlight nodes whose risk levels reach or exceed a specific warning threshold, alerting users to their concerns. The preset risk threshold is a configurable value, indicating when a risk probability reaches that threshold and a warning should be issued. Association tags can be implemented by adding a Boolean field to the node's data structure. When the node's risk weight exceeds the threshold, this field is set to true. For example, when building an index tree or updating node data, the risk weight of each node can be judged; if it exceeds the threshold, a warning label or status field can be added to the node data. Alternatively, different colors, icons, or text prompts can be dynamically rendered based on the node's risk weight during front-end display to visualize the risk warning tags.

[0076] Furthermore, the three-level index tree has a root node corresponding to project-level metrics, intermediate nodes corresponding to review stage-level metrics, and leaf nodes corresponding to incomplete work order-level metrics. Each node stores the aggregated value and risk weight of its corresponding metric. The root node represents the most macroscopic project level, intermediate nodes detail the various review stages of the project, and leaf nodes further detail specific incomplete work orders. Each node stores not only its own metric information but also the aggregated value for that metric (e.g., a project node can aggregate the average risk weight or number of incomplete work orders across all stages; a stage node can aggregate the average risk weight or number of pending work orders across all incomplete work orders) and its own risk weight. For example, a project root node can store aggregated information such as the project's total risk weight, total number of work orders, and number of completed work orders; a review stage intermediate node can store aggregated information such as the stage's risk weight, number of work orders, and average processing time; and an incomplete work order leaf node stores its own risk weight and detailed work order information.

[0077] S53: When responding to a drill-down positioning request, locate the corresponding node in the three-level index tree based on the target dimension identifier corresponding to the drill-down positioning request, and traverse the child nodes downwards along the node, returning the aggregated value of the dimension indicator and the risk warning mark corresponding to the child node as the drill-down positioning result.

[0078] In this embodiment, drill-down positioning is an interactive data exploration function that allows users to gradually delve into lower-level data from higher-level data. When a user initiates a drill-down positioning request, for example, by clicking on a project node, the corresponding node can be found in the three-level index tree based on the target dimension identifier contained in the drill-down positioning request. Then, the child nodes of that node can be traversed further to obtain the aggregated values ​​of the dimension indicators of these child nodes (e.g., all review stage nodes under the project) (such as the risk weight of each stage, the number of work orders) and whether they have risk warning tags. Finally, the obtained information can be organized and returned to the user as the result of drill-down positioning, so that the user can view a more detailed risk distribution and specific issues. For example, when a user clicks on a high-risk project, a list of all review stages under that project is returned, with each stage displaying its risk weight, aggregated data, and warning status, allowing the user to further select a high-risk stage for drill-down.

[0079] For example, as a specific implementation method, suppose a grid connection management department needs to monitor the risks of ongoing grid connection projects. First, it obtains grid connection data records from the grid connection process platform, ITSCADA system, and architecture control system, and integrates them to generate a grid connection review dataset. Then, using a tolerance-overdue risk prediction model, it calculates the tolerance-overdue risk probability for each target grid connection project and generates an overdue risk prediction list.

[0080] Based on this, project-level indicators such as "Project A" and "Project B," review stage-level indicators such as "Document Review Stage" and "On-site Verification Stage," and work order-level indicators such as "Work Order 1001" and "Work Order 1002" with incomplete documentation were extracted from the grid connection review dataset. Simultaneously, the probability of incomplete documentation overdue for "Project A" was extracted as 0.85 from the overdue risk prediction list, and the probability of incomplete documentation overdue for "Project B" was extracted as 0.30. Therefore, 0.85 can be added as a risk weight to the project-level indicators of "Project A," and 0.30 can be added to the project-level indicators of "Project B," thus constructing a multi-dimensional indicator dashboard with risk weights.

[0081] Furthermore, a three-level index tree is constructed based on the multi-dimensional indicator dashboard data. The root nodes of this three-level index tree can be "Project A" and "Project B," which respectively store their respective risk weights (0.85 and 0.30) and other aggregated information (such as the total number of work orders and the number of completed work orders). Under the root node of "Project A," there can be intermediate nodes such as "Document Review Stage" and "On-site Verification Stage." Each stage node also stores the risk weight of that stage (for example, the risk weight of the document review stage is 0.90) and aggregated values ​​(such as the number of pending work orders in that stage). Under the intermediate node of "Document Review Stage," there can be leaf nodes such as "Work Order 1001" and "Work Order 1002." Each work order leaf node stores its own risk weight (for example, the risk weight of work order 1001 is 0.95) and detailed work order information.

[0082] Furthermore, when constructing the index tree, nodes at the same level are sorted in descending order according to their risk weight. For example, at the root node level, "Project A" (risk weight 0.85) will be ranked before "Project B" (risk weight 0.30). At the child node level of "Project A", if the risk weight of "Document Review Stage" is 0.90 and the risk weight of "On-site Verification Stage" is 0.70, then "Document Review Stage" will be ranked before "On-site Verification Stage". Meanwhile, if the preset risk threshold is 0.80, then the risk weights of "Project A", "Document Review Stage", and "Work Order 1001" all exceed the threshold, and a risk warning will be set for each of them.

[0083] Finally, when a user clicks "Project A" on the dashboard interface to drill down, the system responds to this drill-down request by locating the "Project A" node in the three-level index tree and returning information on all its review stage child nodes. The returned results may display the aggregated value and risk warning flag for the "Document Review Stage," as well as the aggregated value and no warning flag for the "On-site Verification Stage." Users can quickly determine that the main risks of "Project A" are concentrated in the "Document Review Stage," and further drill down on the "Document Review Stage" to view the specific incomplete work orders, thereby efficiently locating and handling high-risk issues.

[0084] Through the aforementioned technical solution, this application effectively addresses the challenge of efficiently identifying and locating high-risk projects, high-risk review stages, and high-risk work orders with incomplete documentation in multi-dimensional dashboard analysis of grid connection review content. Specifically, by adding the probability of incomplete documentation exceeding the deadline as a risk weight to each dimension indicator and constructing a three-level index tree with risk perception capabilities, the risk quantification and hierarchical organization of grid connection review data can be achieved. Furthermore, the introduction of a risk perception sorting index allows for the priority presentation of projects, stages, and work orders with higher risks, improving the efficiency and accuracy of risk screening. Simultaneously, the association of risk warning markers provides users with intuitive risk alerts, enabling them to quickly focus on key risk points. Combined with the drill-down positioning function, users can delve from a macro-level project risk overview to specific review stages and details of incomplete documentation work orders, thereby achieving precise positioning, rapid response, and effective management of grid connection review risks and optimizing the risk control capabilities of grid connection review.

[0085] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0086] In one embodiment, a multi-dimensional dashboard analysis system for online review content is provided, which corresponds one-to-one with the multi-dimensional dashboard analysis method for online review content in the above embodiments. The multi-dimensional dashboard analysis system for online review content includes: The data integration module is used to obtain grid connection data records from the grid connection process platform, ITSCADA system and architecture control system, and integrate the grid connection data records to generate a grid connection review dataset with grid connection projects as the basic unit. The time series processing module is used to perform time series alignment processing on the grid connection review dataset, extract the timestamp sequence of each grid connection project, and construct the progress time series feature vector of each grid connection project. The model building module is used to statistically fit the time-series feature distribution corresponding to the defective and overdue events in historical grid-connected projects based on the progress time-series feature vector, extract the preceding feature interval of the defective and overdue events in the time-series feature vector, and build a defective and overdue risk prediction model. The risk prediction module is used to identify target grid-connected projects. The progress time-series feature vector of the target grid-connected project is input into the tolerance for overdue risk prediction model to calculate the tolerance for overdue risk probability of the target grid-connected project and generate an overdue risk prediction list. The index association module is used to generate multi-dimensional indicator dashboard data based on the grid connection review dataset and the overdue risk prediction list, and to establish multi-level index association relationships for the multi-dimensional indicator dashboard data to support the drilling down of indicator data layer by layer. The results generation module is used to establish a correlation mapping between the risk items in the overdue risk prediction list and the progress time sequence feature vectors of the corresponding grid-connected projects, and generate multi-dimensional analysis results of the grid connection review process.

[0087] For specific limitations regarding a multi-dimensional dashboard analysis system for content subject to online review, please refer to the limitations of a multi-dimensional dashboard analysis method for content subject to online review mentioned above, which will not be repeated here. Each module in the aforementioned multi-dimensional dashboard analysis system for content subject to online review can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0088] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a multi-dimensional dashboard analysis method for network-connected review content.

[0089] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements a multi-dimensional dashboard analysis method for network-connected review content.

[0090] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A multi-dimensional dashboard analysis method for content reviewed online, characterized in that, Including the following steps: Grid connection data records are obtained from the grid connection process platform, ITSCADA system and architecture control system, and integrated to generate a grid connection review dataset with grid connection projects as the basic unit. The grid connection review dataset is time-series aligned, and the timestamp sequences of each grid connection project are extracted to construct the progress time-series feature vector of each grid connection project. Based on the progress time-series feature vector, statistical fitting is performed on the time-series feature distribution corresponding to the tolerance delay events in historical grid-connected projects, the preceding feature interval of the tolerance delay events in the time-series feature vector is extracted, and a tolerance delay risk prediction model is constructed. Identify the target grid-connected project, input the progress time sequence feature vector of the target grid-connected project into the tolerance for overdue risk prediction model, calculate the tolerance for overdue risk probability of the target grid-connected project, and generate an overdue risk prediction list; Multidimensional indicator dashboard data is generated based on the grid connection review dataset and the overdue risk prediction list. Multi-level index relationships are established for the multidimensional indicator dashboard data to support drill-down positioning of indicator data. Establish a correlation mapping between the risk items in the overdue risk prediction list and the progress time sequence feature vectors of the corresponding grid-connected projects, and generate multi-dimensional analysis results of the grid connection review process.

2. The multi-dimensional dashboard analysis method for network-connected review content according to claim 1, characterized in that: The steps of acquiring grid connection data records from the grid connection process platform, ITSCADA system, and architecture control system, and integrating these records to generate a grid connection review dataset with grid connection projects as the basic unit, include: Extract the project identifier field and project attribute field from the grid connection process platform, ITSCADA system and architecture control system respectively; Identify grid connection data records with missing or inconsistent project identification fields, and generate corresponding project fingerprint keys based on project attribute fields; The attribution confidence level between grid-connected data records is calculated based on the project fingerprint key, and the grid-connected data records whose attribution confidence level meets the preset conditions are associated with the same grid-connected project to generate a grid-connected review dataset with a unified project primary key.

3. The multi-dimensional dashboard analysis method for network-connected review content according to claim 2, characterized in that: After the steps of calculating the attribution confidence level between grid-connected data records based on the project fingerprint key, and associating grid-connected data records with attribution confidence levels meeting preset conditions to the same grid-connected project to generate a grid-connected review dataset with a unified project primary key, the method further includes: Obtain the architecture application information corresponding to each grid-connected project from the architecture management and control system, and obtain the corresponding asset instance information and operation monitoring information respectively; The architecture declaration information, asset instance information and operation monitoring information are compared for consistency to generate difference information, which includes missing differences, redundancy differences, attribution differences and connectivity differences. The difference information is stored as an architectural consistency difference feature in the grid connection review dataset of the corresponding grid connection project.

4. The multi-dimensional dashboard analysis method for network-connected review content according to claim 1, characterized in that: The steps of performing time-series alignment processing on the grid connection review dataset, extracting the timestamp sequences of each grid connection project, and constructing the progress time-series feature vector of each grid connection project include: Extract the planned time nodes, review and submission time nodes, and work order creation time nodes for each grid-connected project. Use the planned time nodes as a benchmark to perform time sequence alignment and generate a timestamp sequence for each grid-connected project. Based on the timestamp sequence, calculate the progress deviation and node completion rate of each grid-connected project at all time nodes; Based on timestamp sequences, progress deviation values, and node completion rates, a progress time-series feature vector is constructed for each grid-connected project.

5. The multi-dimensional dashboard analysis method for network-connected review content according to claim 1, characterized in that: The steps of statistically fitting the time-series feature distribution corresponding to the tolerance delay events in historical grid-connected projects based on the progress time-series feature vector, extracting the preceding feature interval of the tolerance delay events in the time-series feature vector, and constructing a tolerance delay risk prediction model include: Select a positive sample set from historical grid-connected projects that experienced overdue tolerance events and a negative sample set that did not experience overdue tolerance events; The progress time-series feature vectors corresponding to the tolerance delay events of each grid-connected project in the positive sample set are extracted, and the progress time-series feature vectors are divided into pre-feature sequences and post-feature sequences with the occurrence time of the tolerance delay events as the dividing point. A sliding window scan is performed on the preceding feature sequence to calculate the variance of the progress deviation value, the rate of decrease of the node completion rate, and the rate of change of the frequency of creation of defective work orders within each time window, thus forming the preceding feature distribution. Kernel density estimation is performed on the pre-feature distribution, and the feature value range with distribution density exceeding the preset density threshold is extracted as the pre-feature range of the tolerance overdue event in the time series feature vector; For each grid-connected project in the negative sample set, backtrack to a time window of the same length as the preceding feature sequence of the positive sample set, and extract the negative sample feature sequence. A prediction model for tolerance of overdue risk is constructed by comparing and fitting the preceding feature sequences of the positive sample set with the negative example feature sequences of the negative sample set.

6. The multi-dimensional dashboard analysis method for network-connected review content according to claim 5, characterized in that: The step of performing a sliding window scan on the preceding feature sequence, calculating the variance of the progress deviation value, the rate of decrease in the node completion rate, and the rate of change in the frequency of defective work order creation within each time window, to construct the preceding feature distribution includes: Based on the review stages corresponding to historical grid-connected projects, the pre-requisite feature sequence is divided into phased sub-sequences, which include a document review stage sub-sequence, an on-site verification stage sub-sequence, and a rectification tracking stage sub-sequence. For each staged subsequence, a corresponding baseline window size is set, and a multi-scale sliding window scan is performed on the staged subsequence with a preset ratio of the staged subsequence length as the step size to generate a window feature set. The cross-scale mean values ​​of the variance of the progress deviation value, the rate of decline of the node completion rate, and the rate of change of the frequency of creation of defective work orders in the window feature set are calculated respectively to form the cross-scale mean feature vector of the phased subsequence. Extract overlapping feature windows at the boundaries of adjacent segmented subsequences, calculate the cosine similarity of the cross-scale mean feature vectors on both sides of the boundary, and mark the corresponding boundary as a stage transition point if the cosine similarity is lower than the preset similarity threshold. The cross-scale mean feature vectors of the phased subsequences are concatenated sequentially according to the review stages, and the corresponding stage transition point markers are associated with the concatenated result to form the pre-feature distribution.

7. The multi-dimensional dashboard analysis method for network-connected review content according to claim 1, characterized in that: The steps of generating multi-dimensional indicator dashboard data based on the grid connection review dataset and the overdue risk prediction list, and establishing multi-level index relationships for the multi-dimensional indicator dashboard data to support drill-down positioning of indicator data, include: Extract project-dimensional indicators, review stage-dimensional indicators, and tolerance work order-dimensional indicators from the grid connection review dataset. Extract the tolerance overdue risk probability of each target grid connection project from the overdue risk prediction list. Add the tolerance overdue risk probability as a risk weight to the corresponding project-dimensional indicators to form a multi-dimensional indicator dashboard data. A three-level index tree is constructed based on multi-dimensional indicator dashboard data. Nodes at the same level in the three-level index tree are sorted in descending order of risk weight to generate a risk perception sorting index. Risk warning tags are associated with nodes whose risk weight exceeds the preset risk threshold. When responding to a drill-down positioning request, the corresponding node is located in the three-level index tree based on the target dimension identifier corresponding to the drill-down positioning request. The child nodes are then traversed downwards along the node, and the aggregated value of the dimension indicator corresponding to the child node and the risk warning mark are returned as the drill-down positioning result.

8. A multi-dimensional dashboard analysis system for online content review, characterized in that, include: The data integration module is used to obtain grid connection data records from the grid connection process platform, ITSCADA system and architecture control system, and integrate the grid connection data records to generate a grid connection review dataset with grid connection projects as the basic unit. The time series processing module is used to perform time series alignment processing on the grid connection review dataset, extract the timestamp sequence of each grid connection project, and construct the progress time series feature vector of each grid connection project. The model building module is used to statistically fit the time-series feature distribution corresponding to the defective and overdue events in historical grid-connected projects based on the progress time-series feature vector, extract the preceding feature interval of the defective and overdue events in the time-series feature vector, and build a defective and overdue risk prediction model. The risk prediction module is used to identify target grid-connected projects. The progress time-series feature vector of the target grid-connected project is input into the tolerance for overdue risk prediction model to calculate the tolerance for overdue risk probability of the target grid-connected project and generate an overdue risk prediction list. The index association module is used to generate multi-dimensional indicator dashboard data based on the grid connection review dataset and the overdue risk prediction list, and to establish multi-level index association relationships for the multi-dimensional indicator dashboard data to support the drilling down of indicator data layer by layer. The results generation module is used to establish a correlation mapping between the risk items in the overdue risk prediction list and the progress time sequence feature vectors of the corresponding grid-connected projects, and generate multi-dimensional analysis results of the grid connection review process.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-dimensional dashboard analysis method for network-connected review content as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-dimensional dashboard analysis method for network-connected review content as described in any one of claims 1-7.

Images