AI credible comprehensive measurement method, system and equipment for power grid dispatching and medium

Abstract

The invention provides a power grid dispatching-oriented AI credible comprehensive measurement method, system and equipment and a medium, and the method comprises the steps: constructing an integrated environment, fusing a resource isolated AI training reasoning environment and a power grid regulation and control simulation environment, and forming a business closed-loop verification capability; multi-source heterogeneous data are integrated and retrieved in a unified mode through a middleware chain framework capable of being arranged, and evaluation cases are dynamically organized; establishing a trusted measurement index library covering data and algorithms, performing standardized evaluation by using a calculation engine, and introducing a negative coefficient mechanism to ensure the preciseness of comprehensive measurement; full-period monitoring and traceability analysis are carried out on the power grid dispatching AI, data or model optimization is triggered based on a result, and reevaluation is started; the method has the effects of realizing self-adaption from general verification to a power grid business scene, solving the problems of resource dispersion and non-closed loop verification, and providing systematic platform support for improving the credibility and landing reliability of power grid dispatching AI application.

Description

Technical Field

[0001] This invention belongs to the fields of power dispatching and artificial intelligence technology, specifically relating to an AI-based trusted comprehensive measurement method, system, equipment, and medium for power grid dispatching. Background Technology

[0002] While artificial intelligence (AI) technology brings tremendous opportunities, it also introduces risks and challenges. These include data issues leading to decision-making biases, black-box models resulting in opaque algorithms, and unreliable algorithms causing application risks. Deep learning-based AI technology is inherently vulnerable and susceptible to attacks, making it difficult to fully trust the reliability of AI systems. Power grid dispatching has extremely high requirements for security and reliability; if these risks and challenges of AI lead to decision-making errors, it will cause significant losses to the entire power system. Therefore, the practical application of power grid dispatching AI software must undergo thorough testing and verification, requiring comprehensive validation of the software's reliability metrics.

[0003] Currently, there is no unified platform or system for AI credibility measurement in power grid dispatching. Existing evaluation tools for single dimensions are often isolated and fragmented. Moreover, compared with general artificial intelligence research, AI credibility measurement in power grid dispatching has the following unique characteristics and challenges: the verification of power grid dispatching AI requires a specific business environment; the sample data related to power grid dispatching is complex and diverse, needs to be generated or processed through specific dispatching automation systems, and falls within the scope of business protection, resulting in data barriers; power grid dispatching involves different typical scenarios such as monitoring, analysis, and decision-making, and the differences in scenarios need to be considered when establishing credibility indicators and evaluation models.

[0004] Therefore, there is an urgent need for a systematic and standardized technical solution to address the reliability measurement and scenario verification of power grid dispatch AI. Summary of the Invention

[0005] This invention provides a comprehensive AI credibility measurement method for power grid dispatching. It constructs an integrated environment for AI credibility assessment in the field of power grid dispatching. Through middleware-based data and model integration and retrieval methods, it organizes AI credibility assessment cases for typical scenarios, realizing a comprehensive assessment of the credibility of AI test data and AI algorithm credibility for power grid dispatching.

[0006] In a first aspect, this invention provides an AI-based comprehensive reliability measurement method for power grid dispatching. The method executes an evaluation process based on a pre-built integrated environment and an AI reliability measurement index library. The integrated environment includes a containerized resource management platform deployed in an AI training and inference environment with resource isolation, and a control system simulation environment that dynamically interacts with a power grid dispatching mirror system. The AI ​​reliability measurement index library configures baseline thresholds and weights for each index. The method includes the following steps: S1. In response to the evaluation request, the heterogeneous data resources of power grid dispatch are uniformly managed through the integrated management architecture to retrieve and combine power grid models, AI models and AI test sample data that match the evaluation request from the data source layer, and load them into the integrated environment to form evaluation cases. S2. Based on the evaluation case, select indicators from the AI ​​trustworthiness measurement indicator library to construct an evaluation workflow, calculate the measurement results of each dimension, and perform weighted aggregation of the measurement results of each dimension based on the weight; when the measurement results of the preset key indicators are lower than their corresponding baseline thresholds, use a rejection coefficient to constrain the weighted aggregation results to obtain the comprehensive AI trustworthiness measurement result. S3. Monitor the evaluation process and perform source tracing analysis based on the measurement results of each dimension and the AI ​​trustworthy comprehensive measurement results; when the measurement results do not meet the preset requirements, trigger the optimization operation of AI test data and / or AI model based on the source tracing analysis results, and reload the optimized data and / or model into the integrated environment to perform the evaluation steps again.

[0007] By adopting the above-described scheme, this invention provides a comprehensive AI reliability measurement method for power grid dispatching. This method constructs an integrated verification environment, integrating a resource-isolated AI training and inference platform with a power grid control simulation system to form a closed-loop verification capability. Based on an integrated management architecture with programmable middleware, it achieves unified management and intelligent dispatching of multi-source heterogeneous data. It establishes a measurement index library covering data and algorithm reliability, employing weighted aggregation and rejection mechanisms for comprehensive evaluation. Through full-cycle monitoring, tracing, optimization, and re-evaluation, it forms a closed-loop process for evaluating, diagnosing, optimizing, and re-evaluating power grid dispatching AI. The technical solution of this invention effectively solves key technical challenges such as the lack of a verification environment, dispersed data resources, and inconsistent evaluation standards, providing systematic support for the quality assurance and reliable implementation of power grid dispatching AI products.

[0008] In some embodiments of the present invention, the AI ​​training and inference environment is configured with processor resources, storage resources and graphics processing unit resources according to a preset computing power resource template, and pre-installs a variety of AI algorithm running frameworks and data processing libraries. The AI ​​algorithm running framework includes at least one deep learning framework, and the data processing library includes at least one numerical calculation library and data analysis library, which are used to support AI model training and inference for power grid load forecasting, renewable energy output forecasting and optimized scheduling.

[0009] In some embodiments of the present invention, the simulation operating environment of the control system includes a real-time library, a historical library, a bus service, a scheduling business application service, and an AI agent service; the AI ​​agent service receives the business decision results output by the AI ​​model from the AI ​​training and inference environment, performs format conversion and field mapping on the business decision results, generates scheduling instruction data compatible with the scheduling mirror system, and sends the scheduling instruction data to the scheduling mirror system.

[0010] In some embodiments of the present invention, the integrated management architecture includes an integrated management architecture comprising a data source layer, a middleware layer, and a data application layer; the middleware layer includes a data integration middleware and a data retrieval middleware; the data integration middleware includes a connector middleware, a verification and conversion middleware, and a loading middleware; the connector middleware extracts power grid model, AI model, and AI test sample data from the data source layer; the verification and conversion middleware performs format verification, integrity verification, and conversion processing on the extracted data; the loading middleware loads the verified and converted data into the target data storage; the data retrieval middleware parses the evaluation request and retrieves power grid model, AI model, and AI test sample data matching the evaluation request from the data source layer; the data application layer combines the retrieved power grid model, AI model, and AI test sample data and loads them into the integrated environment.

[0011] In some embodiments of the present invention, the data source layer is used to store multi-scenario power grid topology models, power grid operation condition models, AI models corresponding to different power grid dispatching business scenarios, as well as AI test samples such as historical operation records, simulation samples, and real-time sampling data; The connector middleware in the data integration middleware is configured to establish data connections with the scheduling business system, scheduling mirror system, external simulation system and file system, and supports at least one of relational database interface, file interface and programming interface; The verification and conversion middleware performs format verification, integrity verification, business rule verification, field renaming, unit conversion, and data type conversion on the data extracted by the connector middleware; The loading middleware writes the data processed by the verification and conversion middleware into the target data storage, and updates the data directory information corresponding to the power grid model, AI model, and AI test samples.

[0012] In some embodiments of the present invention, the data retrieval middleware parses the received evaluation request, extracts the target power grid dispatch scenario identifier and the evaluation object identifier, and performs access control verification. The data retrieval middleware, based on a preset query description, retrieves power grid models, AI models, and AI test sample data that match the target power grid dispatch scenario identifier from the data source layer. In the data application layer, these data are categorized and organized into prediction scenarios, optimization dispatch scenarios, and reinforcement learning control scenarios to form corresponding evaluation cases. The evaluation cases are then loaded into the integrated environment.

[0013] In some embodiments of the present invention, the AI ​​trustworthiness metric library includes: A set of data credibility indicators for AI test data, including at least completeness indicators, timeliness indicators, consistency indicators, accuracy indicators, and distribution rationality indicators, is used to characterize the characteristics of power grid dispatch data in terms of data missingness, time alignment, cross-source consistency, and outlier distribution. A set of algorithm credibility indicators for AI algorithms, which includes at least accuracy indicators, robustness indicators, stability indicators and interpretability indicators, used to characterize the deviation level of AI models under multiple operating conditions, sensitivity to disturbances, performance fluctuations across time periods and interpretability of output results; The AI ​​trustworthy metrics library is configured with corresponding subsets of metrics and their baseline thresholds for different power grid dispatching scenarios.

[0014] In some embodiments of the present invention, the computing engine selects a set of target indicators from the AI ​​trustworthiness metric library according to the evaluation case type, and generates an evaluation workflow definition, which includes the calculation order and dependencies of each indicator. The computing engine executes the evaluation workflow in the integrated environment, performs multi-dimensional index calculations on the evaluation cases, obtains the measurement results of each dimension, and performs weighted aggregation of the measurement results of each dimension according to the preset weights for each index. Indicators related to power grid security and dispatch stability are configured as key indicators. These key indicators are associated with veto coefficients. When the measurement result of any key indicator is lower than the corresponding baseline threshold, the veto coefficients are used to downgrade or set to zero on the weighted aggregation result.

[0015] In some embodiments of the present invention, monitoring the evaluation process includes: Collect processor utilization, graphics processing unit utilization, memory usage, storage and network input / output, and container instance running status in the AI ​​training and inference environment and the simulation operation environment of the control system; Collect and evaluate the queuing status, execution start and end times, execution duration, failure reasons, and exception log information of the task. The processor utilization, graphics processing unit utilization, memory usage, storage and network input / output, container instance running status, and the evaluation task information are associated and stored with the measurement results of each dimension obtained from multiple evaluations and the AI ​​trustworthy comprehensive measurement results.

[0016] In some embodiments of the present invention, when the AI ​​trust comprehensive measurement result or the measurement result of any key indicator does not meet the preset requirements, according to the source analysis results based on the measurement results of each dimension and the AI ​​trust comprehensive measurement result in step S4, weak data dimensions and weak algorithm dimensions are identified, and corresponding optimization operations are triggered. The optimization operations include performing at least one of the following operations on AI test data: sample augmentation, sample resampling, and data distribution reconstruction; and performing at least one of the following operations on AI models: hyperparameter tuning, network structure adjustment, and model retraining. The optimized AI test data and / or optimized AI model are loaded into the integrated environment, and step S3 of claim 1 is performed to evaluate the optimized AI test data and / or AI model.

[0017] By adopting the above-mentioned scheme, the AI-based comprehensive trust measurement method for power grid dispatching of this invention solves the problems of lack of verification environment, resource dispersion, and inability to close the loop in power grid dispatching AI trust by constructing an integrated environment that includes an artificial intelligence training environment, an inference environment, and a control system simulation operation environment. It achieves adaptability from general-purpose to power grid business scenarios. By constructing a programmable data processing middleware, it achieves loose coupling and modularity of the system, flexibly meeting the data processing needs of different scenarios. By constructing a comprehensive trust indicator library, it conducts comprehensive trust measurement evaluation and source tracing analysis, realizing a full-cycle trust governance closed loop of evaluation, diagnosis, optimization, and re-evaluation, supporting the sustainable development and application of power grid dispatching AI.

[0018] In a second aspect, the present invention provides a reliable evaluation and verification system for power grid dispatching AI, the system comprising: An integrated environment construction module is used to build an integrated environment for AI-based reliable measurement and evaluation of power grid dispatching; the integrated environment includes a containerized resource management platform deployed in an AI training and inference environment with resource isolation, and a control system simulation operation environment that interacts with the dispatching mirror system; The assessment management and case organization module is used to respond to assessment requests and manage heterogeneous power grid dispatch data resources in a unified manner through an integrated management architecture. It retrieves and combines power grid models, AI models and AI test sample data that match the assessment request from the data source layer and loads them into the integrated environment to form assessment cases. The trustworthy assessment calculation engine module is used to select indicators from the AI ​​trustworthy measurement indicator library according to the assessment case to construct the assessment workflow, calculate the measurement results of each dimension, and perform weighted aggregation of the measurement results of each dimension based on the weight; when the measurement result of the preset key indicator is lower than its corresponding baseline threshold, the weighted aggregation result is constrained by the veto coefficient to obtain the comprehensive AI trustworthy measurement result. The full-cycle governance and source tracing analysis module is used to monitor the evaluation process and perform source tracing analysis based on the measurement results of each dimension and the AI ​​trustworthy comprehensive measurement results. When the measurement results do not meet the preset requirements, the module triggers optimization operations on the AI ​​test data and / or AI model based on the source tracing analysis results, and reloads the optimized data and / or model into the integrated environment to perform the evaluation steps again.

[0019] In some embodiments of the present invention, the AI ​​trustworthiness metric library includes: A set of data credibility indicators for AI test data, including at least completeness indicators, timeliness indicators, consistency indicators, accuracy indicators, and distribution rationality indicators, is used to characterize the characteristics of power grid dispatch data in terms of data missingness, time alignment, cross-source consistency, and outlier distribution. A set of algorithm credibility indicators for AI algorithms, which includes at least accuracy indicators, robustness indicators, stability indicators and interpretability indicators, used to characterize the deviation level of AI models under multiple operating conditions, sensitivity to disturbances, performance fluctuations across time periods and interpretability of output results; The AI ​​trustworthy metrics library is configured with corresponding subsets of metrics and corresponding baseline thresholds for different power grid dispatching scenarios.

[0020] In some embodiments of the present invention, the integrated management architecture includes an integrated management architecture comprising a data source layer, a middleware layer, and a data application layer; the middleware layer includes a data integration middleware and a data retrieval middleware; the data integration middleware includes a connector middleware, a verification and conversion middleware, and a loading middleware; the connector middleware extracts power grid model, AI model, and AI test sample data from the data source layer; the verification and conversion middleware performs format verification, integrity verification, and conversion processing on the extracted data; the loading middleware loads the verified and converted data into the target data storage; the data retrieval middleware parses the evaluation request and retrieves power grid model, AI model, and AI test sample data matching the evaluation request from the data source layer; the data application layer combines the retrieved power grid model, AI model, and AI test sample data and loads them into the integrated environment.

[0021] In some embodiments of the present invention, the data source layer is used to store multi-scenario power grid topology models, power grid operation condition models, AI models corresponding to different power grid dispatching business scenarios, as well as AI test samples such as historical operation records, simulation samples, and real-time sampling data; The connector middleware in the data integration middleware is configured to establish data connections with the scheduling business system, scheduling mirror system, external simulation system and file system, and supports at least one of relational database interface, file interface and programming interface; The verification and conversion middleware performs format verification, integrity verification, business rule verification, field renaming, unit conversion, and data type conversion on the data extracted by the connector middleware; The loading middleware writes the data processed by the verification and conversion middleware into the target data storage, and updates the data directory information corresponding to the power grid model, AI model, and AI test samples.

[0022] In a third aspect, the present invention provides an electronic device including a processor and a memory, the processor being configured to execute a computer program stored in the memory to implement the AI-based trusted integrated measurement method for power grid dispatch.

[0023] In a fourth aspect, the present invention provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the AI-based trusted comprehensive measurement method for power grid dispatch.

[0024] Compared with existing technologies, the beneficial effects of this invention lie in its comprehensive AI credibility measurement method for power grid dispatching. By constructing an integrated verification environment that includes an AI training and inference environment and a simulated operation environment for the control system, it can provide highly realistic, resource-isolated, and closed-loop operational business verification scenarios for power grid dispatching AI, solving the core problems of lack of verification environment, scattered resources, and inability to achieve closed-loop operation in existing technologies. By designing an integrated management architecture that includes a configurable middleware chain (data integration and retrieval middleware), it enhances the system's unified management, flexible arrangement, and efficient scheduling capabilities for multi-source and heterogeneous data in power grid dispatching, enabling rapid organization from data to evaluation cases and significantly improving the automation level of the verification process and its adaptability to different scenarios. By establishing a comprehensive measurement index library covering data and algorithm credibility and introducing a veto mechanism, as well as a standardized evaluation engine, it significantly improves the scientific rigor and comprehensiveness of the evaluation, enhances the ability to detect hidden risks of AI models in key power grid dispatching scenarios, and effectively avoids decision-making errors caused by "black box" models or data bias.

[0025] Additional advantages, objects, and features of the invention will be set forth in part in the description which follows, and will also become apparent in part to those skilled in the art upon studying the text, or may be learned by practice of the invention. The objects and other advantages of the invention will become apparent from the description and the accompanying drawings.

[0026] Those skilled in the art will understand that the objectives and advantages achievable with this invention are not limited to those specifically described above, and that the above and other objectives achievable with this invention will become clearer from the following detailed description. Attached Figure Description

[0027] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0028] In the attached diagram: Figure 1 This is a flowchart illustrating an AI-based trusted comprehensive measurement method for power grid dispatch, provided as an embodiment of the present invention.

[0029] Figure 2 This is a schematic diagram of a power grid dispatching AI credibility assessment and verification system module provided in an embodiment of the present invention.

[0030] Figure 3 This is a system framework diagram of a power grid dispatching AI credibility assessment and verification system provided in an embodiment of the present invention.

[0031] Figure 4This is a schematic diagram of a multi-scenario model and sample integration management for power grid dispatching AI provided in an embodiment of the present invention.

[0032] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0033] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0034] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0035] Figure 1 This is a flowchart illustrating an evaluation method for a reliable measurement of power grid dispatching AI, provided in an embodiment of the present invention.

[0036] Example 1, such as Figure 1 As shown in Figure 3, this invention provides an AI-based trusted comprehensive measurement method for power grid dispatching, the method comprising the following steps: The method executes an evaluation process based on a pre-built integrated environment and an AI trustworthy metrics library. The integrated environment includes a containerized resource management platform deployed in an AI training and inference environment with resource isolation, and a control system simulation environment that dynamically interacts with the power grid dispatch mirror system. The AI ​​trustworthy metrics library configures baseline thresholds and weights for each metric. The method includes the following steps: S1. In response to the evaluation request, the heterogeneous data resources of power grid dispatch are uniformly managed through the integrated management architecture to retrieve and combine power grid models, AI models and AI test sample data that match the evaluation request from the data source layer, and load them into the integrated environment to form evaluation cases. S2. Based on the evaluation case, select indicators from the AI ​​trustworthiness measurement indicator library to construct an evaluation workflow, calculate the measurement results of each dimension, and perform weighted aggregation of the measurement results of each dimension based on the weight; when the measurement results of the preset key indicators are lower than their corresponding baseline thresholds, use a rejection coefficient to constrain the weighted aggregation results to obtain the comprehensive AI trustworthiness measurement result. S3. Monitor the evaluation process and perform source tracing analysis based on the measurement results of each dimension and the comprehensive AI trust measurement results. When the measurement results do not meet the preset requirements, trigger optimization operations on the AI ​​test data and / or AI model based on the source tracing analysis results, and reload the optimized data and / or model into the integrated environment to execute the evaluation steps again. Using the above scheme, the present invention provides an AI trust measurement method for power grid dispatching, offering a method for constructing a comprehensive AI trust measurement and scenario-based verification system for power grid dispatching. It constructs an integrated environment for AI trust assessment in the power grid dispatching field, and through a method of retrieving, integrating, and matching data resources based on an integrated management architecture, organizes AI trust assessment cases for typical scenarios, achieving a comprehensive assessment of the trustworthiness of AI test data and the trustworthiness of AI algorithms. This invention addresses the problems of lack of integrated verification environment and comprehensive evaluation indicators, scattered data resources, poor functional reusability, and inability to close the verification loop in power grid dispatching AI credibility assessment. It helps to establish a flexible, efficient, and loosely coupled verification environment architecture, improve the quality of AI technology products in the power grid control field, promptly identify the shortcomings of AI algorithms, reduce the verification cost of AI results, and effectively improve the reliability and stability of artificial intelligence results in the dispatching field.

[0037] In some embodiments of the present invention, the AI ​​training and inference environment is configured with processor resources, storage resources and graphics processing unit resources according to a preset computing power resource template, and pre-installs a variety of AI algorithm running frameworks and data processing libraries. The AI ​​algorithm running framework includes at least one deep learning framework, and the data processing library includes at least one numerical calculation library and data analysis library, which are used to support AI model training and inference for power grid load forecasting, renewable energy output forecasting and optimized scheduling.

[0038] In some embodiments of the present invention, the simulation operating environment of the control system includes a real-time library, a historical library, a bus service, a scheduling business application service, and an AI agent service; the AI ​​agent service receives the business decision results output by the AI ​​model from the AI ​​training and inference environment, performs format conversion and field mapping on the business decision results, generates scheduling instruction data compatible with the scheduling mirror system, and sends the scheduling instruction data to the scheduling mirror system.

[0039] Optionally, in embodiments of the present invention, an integrated environment for AI-based reliable measurement and evaluation for power grid dispatch is constructed, including an AI training environment and inference environment, and a control system simulation operation environment, as detailed below: 1) The construction of the power grid dispatching AI training and inference environment includes: a. Write a Dockerfile and build a Docker image; b. Set up resource configuration templates and allocate computing resources such as GPU / CPU, memory, and hard disk; c. Load AI frameworks such as TensorFlow, PyTorch, and Keras, as well as various algorithm packages such as NumPy, Pandas, and SciPy, covering algorithms for various scenarios of prediction, optimization, and decision-making; d. Establish a service-oriented container environment with resource isolation, and use Kubernetes for unified container management to support the training, inference, and verification of different core AI algorithms.

[0040] 2) The construction of the simulation operating environment for the control system includes: Deploy real-time and historical databases, bus services, application services, and AI agent services. Design an external interaction mechanism for the control system's simulated operating environment.

[0041] a. The control system simulates the operation environment by receiving data from the power grid dispatch mirror system via real-time forwarding; b. The control system's simulated operating environment feeds back the decision-making results of AI applications to the power grid dispatch mirror system through AI agent services; c. The power grid dispatch mirror system simulates the actual operation of the power grid, forming a dynamic closed loop.

[0042] In some embodiments of the present invention, the integrated management architecture includes an integrated management architecture comprising a data source layer, a middleware layer, and a data application layer; the middleware layer includes a data integration middleware and a data retrieval middleware; the data integration middleware includes a connector middleware, a verification and conversion middleware, and a loading middleware; the connector middleware extracts power grid model, AI model, and AI test sample data from the data source layer; the verification and conversion middleware performs format verification, integrity verification, and conversion processing on the extracted data; the loading middleware loads the verified and converted data into the target data storage; the data retrieval middleware parses the evaluation request and retrieves power grid model, AI model, and AI test sample data matching the evaluation request from the data source layer; the data application layer combines the retrieved power grid model, AI model, and AI test sample data and loads them into the integrated environment.

[0043] In some embodiments of the present invention, the data source layer is used to store multi-scenario power grid topology models, power grid operation condition models, AI models corresponding to different power grid dispatching business scenarios, as well as AI test samples such as historical operation records, simulation samples, and real-time sampling data; The connector middleware in the data integration middleware is configured to establish data connections with the scheduling business system, scheduling mirror system, external simulation system and file system, and supports at least one of relational database interface, file interface and programming interface; The verification and conversion middleware performs format verification, integrity verification, business rule verification, field renaming, unit conversion, and data type conversion on the data extracted by the connector middleware; The loading middleware writes the data processed by the verification and conversion middleware into the target data storage, and updates the data directory information corresponding to the power grid model, AI model, and AI test samples.

[0044] Optionally, in this embodiment, multi-scenario model and sample integration management, such as Figure 4 As shown, the details are as follows: 1) Establish an integrated management layered architecture, which mainly includes the data source layer, middleware layer and data application layer.

[0045] 2) In the data source layer, heterogeneous power grid dispatch data mainly includes power grid models, AI models, and test samples from multiple scenarios. Data formats primarily include relational databases, file systems (CSV, JSON, Text, Excel), API interfaces, and real-time data streams, providing data connection and access interfaces for each data source.

[0046] 3) The middleware layer includes data integration middleware, data retrieval middleware, and common middleware.

[0047] a. The data integration middleware is responsible for extracting data from multiple data source layers, processing it, and then loading it into the target data storage or directly providing it to the data application layer. The core modules are the connector middleware, the verification and transformation middleware, and the loading middleware, and the order of these middleware components is configurable.

[0048] The connector middleware manages connections to different data sources, supports connection pool monitoring, authentication, and retry mechanisms, and supports full and incremental data extraction strategies.

[0049] The verification and conversion middleware validates the data format, integrity, and business rules, while performing conversion operations such as data deduplication, data formatting, and field mapping.

[0050] The loading middleware writes the processed data to the target address, establishes a data resource management directory, and supports batch loading and business flow control.

[0051] b. The data retrieval middleware receives query requests from evaluation tasks, distributes the queries to the underlying data sources, aggregates the results, and returns them to the application layer. The core modules are query parsing middleware, security control middleware, query distribution middleware, and result aggregation middleware; the order of these modules is configurable.

[0052] The query parsing middleware parses the Unified Query Language into an internal abstract structure and performs syntax validation.

[0053] The security control middleware performs identity authentication and permission verification, and supports data access control and data anonymization.

[0054] The query distribution middleware rewrites and optimizes the query plan, matching the query to one or more of the most suitable data sources for execution based on the original data information.

[0055] The result aggregation middleware merges, sorts, and paginates query results from multiple data sources.

[0056] The common middleware is responsible for intermediate processes such as caching, security, and monitoring.

[0057] In some embodiments of the present invention, the data retrieval middleware parses the received evaluation request, extracts the target power grid dispatch scenario identifier and the evaluation object identifier, and performs access control verification. The data retrieval middleware, based on a preset query description, retrieves power grid models, AI models, and AI test sample data that match the target power grid dispatch scenario identifier from the data source layer. In the data application layer, these data are categorized and organized into prediction scenarios, optimization dispatch scenarios, and reinforcement learning control scenarios to form corresponding evaluation cases. The evaluation cases are then loaded into the integrated environment.

[0058] Optionally, in embodiments of the present invention, the generation of evaluation cases involves returning query results through a data retrieval middleware and organizing them into evaluation cases. This includes matching different power grid models, AI models, and test samples to meet the evaluation needs of different scenarios such as prediction and optimized scheduling; and loading the corresponding AI integrated verification environment to organize and form an evaluation verification environment.

[0059] In some embodiments of the present invention, the AI ​​trustworthiness metric library includes: A set of data credibility indicators for AI test data, including at least completeness indicators, timeliness indicators, consistency indicators, accuracy indicators, and distribution rationality indicators, is used to characterize the characteristics of power grid dispatch data in terms of data missingness, time alignment, cross-source consistency, and outlier distribution. A set of algorithm credibility indicators for AI algorithms, which includes at least accuracy indicators, robustness indicators, stability indicators and interpretability indicators, used to characterize the deviation level of AI models under multiple operating conditions, sensitivity to disturbances, performance fluctuations across time periods and interpretability of output results; The AI ​​trustworthy metrics library is configured with corresponding subsets of metrics and their baseline thresholds for different power grid dispatching scenarios.

[0060] In some embodiments of the present invention, the computing engine selects a set of target indicators from the AI ​​trustworthiness metric library according to the evaluation case type, and generates an evaluation workflow definition, which includes the calculation order and dependencies of each indicator. The computing engine executes the evaluation workflow in the integrated environment, performs multi-dimensional index calculations on the evaluation cases, obtains the measurement results of each dimension, and performs weighted aggregation of the measurement results of each dimension according to the preset weights for each index. Indicators related to power grid security and dispatch stability are configured as key indicators. These key indicators are associated with veto coefficients. When the measurement result of any key indicator is lower than the corresponding baseline threshold, the veto coefficients are used to downgrade or set to zero on the weighted aggregation result.

[0061] Optionally, in an embodiment of the present invention, in step S3, an AI trustworthiness measurement index library for various scenarios of power grid dispatching is established, and a computing engine is invoked to conduct standardized trustworthiness assessment, as follows: 1) The construction of the trustworthy metrics library includes: defining and managing a series of trustworthy metrics in a configurable manner, with the evaluation scope covering AI test data and AI algorithms, and scenarios covering power grid dispatch prediction and optimization dispatch, including evaluation of AI test data and evaluation of AI algorithms.

[0062] a. For the evaluation of AI test data, reliable indicators include data integrity, timeliness, consistency, accuracy, and reasonable distribution.

[0063] b. For the evaluation of AI algorithms, reliable metrics include accuracy, recall, robustness, interpretability, mean squared error, and root mean square error.

[0064] 2) Establish baselines for each evaluation indicator. The system pre-sets the evaluation baseline values ​​for various indicators and forms a configurable interface.

[0065] 3) Create a standardized evaluation workflow engine. The engine provides a graphical interface or configuration file method. Based on the type and scenario of the model being tested, users can drag and drop or select appropriate indicators from the indicator library to combine them into a custom evaluation workflow. The engine is responsible for scheduling indicator calculation tasks and managing the dependencies and data flow between tasks.

[0066] 4) The comprehensive credibility index aggregation includes receiving index calculation results from multiple dimensions, designing a weight allocation unit, and realizing index aggregation. Based on the pre-set recommended weights for each credibility dimension within the scenario, upon receiving a new evaluation task, weights are matched based on scenario features. To prevent a weakness in the overall score across a particular key dimension, a veto coefficient is set, and the comprehensive credibility score is calculated comprehensively.

[0067] In some embodiments of the present invention, step S3, monitoring the evaluation process includes: Collect processor utilization, graphics processing unit utilization, memory usage, storage and network input / output, and container instance running status in the AI ​​training and inference environment and the simulation operation environment of the control system; Collect and evaluate the queuing status, execution start and end times, execution duration, failure reasons, and exception log information of the task. The processor utilization, graphics processing unit utilization, memory usage, storage and network input / output, container instance running status, and the evaluation task information are associated and stored with the measurement results of each dimension obtained from multiple evaluations and the AI ​​trustworthy comprehensive measurement results.

[0068] In some embodiments of the present invention, in step S3, when the AI ​​trust comprehensive measurement result or the measurement result of any key indicator does not meet the preset requirements, the weak data dimension and weak algorithm dimension are identified based on the source analysis results of the measurement results of each dimension and the AI ​​trust comprehensive measurement result in step S3, and corresponding optimization operations are triggered. The optimization operations include performing at least one of the following operations on AI test data: sample augmentation, sample resampling, and data distribution reconstruction; and performing at least one of the following operations on AI models: hyperparameter tuning, network structure adjustment, and model retraining. The optimized AI test data and / or optimized AI model are loaded into the integrated environment, and step S2 of claim 1 is performed to evaluate the optimized AI test data and / or AI model.

[0069] Optionally, in embodiments of the present invention, the full-cycle evaluation process monitoring and traceability analysis are specifically as follows: 1) Evaluation process monitoring: During the evaluation process, the system's operating status (GPU, CPU, memory, container nodes, evaluation tasks) is monitored in real time and logs are generated; 2) Credibility traceability analysis: Based on the scores of each dimension of credibility indicators and the comprehensive score, the evaluation results are displayed, and a credibility dimension relationship diagram is established to trace the evolution path of credibility.

[0070] 3) Optimization and re-evaluation: Based on the AI ​​trustworthiness comprehensive metric results, provide optimization suggestions for the AI ​​test data or AI model, and re-trigger the evaluation process. The evaluation process includes: a. For AI test data, call the data expansion or reconstruction interface to reorganize the test data. b. For AI algorithms, retrain them based on the AI ​​training environment in the integrated environment, and then redeploy them to the inference environment to start the evaluation process.

[0071] By adopting the above-mentioned scheme, the AI-based comprehensive trust measurement method for power grid dispatching of this invention solves the problems of lack of verification environment, resource dispersion, and inability to close the loop in power grid dispatching AI trust by constructing an integrated environment that includes an artificial intelligence training environment, an inference environment, and a control system simulation operation environment. It achieves adaptability from general-purpose to power grid business scenarios. By constructing a programmable data processing middleware, it achieves loose coupling and modularity of the system, flexibly meeting the data processing needs of different scenarios. By constructing a comprehensive trust indicator library, it conducts comprehensive trust measurement evaluation and source tracing analysis, realizing a full-cycle trust governance closed loop of evaluation, diagnosis, optimization, and re-evaluation, supporting the sustainable development and application of power grid dispatching AI.

[0072] Example 2, as Figure 2 As shown, the present invention provides a power grid dispatching AI credibility assessment and verification system, the system comprising: an integrated environment construction module S11, an assessment management and case organization module S12, a credibility assessment calculation engine module S13, and a full-cycle governance and source tracing analysis module S14.

[0073] The integrated environment construction module S11 is used to construct an integrated environment for the credibility assessment of AI for power grid dispatching. The integrated environment includes a containerized resource management platform deployed in an AI training and inference environment with resource isolation, and a control system simulation operation environment that interacts with the dispatching mirror system. The control system simulation operation environment receives operation data from the dispatching mirror system and sends the decision results of the AI ​​model back to the dispatching mirror system. The evaluation management and case organization module S12 is used to respond to the received evaluation request for trusted verification of power grid dispatch AI. Through an integrated management architecture comprising a data source layer, middleware layer, and data application layer, it uniformly manages heterogeneous power grid dispatch data. The middleware layer includes a data integration middleware and a data retrieval middleware. The data integration middleware includes a connector middleware, a verification and conversion middleware, and a loading middleware. The connector middleware extracts power grid model, AI model, and AI test sample data from the data source layer. The verification and conversion middleware performs format verification, integrity verification, and conversion processing on the extracted data. The loading middleware loads the verified and converted data into the target data storage. The data retrieval middleware parses the evaluation request and retrieves matching power grid model, AI model, and AI test sample data from the data source layer. The data application layer combines the retrieved power grid model, AI model, and AI test sample data and loads them into the integrated environment described in step S1 as evaluation cases. The trustworthy assessment calculation engine module S13 is used to establish an AI trustworthy measurement index library for power grid dispatching scenarios and call the calculation engine for evaluation. The AI ​​trustworthy measurement index library includes AI test data trustworthiness indexes and AI algorithm trustworthiness indexes, and configures baseline thresholds and weights for each index. The calculation engine selects indicators from the index library according to the evaluation case to construct an evaluation workflow, calculates the measurement results of each dimension, and performs weighted aggregation of the measurement results of each dimension based on the weights. When the measurement result of a preset key index is lower than the baseline threshold corresponding to each index, a veto coefficient is used to constrain the weighted aggregation result to obtain the comprehensive AI trustworthiness measurement result. The full-cycle governance and source tracing analysis module S14 is used to monitor the evaluation process of step S3, and to perform source tracing analysis based on the measurement results of each dimension and the AI ​​trust comprehensive measurement results obtained in step S3. When the measurement results of the AI ​​trust comprehensive measurement results or key indicators do not meet the preset requirements, the module triggers the reconstruction and / or expansion of AI test data and / or the retraining of AI models based on the source tracing analysis results. The optimized AI test data and / or AI models are reloaded into the integrated environment, and the evaluation process of step S3 is executed again.

[0074] In some embodiments of the present invention, the AI ​​trustworthiness metric library includes: A set of data credibility indicators for AI test data, including at least completeness indicators, timeliness indicators, consistency indicators, accuracy indicators, and distribution rationality indicators, is used to characterize the characteristics of power grid dispatch data in terms of data missingness, time alignment, cross-source consistency, and outlier distribution. A set of algorithm credibility indicators for AI algorithms, which includes at least accuracy indicators, robustness indicators, stability indicators and interpretability indicators, used to characterize the deviation level of AI models under multiple operating conditions, sensitivity to disturbances, performance fluctuations across time periods and interpretability of output results; The AI ​​trustworthy metrics library is configured with corresponding subsets of metrics and corresponding baseline thresholds for different power grid dispatching scenarios.

[0075] In some embodiments of the present invention, in the trustworthy assessment computing engine module, the computing engine selects a set of target indicators from the AI ​​trustworthy measurement indicator library according to the assessment case type, and generates an assessment workflow definition, the assessment workflow definition including the calculation order and dependencies of each indicator; The computing engine executes the evaluation workflow in the integrated environment, performs multi-dimensional index calculations on the evaluation cases, obtains the measurement results of each dimension, and performs weighted aggregation of the measurement results of each dimension according to the preset weights for each index. Indicators related to power grid security and dispatch stability are configured as key indicators. These key indicators are associated with veto coefficients. When the measurement result of any key indicator is lower than the corresponding baseline threshold, the veto coefficients are used to downgrade or set to zero on the weighted aggregation result.

[0076] In some embodiments of the present invention, when the AI ​​trust comprehensive measurement result or the measurement result of any key indicator does not meet the preset requirements, based on the source analysis results of the measurement results of each dimension and the AI ​​trust comprehensive measurement result, weak data dimensions and weak algorithm dimensions are identified, and corresponding optimization operations are triggered. The optimization operations include performing at least one of the following operations on AI test data: sample augmentation, sample resampling, and data distribution reconstruction; and performing at least one of the following operations on AI models: hyperparameter tuning, network structure adjustment, and model retraining. The optimized AI test data and / or optimized AI model are loaded into the integrated environment, and the trusted evaluation computing engine module performs an evaluation on the optimized AI test data and / or AI model. In a third aspect, the present invention provides an electronic device including a processor and a memory, the processor being configured to execute a computer program stored in the memory to implement the evaluation method for the trusted measurement of power grid dispatch AI.

[0077] Compared with the prior art, the beneficial effects of the present invention are as follows: By constructing an integrated verification environment that includes an AI training and inference environment and a control system simulation environment, this invention provides highly realistic, resource-isolated, and closed-loop operational verification scenarios for power grid dispatching AI. This achieves adaptability from general verification to power grid business scenarios, solving the core challenges of missing verification environments, scattered resources, and the inability to achieve closed-loop operation in existing technologies. Through the design of an integrated management architecture including a configurable middleware chain (data integration and retrieval middleware), the invention enhances the system's unified management, flexible orchestration, and efficient scheduling capabilities for multi-source and heterogeneous power grid dispatching data. This enables rapid organization from data to evaluation cases, significantly improving the automation level of the verification process and its adaptability to different scenarios. By establishing a comprehensive metric library covering data and algorithm credibility and introducing a veto mechanism, along with a standardized evaluation engine, the invention significantly improves the scientific rigor and comprehensiveness of the evaluation, enhancing the ability to detect hidden risks of AI models in key power grid dispatching scenarios and effectively avoiding decision-making errors caused by "black box" models or data bias. The technical solution of this invention effectively solves key technical challenges such as missing verification environments, scattered data resources, and inconsistent evaluation standards, providing a systematic support platform for the quality assurance and reliable implementation of power grid dispatching AI products.

[0078] Example 3, as Figure 5As shown, the present invention also provides an electronic device 100 for implementing an AI-based trusted comprehensive measurement method for power grid dispatch.

[0079] The electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0080] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the evaluation method for the reliability measurement of power grid dispatch AI as described in the first aspect of the present invention by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0081] The memory 101 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0082] At least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 102 may be a microprocessor or any conventional processor. Processor 102 is the control center of electronic device 100, connecting various parts of electronic device 100 via various interfaces and lines.

[0083] The memory 101 in the electronic device 100 stores multiple instructions to implement an AI-based trusted comprehensive measurement method for power grid dispatch, and the processor 102 can execute multiple instructions to achieve the following: Step S1: Construct an integrated environment for AI credibility assessment of power grid dispatching; the integrated environment includes a containerized resource management platform deployed in an AI training and inference environment with resource isolation, and a control system simulation operation environment that interacts with the dispatching mirror system; the control system simulation operation environment receives operation data from the dispatching mirror system and sends the decision results of the AI ​​model back to the dispatching mirror system. Step S2: In response to the received evaluation request for trusted verification of power grid dispatch AI, a unified management system is implemented for heterogeneous power grid dispatch data through an integrated management architecture comprising a data source layer, a middleware layer, and a data application layer. The middleware layer includes a data integration middleware and a data retrieval middleware. The data integration middleware includes a connector middleware, a verification and conversion middleware, and a loading middleware. The connector middleware extracts power grid model, AI model, and AI test sample data from the data source layer. The verification and conversion middleware performs format verification, integrity verification, and conversion processing on the extracted data. The loading middleware loads the verified and converted data into the target data storage. The data retrieval middleware parses the evaluation request and retrieves matching power grid model, AI model, and AI test sample data from the data source layer. The data application layer combines the retrieved power grid model, AI model, and AI test sample data and loads them into the integrated environment described in step S1 as evaluation cases. Step S3: Establish an AI trustworthiness measurement index library for power grid dispatching scenarios and call the computing engine for evaluation; the AI ​​trustworthiness measurement index library includes AI test data trustworthiness index and AI algorithm trustworthiness index, and configure baseline thresholds and weights for each index; the computing engine selects indicators from the index library according to the evaluation case to construct an evaluation workflow, calculates the measurement results of each dimension, and performs weighted aggregation of the measurement results of each dimension based on the weights; when the measurement result of the preset key index is lower than the baseline threshold corresponding to each index, a veto coefficient is used to constrain the weighted aggregation result to obtain the comprehensive AI trustworthiness measurement result; Step S4: Monitor the evaluation process of step S3, and perform source tracing analysis based on the measurement results of each dimension and the comprehensive AI trust measurement results obtained in step S3; when the comprehensive AI trust measurement results or the measurement results of key indicators do not meet the preset requirements, based on the source tracing analysis results, trigger the reconstruction and / or expansion operation of AI test data, and / or the retraining operation of AI model; reload the optimized AI test data and / or AI model into the integrated environment, and execute the evaluation process of step S3 again.

[0084] Example 4 If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0085] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0086] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0087] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0088] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0089] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A comprehensive AI-based reliable measurement method for power grid dispatching, characterized in that, The evaluation process is executed based on a pre-built integrated environment and an AI trustworthy metrics library; wherein, the integrated environment includes a containerized resource management platform deployed in an AI training and inference environment with resource isolation, and a control system simulation operation environment that dynamically interacts with the power grid dispatch mirror system; the AI ​​trustworthy metrics library is configured with baseline thresholds and weights for each metric; the method includes: In response to the evaluation request, the heterogeneous data resources of power grid dispatch are uniformly managed through the integrated management architecture to retrieve and combine power grid models, AI models and AI test sample data that match the evaluation request from the data source layer, and load them into the integrated environment to form evaluation cases; Based on the evaluation case, indicators are selected from the AI ​​trustworthiness measurement indicator library to construct an evaluation workflow, calculate the measurement results of each dimension, and perform weighted aggregation of the measurement results of each dimension based on the weight; when the measurement results of the preset key indicators are lower than their corresponding baseline thresholds, a veto coefficient is used to constrain the weighted aggregation results to obtain the comprehensive AI trustworthiness measurement result. The evaluation process is monitored, and source analysis is performed based on the measurement results of each dimension and the comprehensive AI credibility measurement results. When the measurement results do not meet the preset requirements, optimization operations on the AI ​​test data and / or AI model are triggered based on the source analysis results, and the optimized data and / or model are reloaded into the integrated environment to perform the evaluation steps again.

2. The AI-based trusted comprehensive measurement method for power grid dispatching according to claim 1, characterized in that, The AI ​​trustworthiness metrics library includes: A set of data credibility indicators for AI test data, including at least completeness indicators, timeliness indicators, consistency indicators, accuracy indicators, and distribution rationality indicators, is used to characterize the characteristics of power grid dispatch data in terms of data missingness, time alignment, cross-source consistency, and outlier distribution. A set of algorithm credibility indicators for AI algorithms, which includes at least accuracy indicators, robustness indicators, stability indicators and interpretability indicators, used to characterize the deviation level of AI models under multiple operating conditions, sensitivity to disturbances, performance fluctuations across time periods and interpretability of output results; The AI ​​trustworthy metrics library is configured with corresponding subsets of metrics and corresponding baseline thresholds for different power grid dispatching scenarios.

3. The AI-based trusted comprehensive measurement method for power grid dispatching according to claim 2, characterized in that, The integrated management architecture comprises a data source layer, a middleware layer, and a data application layer. The middleware layer includes a data integration middleware and a data retrieval middleware. The data integration middleware includes a connector middleware, a verification and transformation middleware, and a loading middleware. The connector middleware extracts power grid model, AI model, and AI test sample data from the data source layer. The verification and transformation middleware performs format verification, integrity verification, and transformation processing on the extracted data. The loading middleware loads the verified and transformed data into the target data storage. The data retrieval middleware parses the evaluation request and retrieves matching power grid model, AI model, and AI test sample data from the data source layer. The data application layer combines the retrieved power grid model, AI model, and AI test sample data and loads them into the integrated environment.

4. The AI-based trusted comprehensive measurement method for power grid dispatching according to claim 3, characterized in that, The data source layer is used to store multi-scenario power grid topology models, power grid operation condition models, AI models corresponding to different power grid dispatching business scenarios, as well as AI test samples such as historical operation records, simulation samples, and real-time sampling data. The connector middleware in the data integration middleware is configured to establish data connections with the scheduling business system, scheduling mirror system, external simulation system and file system, and supports at least one of relational database interface, file interface and programming interface; The verification and conversion middleware performs format verification, integrity verification, business rule verification, field renaming, unit conversion, and data type conversion on the data extracted by the connector middleware; The loading middleware writes the data processed by the verification and conversion middleware into the target data storage, and updates the data directory information corresponding to the power grid model, AI model, and AI test samples.

5. An AI-based trusted comprehensive measurement system for power grid dispatching, characterized in that, The system includes: An integrated environment construction module is used to build an integrated environment for AI credibility assessment of power grid dispatching; the integrated environment includes a containerized resource management platform deployed in an AI training and inference environment with resource isolation, and a control system simulation operation environment that interacts with the dispatching mirror system; The assessment management and case organization module is used to respond to assessment requests and manage heterogeneous power grid dispatch data resources in a unified manner through an integrated management architecture. It retrieves and combines power grid models, AI models and AI test sample data that match the assessment request from the data source layer and loads them into the integrated environment to form assessment cases. The trustworthy assessment calculation engine module is used to select indicators from the AI ​​trustworthy measurement indicator library according to the assessment case to construct the assessment workflow, calculate the measurement results of each dimension, and perform weighted aggregation of the measurement results of each dimension based on the weight; when the measurement result of the preset key indicator is lower than its corresponding baseline threshold, the weighted aggregation result is constrained by the veto coefficient to obtain the comprehensive AI trustworthy measurement result. The full-cycle governance and source tracing analysis module is used to monitor the evaluation process and perform source tracing analysis based on the measurement results of each dimension and the AI ​​trustworthy comprehensive measurement results. When the measurement results do not meet the preset requirements, the module triggers optimization operations on the AI ​​test data and / or AI model based on the source tracing analysis results, and reloads the optimized data and / or model into the integrated environment to perform the evaluation steps again.

6. The AI-based trusted comprehensive measurement system for power grid dispatching according to claim 5, characterized in that, The AI ​​trustworthiness metrics library includes: A set of data credibility indicators for AI test data, including at least completeness indicators, timeliness indicators, consistency indicators, accuracy indicators, and distribution rationality indicators, is used to characterize the characteristics of power grid dispatch data in terms of data missingness, time alignment, cross-source consistency, and outlier distribution. A set of algorithm credibility indicators for AI algorithms, which includes at least accuracy indicators, robustness indicators, stability indicators and interpretability indicators, used to characterize the deviation level of AI models under multiple operating conditions, sensitivity to disturbances, performance fluctuations across time periods and interpretability of output results; The AI ​​trustworthy metrics library is configured with corresponding subsets of metrics and corresponding baseline thresholds for different power grid dispatching scenarios.

7. The power grid dispatching AI credibility assessment and verification system according to claim 6, characterized in that, The integrated management architecture comprises a data source layer, a middleware layer, and a data application layer. The middleware layer includes a data integration middleware and a data retrieval middleware. The data integration middleware includes a connector middleware, a verification and transformation middleware, and a loading middleware. The connector middleware extracts power grid model, AI model, and AI test sample data from the data source layer. The verification and transformation middleware performs format verification, integrity verification, and transformation processing on the extracted data. The loading middleware loads the verified and transformed data into the target data storage. The data retrieval middleware parses the evaluation request and retrieves matching power grid model, AI model, and AI test sample data from the data source layer. The data application layer combines the retrieved power grid model, AI model, and AI test sample data and loads them into the integrated environment.

8. The AI-based trusted comprehensive measurement system for power grid dispatching as described in claim 7, characterized in that, The data source layer is used to store multi-scenario power grid topology models, power grid operation condition models, AI models corresponding to different power grid dispatching business scenarios, as well as AI test samples such as historical operation records, simulation samples, and real-time sampling data. The connector middleware in the data integration middleware is configured to establish data connections with the scheduling business system, scheduling mirror system, external simulation system and file system, and supports at least one of relational database interface, file interface and programming interface; The verification and conversion middleware performs format verification, integrity verification, business rule verification, field renaming, unit conversion, and data type conversion on the data extracted by the connector middleware; The loading middleware writes the data processed by the verification and conversion middleware into the target data storage, and updates the data directory information corresponding to the power grid model, AI model, and AI test samples.

9. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the AI-based trusted integrated measurement method for grid dispatch as described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the AI-based trusted comprehensive measurement method for power grid dispatch as described in any one of claims 1 to 4.

Images