Method for evaluating quality of scheduling data and storage medium

Abstract

The application provides a quality evaluation method for scheduling data and a storage medium. The method is applied to a power grid. The method comprises the following steps: obtaining scheduling data of the power grid; the scheduling data comprises first data and second data; the first data comprises topology data and equipment parameter data of the power grid; the second data comprises measurement point data, alarm event data and scheduling operation data of the power grid; performing evaluation processing on the first data and / or the second data according to at least one preset quality index to obtain at least one quality parameter; determining a weight set corresponding to the at least one quality parameter based on a working state of the power grid; the weight set comprises at least one weight factor corresponding to the at least one quality parameter; determining a quality evaluation score of the power grid based on the at least one weight factor and the at least one quality parameter; and the quality evaluation score is used for representing a quality level of the power grid.

Description

Technical Field

[0001] This application relates to the field of power data management technology, specifically to a method for evaluating the quality of dispatch data and a storage medium. Background Technology

[0002] In the process of power grid dispatching and operation, the quality of dispatching data directly affects the accuracy of power grid dispatching decisions and the safety of power grid operation. Therefore, high-quality evaluation of dispatching data is crucial. However, existing power grid dispatching data quality management technologies have significant limitations in the evaluation stage: some technical solutions use single-dimensional rules and thresholds for inspection, resulting in a limited range of indicators that only focus on extreme anomalies, and the static and rigid threshold settings are difficult to adapt to dynamic changes in power grid operation; other technical solutions, while introducing multi-dimensional indicator systems, often use statically preset weights for each dimension, leading to quality evaluation results that are detached from actual business needs, have limited guiding value, and fail to accurately provide reliable data quality references for power grid dispatching, thus failing to meet the power grid dispatching's need for refined and practical data quality evaluation. Summary of the Invention

[0003] This application provides a method and storage medium for evaluating the quality of dispatch data, aiming to improve the accuracy and adaptability of power grid dispatch data quality evaluation.

[0004] Firstly, this application provides a method for evaluating the quality of scheduling data, applied to a power grid, the method comprising: Obtain the dispatch data of the power grid; the dispatch data includes first data and second data; the first data includes the topology data and equipment parameter data of the power grid, and the second data includes the measurement point data, alarm event data and dispatch operation data of the power grid; The first data and / or the second data are evaluated and processed according to at least one preset quality index to obtain at least one quality parameter; The weight set corresponding to the at least one quality parameter is determined based on the operating state of the power grid; the weight set includes at least one weight factor corresponding to the at least one quality parameter. Based on the at least one weighting factor and the at least one quality parameter, a quality evaluation score for the power grid is determined; the quality evaluation score is used to characterize the quality level of the power grid.

[0005] Secondly, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, is used to implement the method described above.

[0006] In the embodiments of this application, by dividing the scheduling data into first data and second data, it is possible to comprehensively cover the basic and dynamic data related to power grid scheduling, ensuring the integrity of the evaluation object; by pre-setting at least one quality indicator to evaluate the scheduling data, it is possible to reflect the data quality status from multiple dimensions, avoiding the limitations of single-dimensional evaluation; by dynamically determining the weight set based on the working status of the power grid, the quality evaluation can adapt to the needs of different operating scenarios, making the evaluation results more in line with actual business needs; by combining weight factors and quality parameters to calculate the quality evaluation score, the multi-dimensional quality evaluation results can be transformed into intuitive and comparable quantitative indicators, facilitating the rapid judgment of the quality level of power grid scheduling data, thereby improving the technical problems in related technologies such as single evaluation indicator dimensions, static and rigid weight settings, and evaluation results deviating from actual business needs. Attached Figure Description

[0007] Figure 1 A flowchart illustrating a method for evaluating the quality of scheduling data provided in an embodiment of this application; Figure 2 A schematic diagram of another method for evaluating the quality of scheduling data provided in an embodiment of this application; Figure 3 A schematic diagram of the structure of a scheduling data quality evaluation system provided in the embodiments of this application; Figure 4 A schematic diagram of the module interaction timing of the scheduling data quality evaluation system provided in the embodiments of this application; Figure 5 This is a schematic diagram of a hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0008] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0009] The technical solutions adopted in the related technologies include: Option 1, based on single-dimensional rules and thresholds, is the most basic and widely used method, primarily embedded in the front-end acquisition and data processing modules of Supervisory Control and Data Acquisition / Energy Management System (SCADA / EMS). Specific methods include: monitoring communication status to check for data channel interruptions and message continuity (the most basic "presence / absence" judgment); setting fixed upper and lower limits for each measurement point (e.g., voltage 220kV ± 10%), with data exceeding these limits marked as "bad data"; determining whether the variation in data between two adjacent sampling points is within a reasonable physical range (e.g., power variation per second cannot exceed a fixed value); and detecting whether data remains unchanged or zero for extended periods (inconsistent with equipment status).

[0010] Option 2, a comprehensive evaluation method based on a multi-dimensional indicator system, is the mainstream direction in current research and application, aiming to comprehensively characterize data quality. It mainly involves establishing data evaluation dimensions, such as completeness, accuracy, timeliness, and consistency indicators. A calculation formula is designed for each dimension to obtain a quantitative score. Then, a weighted average method with fixed weights (e.g., using the analytic hierarchy process to determine weights) is employed to synthesize a comprehensive quality score or grade (excellent, good, average, poor). The power grid dispatch data quality index is derived through this comprehensive scoring method.

[0011] As with Schemes 1 and 2 above, the data evaluation indicators are singular, focusing only on extreme anomalies; static and rigid thresholds are difficult to adapt to all operating modes; and they cannot detect hidden errors that are "bad but not exceeding limits." The weights are mostly statically preset, unable to adapt to the differentiated needs under different power grid operating conditions; they emphasize "evaluation" itself, with insufficient support for "intelligent diagnosis" and "root cause tracing" of problems; and the calculation methods and fusion models for each dimension of indicators are not highly standardized. They are mostly "post-event checks" and "limit-exceeding alarms," ​​lacking pre-event prediction and early warning capabilities. They can mostly only answer "where it's broken" and "which dimension is bad," failing to automatically and deeply answer "why it's broken" and "what the root cause is." Tracing work heavily relies on human experience. Evaluation standards, thresholds, and weights are mostly statically configured, lacking the intelligence to adaptively adjust according to operating scenarios and business importance. Data quality analysis is often independent of equipment management, communication management, and maintenance logs, resulting in fragmented information and difficulty in forming integrated analysis. Evaluation results are disconnected from governance actions, and the system cannot learn from historical cases, lacking self-optimization capabilities.

[0012] To facilitate understanding of this application, some terms will be explained below: Diagnostic Early Warning: A proactive and intelligent fault prevention and handling capability based on artificial intelligence, integrating "trend prediction, early warning, and automatic root cause analysis." Its goal is to transform data quality management from a passive "post-event firefighting" model to a proactive "pre-event prevention" and "precise policy implementation" model. Intelligent Analysis: A specific concept referring to the process by which a system, after discovering data quality problems (through early warning or routine monitoring), uses artificial intelligence technology to automatically and deeply explore the root causes of the problems and conduct correlation impact assessments. It allows machines to replace the human brain, completing the complex reasoning work from "phenomenon perception" to "cause understanding," far exceeding simple statistics and alarms. Analytic Hierarchy Process (AHP): A multi-criteria decision-making method that decomposes a complex decision problem into levels such as objectives, criteria, and solutions, and then determines the relative importance of each element through pairwise comparisons and mathematical calculations, ultimately providing a quantitative basis for decision-making.

[0013] Knowledge Graph: A technology that uses a graph structure to organize and represent knowledge of the real world. It consists of entities, relationships, and attributes, aiming to enable machines to understand and process complex relationships between things. Common Information Model / Extensible Markup Language (CIM / E) Standard Protocol: Part of the IEC 61970 series of international standards. It can be understood as a "common language" or "universal blueprint" for data exchange and model sharing in the power system field, particularly among energy management systems. Microservice Architecture: A construction method that breaks down large, complex software systems into a set of small, independent, loosely coupled services. Each service is built around a specific business capability and can be independently developed, deployed, scaled, and maintained. Information Technology Service Management (ITSM): A set of process-oriented, customer-centric methodologies and best practices for designing, delivering, managing, and improving how organizations utilize information technology (IT). An ITSM system is the "nervous system" and "command system" connecting the "intelligent analytics brain" and the "physical execution limbs." Q-Score sequence: refers to a set of ordered data formed by arranging comprehensive quality scores in chronological order. It is a single, quantifiable score obtained by integrating multiple dimensions such as "completeness, accuracy, and consistency" through a dynamic weighting model.

[0014] On the one hand, this embodiment provides a method for evaluating the quality of scheduling data, which is applied to the power grid. Figure 1 A flowchart illustrating a method for evaluating the quality of scheduling data provided in an embodiment of this application is shown below. Figure 1As shown, the method includes: Step 101: Obtain the power grid dispatch data; the dispatch data includes first data and second data; the first data includes the power grid topology data and equipment parameter data, and the second data includes the power grid measurement point data, alarm event data, and dispatch operation data.

[0015] The dispatch data can be various types of data generated and associated during the power grid dispatching process. The first type of data can be data characterizing the power grid infrastructure and the inherent attributes of equipment; among these, topology data describes the connection relationships between various devices in the power grid, and equipment parameter data represents the inherent technical parameters of various devices in the power grid. The second type of data can be dynamic data generated or recorded in real time during power grid operation; among these, monitoring point data reflects the operating status of equipment collected by various monitoring points in the power grid, alarm event data records relevant data on abnormal situations during power grid operation, and dispatch operation data records the operations performed by dispatchers for power grid operation.

[0016] In this embodiment, historical scheduling data can be accessed by developing a power grid dispatching data interface and using a standard protocol or data service interface. The standard protocol can be the CIM / E standard protocol, and the data interface can support various time-series database storage (e.g., MySQL / DM). Real-time scheduling data can also be obtained by subscribing to real-time event streams through message queues (e.g., Kafka / RabbitMQ).

[0017] Step 102: Evaluate the first data and / or the second data according to at least one preset quality index to obtain at least one quality parameter.

[0018] Among them, the quality indicators can be pre-defined specific dimensions used to evaluate the quality of scheduling data, and there must be at least one of them, such as completeness, accuracy, consistency, timeliness, reliability, uniqueness, etc. The quality parameters can be the quantitative results obtained after evaluating and processing the scheduling data through the quality indicators, and they correspond one-to-one with the quality indicators.

[0019] In this embodiment, the number of quality indicators can be set according to actual evaluation needs. The evaluation process can adopt a microservice architecture, treating each quality indicator as an independent and scalable service, and using a stream processing framework to perform real-time indicator window calculations or batch distributed processing (e.g., MapReduce or Spark) to obtain quality parameters.

[0020] For example, the quality parameters corresponding to completeness can be calculated by the missing rate. The missing rate can be calculated by dividing the difference between the theoretical number of points and the actual number of points by the theoretical number of points. A sliding window can be used for the calculation, and the window size can be configured. Accuracy is calculated through power imbalance calculation (e.g., physical verification), multi-source data consistency check (e.g., redundancy comparison), and deviation between estimated and measured values ​​(e.g., state estimation error analysis), realizing a topology-aware computing engine; consistency is obtained through switch state-power flow matching verification (e.g., logical rule check), data change smoothness analysis (e.g., time series consistency), and data correlation coefficient calculation (e.g., correlation consistency), combined with a rule engine and statistical methods; timeliness is calculated by the time difference between data entry and collection (delay) and the difference between data and system reference time scale (e.g., time scale deviation), using high-precision clock synchronization and nanosecond-level time recording to ensure accuracy; reliability is calculated by continuous availability duration, jump detection using the Cumulative Sum Control Method (CUSUM) algorithm, and credibility scoring according to data traceability level, using a real-time time series anomaly detection algorithm; uniqueness is calculated by hash value comparison of duplicate records and measurement point identifier (ID) conflict detection, constructing a Bloom filter for fast deduplication.

[0021] Step 103: Determine the weight set corresponding to at least one quality parameter based on the working state of the power grid; the weight set includes at least one weight factor corresponding to at least one quality parameter.

[0022] The power grid's operating status refers to its current operational scenario. The weight set is a collection of weighting factors corresponding to quality parameters, where each weighting factor represents the importance of its corresponding quality parameter in the comprehensive evaluation. The current operating status of the power grid can be automatically determined by analyzing real-time alarms, dispatch instructions, and operation tickets. Alternatively, it can be determined by the user manually switching the system interface mode, thereby dynamically switching the weight set from a pre-defined strategy library.

[0023] In some embodiments, the operating state includes an operational state, a fault state, and an analysis state; the weight set includes a first weight set, a second weight set, and a third weight set; determining the weight set corresponding to at least one quality parameter based on the operating state of the power grid includes: When the power grid is in operation, at least one quality parameter is determined to correspond to a first weight set; When the power grid is in a fault state, at least one quality parameter is determined to correspond to a second weight set; the value of the weight factor in the second weight set is different from the value of the weight factor in the first weight set; When the power grid is in the analysis state, at least one quality parameter is determined to correspond to a third weight set; the value of the weight factor in the third weight set is different from the value of the weight factor in the second weight set; the value of the weight factor in the third weight set is different from the value of the weight factor in the first weight set.

[0024] The operational status can be described as the power grid operating under standard and stable conditions, without any emergencies, and with dispatchers conducting routine monitoring, such as normal monitoring status. The fault status can be described as the power grid experiencing or about to experience a fault, and the system is in an emergency response state, such as fault handling status. The analysis status can be described as the power grid's operational purpose focusing on post-event statistics, analysis, report generation, or model training, such as historical statistics status.

[0025] In this embodiment, when the power grid is in operation, the power grid topology is in normal operating mode, there are no special wiring methods caused by large-scale maintenance operations, and there are no real-time alarms of accidents or faults affecting the operation of the power grid in the dispatching main control alarm window. At this time, the system determines that the quality parameters correspond to the first weight set. When the power grid is in a fault state, it may receive a total fault signal or protection action trip signal from the SCADA system, the system may detect a sudden change in key indicators exceeding the limit, or the user may switch the system interface to the fault handling, fault analysis, or power supply emergency mode. At this time, the system determines that the quality parameters correspond to the second weight set. When the power grid is in the analysis state, it may be that the user opens the daily / monthly report statistics, historical data query, or model training platform, or the system's background timed statistical task starts. At this time, the system determines that the quality parameters correspond to the third weight set.

[0026] As an example, suppose the quality indicators include completeness, accuracy, timeliness, consistency, and reliability. In the first weight set, the weight factor for completeness can be 0.2, accuracy 0.25, timeliness 0.15, consistency 0.2, and reliability 0.2. In the second weight set, to adapt to emergency response needs, the weight factor for accuracy can be adjusted to 0.35, timeliness to 0.25, completeness to 0.1, consistency to 0.2, and reliability to 0.1. Compared to the first weight set, the weight factors for completeness, accuracy, and timeliness are all different. In the third weight set, the weight factor for completeness can be adjusted to 0.3, accuracy 0.25, timeliness 0.1, consistency 0.2, and reliability 0.15. Compared to the first weight set, the weight factors for completeness, timeliness, and reliability are different, and compared to the second weight set, the weight factors for completeness, accuracy, timeliness, and reliability are also different.

[0027] In the embodiments of this application, by subdividing the working state of the power grid into operating state, fault state and analysis state, and configuring a corresponding first weight set, second weight set and third weight set for each state, and with at least one weight factor different between each weight set, the weight allocation of quality parameters can be dynamically adjusted according to the differentiated needs of different business scenarios of the power grid, so that the quality evaluation score can closely match the business needs of different scenarios such as operation monitoring, fault handling and post-event analysis.

[0028] In some embodiments, the quality indicators include timeliness indicators, reliability indicators, integrity indicators, and accuracy indicators; when the power grid is in a fault state, after determining that at least one quality parameter corresponds to a second weight set, the method further includes: In response to a fault signal from the power grid, the weighting factor corresponding to the timeliness index is adjusted upward, and the weighting factor corresponding to the reliability index is also adjusted upward. In response to the operation instructions to the power grid, the weighting factor of the integrity index is increased, and the weighting factor of the accuracy index is also increased.

[0029] Among them, fault signals can be prompts generated and sent by the system when a fault or abnormal situation occurs in the power grid, used to inform relevant systems or personnel of the current fault status of the power grid; operation instructions can be control or operation commands issued by dispatchers or systems for business scenarios such as power grid fault handling, fault analysis, and equipment operation and maintenance. An upward adjustment operation can be an operation that increases the value of the weighting factor corresponding to the quality parameter, used to enhance the importance of that quality indicator in the comprehensive quality evaluation.

[0030] In this embodiment, when the SCADA system issues a general fault signal, a protection trip signal, or a signal generated by a sudden change in a key power grid indicator (e.g., voltage, frequency) exceeding its limit, the power grid is in an emergency response phase, requiring an increase in the weighting factors corresponding to the timeliness and reliability indicators. When dispatchers issue fault analysis commands, fault inversion commands, equipment maintenance control commands, etc., fault troubleshooting and analysis require complete and accurate data support, necessitating an increase in the weighting factors corresponding to the completeness and accuracy indicators.

[0031] In the embodiments of this application, by adjusting the weight factors of timeliness and reliability indicators in response to the fault signal of the power grid and adjusting the weight factors of integrity and accuracy indicators in response to the operation instructions of the power grid, the weights of quality evaluation weights can be accurately and dynamically adjusted in fault scenarios, ensuring that the focus of quality evaluation is highly matched with the needs of different business links such as emergency fault handling and fault analysis.

[0032] Step 104: Determine the power grid quality evaluation score based on at least one weighting factor and at least one quality parameter; the quality evaluation score is used to characterize the quality level of the power grid.

[0033] The quality evaluation score can be a single quantitative score calculated based on weighting factors and quality parameters, used to characterize the quality level of power grid dispatch data, such as excellent, good, qualified, warning, and severe. A weighted aggregation model can be used, incorporating nonlinear adjustment for calculation. For example, each quality parameter is multiplied by its corresponding weighting factor, and the result is summed to obtain the quality evaluation score, denoted as Q-Score. This score can range from 0 to 100 points and can be mapped to different quality levels. For example, 90 to 100 points correspond to excellent, 75 to 90 points to good, 60 to 75 points to qualified, 40 to 60 points to warning, and 0 to 40 points to severe, corresponding to five colors: green, blue, yellow, orange, and red.

[0034] In some embodiments, the method further includes: Acquire historical sample data of the power grid; historical sample data includes historical quality evaluation scores and historical transmission information; Based on historical quality evaluation scores, historical transmission information, and a pre-set target model, the predicted quality evaluation score of the power grid is determined. If the predicted quality evaluation score meets the preset conditions, an early warning message is generated; the early warning message is used to characterize the risk of power grid failure in the future.

[0035] The historical quality evaluation score can be the power grid dispatch data quality evaluation score calculated using the method described in this application over a past period, which can form a continuous score sequence when arranged chronologically. Historical transmission information can be relevant data generated during past power grid data transmission processes, such as communication error rate, network load rate, data transmission delay, and channel stability parameters. The target model can be an algorithmic model used to predict data quality trends, such as a time series model, capable of mining patterns of data quality changes based on historical data. The predicted quality evaluation score can be a quantitative evaluation score obtained by predicting the power grid dispatch data quality for future periods using the target model.

[0036] The preset conditions can be pre-defined threshold conditions used to determine whether an early warning needs to be generated. For example, the predicted quality evaluation score may consistently fall below a preset threshold, such as 70 points; or the predicted score may decrease by more than a set percentage. The early warning information can be a notification to indicate potential data quality failure risks in the power grid in the future, clearly informing maintenance personnel of the relevant information regarding potential risks.

[0037] In this embodiment, the target model can be trained using historical sample data. During training, historical quality assessment score sequences and historical transmission information can be used as joint input features, with the historical quality assessment scores corresponding to subsequent time periods as output labels. The model parameters are adjusted through multiple iterations until the model converges. After training, the model can output predicted quality assessment scores for a specific future time period (e.g., the past 24 hours) by inputting historical quality assessment scores for the most recent period and relevant real-time transmission status data. The target model can be selected based on the actual scenario requirements, such as choosing a Long Short-Term Memory (LSTM) model or a Transformer model, which are suitable for time series prediction. As an example, a yellow trend warning is triggered when the predicted quality assessment score continues to fall below a preset threshold in subsequent time periods (e.g., the predicted quality assessment score is 65 points, but the preset threshold is 70 points).

[0038] In the embodiments of this application, by acquiring historical sample data of the power grid, a predicted quality evaluation score is determined based on historical quality evaluation scores, historical transmission information and a preset target model. When the predicted quality evaluation score meets preset conditions, an early warning information is generated. This can realize early warning of the risk of power grid dispatch data quality failure, allowing operation and maintenance personnel to obtain valuable advance notice of handling and take preventive measures before data quality problems affect power grid operations.

[0039] In some embodiments, the method further includes: If the quality evaluation score or predicted quality evaluation score is less than or equal to a preset threshold, an alarm message is generated; the alarm message includes attribute information of the power grid fault. Based on a pre-defined data knowledge graph, attribute information is transformed into event nodes; By associating event nodes with entity nodes in the data knowledge graph, the association relationships can be obtained. Starting from an event node, a multi-hop traversal is performed based on the relationship and data knowledge graph to obtain at least one root cause data. Probabilistic data for determining root cause data is determined based on a pre-defined Bayesian network. Diagnostic reports for power grid failures are generated based on probability and root cause data; these reports are used to assess the attributable causes of the failures.

[0040] The preset threshold can be a pre-defined quality evaluation score threshold used to determine whether to generate an alarm message. It can be flexibly configured according to power grid business needs and data quality standards. As an example, the preset threshold for the quality evaluation score can be greater than the preset threshold for the predicted quality evaluation score. The alarm message can be a notification indicating the existence of data quality faults or potential fault risks in the power grid. The attribute information can be data describing fault-related characteristics, such as alarm identifier, occurrence time, fault type, severity level, affected objects, and fault description.

[0041] A data knowledge graph can be a graph-structured data used to organize and represent knowledge related to power grid dispatching, including entity nodes and the relationships between them. Entity nodes can be various objects in the power grid, such as equipment, measuring points, communication channels, servers, and applications. Relationships can be the inherent connections between entity nodes, such as physical connections, data flow relationships, and logical dependencies. Event nodes can be dynamic nodes in the knowledge graph obtained by transforming the attribute information of alarm information, used to represent single fault events. Relationships can be the mapping relationships established between event nodes and entity nodes in the data knowledge graph, such as the association between event nodes and affected equipment, measuring points, and other entity nodes.

[0042] In this embodiment, the data knowledge graph can be constructed by collecting node data such as devices, measurement points, communication channels, servers, and applications, as well as the physical connections, data flows, and logical dependencies between these nodes. The transformation process can use the attribute information of alarm information as attributes of event nodes, creating independent event nodes in the knowledge graph. When associating event nodes with entity nodes in the data knowledge graph, the affected object in the alarm information can be used as the matching key to retrieve the corresponding entity node in the data knowledge graph.

[0043] As an example, the physical topology of the power grid is formed through electrical connections of equipment. Information such as the signal source device, board slot, and engineering unit of each measuring point is obtained to establish data acquisition relationships. Information on the access devices and ports at both ends of the communication channel, as well as the server's Internet Protocol (IP), network area, and maintenance team information, is obtained to establish the association between the communication channel and the data system. Through documentation or configuration, a list of input measuring points for key business applications (e.g., state estimation) is determined, establishing the dependency relationship between the server and the data system. Based on the above fundamental data, knowledge is constructed and integrated to form a knowledge graph of power grid dispatch data.

[0044] Multi-hop traversal can be a process of exploring the relationships between entity nodes in a data knowledge graph layer by layer, starting from an event node, to uncover deeply related entities associated with a fault event. In this embodiment, the traversal process can start from the entity nodes associated with the event node and explore other associated entity nodes layer by layer. Root cause data can be entities and related information that may lead to power grid faults, obtained through multi-hop traversal.

[0045] A Bayesian network can be a graphical model based on probabilistic reasoning to handle uncertainty problems, capable of calculating the probability of each root cause data leading to a failure based on known information. The probabilistic data can be the posterior probabilities corresponding to each root cause data calculated by the Bayesian network, used to characterize the credibility of the root cause. In this embodiment, the Bayesian network can be pre-trained using historical failure data; during the calculation process, the prior probability of each root cause data is first obtained, denoted as P(Ci), then the likelihood probability corresponding to each root cause data is calculated, denoted as P(E|Ci), and finally, according to Bayes' theorem P(Ci|E)∝P(E|Ci)×P(Ci), the posterior probability of each root cause data is calculated. Here, the prior probability P(Ci) represents the prior probability of node Ci failing, which can be statistically obtained from historical failure data; the likelihood probability P(E|Ci) is the probability of observing the current alarm event E given that Ci has failed.

[0046] The diagnostic report can be a structured report generated based on probability data and root cause data. It is used to assess the attributable causes of power grid failures and provide a basis for operation and maintenance decisions. Attributable causes can be the root causes leading to power grid failures, including specific fault entities, fault manifestations, and logical connections. In this embodiment, the report may include a list of root causes sorted from high to low probability, supporting evidence for each root cause, the scope of the fault's impact, and troubleshooting recommendations. It should be noted that, based on probability sorting, typically only the top 3-5 most likely causes are output, and causes with probabilities exceeding a certain threshold (e.g., probability greater than 0.5) are highlighted.

[0047] As an example, the alarm ID, time, type, level, object, and description attributes in the alarm message are transformed into event nodes in a knowledge graph. By associating the alarm object with object entities in the knowledge graph, a relationship is established between the event and the entity. More entities and relationships are extracted from the alarm description; for example, "caused by communication channel interruption," the "communication channel" entity can be extracted, and a dependency relationship can be established. Starting from this event node, a multi-hop traversal is performed along the generation and dependency relationships to find all possible sources of influence. A Bayesian network is then used to calculate the probability of data problems, ultimately outputting a data diagnostic report.

[0048] In the embodiments of this application, alarm information containing fault attribute information is generated when the quality evaluation score or predicted quality evaluation score is less than or equal to a preset threshold. The attribute information is converted into event nodes and associated with entity nodes in the data knowledge graph. Root cause data is obtained by multi-hop traversal starting from the event nodes. Then, the probability data of the root cause data is determined based on a Bayesian network and a diagnostic report is generated. This enables automated and in-depth analysis of fault root causes, accurately locating the attribution cause of the fault without relying heavily on human experience, and significantly improving the efficiency and accuracy of root cause location.

[0049] In some embodiments, before converting attribute information into event nodes based on a preset data knowledge graph, the method further includes: Acquire at least two entity nodes in the power grid dispatching scenario; entity nodes include at least equipment nodes, measurement point nodes, communication channel nodes, server nodes, and application terminal nodes in the power grid; A data knowledge graph is constructed based on the mapping relationship between at least two entity nodes; the mapping relationship includes physical connection relationship, data flow relationship and logical dependency relationship.

[0050] Among them, equipment nodes can represent physical operating equipment of the power grid. Measurement point nodes can represent the smallest unit of power grid data acquisition. Communication channel nodes can represent power grid data transmission links. Server nodes can represent power grid data storage and processing equipment. Application terminal nodes can represent power grid business software system modules.

[0051] Mapping relationships can be the inherent connections between entity nodes, including physical connection relationships, data flow relationships, and logical dependencies. Physical connection relationships can be the actual connection between the physical objects corresponding to entity nodes, such as electrical connections between devices or installation connections between devices and measurement points. Data flow relationships can be the connections for data transmission and transfer between entity nodes, such as the flow path of data collected by measurement points being transmitted to the server through a communication channel. Logical dependencies can be the dependency relationships between entity nodes based on business logic or functional implementation, such as the dependency of application terminals on the server's storage or computing power, or the sequential dependency of different stages in the data processing flow.

[0052] In the embodiments of this application, by obtaining at least two entity nodes, such as equipment nodes and measurement point nodes, in the power grid dispatch scenario, sorting out the physical connection relationship, data flow relationship and logical dependency relationship between entity nodes and constructing a data knowledge graph, it is possible to realize the structured and visual presentation of power grid dispatch-related entities and their relationships, forming a complete "entity-relationship" knowledge system.

[0053] In some embodiments, multi-hop traversal is performed according to any of the following priority rules: priority of the relationship type between at least two entity nodes; priority of the node type of at least two entity nodes; priority of the state type of at least two entity nodes; priority of the failure frequency of at least two entity nodes.

[0054] Among these, priority rules serve as the standard for determining the traversal order during multi-hop traversal. Relationship type priority can be a priority order set based on the nature of the associations between entity nodes, used to determine the exploration priority of different associations during traversal. Node type priority can be a priority order set based on the category of the entity node itself, used to determine the exploration priority of different types of entity nodes during traversal. State type priority can be a priority order set based on the current running state of the entity node, used to determine the exploration priority of entity nodes in different running states during traversal. Fault frequency priority can be a priority order set based on the frequency of historical faults of entity nodes, used to determine the exploration priority of entity nodes with different fault probabilities during traversal.

[0055] In this embodiment, when selecting the priority of relation types for traversal, the importance of the relationship in the root cause analysis can be ranked according to the relationship. For example, direct causal relationships have a higher priority than strong dependencies, strong dependencies have a higher priority than weak dependencies, and weak dependencies have a higher priority than general relationships. Direct causal relationships can be relationships where the state of one entity node directly causes the state change of another entity node. Strong dependencies can be business relationships such as data flow. Weak dependencies can be logical dependencies. General relationships can be basic relationships such as physical connections.

[0056] When selecting the priority of node types for traversal, the nodes can be sorted according to the degree of correlation between the physical nodes and data quality issues. For example, physical device nodes have a higher priority than communication channel nodes, communication channel nodes have a higher priority than server nodes, server nodes have a higher priority than application terminal nodes, and application terminal nodes have a higher priority than measurement point nodes. Among them, physical device nodes can be nodes corresponding to equipment that directly participates in the operation of the power grid, such as transformers and remote terminal units (RTUs), while communication channel nodes can be nodes corresponding to data transmission carriers such as fiber optic links.

[0057] When traversing by selecting the priority of the status type, the entity nodes can be sorted according to the health of their current running status. For example, entity nodes with abnormal status have higher priority than entity nodes with normal status. Abnormal status can be due to issues such as excessive running parameters or signal interruption, while normal status can be due to running parameters meeting standards and no abnormal alarms.

[0058] When selecting the priority of fault frequency for traversal, the order can be sorted according to the statistical results of the historical fault records of the entity nodes. For example, entity nodes with frequent historical faults have a higher priority than entity nodes with fewer historical faults, and entity nodes with fewer historical faults have a higher priority than entity nodes with no historical fault records. Among them, frequent historical faults can mean that the number of faults of the entity node in the past period of time exceeds a preset number, and no historical fault records can mean that the entity node has not experienced any faults since it was put into use.

[0059] In the embodiments of this application, by providing four priority rules—relationship type, node type, state type, and fault frequency—any rule can be selected for execution during multi-hop traversal, thereby achieving directional and efficient traversal.

[0060] In some embodiments, after generating a diagnostic report of a power grid fault based on probability data and root cause data, the method further includes: Output the initial work order based on the diagnostic report; Receive the completed work order corresponding to the initial work order; Case data of power grid failures are determined based on initial work orders and completed work orders; Optimize the data knowledge graph, target model, and preset conditions based on case data.

[0061] The initial work order can be a standardized governance task document automatically generated based on the diagnostic report, used to guide maintenance personnel in carrying out fault handling work. The completion work order can be a document recording the entire process and results of task completion after maintenance personnel have implemented governance measures. Case data can be structured learning materials formed by integrating initial work orders, completion work orders, and fault diagnosis-related data, which can fully present the entire process of fault discovery, diagnosis, handling, and verification.

[0062] In this embodiment, based on the initial work hours output from the diagnostic report, the system can automatically call the ITSM system API to extract key fields from the diagnostic report, including problem summary, root cause analysis, scope of impact, urgency, handling recommendations, responsible team, etc., to generate a standardized initial work order. The content of the initial work order may include work order title, details, type, priority, assigned object, inspection items, etc., ensuring that maintenance personnel can clearly obtain the key information needed for fault handling.

[0063] When receiving the completed work order corresponding to the initial work order, the maintenance personnel will fill in information such as the processing procedure, tools or methods used, fault resolution status, and data quality verification results in the work order management system after performing governance measures on-site or remote operations, thus forming a completed work order and submitting it. The system will receive the completed work order in real time through the work order management module.

[0064] When determining case data for power grid faults based on initial and completed work orders, it is necessary to integrate information from both and supplement it with relevant data from the fault diagnosis process. This includes, for example, quality parameters at the time of the fault, root cause data and probability data, knowledge graph association paths, and early warning trigger records, ultimately forming structured case data. Case data can include dimensions such as problem snapshots, diagnostic records, processing records, result verification, and case tags, ensuring comprehensive material support for subsequent system optimization.

[0065] In the embodiments of this application, by outputting an initial work order based on the diagnostic report, receiving and integrating the completed work order to determine the case data, and then using the case data to optimize the data knowledge graph, target model, and preset conditions, the system's capabilities can be continuously self-evolved, making the association relationship of the data knowledge graph more accurate, the prediction effect of the target model better, and the triggering logic of the preset conditions more reasonable.

[0066] In some embodiments, optimizing the data knowledge graph, target model, and preset conditions based on case data includes: Case data is identified as positive samples, and scheduling data without faults is identified as negative samples; The target model is trained based on positive and negative samples to obtain the optimized target model; Extract new faults and their corresponding root cause data from the case data; Based on the newly added faults and their corresponding root cause data, optimize the probability parameters of the Bayesian network and the priority rules for multi-hop traversal. The strength of the mapping relationship between at least two entity nodes in the data knowledge graph is updated based on case data.

[0067] Positive samples can be fault-related data used to train the target model, directly determined from the case data. Negative samples can be fault-free data used to train the target model, determined from grid dispatch data without faults. The optimized target model can be a model whose prediction accuracy and generalization ability are improved after training with both positive and negative samples. New faults can be fault types that appear for the first time in the case data or are not covered by the original rules of the system. Root cause data can be the entities and related information that cause the faults, corresponding to the new faults. The probability parameters of the Bayesian network can be the key parameters used in the Bayesian network to calculate the root cause probability, including prior probability and likelihood probability.

[0068] In this embodiment, when case data is determined as positive samples, fault characteristics in the case data (e.g., trends in quality parameter changes, associated transmission information, fault occurrence time periods, etc.) can all be used as features of positive samples. Dispatch data without faults can be dispatch data collected when the power grid is in normal operation, whose corresponding quality evaluation scores are stable at the excellent or good level, with no alarm records, and can be used as negative samples. When training the target model based on positive and negative samples, the positive and negative samples can be divided into training and validation sets according to a preset ratio (e.g., 7:3). The training set is used to adjust model parameters, and the validation set is used to evaluate model performance. During the training process, the gradient descent algorithm can be used to optimize the model loss function until the model converges, resulting in the optimized target model. For example, the optimized LSTM model can more accurately capture the trend of data quality degradation and reduce false alarms and false negatives.

[0069] When extracting new faults and their corresponding root cause data from case data, unrecorded fault types (e.g., loose interface faults of a new type of communication module) can be identified by comparing the case data with the system's existing fault type library and root cause library. Root cause data such as the root cause entity, fault manifestation, and associated conditions corresponding to the new fault can then be extracted. When optimizing the probability parameters of the Bayesian network based on the new faults and their corresponding root cause data, the prior probabilities of the corresponding root cause entities in the Bayesian network can be updated, while supplementing or correcting the likelihood probabilities. When optimizing the priority rules for multi-hop traversal, the corresponding priorities can be adjusted according to the root cause type of the new fault. For example, if many new faults are caused by communication module failures, the ranking of communication channel nodes in the node type priority can be increased, or the weight of direct causal relationships in the relation type priority can be increased.

[0070] When updating the strength of the mapping relationship between at least two entity nodes in the data knowledge graph based on case data, the mapping relationship strength parameter can be adjusted according to the associated path and impact of the fault in the case. For example, if the data flow relationship between communication channel A and measurement point B is the critical path for fault propagation in a certain case, the strength of the data flow relationship between the two can be enhanced; if a case shows that the logical dependency relationship between device node C and server node D has a small impact on the fault, the strength of the logical dependency relationship can be appropriately reduced.

[0071] In the embodiments of this application, by determining case data as positive samples and fault-free scheduling data as negative samples to train the target model, extracting new faults and corresponding root cause data to optimize the probability parameters of the Bayesian network and the priority rules of multi-hop traversal, and updating the strength of entity node mapping relationships in the knowledge graph based on case data, it is possible to improve the prediction accuracy of the target model, optimize the accuracy of Bayesian network root cause calculation, enhance the efficiency of multi-hop traversal, and improve the accuracy of knowledge graph association relationships.

[0072] In this embodiment, the self-learning process also includes optimizing the governance strategy library. Specifically, this can involve: analyzing historical effective cases, statistically analyzing the efficiency of different work groups in handling various equipment failures, and if it is found that the average resolution time for a certain type of equipment failure assigned to a specific work group (e.g., Automation Work Group 1) is the shortest, then strengthening the assignment rules for that type of failure and the corresponding work group to improve the accuracy of work order assignment; at the same time, extracting the most effective handling steps in the handling of a certain type of failure, solidifying them into a standard checklist, and integrating them into the governance strategy library to provide standardized handling suggestions for the handling of similar failures in the future, further optimizing the efficiency and effectiveness of failure handling.

[0073] Furthermore, when updating the knowledge graph based on case data, the relevant attributes of each entity node can be updated according to the fault handling measures and results. For example, the health status, last maintenance time, and inherent risk level of the equipment node can be updated to enrich the entity attribute information of the knowledge graph and further improve the association logic and data dimensions of the knowledge graph.

[0074] The following describes the method for evaluating the quality of dispatch data provided in the embodiments of this application. As an example, the method for evaluating the quality of dispatch data can be a power grid dispatch data quality evaluation method with diagnostic early warning and intelligent analysis capabilities. This application belongs to the technical field of power system informatization, big data analysis, and artificial intelligence, specifically involving a data quality intelligent management technology solution applied to power grid dispatch control, integrating real-time monitoring, quantitative evaluation, intelligent diagnosis, trend early warning, and root cause analysis. It aims to overcome the lag and shallow analysis defects of related technologies, providing a power grid dispatch data quality evaluation solution with diagnostic early warning and intelligent analysis capabilities. It achieves proactive early warning by predicting quality degradation trends based on historical and real-time data, and uses machine learning and graph computing technologies to deeply mine the complex correlations and root causes of quality problems for intelligent diagnosis. Through case learning, it continuously optimizes the early warning model and diagnostic rules, forming a closed-loop intelligent knowledge automation, and providing a panoramic visualization view from macro-quality status to micro-root cause links.

[0075] This application establishes a systematic evaluation index system, defining a power grid data quality evaluation system covering six dimensions: completeness, accuracy, consistency, timeliness, reliability, and uniqueness. Each dimension has precisely calculable sub-indicators, achieving a "panoramic scan" of data quality. Combined with context-aware dynamic weight allocation, the weight of each dimension in the comprehensive score can be dynamically adjusted based on the real-time power grid operating status (normal, maintenance, post-accident), the critical level of data services (e.g., key section data related to power grid stability), and the evaluation purpose (for monitoring or historical statistics). For example, during accident handling, the weight of "timeliness" and "accuracy" is significantly increased. This overcomes the limitations of a single, static evaluation standard, enabling the quality score results to comprehensively reflect the health status of the data while closely aligning with the business needs of different scenarios, making the evaluation results more instructive.

[0076] The power grid dispatch data quality evaluation method, which has diagnostic, early warning and intelligent analysis capabilities, realizes the quality evaluation and quality trend prediction of power grid dispatch data from data access, quality index calculation, dynamic weight allocation and comprehensive scoring. It also pushes alarms and performs root cause analysis on data with quality problems in advance, and finally forms a knowledge graph of power grid dispatch data quality evaluation.

[0077] Figure 2 This is another flowchart illustrating the method for evaluating the quality of scheduling data provided in the embodiments of this application, such as... Figure 2 As shown, the method specifically includes: after system startup, data access and preprocessing are performed, followed by calculation of multi-dimensional quality indicators. After the multi-dimensional quality indicators are calculated, the process is divided into two branch paths: one is the conventional monitoring path, which sequentially performs dynamic weight allocation and comprehensive scoring, quality level determination, real-time visualization updates, and finally ends the single processing; the other is the intelligent analysis path, which first performs quality trend prediction to determine whether an early warning has been triggered. If not, it proceeds to continuous monitoring; if it has been triggered, it issues a graded early warning and initiates intelligent diagnosis, followed by knowledge graph reasoning analysis, root cause localization and impact assessment, generation of diagnostic reports, creation of governance work orders, governance execution and verification, case entry into the database and model optimization, and finally, it proceeds to continuous monitoring.

[0078] Figure 3 A schematic diagram of the structure of a scheduling data quality evaluation system provided in the embodiments of this application, such as... Figure 3 As shown, a power grid dispatch data quality evaluation method with diagnostic, early warning, and intelligent analysis capabilities is applied to a dispatch data quality evaluation system. This system aims to transform complex backend intelligent algorithms and analysis results into an operational platform that dispatch operation and data maintenance personnel can intuitively perceive, conveniently interact with, and efficiently make decisions through a series of meticulously designed user interfaces. These interfaces are not only information display windows but also command centers driving the entire closed-loop workflow of "monitoring, early warning, diagnosis, and governance."

[0079] The panoramic intelligent cockpit, this interface is the system's "homepage," providing system users and maintenance personnel with a clear overview of the overall network data health. It uses striking colors to display various data quality indicators by level, such as the current total number of data alarms and today's data quality anomaly rate. It displays a 24-hour change curve of the network-wide scheduling data quality score, and allows quick switching to view trends in data quality changes for important data, key sections, and important categories.

[0080] The intelligent early warning and event center displays a timeline-driven list of warnings. Warnings and alarms are arranged chronologically, but a key improvement is that each warning is clearly labeled "predictive" or "real-time," and includes a preview of the predicted quality degradation curve. Color coding (yellow, orange, red) distinguishes severity levels. It supports combined filtering by region, plant / site, equipment type, quality issue dimension (such as "accuracy degradation"), and warning level, allowing for quick focus on key concerns.

[0081] The intelligent diagnostic and root cause analysis platform uses a knowledge graph to dynamically display the data lineage of data quality issues and related devices, links, logical dependencies, etc. It presents multiple possible root causes inferred by the system in a clear list, each with a confidence percentage. It shows the logical reasoning chain by which the system arrives at the diagnostic conclusion in the form of a step flowchart or text summary, and shows the upper-level applications that the data quality issue may affect in the form of a tree diagram or list.

[0082] The data quality reporting and governance dashboard allows users to customize the period (daily / weekly / monthly) and scope, generating structured analysis reports with a single click. These reports include quality trends, top issue lists, root cause classification statistics, and governance recommendations. Reports can be formatted with both text and graphics and can be exported with a single click. The data quality dashboard displays governance task cards in their entirety, from "System Generated," "Distributed," "Processing," to "Verification Closed." Cards are draggable, providing a clear visual representation of the closed-loop process. For closed work orders, it provides comparative curves or bar charts of relevant data quality indicators before and after governance, intuitively verifying the effectiveness of governance actions.

[0083] Figure 4 This is a schematic diagram of the module interaction timing of the scheduling data quality evaluation system provided in the embodiments of this application, as shown below. Figure 4As shown, ① The Data Dashboard initiates a quality panorama analysis request to the User Module (U), displaying a color-changing heatmap of the entire network and popping up a Toast notification. ② Access the Alert Center through the User Module. ③ The Alert Center displays a timeline list of alerts (with prediction curves). ④ Click the "Smart Diagnosis" button through the User Module. ⑤ The Diagnosis Tool displays a knowledge graph source view and a probabilistic root cause list to the Execution Module (E). ⑥ The Execution Module analyzes and confirms the root causes, then clicks "Generate Governance Work Order" and sends it to the Diagnosis Tool. ⑦ The Diagnosis Tool automatically creates a governance task card and sends it to the Governance Dashboard. ⑧ The Governance Dashboard updates its status and pushes the task to the Work Order Module (M). ⑨ The Work Order Module assigns tasks to the Governance Dashboard and tracks their processing status. ⑩ The task card from the Work Order Module is dragged to "Verification in Progress". The data dashboard obtains post-governance quality data from the governance dashboard. Data dashboard feedback quality recovery trend chart. Confirm the effect in the work order module and close the work order. The governance dashboard stores closed-loop cases in a learning library. The configuration center automatically optimizes the model and knowledge base.

[0084] This application's technical solution aims to address a series of systemic and fundamental problems in current power grid dispatch data quality management. It will drive a fundamental shift in data governance models from "passive response, manual investigation, and experience-driven" to "proactive early warning, intelligent diagnosis, and closed-loop optimization." Specifically, this application will solve the following four key problems: Addressing the issue of "passive lag": Moving from "post-event alerts" to "pre-event warnings," based on historical data quality scoring sequences, predicting future quality degradation trends, and issuing "yellow alerts" before data impacts business operations. Utilizing unsupervised learning to identify implicit anomalies in multi-dimensional indicators, uncovering "sub-healthy" states where individual indicators appear "qualified" but the overall pattern is abnormal. This significantly advances the data quality risk management window, giving maintenance personnel valuable lead time to address issues, transforming "fault handling" into "preventive maintenance," and significantly enhancing the proactive defense capabilities of the power grid.

[0085] To address the issue of "rigid evaluation," the approach shifts from "static universality" to "dynamic adaptation." Based on the real-time operating status of the power grid, the importance of data services, and the evaluation application scenario, the weight of various indicators (e.g., completeness, accuracy) in the comprehensive score is dynamically adjusted from a pre-set strategy library. This transforms data quality scoring from a "numbers game" detached from business operations into a "flexible benchmark" closely aligned with the actual needs of dispatching operations, making the evaluation results more business-oriented and practical.

[0086] Addressing the issue of "fragmented management": From "information silos" to "governance closed loops," achieving end-to-end digital integration from "monitoring, evaluation, early warning, diagnosis, report or work order generation, governance push, and effect verification." Through panoramic visualizations such as quality heatmaps and root cause diagrams, complex analytical conclusions are presented intuitively, directly supporting decision-making. This bridges the "last mile" of data quality management, establishing a complete Plan-Do-Check-Act (PDCA) cycle, ensuring that every identified problem is tracked, addressed, and verified, effectively transforming technical capabilities into management efficiency.

[0087] Addressing the issue of "system rigidity": Moving from a "static tool" to a "self-evolving system," each successfully handled early warning and diagnostic case (including data characteristics, analysis process, and root cause verification) is stored in a case library for continuous training and optimization of the early warning and diagnostic reasoning models, as well as enriching the knowledge graph. This imbues the system with the intelligence to continuously learn and improve itself. As operating time increases, its early warnings for various quality issues become increasingly accurate, and its diagnoses become more precise, truly achieving "getting smarter with use" and realizing the sustainable evolution of its capabilities.

[0088] To implement the method of the embodiments of this application, Figure 5 A schematic diagram of a hardware structure of an electronic device provided in an embodiment of this application, such as... Figure 5 As shown in the illustration, this application embodiment also provides an electronic device 50 that may include: a memory 501 for storing a computer program; and a processor 502 for implementing the method described above when executing the computer program. The processor 502 can implement the steps of any of the methods described above, which will not be elaborated further here.

[0089] Of course, in practical applications, such as Figure 5 As shown, the electronic device 50 may further include at least one network interface 503. Various components in the electronic device are coupled together via a bus system 504. It is understood that the bus system 504 is used to implement communication between these components. In addition to a data bus, the bus system 504 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 5Various buses are labeled as bus systems 504. The number of processors 502 can be at least one. Network interface 503 is used for wired or wireless communication between electronic devices and other devices. Memory 501 in this embodiment is used to store various types of data to support the operation of the electronic device. The methods disclosed in the above embodiments can be applied to processor 502, or implemented by processor 502. Processor 502 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 502 or by instructions in software form. The processor 502 can be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 502 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. A general-purpose processor can be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly reflected in the combined execution of hardware and software modules in a microcontroller. The software module may reside in a storage medium located in memory 501. Processor 502 reads information from memory 501 and, in conjunction with its hardware, completes the steps of the aforementioned method. In an exemplary embodiment, electronic device 50 may be implemented using one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to execute the aforementioned method.

[0090] Specifically, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, such as a memory 501 storing the computer program, which can be executed by a processor 502 to complete the aforementioned method steps. The computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.

[0091] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0092] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0093] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0094] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for evaluating the quality of scheduling data, characterized in that, Applied to power grids, the method includes: Obtain the dispatch data of the power grid; the dispatch data includes first data and second data; the first data includes the topology data and equipment parameter data of the power grid, and the second data includes the measurement point data, alarm event data and dispatch operation data of the power grid; The first data and / or the second data are evaluated and processed according to at least one preset quality index to obtain at least one quality parameter; The weight set corresponding to the at least one quality parameter is determined based on the operating state of the power grid; the weight set includes at least one weight factor corresponding to the at least one quality parameter. Based on the at least one weighting factor and the at least one quality parameter, a quality evaluation score for the power grid is determined; the quality evaluation score is used to characterize the quality level of the power grid.

2. The method according to claim 1, characterized in that, The operating state includes running state, fault state, and analysis state; the weight set includes a first weight set, a second weight set, and a third weight set; determining the weight set corresponding to the at least one quality parameter based on the operating state of the power grid includes: When the power grid is in the operating state, it is determined that the at least one quality parameter corresponds to the first weight set; When the power grid is in the fault state, it is determined that the at least one quality parameter corresponds to the second weight set; the value of the weight factor in the second weight set is different from the value of the weight factor in the first weight set; When the power grid is in the analysis state, it is determined that the at least one quality parameter corresponds to the third weight set; the value of the weight factor in the third weight set is different from the value of the weight factor in the second weight set; the value of the weight factor in the third weight set is different from the value of the weight factor in the first weight set.

3. The method according to claim 2, characterized in that, The quality indicators include timeliness indicators, reliability indicators, integrity indicators, and accuracy indicators; after determining that the at least one quality parameter corresponds to the second weight set when the power grid is in the fault state, the method further includes: In response to a fault signal from the power grid, an upward adjustment operation is performed on the weight factor corresponding to the timeliness index and the weight factor corresponding to the reliability index. In response to the operation command for the power grid, the weight factor of the integrity index is increased, and the weight factor corresponding to the accuracy index is increased.

4. The method according to claim 1, characterized in that, The method further includes: Acquire historical sample data of the power grid; the historical sample data includes historical quality evaluation scores and historical transmission information; Based on the historical quality evaluation score, the historical transmission information, and the preset target model, the predicted quality evaluation score of the power grid is determined; If the predicted quality evaluation score meets the preset conditions, an early warning message is generated; the early warning message is used to characterize the risk of faults in the power grid in the future.

5. The method according to claim 4, characterized in that, The method further includes: If the quality evaluation score or the predicted quality evaluation score is less than or equal to a preset threshold, an alarm message is generated; the alarm message includes attribute information of the power grid failure. The attribute information is transformed into event nodes based on a preset data knowledge graph; The event nodes and entity nodes in the data knowledge graph are associated to obtain the association relationship; Starting from the event node, a multi-hop traversal is performed based on the association and the data knowledge graph to obtain at least one root cause data. The probability data of the root cause data is determined based on a preset Bayesian network; A diagnostic report on the power grid failure is generated based on the probability data and the root cause data; the diagnostic report is used to assess the attributable causes of the power grid failure.

6. The method according to claim 5, characterized in that, Before converting the attribute information into event nodes based on a preset data knowledge graph, the method further includes: Obtain at least two entity nodes in the power grid dispatch scenario; the entity nodes include at least equipment nodes, measurement point nodes, communication channel nodes, server nodes, and application terminal nodes in the power grid; The data knowledge graph is constructed based on the mapping relationship between the at least two entity nodes; the mapping relationship includes physical connection relationship, data flow relationship and logical dependency relationship.

7. The method according to claim 5 or 6, characterized in that, Perform the multi-hop traversal according to any of the following priority rules: Priority of the relationship type between at least two of the entity nodes; Priority of the node types of at least two of the said entity nodes; Priority of the state types of at least two of the entity nodes; Prioritization of the failure frequencies of at least two of the said entity nodes.

8. The method according to claim 5, characterized in that, After generating a diagnostic report of the power grid fault based on the probability data and the root cause data, the method further includes: An initial work order is generated based on the diagnostic report; Receive the completed work order corresponding to the initial work order; The case data of power grid failures are determined based on the initial work order and the completed work order; The data knowledge graph, the target model, and the preset conditions are optimized based on the case data.

9. The method according to claim 8, characterized in that, The optimization of the data knowledge graph, the target model, and the preset conditions based on the case data includes: The case data is identified as positive samples, and the scheduling data without faults is identified as negative samples. The target model is trained based on the positive samples and the negative samples to obtain the optimized target model; Extract the newly added faults and their corresponding root cause data from the case data; Based on the newly added faults and the root cause data corresponding to the newly added faults, optimize the probability parameters of the Bayesian network and the priority rules of the multi-hop traversal. The strength of the mapping relationship between at least two entity nodes in the data knowledge graph is updated based on the case data.

10. A computer-readable storage medium, characterized in that, The computer-readable medium stores a computer program that, when executed by a processor, is used to implement the method according to any one of claims 1 to 9.

Images