Wide-area power data load prediction method and system based on multiple time scales

Abstract

The invention relates to the technical field of power load prediction, and discloses a wide-area power data load prediction method and system based on multiple time scales. A data acquisition module of the system acquires historical and real-time load data of wide-area power, analyzes a load change trend, calculates a fluctuation frequency, compares data differences to judge an abnormal fluctuation risk and generates a load abnormal index; the multi-time-scale analysis module is used for positioning an abnormal load time node based on an abnormal index, analyzing a matching relationship between the node and a load event, evaluating a risk level, calculating a correlation degree, predicting a load deviation and generating a risk evaluation result; the dynamic adjustment module identifies a risk storage node according to an evaluation result, analyzes data distribution, calculates an adjustment priority, plans a path, and migrates data to a low-risk node to obtain adjustment configuration; the behavior prediction module compares access frequency based on configuration, identifies abnormal activity, judges abnormal characteristics of a behavior mode, and positions a deviation source to obtain an analysis result.

Description

Technical Field

[0001] This invention relates to the field of power load forecasting technology, specifically to a method and system for wide-area power data load forecasting based on multiple time scales. Background Technology

[0002] As the scale of power systems continues to expand and the coverage of wide-area power networks gradually extends, users' demands for the stability and reliability of power supply are increasing. Load forecasting, as a crucial component of power dispatching and operation management, directly impacts the overall operational efficiency of the power system in terms of accuracy and timeliness. Currently, traditional power load forecasting methods mostly rely on a single time scale for data processing and analysis, depending solely on historical load data for trend extrapolation. This neglects the dynamic changes in real-time load data, making it difficult to capture sudden anomalies during load fluctuations, resulting in significant discrepancies between forecast results and actual load demand. In wide-area power data scenarios, load data comes from diverse sources and is massive in volume. Load data from different regions and time periods exhibit significant differences in time scale, such as short-term fluctuations in daily load, periodic changes in weekly load, and long-term trends in seasonal load. Single-time-scale analysis methods cannot fully cover these complex load variation patterns, thus affecting the accurate identification of abnormal loads. Furthermore, existing systems, after detecting load anomalies, lack in-depth assessment of the correlation between abnormal time points and load events, making it impossible to accurately determine the risk level of abnormal loads. This results in subsequent load data adjustments lacking specificity and failing to effectively reduce the impact of load fluctuations on power system operation. Traditional load forecasting systems, in terms of data storage and management, fail to consider the risk attributes of load data, centralizing load data of different risk levels on the same node. When a load anomaly occurs at one node, it can easily trigger a chain reaction, affecting the stability of the entire data system. In the load behavior forecasting stage, existing methods only focus on the numerical changes in load data before and after adjustments, without in-depth analysis of the abnormal characteristics of load access frequency and the sources of load deviations. This fails to provide comprehensive behavioral analysis basis for subsequent power dispatching, further reducing the overall effectiveness of load forecasting. These problems make current wide-area power load forecasting systems unable to meet the needs of refined and intelligent power system operation, necessitating a comprehensive analysis system based on multiple time scales to address these shortcomings. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for wide-area power data load forecasting based on multiple time scales, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides a wide-area power data load forecasting system based on multiple time scales, the system comprising: The data acquisition module collects historical and real-time load data from the wide-area power data, analyzes the changing trends of the load data, calculates the load fluctuation frequency, compares the differences between historical and real-time loads, judges the risk of abnormal load fluctuations, and generates load anomaly indicators. Based on the load anomaly index, the multi-timescale analysis module locates the time nodes of abnormal loads, analyzes the matching relationship between time nodes and load events, assesses the risk level of time nodes, calculates the correlation between load frequency and load events, predicts the load deviation that will occur at time nodes, and generates load deviation risk assessment results. Based on the load deviation risk assessment results, the dynamic adjustment module identifies risky load data storage nodes, analyzes load data distribution, calculates adjustment priorities, plans adjustment paths, and migrates load data to low-risk storage nodes to obtain load data adjustment configurations. The behavior prediction module adjusts the configuration based on the load data, compares the access frequency of the load data before and after the adjustment, identifies abnormal load activities, judges the abnormal characteristics of the load behavior pattern, locates the source of load deviation, and obtains the load characteristic analysis results.

[0005] Preferably, the load anomaly indicators include load change frequency, fluctuation frequency difference, and load change frequency; the load deviation risk assessment results include time node risk level, load and event matching degree, and deviation identification results; the load data adjustment configuration includes data adjustment priority, data adjustment path, and load update standard; and the load characteristic analysis results include load pattern change indicators, access frequency comparison results, and abnormal activity identifiers.

[0006] Preferably, the data acquisition module includes: The load trend analysis submodule collects historical and real-time load data from wide-area power data, analyzes the load change time series of the load data, calculates the time interval between consecutive loads, counts the frequency change of load changes, compares the load rise and fall ratios, identifies the time points of abnormal load changes, and obtains load variability indicators. Based on the load variability index, the load access monitoring submodule retrieves access data at abnormal load time points, analyzes the distribution of access times over different time periods, calculates the degree of load access fluctuation in the short term, determines whether there is abnormal access behavior of the load, and obtains the load access fluctuation index. The load change statistics submodule, based on the load access fluctuation index, calls the load change records, counts the number of changes, and calculates the degree of abnormality of the change by combining the load changes and access monitoring data, and generates a load anomaly index.

[0007] Preferably, the multi-timescale analysis module includes: The abnormal load identification submodule filters the time data of abnormal loads based on the load anomaly index, analyzes the correlation between time points and load volume and load operation mode, calculates the distribution density of abnormal loads and classifies them, identifies abnormal load time nodes, and generates a set of abnormal load time nodes. The time node matching submodule calls the set of abnormal load time nodes, parses the load behavior characteristics, compares the patterns of identified load events, calculates the matching degree between nodes and load events, assesses the risk level of time nodes, and generates a load risk matching index. The load deviation risk assessment submodule analyzes the load frequency of abnormal nodes based on the load risk matching index, extracts the load time interval, calculates the load fluctuation range within a short period, predicts the probability of load deviation based on the risk matching degree of the nodes, and generates a load deviation risk assessment result.

[0008] Preferably, the dynamic adjustment module includes: Based on the load deviation risk assessment results, the risk node identification submodule detects risk nodes in the data storage network, analyzes the data types and sensitivity of the load stored by the nodes, filters storage nodes with data deviations, determines the range of data nodes that need to be adjusted, and obtains a list of risk nodes. Based on the risk node list, the data adjustment and analysis submodule analyzes the data distribution among the affected nodes, calculates the degree of data correlation and interaction frequency between nodes, determines the impact scope and priority of data adjustment, and generates a data adjustment priority index. The data storage reconstruction submodule analyzes the data flow path between storage nodes based on the data adjustment priority index, allocates storage resources, plans the optimal data adjustment path, and adjusts the access permissions of the secure storage nodes to obtain the load data adjustment configuration.

[0009] Preferably, the behavior prediction module includes: The behavior pattern change analysis submodule adjusts the configuration based on the load data, calls the behavior records of the adjusted load, compares the load activity characteristics before and after the adjustment, analyzes the magnitude and frequency of load behavior changes, calculates the degree of behavior pattern deviation, and obtains the behavior pattern deviation index. The data access frequency comparison submodule compares the data access frequency before and after the adjustment based on the behavior pattern offset index, analyzes the changes in access time, access duration and access frequency, judges the fluctuation of access frequency, and generates access frequency fluctuation index. The abnormal behavior identification submodule identifies load activities that deviate from the normal pattern based on the access frequency fluctuation index, analyzes the characteristics of abnormal load behavior patterns, matches the relationship between load activity characteristics and known deviations, locates the source of load deviations, and generates load characteristic analysis results.

[0010] Preferably, the system further includes; The dynamic topology update module establishes a power network association map based on the preprocessed data and dynamically updates the power network association map. The load status monitoring module monitors and analyzes the operating status of wide-area power loads based on the preprocessed data and power network correlation map. The load forecasting module predicts potential deviations in the wide-area power load based on real-time analysis of the operating status of the wide-area power load and load changes.

[0011] Preferably, the system further includes; Based on the load characteristic analysis results, the predictive response module identifies abnormal load points, adjusts load access permissions, allocates load quota ratios, updates load verification methods, and generates load protection measures. The load protection measures specifically include the load permission adjustment results, load quota configuration, and updated verification methods.

[0012] Preferably, the predictive response module includes: Based on the load characteristic analysis results, the load access control submodule analyzes the frequency of risky operations on the load, calculates the impact range of load access changes, identifies loads with frequent abnormal operations, reconfigures the access permissions of the load, implements access restrictions on high-risk loads, and generates load access adjustment configurations. The load quota management submodule calls the load permission adjustment configuration, calls the transaction records of abnormal loads, calculates the fluctuation range of load quotas, analyzes the short-term load quota change trend, judges the degree of deviation between the load amount and the normal behavior of the load, adjusts the upper limit of the load transaction quota, allocates the load quota ratio, and generates the optimized load quota. The verification method optimization submodule, based on the optimized load limit, filters load frequency anomalies, extracts the identity verification records of the anomalies, analyzes the security level of the abnormal load verification, determines whether the verification matches the load risk level, optimizes the load verification method, and generates load protection measures.

[0013] Preferably, the present invention also includes a method for wide-area power data load forecasting based on multiple time scales, comprising all the modules and method flows of the aforementioned wide-area power data load forecasting system based on multiple time scales.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This wide-area power data load forecasting system based on multiple time scales synchronously collects and analyzes historical and real-time load data from the wide-area power data through a data acquisition module. It can comprehensively capture the dynamic trend of load data changes, calculate the load fluctuation frequency, and compare the differences between historical and real-time load data. This enables timely judgment of the risk of abnormal load fluctuations and the generation of abnormal indicators, effectively making up for the shortcomings of traditional forecasting methods that rely on only a single data type and ignore the dynamic changes of data. This makes the preliminary identification of load anomalies more comprehensive and timely. The multi-timescale analysis module, based on load anomaly indicators, accurately locates the time nodes of abnormal loads. By analyzing the matching relationship between time nodes and load events, it can deeply explore the potential causes of abnormal loads. At the same time, it assesses the risk level of time nodes and calculates the correlation between load frequency and load events, thereby predicting the load deviation that will occur at the time nodes and generating risk assessment results. Compared with the problem that traditional single-timescale analysis cannot cover complex load change patterns, this module can analyze load anomaly characteristics from different time dimensions, making the risk assessment of abnormal loads more targeted and accurate, and providing clear directional guidance for subsequent load adjustments. Based on the load deviation risk assessment results, the dynamic adjustment module accurately identifies the data storage nodes of risky loads. By analyzing the distribution of load data, it calculates the adjustment priority and plans a reasonable adjustment path, migrating the load data to low-risk storage nodes to form an adjustment configuration. This avoids the system stability problems caused by the mixed storage of data of different risk levels in traditional data storage. It can achieve differentiated storage management according to the risk attributes of load data, reduce the impact of abnormal load data on the overall data system, improve the security and reliability of load data storage, and ensure the stable operation of the power data system. The behavior prediction module adjusts configurations based on load data, compares the access frequency of load data before and after adjustment, and can keenly identify abnormal load activities. At the same time, it judges the abnormal characteristics of load behavior patterns and locates the source of load deviation, forming a comprehensive load characteristic analysis result. It breaks through the limitations of traditional behavior prediction that only focuses on numerical changes and ignores access frequency and deviation sources. It can deeply analyze the load change pattern from the behavior level, provide richer decision reference information for power dispatch and operation management, help the power system respond more accurately to load fluctuations, optimize the overall operation strategy, and adapt to the refined and intelligent operation needs of wide-area power networks. Attached Figure Description

[0015] Figure 1 This is a schematic diagram illustrating the working principle of a wide-area power data load forecasting system based on multiple time scales as described in this invention. Figure 2 A flowchart illustrating the correspondence between key load indicators and evaluation results; Figure 3The flowchart of the multi-timescale analysis module; Figure 4 This is a flowchart of the system's extended modules. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 This invention provides a method and system for wide-area power data load forecasting based on multiple time scales, the system comprising: The data acquisition module is responsible for collecting historical and real-time load data from the wide-area power data, analyzing load data trends, calculating load fluctuation frequency, comparing historical and real-time load differences, assessing the risk of abnormal load fluctuations, and generating load anomaly indicators. The multi-timescale analysis module, based on load anomaly indicators, locates the time nodes of abnormal loads, analyzes the matching relationship between time nodes and load events, assesses the risk level of time nodes, calculates the correlation between load frequency and load events, predicts load deviations occurring at time nodes, and generates load deviation risk assessment results. The dynamic adjustment module, based on the load deviation risk assessment results, identifies risky load data storage nodes, analyzes load data distribution, calculates adjustment priorities, plans adjustment paths, migrates load data to low-risk storage nodes, and obtains the load data adjustment configuration. The behavior prediction module, based on the load data adjustment configuration, compares the access frequency of load data before and after adjustment, identifies abnormal load activities, determines the abnormal characteristics of load behavior patterns, locates the source of load deviation, and obtains load characteristic analysis results.

[0018] Example 1: See Figure 2The load trend analysis submodule first preprocesses the raw data, including data cleaning, missing value imputation, and preliminary outlier screening. Then, this submodule constructs a time-series model of load changes, calculating the time interval between consecutive load data points using a sliding window technique. For example, when analyzing load data for a regional power grid, the module uses one minute as a sampling point, continuously analyzing 1440 data points over 24 hours, and calculating the rate of change of the difference between adjacent data points. By statistically analyzing the distribution characteristics of these rates of change, the module identifies the frequency and intensity of load changes. This submodule pays particular attention to the ratio of load increases to decreases, detecting abnormal time points that deviate from the normal pattern by comparing the change patterns of historical and real-time data within the same time period. These calculation and analysis results are integrated into a load variability index, which quantifies the fluctuation characteristics and anomalies of the load data.

[0019] The load access monitoring submodule further analyzes the aforementioned variability indicators. This submodule focuses on time points marked as abnormal, deeply retrieving all related access operation data. This access data includes records of queries, calls, and modifications to load data within the system. The module analyzes the distribution of access operation frequency, operation type, and operation source near these abnormal time points. For example, when a sharp fluctuation in load data is detected at a certain time point, the module traces all data access records within a period before and after this time point, analyzing the temporal distribution characteristics and operation patterns of these accesses. By calculating the variance and coefficient of variation of access frequency within a short time window, the module assesses the degree of fluctuation in access behavior and determines whether abnormal access patterns exist. These analytical results are quantified as load access fluctuation indicators, reflecting the degree of abnormality in data access behavior.

[0020] The load change statistics submodule receives the output from the first two submodules and calls the load data change logs recorded by the system. This submodule counts the number of load data changes within a specific time range, including modifications to data values ​​and additions / deletions of data records. The module performs correlation analysis with load variability indicators and access fluctuation indicators to calculate the degree of anomaly in the change operations. For example, the module may find that load data changes frequently within a certain time period, while the load variability indicator shows abnormal fluctuations and the access fluctuation indicator also indicates abnormal access patterns. This overlap of multiple anomalies indicates significant anomalies in that time period. By weighted summarizing these anomalies, the submodule finally generates a load anomaly indicator, which includes three dimensions: load change frequency, fluctuation frequency difference, and load change frequency, providing a quantitative basis for subsequent analysis.

[0021] The data acquisition module employs a distributed computing architecture to process massive amounts of data. Data acquisition nodes are deployed in different areas of the power grid, and the acquired data is transmitted to the central processing platform in real time via a dedicated communication network. The platform uses a streaming computing framework to process real-time data, while simultaneously using a batch processing framework to analyze historical data. This hybrid computing model ensures the timeliness and integrity of data processing. An internal data quality monitoring mechanism is established within the module to perform integrity verification and consistency checks on the acquired data. When data quality issues are detected, the module automatically triggers data re-acquisition or data repair processes. Furthermore, the module implements a data caching mechanism to provide rapid response capabilities for frequently accessed load data.

[0022] The calculation process for load anomaly indicators employs a multi-level verification mechanism. The calculation results of each submodule undergo consistency verification and rationality checks to ensure the accuracy and reliability of the indicators. After generation, the indicators are stored in a distributed database for use by subsequent modules. The database uses a time-series data optimization storage scheme, supporting efficient data query and retrieval operations. The implementation of the entire data acquisition module emphasizes system scalability and fault tolerance. The module adopts a microservice architecture design, allowing each submodule to be deployed and expanded independently. When a submodule fails, the system automatically starts a backup instance to take over, ensuring the continuity of the data acquisition process. Simultaneously, the module implements a robust data backup and recovery mechanism to ensure data security.

[0023] Example 2: See Figure 3 In the implementation of a wide-area power data load forecasting system based on multiple time scales, the execution of the multi-time scale analysis module constitutes the core of the system's operation. This module, through the coordinated operation of its three internal sub-modules, completes the entire analysis process from anomaly indicators to risk assessment. The anomaly load identification sub-module first performs spatiotemporal analysis on these indicator data. This sub-module establishes a multi-level time window analysis mechanism, scanning and screening anomaly load data at minute, hour, and day time granularities. In minute-level analysis, the module detects instantaneous load mutations; for example, in monitoring a regional power grid, it finds that a specific feeder experiences power fluctuations exceeding a threshold within a 3-minute time window. In hourly-level analysis, the module analyzes the persistence and periodicity of load changes, identifying the duration patterns of anomaly loads. In daily-level analysis, the module compares historical data from the same period to discover the date regularity of anomaly load occurrences. Through comprehensive analysis of these multi-time scale characteristics, the module calculates the distribution density of anomaly loads and uses clustering algorithms to classify anomaly loads into different categories such as sudden, continuous, and intermittent types. Finally, a set of abnormal load time nodes containing information such as timestamps, load values, and anomaly categories is generated.

[0024] After receiving the set of abnormal load time nodes, the time node matching submodule initiates the matching analysis process of the load event pattern library. This module incorporates feature patterns of various typical load events, including equipment start-up / shutdown, line switching, and fault disturbances. For each abnormal time node, the module extracts its load behavior feature vector, including parameters such as load change rate, fluctuation amplitude, and duration. By calculating the similarity distance between these feature vectors and standard event patterns, the module evaluates the degree of matching between each node and known load events. During implementation, the module employs a multi-dimensional weighted matching algorithm, comprehensively considering factors such as time consistency, load feature similarity, and spatial correlation. For example, when an abnormal fluctuation in the outgoing load of a substation is detected at a specific time point, the module simultaneously queries the station's operation records, equipment status, and the surrounding power grid operation to comprehensively determine whether the anomaly is related to a planned equipment operation event. Based on the matching results, the module labels each time node with a risk level, generating a load risk matching index that includes a risk score and event correlation.

[0025] The load deviation risk assessment submodule conducts in-depth predictive analysis based on the load risk matching index. This submodule employs time series forecasting technology to establish a multivariate load forecasting model. The model comprehensively considers multiple influencing factors, including historical load patterns, real-time operating status, and environmental factors. For each identified abnormal node, the module analyzes its load frequency characteristics, extracting the interval patterns and fluctuation characteristics of load changes. By establishing a short-term load forecast curve, the module calculates the probability distribution of load fluctuation range within future time windows. During implementation, the module pays particular attention to deviation prediction for high-risk nodes, using Monte Carlo simulation to generate multiple possible load development scenarios. For example, for a fault disturbance event node with a high degree of matching, the module simulates the chain reaction that the event may trigger, predicting the deviation development trend of the system load. Based on the node's risk matching degree and predicted deviation probability, the module generates a comprehensive assessment report that includes the risk level of the time node, the load-event matching degree, and the deviation identification results.

[0026] The multi-timescale analysis module employs a distributed in-memory computing architecture to process massive amounts of time-series data. Deployed on a high-speed computing cluster, each submodule utilizes parallel computing to achieve synchronous analysis of power grid data from multiple regions. Data storage employs an optimized time-series database scheme, supporting high-speed time-range queries and complex conditional retrievals. An internal data quality monitoring mechanism is established to perform integrity checks and logical consistency checks on input data. When data anomalies are detected, the module automatically triggers a data review process to ensure the accuracy of the analysis results.

[0027] The module implements a multi-level caching mechanism, providing memory-level fast response for frequently accessed load pattern data and event characteristic data. An incremental update strategy is used during calculation; when new load data arrives, only the affected time window needs to be recalculated, significantly improving analysis efficiency. The module also implements a result backtracking mechanism; all intermediate calculation results and analysis process data are persistently stored, supporting the traceability and verifiability of analysis results. Fault tolerance is a crucial aspect of the module's implementation. When a sub-computing node fails, the system automatically migrates the computation task to a backup node, ensuring the continuity of the analysis process. The module implements a timed saving function for computation status, enabling recovery from the nearest computation point after an abnormal system interruption. Data consistency is guaranteed through a distributed transaction mechanism, ensuring the accuracy and reliability of the analysis results. The module's output uses a standardized data format, facilitating subsequent module calls and use. Load deviation risk assessment results include structured risk assessment data and metadata information, supporting multiple data exchange protocols and interface standards. Results data is pushed to the message middleware in real time for dynamic adjustment module subscriptions.

[0028] Example 3: In the implementation of a wide-area power data load forecasting system based on multiple time scales, the execution of the dynamic adjustment module constitutes a key link in the system's operation. This module, through the coordinated operation of its three internal sub-modules, completes the entire processing flow from risk assessment to data adjustment. Its implementation begins with receiving and parsing the load deviation risk assessment results generated by the multi-time scale analysis module. These assessment results contain important information such as the risk level at specific time points, load-event matching degree, and deviation identification results, providing a basis for decision-making in subsequent data storage optimization.

[0029] The risk node identification submodule first performs in-depth analysis of the assessment data. This submodule establishes a topology mapping model of the data storage network, associating the risk information in the assessment results with the actual storage nodes. By scanning the node status information of the data storage network, the module collects operational parameters such as real-time load, storage capacity utilization, and data access pressure for each node. The module employs a detection algorithm based on a risk propagation model to analyze the data types and sensitivity characteristics of the load data stored by the nodes. For nodes storing highly sensitive load data, the module increases the sensitivity threshold for risk detection. During the detection process, the module pays special attention to nodes storing historically abnormal data or currently in an abnormal state. Through a comprehensive evaluation of multi-dimensional risk indicators, it filters out storage nodes with data deviation risks. Finally, it generates a list of risk nodes containing information such as node identifier, risk level, and data characteristics.

[0030] The data adjustment and analysis submodule conducts in-depth data distribution analysis based on the risk node list. This submodule establishes a data association graph model to analyze the data distribution characteristics and interaction relationships among affected nodes. The module calculates the degree of data association and interaction frequency indicators between nodes by collecting data flow records and access logs. During the analysis, the module employs a graph theory-based community detection algorithm to identify closely related groups of nodes in the data storage network. For each node requiring adjustment, the module assesses the potential cascading effects of its data adjustment, including the degree of impact on related nodes and the scope of impact on the overall system operation.

[0031] The data adjustment priority is calculated using the following formula: in: This indicates a data adjustment priority index. The value represents the quantified risk level of the node. Represents the data sensitivity coefficient. To influence the propagation factor, , , Each indicator has its own weighting coefficient. These coefficients are dynamically adjusted based on the system's operational status to reflect risk preferences and adjustment strategies at different times. Through this comprehensive calculation model, the module generates a data adjustment priority index, providing execution order guidance for subsequent data migration operations.

[0032] The data storage reconstruction submodule implements specific storage optimization operations based on the data adjustment priority index. This submodule first analyzes the data flow path characteristics between storage nodes and establishes a data migration cost model. The model comprehensively considers multiple factors such as network bandwidth resources, storage space distribution, and data access patterns. The module employs a path optimization algorithm to plan the optimal data adjustment path, ensuring the efficiency and reliability of the data migration process. Before implementing data migration, the module conducts a security assessment of the target storage nodes, including testing their security protection capabilities, access control mechanisms, and fault recovery capabilities. Based on the assessment results, the module dynamically adjusts the access permission settings of the secure storage nodes to ensure the security and availability of the migrated data.

[0033] The dynamic adjustment module employs a distributed transaction processing mechanism to ensure the consistency of data adjustment operations. The module establishes an operation log recording system to record the execution process and result status of each adjustment operation in detail. In the event of operational anomalies or system failures, the module can roll back or recover operations based on the log records, ensuring that data integrity is not affected. The module implements a real-time monitoring mechanism to continuously track key indicators during the data adjustment process. These indicators include data transmission rate, resource utilization, and operation completion progress. Monitoring data is fed back to the decision-making system in real time, supporting the dynamic optimization of adjustment strategies. The module also establishes an anomaly handling mechanism that automatically triggers corrective measures or manual intervention procedures when an adjustment operation deviates from expectations.

[0034] Fault tolerance design and fault recovery are crucial aspects of module implementation. The module employs a redundant deployment architecture, with backup instances running for critical functional components. When the primary instance fails, the backup instance seamlessly takes over. A breakpoint resumption mechanism is used during data migration to ensure reliable data transmission in the event of network interruptions or other anomalies. The module's output uses a standardized data format, containing complete information such as data adjustment priority, data adjustment path, and load update standards. These results are pushed to the system's message bus for subsequent behavior prediction modules to subscribe to. Simultaneously, adjustment configuration information is persistently stored in the system database, supporting historical queries and audit trails. The dynamic adjustment module effectively identifies and manages data storage risks, optimizes the data distribution structure, and improves the overall system efficiency and security. The module's operation demonstrates a proactive response to and effective execution of risk assessment results, providing reliable data storage assurance for the entire prediction system.

[0035] Example 4: See Figure 4 The behavior pattern change analysis submodule first retrieves load behavior records before and after the adjustment. This submodule establishes a multi-dimensional behavior feature extraction mechanism, analyzing load activity characteristics from the perspectives of time distribution, operation frequency, and access patterns. The module employs a time window sliding analysis technique to compare load behavior data for the same time period before and after the adjustment. By calculating the magnitude and frequency of behavior changes, the module quantifies the degree of shift in behavior patterns. During the analysis, the module pays particular attention to load nodes with large adjustment amplitudes, recording in detail the trajectory of their behavioral characteristics changes. These analysis results are quantified into a behavior pattern shift index, which reflects the degree of change in load behavior relative to historical patterns.

[0036] The data access frequency comparison submodule conducts in-depth analysis based on the behavior pattern offset index. This submodule establishes a time-series database of access behavior, recording detailed access logs for each load node. The module analyzes the distribution patterns of access time points and identifies the temporal clustering characteristics of access behavior. By comparing access data within the same time interval before and after adjustments, the module calculates the difference rate of access duration and the magnitude of change in access frequency. During the analysis, the module employs statistical analysis methods to identify abnormal fluctuation patterns in access frequency. These analytical results are synthesized into an access frequency fluctuation index, which includes multi-dimensional information such as access time consistency and access intensity change rate.

[0037] The abnormal behavior identification submodule performs anomaly detection based on the access frequency fluctuation index. This submodule establishes a load behavior characteristic pattern library, containing feature templates for various typical normal and abnormal behaviors. The module uses a pattern matching algorithm to calculate the similarity between the current load activity characteristics and known behavior patterns. For detected abnormal behaviors, the module further analyzes their behavioral pattern characteristics, including operational sequence regularity and time distribution characteristics. By tracing the origin node and propagation path of abnormal behaviors, the module accurately locates the source of load deviations (see Table 1).

[0038] Table 1: Comparison data of load activity patterns.

[0039]

[0040] The dynamic topology update module establishes a power network relational graph based on preprocessed power grid operation data. This module collects power grid topology data in real time, including node connections, equipment parameters, and operating status. Using graph database technology, the module constructs a network model that reflects the actual interconnections of the power grid. The module implements a dynamic update mechanism, promptly updating node relationships and equipment parameters in the relational graph when changes in the power grid topology are detected. These update operations ensure that the network model always remains consistent with the actual situation.

[0041] The load status monitoring module implements real-time monitoring of the operating status of wide-area power loads based on preprocessed data and power network correlation maps. This module establishes a multi-level monitoring system, comprehensively monitoring load operating status from the equipment level, node level, to the system level. By collecting real-time load data and combining it with network topology information, the module analyzes the spatiotemporal characteristics of load distribution. During monitoring, the module pays particular attention to load data associated with key nodes in the network topology, analyzing the impact of their operating status on the overall system.

[0042] The load forecasting module predicts load deviations based on real-time analysis of the wide-area power load operation status and load changes. This module establishes a multi-factor forecasting model, comprehensively considering historical load patterns, real-time operating status, network topology characteristics, and other factors. Through time series analysis, the module predicts potential load deviation trends over future periods. During the forecasting process, the module pays particular attention to load nodes associated with abnormal behavior, increasing the monitoring frequency of their forecast results. The behavior prediction module employs a distributed stream processing architecture to process real-time data streams. The module is deployed on a high-speed computing cluster, with each submodule possessing independent data processing capabilities. Data storage utilizes a time-series database optimization scheme, supporting efficient time-range queries and complex condition retrieval. An internal data quality verification mechanism is established within the module to verify the integrity and logical consistency of input data.

[0043] The module implements a multi-level caching mechanism, providing fast memory-level response for frequently accessed behavioral pattern data and feature template data. An incremental update strategy is used during computation; when new load behavior data arrives, only the affected analysis windows need to be recalculated. The module also implements a result traceability mechanism; all analysis process data is persistently stored, supporting the verifiability of analysis results. Fault tolerance is a crucial aspect of the module's implementation. When a computing node fails, the system automatically migrates the computation task to a backup node, ensuring the continuity of the analysis process. The module implements a timed saving function for computation status, enabling recovery from the nearest computing point after an abnormal system interruption. Data consistency is guaranteed through a distributed transaction mechanism, ensuring the accuracy of the analysis results.

[0044] The module's output uses a standardized data format, including complete information such as load pattern change indicators, access frequency comparison results, and abnormal activity identifiers. These results are pushed to the system's message middleware in real time for subsequent predictive response modules to subscribe to and use. Simultaneously, the analysis results are persistently stored in the system database, supporting historical queries and trend analysis. The behavior prediction module can deeply analyze load behavior characteristics and accurately identify abnormal behavior patterns.

[0045] Example 5: The load access control submodule first performs in-depth processing on these feature analysis data. This submodule establishes a risk operation identification model and analyzes the operational behavior characteristics of load nodes. By statistically analyzing the frequency of risk operations for each load node, it calculates the distribution pattern and probability of abnormal operations. The module uses clustering analysis based on behavioral features to identify load nodes exhibiting frequent abnormal operation patterns. For the identified high-risk load nodes, the module assesses the potential impact of permission changes, including the degree of impact on related system functions and the effect on user experience. Based on the assessment, the module reconfigures the access permissions of these load nodes, implementing differentiated access control policies. For load nodes with higher risk levels, the module implements strict access restrictions, including access time restrictions, operation scope restrictions, and functional permission restrictions. These permission adjustment operations generate detailed load access configuration records.

[0046] The load quota management submodule performs quota optimization analysis based on permission adjustment configuration. This submodule calls upon historical transaction record data of abnormal load nodes to establish a quota usage analysis model. By analyzing transaction time distribution, transaction amount distribution, and transaction frequency characteristics, the module calculates the fluctuation range and trend of load quotas. The module uses time series analysis to identify the periodicity and abnormality of quota usage. For detected abnormal quota usage patterns, the module further analyzes the degree of deviation from normal behavior patterns. Based on the analysis, the module adjusts the transaction quota upper limit parameter of load nodes, allocating differentiated quota ratios according to the node's risk level and behavioral characteristics. The quota allocation process comprehensively considers multiple factors such as the node's historical behavior, current risk status, and business needs to ensure that the quota configuration meets security requirements without affecting normal business operations. These quota optimization operations generate an optimized load quota scheme containing quota parameters and allocation rules.

[0047] The verification method optimization submodule upgrades the verification mechanism based on the optimized quota scheme. This submodule identifies nodes with abnormal load frequencies and extracts their authentication history. By analyzing data such as verification time, verification method, and verification success rate, the module evaluates the security and effectiveness of the existing verification mechanism. The module establishes a verification security level assessment model, matching corresponding verification strength requirements based on the node's risk level and behavioral characteristics. For nodes with higher risk levels, the module increases the verification security level, adopting multi-factor authentication or enhanced verification methods. During the verification method optimization process, the module comprehensively considers user experience factors, optimizing the convenience of the verification process as much as possible while ensuring security. These optimization measures generate a load protection scheme that includes verification rules and security parameters.

[0048] The predictive response module employs a distributed decision-making architecture to handle various protection decision requests. Deployed across multiple computing nodes, each submodule possesses independent data processing and decision-making capabilities. Data exchange utilizes a unified interface standard to ensure the accuracy and consistency of data transmission between modules. An internal decision log recording mechanism is established to meticulously record the generation process and supporting data for each protection decision. The module implements a real-time monitoring and feedback mechanism to continuously evaluate the effectiveness of implemented protection measures. By collecting load behavior data after the execution of protection measures, the module analyzes the actual effectiveness and scope of impact. Evaluation results are fed back to the decision-making system, supporting the dynamic adjustment and optimization of protection strategies. The module also establishes an abnormal decision detection mechanism, automatically initiating a correction process when deviations in protection decisions or unexpected impacts are detected.

[0049] Fault tolerance design and fault recovery are crucial aspects of module implementation. The module employs a redundant deployment scheme, with backup instances running for critical decision-making components. When the primary decision-making node fails, the backup node can promptly take over the decision-making process. A transaction consistency guarantee mechanism is used during the decision-making process to ensure the atomicity and consistency of the protective measures.

[0050] The module's output uses a standardized format, including complete information such as load permission adjustment results, load quota configuration, and updated verification methods. These protective measures are pushed to the system's execution components and persistently stored in the system database. The implementation status and effectiveness data of the protective measures are monitored and recorded in real time, providing a reference for subsequent decision-making and optimization. The predictive response module can generate effective security protection measures based on load characteristic analysis results, enabling timely response and risk control to abnormal load behavior. The module's operation demonstrates a complete closed-loop processing capability from analysis to execution, providing crucial security protection functions for the entire predictive system. The protection decision-making process fully considers the balance between security and practicality, ensuring the effectiveness and adaptability of the protective measures.

[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0052] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A wide-area power data load forecasting system based on multiple time scales, characterized in that, The system includes: The data acquisition module collects historical and real-time load data from the wide-area power data, analyzes the changing trends of the load data, calculates the load fluctuation frequency, compares the differences between historical and real-time loads, judges the risk of abnormal load fluctuations, and generates load anomaly indicators. Based on the load anomaly index, the multi-timescale analysis module locates the time nodes of abnormal loads, analyzes the matching relationship between time nodes and load events, assesses the risk level of time nodes, calculates the correlation between load frequency and load events, predicts the load deviation that will occur at time nodes, and generates load deviation risk assessment results. Based on the load deviation risk assessment results, the dynamic adjustment module identifies risky load data storage nodes, analyzes load data distribution, calculates adjustment priorities, plans adjustment paths, and migrates load data to low-risk storage nodes to obtain load data adjustment configurations. The behavior prediction module adjusts the configuration based on the load data, compares the access frequency of the load data before and after the adjustment, identifies abnormal load activities, judges the abnormal characteristics of the load behavior pattern, locates the source of load deviation, and obtains the load characteristic analysis results.

2. The wide-area power data load forecasting system based on multiple time scales according to claim 1, characterized in that, The load anomaly indicators include load change frequency, fluctuation frequency difference, and load change frequency. The load deviation risk assessment results include time node risk level, load and event matching degree, and deviation identification results. The load data adjustment configuration includes data adjustment priority, data adjustment path, and load update standard. The load characteristic analysis results include load pattern change indicators, access frequency comparison results, and abnormal activity identifiers.

3. The wide-area power data load forecasting system based on multiple time scales according to claim 1, characterized in that, The data acquisition module includes: The load trend analysis submodule collects historical and real-time load data from wide-area power data, analyzes the load change time series of the load data, calculates the time interval between consecutive loads, counts the frequency change of load changes, compares the load rise and fall ratios, identifies the time points of abnormal load changes, and obtains load variability indicators. Based on the load variability index, the load access monitoring submodule retrieves access data at abnormal load time points, analyzes the distribution of access times over different time periods, calculates the degree of load access fluctuation in the short term, determines whether there is abnormal access behavior of the load, and obtains the load access fluctuation index. The load change statistics submodule, based on the load access fluctuation index, calls the load change records, counts the number of changes, and calculates the degree of abnormality of the change by combining the load changes and access monitoring data, and generates a load anomaly index.

4. The wide-area power data load forecasting system based on multiple time scales according to claim 1, characterized in that, The multi-timescale analysis module includes: The abnormal load identification submodule filters the time data of abnormal loads based on the load anomaly index, analyzes the correlation between time points and load volume and load operation mode, calculates the distribution density of abnormal loads and classifies them, identifies abnormal load time nodes, and generates a set of abnormal load time nodes. The time node matching submodule calls the set of abnormal load time nodes, parses the load behavior characteristics, compares the patterns of identified load events, calculates the matching degree between nodes and load events, assesses the risk level of time nodes, and generates a load risk matching index. The load deviation risk assessment submodule analyzes the load frequency of abnormal nodes based on the load risk matching index, extracts the load time interval, calculates the load fluctuation range within a short period, predicts the probability of load deviation based on the risk matching degree of the nodes, and generates a load deviation risk assessment result.

5. The wide-area power data load forecasting system based on multiple time scales according to claim 1, characterized in that, The dynamic adjustment module includes: Based on the load deviation risk assessment results, the risk node identification submodule detects risk nodes in the data storage network, analyzes the data types and sensitivity of the load stored by the nodes, filters storage nodes containing data deviations, determines the range of load data nodes that need to be adjusted, and obtains a list of risk nodes. Based on the risk node list, the data adjustment and analysis submodule analyzes the data distribution among the affected nodes, calculates the degree of data correlation and interaction frequency between nodes, determines the impact scope and priority of data adjustment, and generates a data adjustment priority index. The data storage reconstruction submodule analyzes the data flow path between storage nodes based on the data adjustment priority index, allocates storage resources, plans the optimal data adjustment path, and adjusts the access permissions of the secure storage nodes to obtain the load data adjustment configuration.

6. The wide-area power data load forecasting system based on multiple time scales according to claim 1, characterized in that, The behavior prediction module includes: The behavior pattern change analysis submodule adjusts the configuration based on the load data, calls the behavior records of the adjusted load, compares the load activity characteristics before and after the adjustment, analyzes the magnitude and frequency of load behavior changes, calculates the degree of behavior pattern deviation, and obtains the behavior pattern deviation index. The data access frequency comparison submodule compares the data access frequency before and after the adjustment based on the behavior pattern offset index, analyzes the changes in access time, access duration and access frequency, judges the fluctuation of access frequency, and generates access frequency fluctuation index. The abnormal behavior identification submodule identifies load activities that deviate from the normal pattern based on the access frequency fluctuation index, analyzes the characteristics of abnormal load behavior patterns, matches the relationship between load activity characteristics and known deviations, locates the source of load deviations, and generates load characteristic analysis results.

7. The wide-area power data load forecasting system based on multiple time scales according to claim 1, characterized in that, The system also includes; The dynamic topology update module establishes a power network association map based on the preprocessed data and dynamically updates the power network association map. The load status monitoring module monitors and analyzes the operating status of wide-area power loads based on the preprocessed data and power network correlation map. The load forecasting module predicts potential deviations in the wide-area power load based on real-time analysis of the operating status of the wide-area power load and load changes.

8. The wide-area power data load forecasting system based on multiple time scales according to claim 1, characterized in that, The system also includes; Based on the load characteristic analysis results, the predictive response module identifies abnormal load points, adjusts load access permissions, allocates load quota ratios, updates load verification methods, and generates load protection measures. The load protection measures specifically include the load permission adjustment results, load quota configuration, and updated verification methods.

9. The wide-area power data load forecasting system based on multiple time scales according to claim 8, characterized in that, The predictive response module includes: Based on the load characteristic analysis results, the load access control submodule analyzes the frequency of risky operations on the load, calculates the impact range of load access changes, identifies loads with frequent abnormal operations, reconfigures the access permissions of the load, implements access restrictions on high-risk loads, and generates load access adjustment configurations. The load quota management submodule calls the load permission adjustment configuration, calls the transaction records of abnormal loads, calculates the fluctuation range of load quotas, analyzes the short-term load quota change trend, judges the degree of deviation between the load amount and the normal behavior of the load, adjusts the upper limit of the load transaction quota, allocates the load quota ratio, and generates the optimized load quota. The verification method optimization submodule, based on the optimized load limit, filters load frequency anomalies, extracts the identity verification records of the anomalies, analyzes the security level of the abnormal load verification, determines whether the verification matches the load risk level, optimizes the load verification method, and generates load protection measures.

10. A method for wide-area power data load forecasting based on multiple time scales, characterized in that, It includes all modules and method flows of the wide-area power data load forecasting system based on multiple time scales as described in any one of claims 1 to 9.

Images