Frequency and voltage regulation coordinated control method and device applied to the interaction between virtual power plants and large power grids

Abstract

This invention provides a method and apparatus for frequency and voltage regulation coordinated control applied to the interaction between a virtual power plant and a large power grid. By acquiring the real-time operational data set generated during the interaction between the virtual power plant and the large power grid, and based on the statistical distribution attributes carried, a dedicated sparse dictionary adapted to the frequency and voltage regulation scenario of the virtual power plant and the large power grid interaction is generated. Using this dictionary as a processing benchmark, a sparse representation operation is performed on the real-time operational data set to obtain a set of sparse components with responsiveness to frequency and voltage regulation coordinated control. Simultaneously, redundant data components unrelated to frequency and voltage regulation coordinated control are filtered out from the real-time operational data set. An inverse transformation operation is performed on the sparse component set with responsiveness to reconstruct key operational data. Error verification is performed to obtain the key operational data that passes the verification. Based on this, a coordinated control deduction operation is performed to obtain the frequency and voltage regulation coordinated control command between the virtual power plant and the large power grid. This invention improves the accuracy and effectiveness of frequency and voltage regulation coordinated control.

Description

Technical Field

[0001] This invention relates to the field of intelligent control, and more specifically, to a frequency and voltage regulation coordinated control method and apparatus for interaction between a virtual power plant and a large power grid. Background Technology

[0002] Frequency and voltage regulation coordinated control between virtual power plants and the main power grid is a process of coordinating the operation of distributed resources within the virtual power plant and the main power grid to achieve coordinated matching of their frequency and voltage regulation actions. This is primarily used to improve the operational stability and control flexibility of the main power grid. Currently, control is typically based on data collected from the operation of distributed resources within the virtual power plant or from the overall operation of the main power grid. After sparsification processing, frequency and voltage regulation control commands are generated and sent to the control nodes. However, the collected data often fails to accurately cover the core interactive information during the interaction process. The compatibility between a general dictionary and real-world scenario data is difficult to meet control requirements, and the reliability of the processed data lacks targeted verification. Consequently, the generated control commands do not match the actual operating state of the interactive scenario well, making it difficult to fully leverage the effectiveness of coordinated control between the virtual power plant and the main power grid. Summary of the Invention

[0003] In view of this, the present invention provides a frequency regulation and voltage regulation coordinated control method and device for interaction between virtual power plants and large power grids.

[0004] According to one aspect of the present invention, a frequency and voltage regulation coordinated control method for interaction between a virtual power plant and a large power grid is provided, comprising: acquiring a real-time operating data set generated during the interaction between the virtual power plant and the large power grid, wherein the real-time operating data set carries electrical status information and interaction response information of each distributed resource of the virtual power plant during the interaction with the large power grid; generating a dedicated sparse dictionary adapted to the frequency and voltage regulation scenario of the interaction between the virtual power plant and the large power grid based on the statistical distribution attributes carried by the real-time operating data set, wherein the dedicated sparse dictionary and the statistical distribution attributes of the real-time operating data set form a unique correspondence; and performing a sparse representation operation on the real-time operating data set using the dedicated sparse dictionary as a processing benchmark to obtain a frequency and voltage regulation coordinated control method. The frequency and voltage regulation coordinated control system uses a set of sparse components with responsive directionality, while simultaneously filtering out redundant data components in the real-time operating data set that are not related to the frequency and voltage regulation coordinated control. An inverse transformation operation is performed on the set of sparse components with responsive directionality to reconstruct the key operating data required for the coordinated control of frequency and voltage regulation between the virtual power plant and the main power grid. Error verification is performed on the key operating data, and abnormal data units with reconstruction deviations exceeding a preset range are removed to obtain the verified key operating data. Based on the key operating data, a coordinated control simulation operation is performed to obtain the coordinated control commands for frequency and voltage regulation between the virtual power plant and the main power grid. These commands are then transmitted to the corresponding control nodes of the virtual power plant and the main power grid to complete the coordinated control of frequency and voltage regulation.

[0005] According to another aspect of the present invention, a control device is provided, comprising: a processor; and a memory, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the method as described above.

[0006] This invention collects real-time operational data generated during the interaction between a virtual power plant and a large power grid, covering the electrical status and response information of the interaction between the two, avoiding information loss and bias caused by collecting data from a single subject. Based on the statistical distribution attributes of the data, a dedicated sparse dictionary is generated to adapt to the specific scenario, matching the inherent characteristics of the target data and solving the problem of insufficient scenario adaptation of predefined dictionaries. Sparse representation is performed based on the dedicated sparse dictionary, simultaneously filtering out core data pointing to frequency and voltage regulation control and eliminating redundant content to reduce interference from invalid data in subsequent processes. Error verification is performed on the reconstructed key operational data, eliminating abnormal data units to ensure the reliability of data entering the control stage. Finally, collaborative control commands are generated based on the verified key data, and these commands perfectly match the actual operating state of the interactive scenario, improving the accuracy and effectiveness of frequency and voltage regulation collaborative control.

[0007] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present invention. Attached Figure Description

[0008] Figure 1 This is a schematic diagram illustrating an application scenario provided by an embodiment of the present invention.

[0009] Figure 2 This is a schematic diagram illustrating the implementation process of a frequency and voltage regulation coordinated control method for interaction between a virtual power plant and a large power grid, provided in an embodiment of the present invention.

[0010] Figure 3 This is a schematic diagram of the hardware entity of a control device provided in an embodiment of the present invention. Detailed Implementation

[0011] The frequency and voltage regulation coordinated control method for interaction between virtual power plants and large power grids provided in this invention can be applied to, for example... Figure 1In the application environment shown, the distributed resource node 102 communicates with the control device 104 via a network. A data storage system can store the data that the control device 104 needs to process. The data storage system can be integrated into the control device 104 or located in the cloud or on other network servers. The operational data can be stored in the local storage of the distributed resource node 102, or in the data storage system or cloud storage associated with the control device 104. When data processing is required, the control device 104 can retrieve the operational data from the local storage of the distributed resource node 102, the data storage system, or the cloud storage. The control device 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0012] Specifically, data acquisition and monitoring units are installed at each of the distributed resource nodes 102 aggregated in the virtual power plant, such as the inverters of a distributed photovoltaic power station, the battery management system of an energy storage system, and the intelligent control terminal of an interruptible load. These units include transformers for measuring voltage and current, power transmitters for calculating power, and GPS timing modules for recording time information. These acquisition units continuously collect instantaneous voltage and current waveform data at the nodes at a preset sampling frequency. The acquired analog signals are then converted into digital signals by an analog-to-digital converter and preliminarily encapsulated by an embedded processor according to a communication protocol. For example, for an energy storage system, in addition to collecting voltage and current, its battery management system also collects state of charge, temperature, and other status information. This information, along with interactive response information such as the start time of responding to the virtual power plant's automatic power generation control commands and the time required to reach the target power ratio, is packaged into a data packet containing a timestamp. Then, the data from all distributed resource nodes is aggregated in real time to the virtual power plant's data concentrator or cloud data processing platform via high-speed communication networks such as industrial Ethernet, 5G wireless communication, or fiber optic communication. In this platform, heterogeneous data streams from different sources with different time labels first need to undergo time synchronization calibration to ensure that all data are based on a unified time reference. Finally, the cleaned and preliminarily sorted real-time operational data is organized according to a preset data structure to form a real-time operational data set containing all electrical status information and interaction response information, which is stored in a real-time database for use by upper-layer applications.

[0013] Please refer to Figure 2 The frequency and voltage regulation coordinated control method for interaction between virtual power plants and large power grids provided in this embodiment of the invention specifically includes the following steps: Step S100: Obtain the real-time operation data set generated during the interaction between the virtual power plant and the large power grid. The real-time operation data set carries the electrical status information and interaction response information of each distributed resource of the virtual power plant during the interaction with the large power grid.

[0014] A virtual power plant is an entity that aggregates, coordinates, and optimizes distributed resources such as distributed generation, energy storage, and controllable loads, participating in grid operation and the electricity market as a special type of power plant. A large power grid refers to a vast power network system composed of generation, transmission, transformation, and distribution equipment used for power transmission and distribution. Real-time operational data sets refer to the data sets collected and aggregated in real time by measuring devices deployed at various distributed resource nodes and grid-connected interfaces during energy exchange and information interaction between the virtual power plant and the large power grid. Electrical status information specifically describes the electrical quantity characteristics of distributed resources during operation, including but not limited to core electrical parameters characterizing the system's operating status such as voltage, current, frequency, active power, and reactive power. Interactive response information specifically describes the response behavior and results of distributed resources after receiving regulation commands, including data reflecting the interaction effect such as response time, regulation rate, regulation accuracy, and response duration.

[0015] Step S200: Based on the statistical distribution attributes carried by the real-time operating data set, generate a dedicated sparse dictionary that adapts to the frequency and voltage regulation scenarios of the interaction between the virtual power plant and the large power grid. The dedicated sparse dictionary and the statistical distribution attributes of the real-time operating data set form a unique correspondence.

[0016] Statistical distribution attributes refer to the statistical characteristics of various types of data in a real-time operating dataset in terms of numerical distribution, variation patterns, correlations, and probabilistic features. Examples include the fluctuation range of voltage data, the central value and random fluctuation characteristics of frequency data, and the periodicity of load power data. A dedicated sparse dictionary is a structured data representation system specifically built for frequency and voltage regulation scenarios in virtual power plants and large power grids. Its sparsity means that the dictionary aims to represent most of the key information in the original dataset using as few linear combinations of primitives as possible. Adapting to frequency and voltage regulation scenarios means that the dictionary's primitive design and organization closely revolve around the two core tasks of frequency regulation and voltage control, efficiently capturing and expressing data features related to frequency and voltage changes. Forming a unique correspondence means that the structure, primitive content, and organization of the generated dedicated sparse dictionary strictly match the inherent statistical distribution attributes of the current real-time operating dataset.

[0017] In an optional embodiment, step S200 may specifically include the following steps S210 to S260: Step S210: Perform time-series correlation on the real-time operation data set, connect the data points of each distributed resource of the virtual power plant and the large power grid throughout the entire time period, mark the causal relationship paths between data in different time periods, and obtain the real-time operation data distribution correlation map. The nodes of the real-time operation data distribution correlation map correspond to a single real-time operation data, and the edge data corresponds to the causal relationship description between the data.

[0018] Specifically, firstly, in the virtual power plant data processing platform, all real-time operational data with timestamps collected in step S100 are globally sorted according to the ascending order of timestamps, forming a basic time series. Then, a pre-built knowledge base of the physical laws and logical rules of power system operation is used to initially identify potential causal pairs in the data series. For the initially determined causal relationships, statistical methods such as the Granger causality test can be used for verification, confirming causality by analyzing the predictive power of the lagged value of one time series to the current value of another time series. After confirming the causal relationship, two corresponding nodes are created in the real-time operational data distribution correlation graph, and a directed edge is drawn from the cause node to the result node, with the corresponding causal description labeled on the edge. By traversing all data points and repeating this process, a real-time operational data distribution correlation graph composed of nodes and directed edges is gradually constructed. This graph not only shows the changes in data over time, but more importantly, reveals the driving logic behind these changes.

[0019] Step S220: Locate scene anchor points in the real-time running data distribution association map, extract the map nodes and associated paths directly related to frequency modulation and voltage regulation actions, mark the interaction action triggering conditions and response results corresponding to each path, and obtain the frequency modulation and voltage regulation scene interaction logic anchor point set. The content of the frequency modulation and voltage regulation scene interaction logic anchor point set covers all triggering and response links of frequency modulation and voltage regulation.

[0020] Specifically, a feature template for frequency and voltage regulation scenarios is first defined, containing rules and keywords for identifying scenario anchor points. For example, node data labels may contain key fields such as "frequency," "voltage," "AGC," and "AVC," or the data value represented by a node may exceed the preset frequency and voltage regulation dead zone. All nodes are traversed in the real-time running data distribution correlation graph, and node attributes are matched against the feature template. When a matching node is found, it is marked as a trigger anchor point. Then, starting from this anchor point, a series of subsequent nodes are traced along its outgoing edges, while simultaneously tracing back along its incoming edges to find the cause of the event. All traced nodes directly related to frequency and voltage regulation actions, as well as the paths connecting them, are marked as action anchor points and response anchor points. For each marked path, the state of the starting node is extracted as the trigger condition, and the state change of the ending node is extracted as the response result. All marked anchor points and anchor paths, along with their trigger conditions and response results, are organized and cataloged according to the logical order of event occurrence, ultimately forming a set of interactive logic anchor points for frequency and voltage regulation scenarios, fully covering all response stages in the frequency and voltage regulation process.

[0021] Step S230: Perform primitive transformation on the frequency modulation and voltage regulation scenario interaction logic anchor point set, and decompose the interaction logic corresponding to each anchor point into the smallest unit that can be matched repeatedly. Each unit corresponds to the feature description of a type of interaction action. All the smallest units are sorted together to obtain a dedicated sparse dictionary primitive library. The number of its units corresponds to the number of interaction categories in the frequency modulation and voltage regulation scenario interaction logic anchor point set.

[0022] For example, we first analyze each anchor node and each anchor path in the frequency and voltage regulation scenario interaction logic anchor point set, decomposing the complex interaction logic into a series of logical fragments such as trigger conditions, action commands, device responses, and system state changes. For each logical fragment, we further analyze its data characteristics and abstract these characteristics into a mathematical expression or template composed of relevant parameters, i.e., the prototype of a primitive. In this way, we decompose and abstract the features of all nodes and paths in the anchor point set one by one, generating a large number of templates describing single-type interaction actions or system state characteristics. To ensure the universality and repeatability of these templates, we use a clustering algorithm to merge templates with very similar features into a standardized primitive. All the final indivisible and representative standardized primitives are classified, organized, and stored in an orderly manner according to the interaction action category they describe, thus constructing a dedicated sparse dictionary primitive library.

[0023] Step S240: Perform hierarchical mapping on the dedicated sparse dictionary primitive library. Set the hierarchical structure of the primitive library according to the distribution hierarchy of the real-time running data, so that the global coverage level primitive corresponds to the global data distribution, and the single data level primitive corresponds to the single data detail. After completing the hierarchical mapping, the initial dedicated sparse dictionary is obtained. The hierarchical structure of the initial dedicated sparse dictionary matches the distribution hierarchy of the real-time running data.

[0024] For example, the hierarchical structure of the real-time running data set obtained in step S100 is analyzed. Methods such as hierarchical clustering or wavelet analysis are used to decompose the data at multiple scales, breaking down the original time series data into approximate parts reflecting long-term trends and detailed parts reflecting instantaneous fluctuations, each corresponding to a different data level. After clarifying the hierarchical structure of the data, a hierarchical mapping is performed on the primitive library. Primitives that describe macroscopic trends and overall structure are assigned to the top level of the dictionary, while primitives that describe specific device behavior and detailed characteristics are assigned to the bottom level. Intermediate levels are set between the top and bottom levels as needed. Through this top-down mapping, the originally flat primitive library is organized into a tree-like hierarchical structure. Each upper-level primitive can be associated with multiple lower-level primitives, and the lower-level primitives are specific expansions of the macroscopic patterns represented by the upper-level primitives in different dimensions or details.

[0025] Step S250: Bind the initial dedicated sparse dictionary to the statistical distribution attributes of the real-time running data set one-to-one, so that each dictionary level corresponds to a feature of a statistical distribution attribute, and each dictionary primitive corresponds to a detailed description of a statistical distribution attribute. After completing the binding operation, the bound dedicated sparse dictionary is obtained. The binding relationship of the bound dedicated sparse dictionary covers all levels and primitives of the dictionary.

[0026] For example, a thorough statistical analysis is first performed on the real-time running dataset to quantify its statistical distribution attributes. This includes calculating the multidimensional probability density function and covariance matrix of the entire dataset, calculating the statistical characteristics of each layer of data after decomposition according to the time scale, and extracting statistical features such as duration, amplitude, and rate of change from typical event segments in the data. Next, binding relationships are established. For the top-level primitives of the initial dedicated sparse dictionary, they are bound to the global statistical features of the corresponding data in the entire dataset, and statistical attribute labels are attached to the top-level primitives in the dictionary's metadata. For the intermediate and bottom-level primitives of the dictionary, they are bound to the statistical attributes of the corresponding level of data or event segments. This binding process is a systematic operation that traverses all levels and all primitives of the dictionary, determining the specific statistical attribute to be bound by calculating the matching degree between the data pattern represented by each primitive and the corresponding feature in the dataset. Finally, after completing all bindings, the resulting dedicated sparse dictionary has each node, whether level or primitive, becoming a quantifiable data feature descriptor with clear statistical connotations.

[0027] Step S260: Apply the bound dedicated sparse dictionary to match the real-time running data, adjust the description content and hierarchical association of the dictionary primitives according to the matching results, until the matching coverage of the dictionary and the real-time running data reaches the scenario requirements, and obtain a dedicated sparse dictionary adapted to the frequency and voltage regulation scenario of virtual power plant and large power grid interaction.

[0028] In an optional embodiment, step S260 may specifically include the following steps S261 to S266: Step S261: Perform element-by-element matching on the dedicated sparse dictionary and real-time operation data set adapted to the frequency and voltage regulation scenario of virtual power plant and large power grid interaction, record the data matching status and coverage of each dictionary primitive, arrange all the recorded contents to obtain the dictionary primitive matching degree description set, and each record in the dictionary primitive matching degree description set corresponds to the matching information of a dictionary primitive.

[0029] Specifically, each primitive in the dedicated sparse dictionary is defined as a feature template, which can be in the form of a numerical vector, waveform fragment, or logical rule. Then, the real-time running dataset is traversed. For each data point or data fragment captured by a sliding window, its similarity is calculated with all primitives in the dictionary. Similarity calculation methods can include Euclidean distance, cosine similarity, or correlation coefficient. If the calculated similarity exceeds a preset matching threshold, the data point or data fragment is considered a successful match with the primitive. The index of each successfully matched primitive, the number of matches, and the similarity value for each match are recorded. After traversing the entire dataset, the matching records for each primitive are statistically summarized. The matching coverage of each primitive is calculated, i.e., the proportion of its matched data points to the total number of data points, as well as the distribution of matched data on the time axis. Key information such as the identifier, number of matches, average matching similarity, and coverage percentage of each primitive is compiled into a record. The sum of all primitive records yields the dictionary primitive matching description set.

[0030] Step S262: Perform gap identification on the dictionary primitive matching degree description set, identify the dictionary primitives whose matching coverage does not meet the requirements, record the real-time running data category and interaction scenario corresponding to each gap primitive, and collect all gap information in an orderly manner to obtain a primitive optimization requirement list. The primitive optimization requirement list clearly defines the optimization direction and corresponding scenario for each gap primitive.

[0031] For example, the coverage range value of each primitive is first extracted from the dictionary primitive matching degree description set. Coverage range requirements are preset for different types of primitives; for example, the coverage range requirement may be higher for primitives describing normal operating data, while the requirement may be relatively lower for primitives describing rare disturbance events. The actual coverage range of each primitive is compared with its corresponding requirement threshold. If the actual coverage range of a primitive is lower than the requirement threshold, it is marked as a gap primitive. For each marked gap primitive, the relevant information bound in steps S230 and S250 is further traced, including its corresponding real-time operating data category, such as frequency deviation data or reactive power data, and its associated interaction scenario, such as whether it is related to frequency rise or voltage drop. This information, along with the primitive's identifier and the current coverage range value, is recorded to form a complete record containing primitive ID, data category, interaction scenario, current coverage, and optimization direction suggestions. All gap primitive records are systematically collected according to data category or interaction scenario, ultimately forming a primitive optimization requirement list.

[0032] Step S263: Extract details from the real-time running data set corresponding to the primitive optimization requirement list, extract all interaction data details in the scenario corresponding to the missing primitive, transform the extracted details into the smallest matching unit, and sort all the supplementary units to obtain the supplementary primitive set. The representation of the supplementary primitive set is consistent with the original dictionary primitives.

[0033] For example, based on the interaction scenario information recorded in the primitive optimization requirement list, such as the scenario of "excessive delay in automatic voltage control response," all data records related to this scenario are filtered from the raw real-time running data set obtained in step S100. Filtering conditions can include combinations of timestamp range, event tags, and data types. For example, voltage, reactive power, and related equipment status data within 10 seconds before and after the "AVC command issuance" event can be extracted. For each extracted scenario data category, primitive transformation is performed using the same method as in step S230. These data fragments undergo preprocessing such as alignment and normalization. Then, feature extraction algorithms, such as dynamic time warping or principal component analysis, are used to extract feature vectors or waveform templates that represent the common patterns of this type of scenario data. Next, these feature templates are encapsulated according to the original dictionary primitive format, assigned unique identifiers and feature descriptions, forming new primitives. By processing the data corresponding to all gap scenarios, a batch of new primitives is generated. All newly generated primitives are systematically grouped according to their data category or interaction scenario to form a supplementary primitive set.

[0034] Step S264: Embed the supplementary primitive set into the dedicated sparse dictionary adapted to the frequency and voltage regulation scenario of virtual power plant and large power grid interaction, and supplement it to the level and position specified in the primitive optimization requirement list. Adjust the association path between the supplementary primitive and the original primitive so that the supplementary primitive is integrated into the hierarchical structure of the dictionary, and obtain the expanded dedicated sparse dictionary. The hierarchical structure of the expanded dedicated sparse dictionary remains intact.

[0035] For example, the primitive optimization requirement list is first parsed to determine the level and category of each missing primitive in the original dictionary. Then, each new primitive in the supplementary primitive set is matched with these levels and categories to determine its specific insertion position. For instance, if the gap occurs on a bottom-level primitive describing a "primary frequency modulation" scenario, the newly generated primitive describing the newly emerging response pattern in that scenario will be inserted into the bottom-level position under the "primary frequency modulation" category in the dictionary. After insertion, the dictionary index and association relationships need to be updated, the data characteristics of the new primitive are analyzed, and its similarity and logical relationship with other primitives in the same, upper, and lower layers are calculated. If the new primitive and the macroscopic pattern described by a certain upper-level primitive have a part-whole relationship, then an inclusion relationship is established from the upper-level primitive to the new primitive. If the new primitive and the primitive in the same layer describe a sequential relationship, such as describing the "power change" process after "command reception", then a temporal association path is established between the two. By establishing these associations one by one, the new primitive is no longer an isolated point.

[0036] Step S265: Perform element-by-element matching again on the expanded dedicated sparse dictionary and the real-time running data set, record the matching status and coverage of each dictionary primitive and supplementary primitive, and arrange all the recorded contents to obtain the expanded dictionary matching degree description set. The recording format of the expanded dictionary matching degree description set is consistent with the initial matching degree description set.

[0037] The second element-by-element matching refers to performing a completely new matching evaluation on the expanded dedicated sparse dictionary and the same real-time running dataset using the exact same method and parameters as in step S261, after dictionary expansion. Each dictionary primitive and supplementary primitive participates in this matching, and their respective matching status and coverage are recorded. The expanded dictionary matching degree description set is the record set of this matching result, and its record format and included fields are completely consistent with the initial matching degree description set obtained in step S261, thus ensuring the consistency of the comparison before and after. The execution process of this step aims to objectively evaluate the actual effect of the dictionary expansion operation in step S264, verify whether the supplementary primitives have successfully resolved the original matching gaps, and whether the expansion operation has affected the matching performance of the original primitives.

[0038] Step S266: Perform a compliance check on the expanded dictionary matching degree description set to confirm that the matching coverage of all dictionary primitives meets the requirements. If there are primitives that do not meet the requirements, repeat the above optimization process until all primitives meet the requirements. Then, the optimized dedicated sparse dictionary is obtained. The matching coverage of the optimized dedicated sparse dictionary and the real-time running data meets the scenario requirements.

[0039] In an optional embodiment, step S266 may specifically include the following steps S2661 to S2666: Step S2661: Compare the statistical distribution attributes corresponding to the optimized dedicated sparse dictionary with the latest statistical distribution attributes of the real-time running data set, record the changes and ranges of each type of attribute, and arrange all comparison results to obtain a distribution attribute difference description set. Each record in the distribution attribute difference description set corresponds to the change information of a type of distribution attribute.

[0040] For example, firstly, the metadata of the optimized dedicated sparse dictionary is extracted from the database. This metadata includes the statistical distribution attributes of the real-time running dataset to which the dictionary was generated, denoted as the baseline statistical attributes. Then, the latest collected real-time running dataset with the most recent time window is subjected to the same comprehensive statistical analysis as in step S200 to obtain the latest statistical distribution attributes. Next, the class-by-class comparison module is activated. This module compares the latest statistical attributes with the baseline statistical attributes one by one according to a preset attribute list, such as voltage mean, frequency variance, and correlation coefficient between active power and frequency. For each type of attribute, its change or rate of change is calculated. For example, the mean offset is obtained by subtracting the baseline frequency mean from the latest frequency mean, and the variance change factor is obtained by dividing the latest frequency variance by the baseline frequency variance. The name, baseline value, latest value, change / rate of change, and whether the change is significant are recorded as a record for each type of attribute. The records of all attributes are summarized together to obtain the distribution attribute difference description set.

[0041] Step S2662: Locate the dictionary primitives corresponding to the distribution attribute difference description set, determine the dictionary primitives directly associated with the changed attributes, record the hierarchical position and association path of each primitive to be adjusted, and sort all the primitives to be adjusted to obtain the dictionary primitive set to be adjusted. The content of the dictionary primitive set to be adjusted covers all primitives related to attribute changes.

[0042] For example, first analyze the distribution attribute difference description set, extract statistical attributes whose change amount or rate of change exceeds a preset significant threshold, and mark them as significant change attributes. Then, traverse all primitives in the optimized dedicated sparse dictionary, and check the statistical attribute labels bound to the metadata of each primitive. If the attribute bound to a primitive is a significant change attribute, or if the primitive's own feature description parameters, such as frequency threshold or duration range, are directly related to the numerical range of the significant change attribute, then the primitive is marked as a primitive to be adjusted. For example, if the latest statistical data shows that the "90% probability interval of frequency deviation" has expanded from ±0.03 Hz to ±0.05 Hz, then all primitives related to the frequency deviation threshold, such as the "low frequency limit trigger" primitive, will have their internal threshold parameters adjusted. Record the unique identifier of each primitive to be adjusted, its complete path in the dictionary hierarchy, such as "top level / frequency class / trigger condition subclass / low frequency limit trigger", and its association with the primitives above and below.

[0043] Step S2663: Extract the latest statistical distribution attributes of the real-time running data set corresponding to the primitive set of the dictionary to be adjusted, extract the interaction data details corresponding to the changed attributes, replace the description of the primitive to be adjusted with the smallest unit form of the corresponding details, and obtain the adjusted primitive set of the dictionary. The description form of the adjusted primitive set of the dictionary is consistent with the original primitive.

[0044] Specifically, first, based on the interaction scenario associated with each primitive in the set of primitives to be adjusted, such as the "frequency modulation action" scenario, the latest data fragments belonging to that scenario are filtered from the latest real-time running data set. The filtering time window should cover the data from the last dictionary update to the current moment. For the filtered data fragments, the same primitive transformation method as in step S263 is used for processing. Taking the "low frequency limit trigger" primitive, which requires threshold adjustment, as an example, all data fragments marked as "low frequency limit events" are extracted from the latest data, and the distribution of the actual frequency deviation values ​​when these events are triggered is analyzed. By statistically averaging or specifying quantiles of these deviation values, the representative threshold of this triggering condition under the new data distribution is calculated, for example, adjusting the original -0.03 Hz to -0.04 Hz. This new threshold parameter, along with other potentially changing features, such as event duration, is encapsulated into a new primitive template, the format of which is completely consistent with the original primitives. After performing this operation on all primitives to be adjusted, a set of adjusted dictionary primitives matching the latest data distribution is obtained.

[0045] Step S2664: Replace the corresponding primitives to be adjusted in the optimized dedicated sparse dictionary with the adjusted dictionary primitive set, adjust the association path between the replaced primitives and other primitives, so that the replaced primitives are integrated into the hierarchical structure of the dictionary, and obtain the dynamically adjusted dedicated sparse dictionary. The hierarchical structure of the dynamically adjusted dedicated sparse dictionary remains intact.

[0046] Specifically, the process begins by locating the exact position of each primitive within the set of primitives to be adjusted in the optimized dedicated sparse dictionary. Then, the content of the original primitive is directly overwritten with the content of the corresponding adjusted primitive. After content replacement, the association paths of the primitives need to be checked and adjusted. Since the key feature parameters of the primitives may have changed, the inclusion relationship with the parent primitive and the instantiation relationship with the child primitives need to be re-evaluated. For example, if a parent primitive representing "severe frequency violation" originally contained a child primitive with a threshold of -0.05 Hz, and this child primitive's threshold is adjusted to -0.06 Hz, it needs to be confirmed whether this new threshold still falls within the "severe frequency violation" category. If not, it is moved to another parent primitive. These associations are automatically adjusted or manually confirmed by calculating the similarity of the new primitive's features with the candidate parent primitives and its compatibility with the candidate child primitives. By establishing new and reasonable association paths, it is ensured that the new primitive can be correctly integrated into the dictionary's knowledge network. After replacing and associating all primitives to be adjusted, a dynamically adjusted dedicated sparse dictionary is obtained. The contents of this dictionary have been updated and can better represent the characteristics of the current running data.

[0047] Step S2665: Perform element-by-element matching again on the dynamically adjusted dedicated sparse dictionary and the real-time running data set, record the matching status and coverage of each dictionary primitive, and arrange all the recorded contents to obtain the dynamically adjusted dictionary matching degree description set. The recording format of the dynamically adjusted dictionary matching degree description set is consistent with the previous matching degree description set.

[0048] The next element-by-element matching step refers to performing a completely new matching evaluation between the dynamically adjusted sparse dictionary and the latest real-time running dataset, using the exact same methods and parameters as steps S261 and S265, after the dictionary has been dynamically adjusted. Each dictionary primitive, including those that have been replaced and those that haven't changed, participates in this matching, and their matching status and coverage are recorded. The dynamically adjusted dictionary matching degree description set is the record set of this matching result, and its recording format is completely consistent with previous matching degree description sets, such as the initial matching degree description set and the expanded dictionary matching degree description set. The execution of this step aims to objectively evaluate the actual effect of the dictionary dynamic adjustment operation in step S2664, verify whether the adjusted primitives have successfully adapted to the new data distribution, and whether the adjustment operation has had the expected impact on the overall matching performance of the dictionary.

[0049] Step S2666: Perform a compliance check on the dynamically adjusted dictionary matching degree description set to confirm that the matching coverage of all dictionary primitives meets the requirements. If there are primitives that do not meet the requirements, repeat the above adjustment process until all primitives meet the requirements. Then, a dedicated sparse dictionary adapted to the latest distribution attributes is obtained. The dedicated sparse dictionary adapted to the latest distribution attributes corresponds completely to the latest statistical distribution attributes of the real-time running data.

[0050] The compliance determination refers to comparing the coverage of each primitive recorded in the dynamically adjusted dictionary matching description set with the preset scenario requirement threshold. Confirming that the matching coverage of all dictionary primitives meets the requirements means that after dynamic adjustment, each primitive in the dictionary demonstrates sufficient coverage capability on the latest real-time running data set. If there are still primitives that do not meet the requirements, the dynamic adjustment process starting from step S2662 needs to be repeated, i.e., relocating primitives related to the latest distribution differences, performing detailed extraction and replacement, and re-matching and determining, forming a dynamic closed-loop optimization. The dedicated sparse dictionary adapted to the latest distribution attributes is a dictionary version that, after one or more rounds of dynamic adjustment, ultimately ensures that all primitives perfectly match the latest data statistical characteristics, forming a complete correspondence with the latest statistical distribution attributes of the real-time running data. The execution process of this step aims to finally confirm the dynamic adaptability of the dictionary, ensuring that the dictionary can be updated synchronously as the system's running data evolves.

[0051] Step S300: Using a dedicated sparse dictionary as the processing benchmark, perform sparse representation operations on the real-time running data set to obtain a set of sparse components that have responsiveness to frequency modulation and voltage regulation coordinated control, while filtering out redundant data components in the real-time running data set that are not related to frequency modulation and voltage regulation coordinated control.

[0052] In an optional embodiment, step S300 may specifically include the following steps S310 to S360: Step S310: Match each real-time running data set with the dedicated sparse dictionary adapted to the latest distribution attributes, map each real-time running data to the dictionary primitive with the highest matching degree, record the primitive correspondence and matching details of each data, arrange all the records to obtain the data-primary association mapping table, and each record in the data-primary association mapping table corresponds to the primitive matching status of one real-time running data.

[0053] Specifically, the process begins by acquiring the latest real-time running dataset and loading a dedicated sparse dictionary adapted to the latest distribution attributes. A matching degree calculation function is defined; for example, for data and primitives in vector form, their cosine similarity is calculated, which is the dot product of the data vector and the primitive vector divided by the product of their respective magnitudes. This process iterates through each data record in the real-time running dataset. For the current data record, it is represented as a vector. Then, each primitive in the dedicated sparse dictionary is represented as a vector, and the cosine similarity between the data vector and the primitive vector is calculated. After iterating through all primitives, the maximum similarity value and its corresponding primitive identifier are found. If the maximum value exceeds a preset minimum matching threshold, the unique identifier of this data, the identifier of the matched primitive, and the matching similarity score are recorded. If the maximum value is below the threshold, it may be marked as "no match" or covered by a specific primitive. All this information is then organized into a table according to the original data order or index. Each row of the table corresponds to one original data record, containing fields such as data ID, best matching primitive ID, and matching similarity.

[0054] Step S320: Perform scenario matching on the primitives in the data-primitive association mapping table, map the primitives to the anchor points in the frequency modulation and voltage regulation scenario interaction logic anchor point set, mark the anchor point category and interaction action corresponding to each data, arrange all the marked contents to obtain the response orientation data tag table, and each record in the response orientation data tag table corresponds to the response orientation information of one data.

[0055] For example, first load the anchor point set for the frequency modulation and voltage regulation scenario interaction logic. This anchor point set itself is an information network containing nodes and paths. Establish a mapping dictionary, associating each primitive ID in the dedicated sparse dictionary with the anchor ID or path ID in the anchor point set. This association can be initially established during dictionary generation and optimization, and explicitly defined in the binding operation of step S250. For example, the "frequency low limit trigger" primitive generated in step S230 is associated with the anchor node "trigger condition: frequency low limit" in the anchor point set. Traverse each record in the data-primitive association mapping table, and based on the "best matching primitive ID" in the record, search for its corresponding anchor information in the aforementioned mapping dictionary. The found information may include the scenario category to which the primitive belongs, such as "frequency modulation," and a specific interaction action description, such as "trigger one frequency modulation." This anchor information, including the anchor ID, scenario category, and interaction action description, is appended as a new field to the corresponding record in the data-primitive association mapping table. After all records are processed, the expanded new table is the response-oriented data tag table.

[0056] Step S330: Filter the response directional data label table, extract the real-time operation data that is directly related to frequency and voltage regulation actions, and collect all the filtered data in an orderly manner to obtain a subset of real-time operation data with response directional characteristics. The content of the subset of real-time operation data with response directional characteristics covers all real-time operation data directly related to frequency and voltage regulation actions.

[0057] For example, first define filtering rules based on the field content in the response-oriented data tag table. For instance, set the filtering conditions to "scene category field value is 'frequency modulation' or 'voltage regulation'" and "interaction action field value does not contain keywords such as 'monitoring', 'recording', or 'environment'". Create an empty data container to store the filtering results. Then, iterate through each record in the response-oriented data tag table. For the current record, read its "scene category" and "interaction action" fields and match them against the preset filtering rules. If the record's field values ​​completely match the filtering conditions, such as the scene category being "frequency modulation" and the interaction action being "speed controller action", then extract the corresponding complete data content from the original real-time running data set based on the record's data ID and store it in the result container. If the record does not meet the filtering conditions, such as the scene category being "monitoring" or the interaction action being "ambient temperature", then skip the record and do not extract data from it. After traversing the entire response-oriented data labeling table, all the data collected in the result container—that is, the data directly related to frequency and voltage regulation actions—is ordered and rearranged according to time sequence or data category to form a real-time operational data subset with response orientation. This subset is much smaller than the original dataset, but retains all the core information needed for frequency and voltage regulation control decisions.

[0058] Step S340: Perform sparse transformation on the real-time running data subset with response orientation, and correspond each data to a combination of dictionary primitives so that the transformed content retains only the part related to response orientation. Sort all the transformed content into an ordered collection to obtain a sparse component set with response orientation. The content structure of the sparse component set with response orientation is consistent with the hierarchical structure of the dedicated sparse dictionary that adapts to the latest distribution attributes.

[0059] In an optional embodiment, step S340 may specifically include the following steps S341 to S346: Step S341: Perform component-by-component matching on the set of sparse components with responsive orientation and the dedicated sparse dictionary adapted to the latest distribution attributes. Record the dictionary primitive combination and association path corresponding to each sparse component. Arrange all the records to obtain the component-primary correspondence table. Each record in the component-primary correspondence table corresponds to the primitive association of a sparse component.

[0060] Specifically, first, a set of sparse components with response-oriented characteristics and a dedicated sparse dictionary adapted to the latest distribution attributes are loaded. Each component in the sparse component set may be represented at the data level as a vector or list containing primitive indices and coefficient values. Each component in the sparse component set is traversed. For the current component, its data structure is parsed, extracting a list of unique identifiers for all primitives constituting the component, as well as the coefficient value corresponding to each primitive. Then, for each primitive identifier in the list, the detailed hierarchical path of the primitive is searched in the dedicated sparse dictionary, such as the primitive "top-level / frequency class / response result subclass / frequency recovery," and the predefined associations of this primitive with other primitives, such as its parent node being "response result." The identifier of the current component, the list of IDs constituting the primitives and their coefficients, the hierarchical path of each primitive, and the combination relationships between primitives are compiled into a record. All the records for all components are summarized together to obtain the component-primary correspondence table.

[0061] Step S342: Perform scene mapping on the primitive association paths in the component-primary correspondence table, map the paths to the interaction paths in the frequency modulation and voltage regulation scene interaction logic anchor point set, mark the complete interaction action chain corresponding to each component, arrange all the marked contents to obtain the component interaction path description set, and each record in the component interaction path description set corresponds to the interaction path information of a component.

[0062] Specifically, first, the set of anchor points for the frequency and voltage regulation scenario interaction logic is loaded, containing a series of directed paths such as "frequency exceeding limits," "issuing adjustment commands," "device response," and "frequency recovery." Then, each record in the component-primary correspondence table, i.e., each sparse component, is traversed. For the current component, a list of its constituent primitives is obtained. Next, based on the anchor nodes associated with each primitive in the anchor point set, an attempt is made to arrange these nodes according to the predefined logical paths in the anchor point set. For example, if a component's primitive combination contains a "low frequency limit violation" primitive and a "energy storage discharge power increase" primitive, the anchor set will be searched for a path connecting these two anchor nodes. This might result in a path like "low frequency limit violation", "trigger frequency modulation command", or "energy storage discharge power increase". The primitives in the component are then filled into the corresponding links of this path, thus constructing a specific interaction action chain belonging to that component, such as "low frequency limit violation (0.2 weight), trigger frequency modulation command, energy storage discharge power increase (0.9 weight)". This constructed interaction action chain, along with the primitive weights corresponding to each link in the chain, is recorded as the interaction path information of that component.

[0063] Step S343: Perform frequency statistics on the interaction action chains in the component interaction path description set, record the number of action triggers and responses in each chain, convert them into frequency descriptions of the interaction actions corresponding to the components, and arrange all statistical results to obtain the component interaction intensity description set. Each record in the component interaction intensity description set corresponds to the interaction intensity information of a component.

[0064] Specifically, a statistical time window is first determined, such as the past 5 minutes or 1 hour. The component interaction path description set is then traversed. Each record in this set corresponds to a sparse component and contains information about its interaction action chain. Within the selected time window, the unique interaction action chain pattern is counted. For example, all components consisting of "low frequency exceeding limit" and "increased energy storage discharge power" are considered to be of the same frequency modulation response pattern, and the total number of times this type of pattern appears within the time window is counted. Simultaneously, the relative frequency of each link in each interaction chain can also be counted, such as the ratio of trigger events to response events. Key indicators such as the sparse component or its represented pattern category, the total frequency of occurrence within the statistical window, the number of triggers, and the number of responses are recorded as the interaction strength information of that component. All this information for all components is then summarized to obtain the component interaction strength description set.

[0065] Step S344: Assign corresponding frequency descriptions to the component interaction intensity description set, convert the frequency description of each component into its proportion description in collaborative control, record the proportion value of each component and the corresponding interaction path, arrange all the assignment results to obtain the component proportion allocation table, and each record in the component proportion allocation table corresponds to the proportion information of a component.

[0066] For example, first, obtain the occurrence frequency of all sparse components from the component interaction intensity description set, and calculate the sum of the occurrence frequencies of all components, which is the total number of all frequency modulation and voltage regulation related events occurring within the statistical window. Then, for each sparse component, divide its own frequency by this sum to obtain a ratio between 0 and 1. This ratio is the proportion value of that component. For example, if the total frequency is 100 and the frequency of a certain component is 20, then its proportion is 0.2. Record this proportion value and associate it with the identifier of that component and the interaction path information determined in step S342. After performing this calculation for all components, a set of records containing component ID, interaction path, and proportion value is obtained. Summarize these records and sort them in descending order of proportion to obtain the component proportion allocation table.

[0067] Step S345: Weight the set of sparse components with response orientation, bind the content of each component to the corresponding proportion in the component proportion allocation table, so that the weight information becomes the inherent attribute of the component, and sort all the weighted components to obtain the weighted set of sparse components with response orientation. The content structure of the weighted set of sparse components with response orientation is consistent with the hierarchical structure of the dedicated sparse dictionary that adapts to the latest distribution attributes.

[0068] For example, first load the sparse component set with response orientation obtained in step S340, and the component proportion allocation table generated in step S344, and iterate through each component in the sparse component set. For the current component, look up its corresponding proportion value in the component proportion allocation table based on its unique identifier. Once found, add this proportion value as a new attribute field, such as "priority_weight", to the component's metadata. All components that have completed weight binding are then ordered and aggregated according to their original order to obtain a weighted sparse component set with response orientation.

[0069] Step S346: Perform range unification processing on the weighted sparse component set with response orientation, adjust the proportion values ​​of all components to a uniform range, ensure that the proportion comparison between components is clear and directly comparable, and orderly collect all adjusted components to obtain the weighted optimized sparse component set with response orientation. The proportion comparison of the weighted optimized sparse component set with response orientation meets the decision requirements of collaborative control.

[0070] In an optional embodiment, step S346 may specifically include the following steps S3461 to S3466: Step S3461: Perform primitive combination comparison on the weighted optimized sparse component set with response orientation to determine the sparse components with completely consistent primitive combinations, record the number of members and proportion of each repeating component group, arrange all the identification results to obtain the repeating component identification list, and each record in the repeating component identification list corresponds to the information of a repeating component group.

[0071] Specifically, first, all components in the weighted optimized sparse component set are formatted, and their primitive combinations are represented as a unique feature string. For example, after sorting by primitive ID, the ID and coefficient are concatenated into a hash value, and an empty dictionary is created to store the grouping information. Then, each sparse component in the set is traversed. For the current component, its feature string is calculated, and it is checked whether a record with that feature string as the key already exists in the dictionary. If it does not exist, a new entry is created in the dictionary with that feature string as the key, and the ID of the current component and its weight percentage are stored in the list under that entry. If it exists, the ID of the current component and its weight percentage are directly appended to the list of the existing entry. After traversing all components, the value list of each key-value pair in the dictionary represents a group of sparse components with completely identical primitive combinations. Traversing this dictionary, for each key, if the length of its corresponding value list is greater than 1, it indicates a duplicate. This record, including the unique identifier of the group (which can be abbreviated from the feature string), the number of members in the group, and a list of weight percentages for each member, is written into the duplicate component identification list.

[0072] Step S3462: Merge the proportions of each group of duplicate components in the duplicate component identification list, sum up the proportion values ​​of all components in the group, retain the primitive combination content of any component in the group, arrange all merging schemes to obtain the duplicate component merging scheme, and each record of the duplicate component merging scheme corresponds to a set of duplicate component merging rules.

[0073] In this embodiment of the invention, each record in the duplicate component identification list generated in step S3461, i.e., each group of duplicate components, can be traversed. For the group of duplicate components pointed to by the current record, the weight percentage values ​​of all components in the group are extracted, and these values ​​are summed to obtain a total percentage value. At the same time, a component is randomly selected from the group, and its complete primitive combination content, including the primitive ID list and the corresponding coefficients, is extracted as the standard content of the merged component. The unique identifier of this group of duplicate components, the merged standard primitive combination content, and the calculated total percentage value are combined into a merging scheme record. After performing this operation on all duplicate groups, a set of merging scheme records is obtained. By summarizing these records, the duplicate component merging scheme is obtained.

[0074] Step S3463: Merge duplicate components in the weighted optimized sparse component set with response orientation. Replace duplicate components with single components after accumulating proportions according to the rules in the duplicate component merging scheme. Orderly collect all merged components to obtain the merged sparse component set with response orientation. The number of components in the merged sparse component set with response orientation is the number of components in the original set after removing duplicates.

[0075] For example, first create an empty result container to store the merged components. Then, iterate through each component in the weighted, optimized set of responsive sparse components. Simultaneously, maintain a list of labeled processed groups. For the currently traversed component, check if it belongs to a group in the duplicate component identification list. If it doesn't belong to any duplicate group, it's unique and is directly copied to the result container. If it belongs to a duplicate group, check if that group has already been processed. If not, create a new sparse component according to the rules defined for that group in the duplicate component merging scheme. The primitive combination content of this new component uses the standard content defined in the scheme, and its weight percentage uses the cumulative percentage defined in the scheme. Add this newly created component to the result container and mark the group as processed. If the group has already been processed, skip the current component and perform no operation. After traversing all components in the original set in this way, each component in the result container is a unique mathematical pattern. This result container is the merged set of responsive sparse components, significantly smaller in size and with higher information density than the original set.

[0076] Step S3464: Compare the interaction paths of the merged sparse component set with response directionality, identify the sparse components with conflicting interaction paths, record the interaction path content and proportion of each conflicting component, and sort all the conflicting components to obtain the interaction path conflict component set. The content of the interaction path conflict component set covers all the components with conflicting interaction paths.

[0077] For example, first load the merged set of sparse components with response directionality, and the set of component interaction path descriptions generated in step S342. Define a conflict detection rule base based on physical laws, such as "the active power of the same device cannot be both positive and negative at the same time" and "the rate of frequency change and the direction of power imbalance must be consistent." Traverse the component set, and for every two or more potentially related components, apply the conflict detection rules to perform a logical consistency check based on their associated devices, timestamps, and interaction path content. For example, compare two components that both involve the same group of energy storage units, one with an interaction path of "discharging" and the other with "charging." If their timestamps overlap, they are determined to be in conflict. Record each component determined to be in conflict, including its identifier, complete interaction path description, its weight percentage, and information about other conflicting components. All recorded conflicting components are summarized to obtain the interaction path conflict component set.

[0078] Step S3465: Perform path verification on the frequency modulation and voltage regulation scenario interaction logic anchor point set corresponding to the interaction path conflict component set, determine the correct interaction logic of each conflict path, record the adjustment direction and target content of each conflict component, arrange all adjustment schemes to obtain the conflict component adjustment scheme, and each record of the conflict component adjustment scheme corresponds to the adjustment rule of a conflict component.

[0079] For example, first, the frequency and voltage regulation scenario interaction logic anchor point set is loaded. This is a knowledge base storing all standard and correct interaction logic. Each conflicting component in the interaction path conflict component set is traversed. For the current conflicting component, its interaction path content is analyzed, and combined with information from other conflicting components, the focus of the conflict is determined. Then, a precise search is performed in the frequency and voltage regulation scenario interaction logic anchor point set to find a standard interaction path that covers the scenario. For example, if the focus of the conflict is that the energy storage unit is both discharging and charging at the same time, the behavior pattern of the energy storage unit in the anchor point set during the same period is checked. Standard logic usually stipulates that the energy storage unit can only be in one of the following states within a control cycle: charging, discharging, or idle. Based on the standard of the anchor point set, it is determined which component's path is misjudged, such as misjudging the "idle" state as "discharging." An adjustment scheme is then formulated for this conflicting component: its interaction path content is corrected from "discharging" to "idle," and its primitive combination or weight is adjusted accordingly. The identifier of each conflicting component and its corresponding adjustment scheme, including the corrected target path content, are recorded. The adjustment schemes for all conflicting components are summarized to obtain the conflicting component adjustment scheme.

[0080] Step S3466: Adjust the conflicting components of the merged sparse component set with responsiveness, modify the interaction path content of the conflicting components according to the rules in the conflicting component adjustment scheme, and sort all the adjusted components to obtain the deduplicated and optimized sparse component set with responsiveness. The content of the deduplicated and optimized sparse component set with responsiveness is not repeated and there is no interaction path conflict.

[0081] For example, first create a result container to store the adjusted components, and then iterate through each component in the merged set of sparse components with responsiveness. For the currently iterated component, check if it exists in the set of conflicting components along the interaction path. If it doesn't exist, the component is logically consistent, and it is directly copied into the result container. If it exists, modify it according to the rules specified for that component in the conflict component adjustment scheme. Modifications may include: regenerating the primitive combination of the component based on the new correct path, which involves re-running sparse coding; or directly modifying the path description field in its metadata; or, in extreme cases, discarding the component if its existence is itself incorrect, without placing it in the result container. After modification, place the modified new component into the result container. After iterating through all components, the result container contains the final cleaned set, which is then ordered, for example, re-sorted by time or by weight, to obtain the deduplicated and optimized set of sparse components with responsiveness.

[0082] Step S350: Perform reverse filtering on the response directional data labeling table, extract real-time operating data that is marked as unrelated to frequency and voltage regulation actions, and orderly collect all the filtered data to obtain a redundant real-time operating data subset. The content of the redundant real-time operating data subset covers all real-time operating data that is unrelated to frequency and voltage regulation actions.

[0083] For example, first define a filtering rule complementary to step S330, such as setting the filtering condition as "the scene category field value is not 'frequency regulation' and not 'voltage regulation'", or "the interaction action field value contains 'monitoring' or 'environment'", and create an empty data container to store redundant data. Then, iterate through each record in the response orientation data tag table. For the current record, read its "scene category" and "interaction action" fields and match them with the reverse filtering rule. If the field value of the record meets the reverse filtering condition, such as the scene category being "monitoring", then extract the corresponding complete data content from the original real-time running data set according to the data ID in the record and store it in the result container. If the record does not meet the reverse filtering condition, i.e., it belongs to core data, then skip it. After iterating through the entire response orientation data tag table, all the data collected in the result container, i.e., the redundant data that is not directly related to frequency regulation and voltage regulation control, are ordered and aggregated to form a redundant real-time running data subset. Although this subset is not used for the current control decision, it can be archived or used for other non-real-time analysis tasks.

[0084] Step S360: Isolate and remove the redundant real-time running data subset, completely removing it from the processing flow of the real-time running data set, ensuring that subsequent processing flows only target the real-time running data subset with responsiveness, and completing the screening operation of redundant data components in the real-time running data set.

[0085] This process is a logical operation rather than a physical deletion. A data splitter is set up in memory or the data processing pipeline. After the original real-time running data set is obtained in step S100, the data first enters the splitter. Based on the index or identifier of the redundant real-time running data subset generated in step S350, the splitter divides the data stream into two branches. One branch is the "core data stream," which contains a real-time running data subset with responsive direction and will be sent to step S400 for inverse transformation and other subsequent processing. The other branch is the "redundant data stream," which contains redundant real-time running data subsets. This branch can be directed to the data storage system for archiving or to other non-real-time monitoring and analysis modules. Through this splitting mechanism, the core processing flow only encounters the data most relevant to frequency and voltage regulation coordinated control from beginning to end, ensuring the efficiency and accuracy of subsequent stages. At this point, the filtering operation of redundant data components in the real-time running data set is complete.

[0086] Step S400: Perform an inverse transformation operation on the sparse component set with response directionality to reconstruct the key operating data required for the coordinated control of frequency and voltage regulation of the virtual power plant and the large power grid. Perform error verification on the key operating data, remove abnormal data units whose reconstruction deviation exceeds the preset range, and obtain the key operating data that has passed the verification.

[0087] In an optional embodiment, step S400 may specifically include the following steps S410 to S460: Step S410: Perform an inverse transformation on the set of sparse components with response orientation after deduplication and optimization, restore the primitive combination corresponding to each sparse component to the original representation of the real-time running data, and collect all the restored content in an orderly manner to obtain the initial reconstruction key running data. The representation of the initial reconstruction key running data is consistent with the representation of the real-time running data set.

[0088] For example, first, a dedicated sparse dictionary adapted to the latest distribution attributes is loaded, containing the specific numerical vector of each primitive. Simultaneously, the deduplicated and optimized set of responsive sparse components generated in step S300 is loaded, and each sparse component in the set is traversed. For the current component, the stored information includes a list of primitive IDs that make it up, such as [primitive A, primitive B], and the weight coefficient corresponding to each primitive, such as [0.5, 0.8]. The corresponding primitive vector is retrieved from the dictionary based on the primitive ID; for example, the primitive A vector is [a1, a2, ..., an], and the primitive B vector is [b1, b2, ..., bn]. Then, a linear combination operation is performed: reconstructed vector = 0.5 * [a1, a2, ..., an] + 0.8 * [b1, b2, ..., bn]. The result is a new vector with the same dimension as the primitive vectors. This vector represents the original data fragment corresponding to the sparse component. This reconstructed vector is encapsulated according to its original timestamp and data label to form a standard running data record. After performing this operation on all sparse components, a series of such data records are obtained. These records are then ordered by timestamp to obtain the initial key operational data for reconstruction.

[0089] Step S420: Compare the initial key reconstructed running data with the real-time running data subset that has responsiveness one by one, record the difference content and position of each data, arrange all the comparison results to obtain the reconstructed data difference description set, and each record in the reconstructed data difference description set corresponds to the difference information of one data.

[0090] First, the initial key operational data and the real-time operational data subset with responsiveness are precisely aligned according to timestamps to ensure that the same physical quantity at the same moment is being compared, creating an empty result list. Then, the aligned data pairs are traversed, with each pair containing one reconstructed data entry and one corresponding original data entry. For the current data pair, if the data is a scalar, such as a frequency value, the difference is directly calculated and recorded as the difference content. If the data is a vector or multidimensional, such as a composite record containing voltage, current, and power, the difference for each dimension is calculated separately, and the positions and specific values ​​of all differences that are not zero or exceed a minimum threshold are recorded. For example, for a record containing voltage U and current I, the comparison result might be "U difference: +0.1kV, position: voltage field; I difference: 0A, position: current field". The timestamp, reconstructed data ID, original data ID, and all calculated difference content and position of the current data pair are compiled into a single record and appended to the result list.

[0091] Step S430: Filter the set of reconstructed data differences, extract the initial key reconstructed operation data whose differences exceed the preset range, and sort all the filtered data to obtain a subset of initial key reconstructed operation data whose deviations exceed the range. The content of the subset of initial key reconstructed operation data whose deviations exceed the range covers all reconstructed data whose deviations exceed the standard.

[0092] Specifically, first, an allowable error range, or preset range, is defined for each type of data. For example, for voltage data, the maximum allowable absolute error is ±0.5 kV; for frequency data, the maximum allowable absolute error is ±0.01 Hz. An empty result container is created. Then, each record in the reconstructed data difference description set generated in step S420 is traversed. For the current record, each difference content is checked. If the absolute value of any difference content exceeds the preset range threshold corresponding to that type of data (e.g., a voltage difference of +0.6 kV, while the threshold is ±0.5 kV), the reconstructed data corresponding to that record is determined to have exceeded the deviation limit. Based on the timestamp and reconstructed data ID in this record, a complete data record is extracted from the initial reconstructed key operation data set and placed into the result container. If all difference content in a record is within the preset range, the record is skipped. After traversing the entire difference description set, all the data collected in the result container, i.e., all abnormal data units with exceeded reconstruction errors, together form the initial reconstructed key operation data subset with deviations exceeding the range.

[0093] Step S440: Locate the deduplicated and optimized sparse components with responsiveness corresponding to the initial reconstruction key running data subset where the deviation exceeds the range, adjust the primitive combination content corresponding to the component, and collect all the adjusted components in an orderly manner to obtain the corrected sparse component subset with responsiveness. The primitive combination of the corrected sparse component subset with responsiveness corresponds completely with the corrected data.

[0094] For example, based on the timestamp or index of each data point in the initial key data subset where the deviation exceeds the range, a reverse search is performed on the deduplicated and optimized sparse component set with responsive directionality obtained in step S300 to find the sparse component that generated the data point. These found sparse components are then extracted. For each extracted sparse component, its difference from the original data is analyzed. A more precise sparse coding algorithm can be used, such as using stricter convergence conditions, to recalculate the sparse representation for the original data segment corresponding to that component. The new calculation may yield a slightly different set of primitive weights or a different combination of primitives. This new combination can reconstruct the original data with a smaller error. The original sparse component content is replaced with this newly calculated sparse representation, i.e., the new combination of primitives and weights. By performing this adjustment operation on all components corresponding to data with excessive deviation, a batch of corrected sparse components is obtained. These components are then ordered and grouped to obtain a corrected subset of sparse components with responsive directionality.

[0095] Step S450: Perform an inverse transformation on the modified sparse component subset with response orientation to restore the adjusted primitive combination to the original representation of the real-time running data. Orderly collect all the restored contents to obtain the modified key running data. The representation of the modified key running data is consistent with the representation of the real-time running data set.

[0096] The inverse transformation refers to performing the exact same inverse transformation operation as in step S410 on the modified sparse component subset generated in step S440. This restores the adjusted primitive combination to the real-time running data, essentially recombining the modified new sparse representation into a data vector. The modified key running data is a set of new data obtained from the inverse transformation of all modified sparse components. The purpose of this step is to regenerate more accurate running data based on the modified, more precise sparse representation.

[0097] Step S460: Merge the corrected key operating data with the initial reconstructed key operating data, replace the original deviation exceeding the standard data, and collect all the merged data in an orderly manner to obtain the key operating data that has passed the verification. All contents of the key operating data that has passed the verification do not exceed the preset deviation range and fully meet the decision requirements of collaborative control.

[0098] In an optional embodiment, step S460 may specifically include the following steps S461 to S466: Step S461: Perform time-series correlation on the key operational data that has passed the verification. Connect each data point according to the time-series correlation path of the real-time operational data set, mark the time-series position of each data point and its correlation with adjacent data, and arrange all the correlation content to obtain the time-series correlation map of the key operational data. The structure of the time-series correlation map of the key operational data is consistent with the time-series correlation structure of the real-time operational data set.

[0099] For example, first, load the real-time running data distribution association graph generated in step S210 for the original real-time running data set. This graph contains the association paths between all original data points. Then, load the validated key running data set and traverse each data point in the key running data. For the current data, find the corresponding node in the original association graph based on its timestamp and data label. Then, extract all edges in the original graph that are connected to this node and whose other end node also exists in the key running data set, as well as the node information connected to these edges. Using the nodes in the key running data set, redraw these edges and nodes in memory to construct a new, smaller graph. In this way, the position and relationship of all data in the key running data set in the original logical network are found, ultimately constructing a complete key running data time-series association graph. This new graph is a projection of the original graph onto the key data, and its topology, such as the direction of the edges and the relative positions of the nodes, is strictly consistent with the original graph, thus ensuring that the logical relationships between the data are not destroyed.

[0100] Step S462: Perform time series consistency verification on the time series correlation graph of key operational data, identify the content where the correlation relationship between adjacent data is inconsistent with the correlation relationship of the real-time operational data set, record the time series information and difference content of each abnormal position, arrange all verification results to obtain a time series correlation anomaly description set, and each record in the time series correlation anomaly description set corresponds to the position and content of a time series anomaly.

[0101] Specifically, the key operational data time-series correlation graph constructed in step S461 is overlaid and compared with the original real-time operational data distribution correlation graph generated in step S210. Each edge in the key data graph, i.e., each pair of adjacent key data nodes and their relationships, is traversed. For the current edge, the two nodes are located in the original graph. It is checked whether there is a directly connected edge between these two nodes in the original graph, and whether the edge attributes, such as causal description, are consistent. If there is no direct edge between these two nodes in the original graph, but they are indirectly connected through other intermediate nodes, or if there is a direct edge but the attributes are different, it is determined that the correlation relationship is inconsistent. This anomaly is recorded, including the time point of the anomaly, the two data nodes involved, and the correct correlation relationship description that should exist in the original graph, such as "In the original graph, node A and node B should be indirectly associated through node C, with the relationship chain being A, C, B". All detected anomaly records are summarized to form a time-series correlation anomaly description set.

[0102] Step S463: Locate the time-series correlation anomaly description set, extract the key operational data that has passed the verification at the corresponding position, and collect all the extracted data in an orderly manner to obtain a subset of key operational data for time-series anomalies. The content of the subset of key operational data for time-series anomalies covers all key operational data with time-series correlation anomalies.

[0103] For example, each record in the time-series correlation anomaly description set generated in step S462 is traversed. Each record indicates the timestamps of two or more data nodes with inconsistent correlations. Based on these timestamps, the complete data records corresponding to the verified key operational data set are extracted and placed into a temporary result container. To avoid duplicate extraction, a set of extracted timestamps can be maintained. After traversing all anomaly records, all data collected in the result container—that is, all data points related to correlation anomalies—are ordered and grouped, for example, by time, to form a subset of key operational data for time-series anomalies.

[0104] Step S464: Extract the details of the time-series correlation path of the real-time running data set corresponding to the subset of time-series abnormal key running data, extract the correct correlation and data content corresponding to the abnormal location, record the correction direction and target content of each abnormal data, arrange all correction schemes to obtain the time-series correction scheme, and each record of the time-series correction scheme corresponds to a correction rule for an abnormal data.

[0105] Specifically, each data point in the critical runtime data subset of time-series anomalies is traversed. For the current data, based on its timestamp, the corresponding node is located in the original real-time runtime data distribution association graph generated in step S210. Then, all incoming and outgoing edges of the node in the original graph are analyzed to find all its predecessor and successor nodes, and these correct associations are recorded. For example, for the current data node, its correct association might be "its predecessor should be node X with timestamp T-1, and its successor should be node Y with timestamp T+2." The identifier of the current data and this correct association that needs to be established are combined into a correction scheme. If the correction scheme requires the introduction of a data node that does not exist in the current critical data set, such as node X, then the data content of node X also needs to be extracted from the original data as a supplement. After the correction schemes are formulated for all the anomalies, these schemes are summarized to obtain the time-series correction scheme.

[0106] Step S465: Perform time-series correction on the critical operational data subset with time-series anomalies. Modify the time-series association content and self-description of the data according to the rules in the time-series correction scheme. Collect all the corrected data in an orderly manner to obtain the time-series corrected critical operational data subset. The time-series association of the time-series corrected critical operational data subset is completely consistent with the time-series association of the real-time operational data set.

[0107] For example, based on the timing correction scheme generated in step S464, operations are performed on the critical runtime data subsets with timing anomalies. For each anomalous data point, the rules defined in the correction scheme are executed. This may include: creating a new data node in the critical data set, derived from the original data, to act as a missing predecessor or successor, and establishing the corresponding directed edge; or correcting the metadata of existing data nodes, updating their association pointers with predecessors and successors; or replacing a data node with the correct data content if its content causes it to be incorrectly connected. By performing all these correction operations, it is ensured that all data points that previously had logical breaks are now in a complete causal chain that is fully consistent with the original graph.

[0108] Step S466: Merge the timing-corrected subset of critical operational data with the verified critical operational data, replace the original timing-abnormal data, and sort all the merged data to obtain timing-consistent critical operational data.

[0109] In an optional embodiment, step S466 may specifically include the following steps S4661 to S4666: Step S4661: Perform scenario adaptability verification on key operational data with consistent timing. Map each data point to an anchor point in the frequency and voltage regulation scenario interaction logic anchor point set. Record the anchor point matching status and interaction action consistency for each data point. Arrange all verification results to obtain a scenario adaptability description set. Each record in the scenario adaptability description set corresponds to the scenario adaptability information of one data point.

[0110] Specifically, the set of frequency and voltage regulation scenario interaction logic anchor points generated in step S220 is loaded, and each piece of data in the key operational data set with consistent timing is traversed. For the current data, based on its data content and timestamp, an attempt is made to find a matching anchor node or anchor path in the anchor point set. For example, a piece of data containing a frequency value and a downward trend may match the "frequency drop trigger" anchor point, and the matching anchor point ID is recorded. Then, it is checked whether the sequence formed by this data and the data before and after it conforms to the logical path of the anchor point in the anchor point set. For example, after the "frequency drop trigger" anchor point, there should be a "issue frequency regulation command" anchor point. If the current data is a frequency drop and its preceding data is a load surge data point, which is logically consistent, then the consistency is marked as "yes," and the identifier of each piece of data, the matched anchor point ID, the matching degree score, and the consistency judgment result with the preceding and following data are compiled into a record.

[0111] Step S4662: Filter the scene adaptability description set, extract key operational data that are consistent with the time sequence but inconsistent with the interaction actions of the anchor point, and orderly collect all the filtered data to obtain a subset of key operational data with scene adaptability anomalies. The content of the subset of key operational data with scene adaptability anomalies covers all key operational data with scene adaptability anomalies.

[0112] Specifically, the process iterates through each record in the scenario adaptability description set. For the current record, it reads the "interaction consistency" field. If the value of this field is "no" or "inconsistent," it determines that the data corresponding to this record has a scenario adaptability anomaly. Based on the data identifier or timestamp in this record, it extracts the complete data record from the time-series consistent key runtime data set and places it into a temporary result container. If the value of this field is "yes" or "consistent," it skips the step. After iterating through the entire description set, all the data collected in the result container—that is, all the data points that do not conform to the standard business logic—together form the critical runtime data subset of scenario adaptability anomalies.

[0113] Step S4663: Extract details of the frequency modulation and voltage regulation scene interaction logic anchor point set corresponding to the subset of critical running data of scene adaptation anomalies, extract the correct interaction actions and data features of the anchor points corresponding to the anomalies, record the correction direction and target content of each anomaly, arrange all correction schemes to obtain the scene correction scheme, and each record of the scene correction scheme corresponds to a correction rule for anomalies.

[0114] For example, traversing each data point in the critical runtime data subset for scene adaptation anomalies, for the current data, firstly, based on the anchor point ID matched in step S4661, the anchor point node is found in the anchor point set. Then, the standard path centered on this node in the anchor point set is analyzed. For example, if the node is "a single frequency modulation action," its standard path should be "frequency limit violation" first, followed by "frequency recovery." However, the predecessor of the current data in the sequence might be "voltage limit violation," which is inconsistent. Therefore, a correction scheme is formulated: either the current data is reinterpreted as another anchor point, such as a misjudgment, or its predecessor data is marked as anomaly. For the current data, the correction scheme might be "adjust its predecessor data to be frequency limit violation data; if missing, it needs to be supplemented from the original data." The identifier of the current data and this specific correction rule are recorded. After formulating correction schemes for all abnormal data, these schemes are summarized to obtain the scene correction scheme.

[0115] Step S4664: Perform scene correction on the critical runtime data subset with scene adaptation anomalies. Modify the data description and interaction association according to the rules in the scene correction scheme. Collect all the corrected data in an orderly manner to obtain the critical runtime data subset after scene correction.

[0116] Specifically, based on the scenario correction scheme generated in step S4663, operations are performed on the subset of critical operational data that are not aligned with the scenario adaptation. For each piece of anomalous data, the rules defined in the correction scheme are executed. This may include: fine-tuning certain numerical parameters of the data based on the feature template of the standard anchor point; or changing the correlation of the data in the sequence, connecting it to the correct predecessor or successor data node; or, if required by the correction scheme, extracting the missing data fragments used to form a complete logical chain from the original data set and inserting them as new data into the current critical data set. By performing all these correction operations, it is ensured that all data points that previously did not conform to standard business logic are now in a correct scenario that conforms to physical laws.

[0117] Step S4665: Merge the subset of critical runtime data after scene correction with the critical runtime data that is consistent with the time series, replace the original data with abnormal scene adaptation, and collect all the merged data in an orderly manner to obtain the critical runtime data adapted to the scene.

[0118] For example, first, copy the time-consistent set of critical operational data into a new result container. Then, iterate through each data entry in the subset of critical operational data after scenario correction. For each data entry in the subset, check if a data record with the same timestamp exists in the result container. If it exists, replace the old data in the container with the new data from the subset. If it doesn't exist, it means this is a new data entry introduced to fix the logic, so add it as a new record to the result container. After all replacement and addition operations are completed, all the data in the result container together constitute a dataset that is completely self-consistent with the business scenario. Perform a final ordered aggregation on this dataset to obtain the scenario-adapted critical operational data.

[0119] Step S4666: Perform a final consistency check on the key operational data for scenario adaptation, confirming that the representation, temporal correlation, and scenario adaptation of each data item are consistent with the corresponding content of the real-time operational data set. Collect all the data that have passed the check in an orderly manner to obtain the key operational data that has passed the verification.

[0120] Specifically, the key operational data for scenario adaptation obtained in step S4665 is used as input. First, a format consistency check is performed to ensure that the fields, units, and data types of each data entry are completely consistent with the original data format defined in step S100. Second, a temporal consistency check is performed, which can reuse the comparison logic of step S462, but this time it is compared with the original graph to ensure that all relationships are correct. Finally, a scenario adaptability check is performed, which can reuse the logic of step S4661, but this time it ensures that each data entry and the scenario it forms perfectly match the standard anchor set. Only data that passes all three checks is finally released, and all the data that passes the checks are systematically aggregated to obtain the final verified key operational data.

[0121] Step S500: Based on key operating data, perform collaborative control simulation to obtain frequency and voltage regulation collaborative control commands between the virtual power plant and the main power grid. Transmit the frequency and voltage regulation collaborative control commands to the corresponding control nodes of the virtual power plant and the main power grid to complete the frequency and voltage regulation collaborative control.

[0122] In an optional embodiment, step S500 may specifically include the following steps S510 to S560: Step S510: Extract the trigger conditions for the key operational data that has passed the verification, convert the anchor point corresponding to each data point into the trigger condition for collaborative control simulation, record the interactive actions and response requirements corresponding to each trigger condition, and arrange all the trigger conditions to obtain the set of trigger conditions for collaborative control simulation.

[0123] Specifically, the key operational data that passed verification obtained in step S4666 and the frequency and voltage regulation scenario interaction logic anchor point set generated in step S220 are loaded first, and each piece of data in the key operational data is traversed. For the current data, according to the scenario adaptation information recorded in step S4661, its corresponding anchor point ID is obtained, and the anchor point is searched in the anchor point set. If the anchor point type is "trigger type", such as "frequency low limit" or "voltage high limit", it is identified as a trigger condition that needs to be responded to. The standard interaction action associated with the trigger condition is extracted from the anchor point set, such as "frequency regulation should be started to increase active power output", and the expected response requirement, such as "the frequency deviation should be reduced to within 0.02Hz within 2 seconds". This trigger condition, associated interaction action, and response requirement, along with the data timestamp, are organized into a record. After traversing all the data, all such records are summarized to obtain the collaborative control simulation trigger condition set.

[0124] Step S520: Build a scenario for the set of triggering conditions for collaborative control simulation, map all triggering conditions to the time-series association structure of the key running data that has passed verification, and construct a simulation environment that includes all triggering conditions and interactive actions to obtain the collaborative control simulation scenario.

[0125] For example, based on the verified key operational data and its inherent time-series correlation graph obtained in step S4666, each trigger condition in the collaborative control simulation trigger condition set generated in step S510 is taken as a "hotspot" event in the current scenario and anchored to the corresponding node in the time-series correlation graph. Then, with these hotspot events as the center, the simulation extends forward and backward along the correlation path in the graph, incorporating related preceding states and expected subsequent response steps into the scenario scope. For example, a "frequency low limit violation" trigger condition will include its preceding "load surge" data node, as well as subsequent target state nodes such as "frequency adjustment action" and "frequency recovery," into the current simulation scenario. Furthermore, the real-time status and adjustability of all controllable resources are extracted from the key operational data as the initial parameters of the scenario. The final constructed collaborative control simulation scenario is an information complex encompassing all unresolved issues, problem background, available resources, and expected goals.

[0126] Step S530: Perform simulation operation on the collaborative control scenario, trigger control instructions according to the interaction action chain of the frequency modulation and voltage regulation scenario interaction logic anchor point set, record the instruction content and execution object corresponding to each trigger action, and arrange all the simulation obtained instructions to obtain the initial collaborative control instruction set.

[0127] Specifically, the collaborative control simulation scenario constructed in step S520 is input into a rule-based or optimization-based simulation system. If a rule-based method is used, all triggering conditions in the scenario are traversed. For each triggering condition, a matching action chain is found in the frequency and voltage regulation scenario interaction logic anchor point set. Then, based on the steps defined in the action chain and the actual state of resources in the current scenario, specific control commands are instantiated. For example, for the "low frequency limit violation" triggering condition, the action chain might specify "call the top 3 fastest response resources." Based on the current state of charge and response speed ranking of each energy storage unit, it is determined that it specifically refers to energy storage unit 3 and energy storage unit 5, and commands such as "energy storage unit 3, increase discharge power by 2MW" and "energy storage unit 5, increase discharge power by 1.5MW" are generated. If an optimization-based method is used, a dynamic optimization problem is solved using the scenario state as the initial value and the target defined in the anchor point set as the optimization objective. The output is a sequence of control commands.

[0128] Step S540: Perform feasibility verification on the initial set of collaborative control instructions, map each instruction to the key operational data that has passed verification, record the execution feasibility of each instruction and its impact on the operational data, and arrange all verification results to obtain a set of instruction rationality descriptions.

[0129] Specifically, the initial set of coordinated control instructions and verified key operational data are loaded. For each instruction in the instruction set, its target execution object and expected action are first parsed, such as "issue an order to increase the discharge power of energy storage station A by 2MW". Then, the current operating status of energy storage station A is found in the key operational data, especially its "maximum discharge power". If 2MW is less than or equal to the maximum discharge power, the execution feasibility is marked as "feasible"; otherwise, it is marked as "infeasible". Next, using an embedded, simplified power system model, the impact of executing this instruction on the key operational data is simulated, such as predicting the change in grid connection point voltage. The predicted voltage value is compared with the voltage safety limit. If the predicted value is within the safety range, the impact is marked as "positive" or "acceptable"; if the predicted value will exceed the limit, the impact is marked as "negative". The ID, feasibility mark, expected impact assessment result, and other information of each instruction are compiled into a record.

[0130] Step S550: Filter the instruction rationality description set, extract the initial cooperative control instructions that meet the execution feasibility requirements, and arrange all the filtered instructions to obtain the cooperative control instruction set.

[0131] In an optional embodiment, step S550 may further include the following steps S551 to S556: Step S551: Evaluate the impact of the set of collaborative control instructions, map each instruction to the key operational data that has passed verification, record the changes and scope of the operational data caused by the execution of each instruction, and arrange all the evaluation results to obtain a description set of the impact of the instructions.

[0132] For example, using a set of coordinated control commands and validated key operational data as input, an impact assessment model, such as a sensitivity-based linearized model or a trained neural network model, is employed. For each command, such as "issue an order to increase discharge power by 2MW to energy storage A," this command is used as input. Based on current operational data, the model predicts the changes in key quantities such as voltage at each critical node and system frequency, and records these changes, for example, "voltage U1 change: +0.015kV, voltage U2 change: -0.002kV, frequency F change: +0.008Hz." The scope of the impact may also be assessed, such as identifying which node is most severely affected. Each command's ID, a detailed list of its expected impact (including affected variables, direction and magnitude of change), and a description of the impact's scope) are compiled into a single record.

[0133] Step S552: Sort the instruction influence description set, arrange the collaborative control instructions from widest to narrowest according to the influence range of the instructions, record the sorting position and influence range of each instruction, and arrange all sorting results to obtain the initial instruction priority sorting table.

[0134] Specifically, each record in the instruction impact description set generated in step S551 is traversed. For each record, a quantifiable "impact range" indicator needs to be extracted from its impact description. This indicator can be the number of nodes affected by the instruction, such as the number of nodes causing voltage changes exceeding 0.01kV; or it can be the degree of improvement of a key affected indicator, such as the reduction in frequency deviation. The ID of each instruction is associated with its corresponding quantified impact range value, such as "affected 3 nodes" or "frequency improved by 0.008Hz". Then, based on this quantified impact range value, all instructions are sorted from largest to smallest. After sorting, a record containing the instruction ID, sort position, and its quantified impact range value is generated for each instruction. By summarizing the sorted records of all instructions, the initial instruction priority sorting table is obtained.

[0135] Step S553: ​​Adjust the initial sorting table of instruction priorities. Adjust the sorting position of instructions according to the corresponding position of the anchor point in the set of anchor points for frequency modulation and voltage regulation scenario interaction logic. Record the adjustment direction and the sorting after adjustment for each instruction. Arrange all adjustment schemes to obtain the instruction priority adjustment scheme.

[0136] For example, first load the set of collaborative control simulation trigger conditions generated in step S510, and the set of frequency and voltage regulation scenario interaction logic anchor points from step S220. For each instruction in the initial instruction priority sorting table, it is necessary to trace back which trigger condition it was generated from and which interaction action chain it was based on. Based on this information, find the anchor point corresponding to the instruction in the anchor point set, and obtain the "urgency" or "importance" level of the anchor point. For example, the anchor point may be marked as "Level 1 critical event" or "Level 2 general event". Take into account the importance level of this anchor point and the initial sorting position of the instruction to formulate adjustment rules. For example, the rule can be set as "unconditionally promote all instructions corresponding to 'Level 1 critical event' to the top 30% of the sorting list". Record the adjustment direction for each instruction, such as "promote by 5 places" or "remain unchanged", and the expected new sorting position after adjustment. Summarize the adjustment rules for all instructions to obtain the instruction priority adjustment scheme, which clarifies how to integrate business logic into the sorting.

[0137] Step S554: Adjust the initial instruction priority sorting table, modify the sorting position of the instructions according to the rules in the instruction priority adjustment scheme, and arrange all the adjusted sorting results to obtain the final instruction priority sorting table. Each record in the final instruction priority sorting table corresponds to the final sorting information of one instruction.

[0138] For example, taking the initial instruction priority sorting table and the instruction priority adjustment scheme generated in step S553 as input, a copy of the initial sorting table is first made. Then, each rule in the adjustment scheme is traversed. For each rule, the current position of the instruction is found in the copied sorting table according to its specified instruction ID. Then, according to the "adjustment direction" and "adjustment magnitude" specified in the rule, such as "promote by 5 places", the position of the instruction is moved in the list. During the movement, conflicts with other instructions are considered, which can be resolved by using stable sorting or recalculating the position. After all adjustment rules have been executed, a brand new, updated sorting list is obtained. Each instruction in this new list and its final sorting position are extracted to form a record. The set of all records is the final instruction priority sorting table.

[0139] Step S555: Prioritize the set of cooperative control instructions, rearrange the instructions according to their positions in the final priority sorting table, and collect all sorted instructions in an orderly manner to obtain a set of cooperative control instructions sorted by priority.

[0140] For example, taking the set of cooperative control instructions and the final ordering table of instruction priorities as input, each sorting position is traversed according to the order in the final ordering table. For each position, the instruction ID corresponding to that position is obtained from the ordering table. Then, based on this ID, the complete instruction content is extracted from the original set of cooperative control instructions. The extracted instructions are then placed into a new result container in the order of traversal. After traversing all positions in the final ordering table, the order of instructions stored in the result container is completely consistent with the final ordering table. This result container is then sorted to obtain the set of cooperative control instructions ordered by priority.

[0141] Step S556: Perform sequential transmission operation on the set of coordinated control instructions sorted by priority, send the instructions to the corresponding control nodes in order from first to last according to the sorting position, receive the execution feedback information from the receiving nodes, and collect all the feedback information in an orderly manner to obtain the frequency modulation and voltage regulation coordinated control completion signal executed according to priority.

[0142] In optional embodiments, the method provided by the present invention may further include the following steps S557~S5512: Step S557: Evaluate the effect of the frequency modulation and voltage regulation coordinated control completion signals executed according to priority, map the execution feedback of each instruction to the latest statistical distribution attribute of the real-time running data set, record the execution effect and attribute changes of each instruction, and arrange all evaluation results to obtain the instruction execution effect description set.

[0143] For example, first, obtain the latest real-time running data set within a short time window after the instruction is executed, and perform the same statistical analysis as in step S200 on this new data set to obtain the latest statistical distribution attributes, such as the latest average frequency and voltage qualification rate. Then, iterate through the execution feedback of each instruction recorded in step S556. For each instruction, based on its execution object and time, find the related attribute changes in the latest statistical distribution attributes. For example, for an instruction to increase the energy storage discharge power, check the improvement of frequency deviation in the latest statistics. The expected effect of the instruction can be obtained from the impact description set in step S551, compared with the actual observed effect, and recorded evaluation results such as "actual frequency improvement of 0.01Hz, better than the expected 0.008Hz" or "actual voltage increase of 0.03kV, exceeding the expected 0.02kV". The evaluation results of all instructions are compiled into records and summarized to obtain the instruction execution effect description set.

[0144] Step S558: Filter the instruction execution effect description set, extract the collaborative control instructions whose execution effect does not match the expectation, and sort all the filtered instructions to obtain the subset of collaborative control instructions to be adjusted.

[0145] Specifically, an acceptable deviation range is defined for each control effect, such as "the deviation of the frequency improvement effect should not exceed 20% of the expected value." Each record in the instruction execution effect description set is traversed. For the current record, the comparison information between its actual effect and expected effect is read. If the deviation between the actual effect and the expected effect exceeds the preset acceptable range, or if the actual effect leads to negative consequences, such as voltage exceeding limits, then the instruction execution effect is determined to be unsatisfactory. Based on the instruction ID in this record, the corresponding original instruction content is extracted from the priority-sorted set of collaborative control instructions and placed into a temporary container. After traversing all records, the instructions collected in the result container—that is, all instructions that need to be analyzed and adjusted—together form the subset of collaborative control instructions to be adjusted.

[0146] Step S559: Extract the latest statistical distribution attributes of the real-time running data set corresponding to the subset of collaborative control instructions to be adjusted, extract the interaction action details corresponding to the attribute changes, record the adjustment direction and target content of each instruction to be adjusted, and arrange all adjustment schemes to obtain the instruction adjustment scheme.

[0147] For example, the input consists of a subset of coordinated control commands to be adjusted and real-time operational data before and after command execution. For each command to be adjusted, detailed operational data for a period of time before and after its execution is first obtained, and this data is analyzed to try to find the cause of the effect deviation. For example, if an energy storage discharge command fails to increase the frequency as expected, data analysis reveals that a large industrial load was also started at the same time as the energy storage discharge, offsetting the discharge effect. Based on this finding, an adjustment plan is formulated: in the future, when encountering similar scenarios, commands should be issued simultaneously to cut off some interruptible loads, or the discharge power reserve of energy storage should be increased. The ID of this command to be adjusted, as well as the specific adjustment plan, such as "add a coordinated command: while issuing a discharge command to energy storage, check and issue a cut-off command for the corresponding area of ​​interruptible loads," is recorded.

[0148] Step S5510: Adjust the subset of collaborative control instructions to be adjusted, modify the content and execution object of the instructions according to the rules in the instruction adjustment scheme, and collect all the adjusted instructions in an orderly manner to obtain the adjusted subset of collaborative control instructions.

[0149] For example, based on the instruction adjustment scheme generated in step S559, each instruction in the subset of cooperative control instructions to be adjusted is operated on. For each instruction, its corresponding adjustment scheme is parsed. If the scheme requires modification of parameters, the numerical fields in the original instruction are directly modified. If the scheme requires adding cooperative instructions, one or more new, matching instructions are generated and associated with the original instructions. If the scheme requires changing the execution object, the target address of the instruction is modified, and all modified instructions, as well as any newly generated instructions, are collected to form the adjusted subset of cooperative control instructions.

[0150] Step S5511: Merge the adjusted subset of cooperative control instructions with the priority-sorted set of cooperative control instructions, replace the original instructions to be adjusted, and rearrange the instructions according to the final priority sorting table to obtain the adjusted priority-sorted set of cooperative control instructions.

[0151] For example, first, copy the original priority-sorted set of cooperative control instructions used in step S556. Then, iterate through the adjusted subset of cooperative control instructions. For each instruction in the subset, find the old instruction to be replaced in the copied set based on its associated original instruction ID, and overwrite it with the new instruction in the subset. If the subset contains newly added instructions, add them as new entries to the set. After all replacements and additions are completed, an updated instruction set is obtained. Finally, call the priority sorting process from steps S552 to S555, or directly reuse the latest instruction priority final sorting table, to reorder this updated set, ensuring that newly added or modified instructions are placed in the correct positions.

[0152] Step S5512: Perform a transmission operation on the adjusted set of coordinated control instructions sorted by priority, send the instructions to the corresponding control nodes, receive the execution feedback information from the receiving nodes, confirm that the execution effect of all instructions meets expectations, and obtain the frequency modulation and voltage regulation coordinated control optimization completion signal.

[0153] Specifically, repeat step S556, sending the adjusted, priority-sorted set of coordinated control instructions one by one to the corresponding control nodes according to priority, and receiving execution feedback. However, an additional verification step is added after this: real-time operational data after instruction execution is acquired again, and a rapid effect evaluation is performed, similar to step S557. If the evaluation results show that all key indicators, such as frequency and voltage, are within the expected range and no new problems arise, then this round of optimization is considered successful, this success status is recorded, and a frequency and voltage regulation coordinated control optimization completion signal is generated. This signal marks the end of a successful feedback-based adaptive control loop and provides a better starting point for the next control loop.

[0154] Step S560: Perform transmission operation on the set of coordinated control commands, send each command to the corresponding virtual power plant or large power grid control node, receive command execution feedback information from the receiving node, and collect all feedback information in an orderly manner to obtain the frequency regulation and voltage regulation coordinated control completion signal.

[0155] Taking the priority-ordered set of coordinated control commands generated in step S555 or S5511 as input, the virtual power plant's communication front-end processor encapsulates each command into a data packet conforming to the corresponding control node's communication protocol, such as 104 protocol or Modbus message, according to the command priority order, and sends it out through communication links such as fiber optic, 4G / 5G, etc. For each sent command, it waits for and receives an acknowledgment frame or execution result message from the target node, recording the sending time of each command, the returned acknowledgment information, and any post-execution status data that may be included. After all commands have been sent and corresponding feedback has been received, all feedback information is organized and packaged according to the original command order to form a complete execution report. The generation of this report indicates that the frequency and voltage regulation coordinated control task has been completed, and the data acquisition and processing for the next cycle can begin.

[0156] In one embodiment, a control device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, the control device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a frequency and voltage regulation coordinated control method applied to the interaction between a virtual power plant and a large power grid.

[0157] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the control device to which the present invention is applied. The specific control device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

Claims

1. A frequency and voltage regulation coordinated control method applied to the interaction between a virtual power plant and a large power grid, characterized in that, The method includes: Acquire a set of real-time operational data generated during the interaction between the virtual power plant and the large power grid. The set of real-time operational data carries electrical status information and interaction response information of each distributed resource of the virtual power plant during the interaction with the large power grid. Based on the statistical distribution attributes carried by the real-time operating data set, a dedicated sparse dictionary is generated to adapt to the frequency and voltage regulation scenarios of the interaction between the virtual power plant and the large power grid. The dedicated sparse dictionary and the statistical distribution attributes of the real-time operating data set form a unique correspondence. Using the dedicated sparse dictionary as the processing benchmark, a sparse representation operation is performed on the real-time running data set to obtain a sparse component set that has a responsiveness to frequency modulation and voltage regulation coordinated control, while filtering out redundant data components in the real-time running data set that are not related to frequency modulation and voltage regulation coordinated control. An inverse transformation operation is performed on the sparse component set with response directionality to reconstruct the key operating data required for the coordinated control of frequency and voltage regulation of the virtual power plant and the large power grid. The key operating data is then subjected to error verification, and abnormal data units with reconstruction deviations exceeding the preset range are removed to obtain the key operating data that has passed the verification. Based on the key operating data, a collaborative control simulation operation is performed to obtain frequency and voltage regulation collaborative control commands between the virtual power plant and the main power grid. These commands are then transmitted to the corresponding control nodes of the virtual power plant and the main power grid to complete the frequency and voltage regulation collaborative control.

2. The method as described in claim 1, characterized in that, Based on the statistical distribution attributes carried by the real-time operating data set, a dedicated sparse dictionary is generated to adapt to the frequency and voltage regulation scenarios of virtual power plants and large power grids. This dedicated sparse dictionary has a unique correspondence with the statistical distribution attributes of the real-time operating data set, including: The real-time operation data set is correlated in time series, and the data points of the interaction between the distributed resources of the virtual power plant and the power grid at all times are connected. The causal relationship paths between data in different time periods are marked to obtain the real-time operation data distribution correlation map. Scene anchor point localization is performed on the real-time operation data distribution association map, and map nodes and associated paths directly related to frequency modulation and voltage regulation actions are extracted. The interaction action triggering conditions and response results corresponding to each path are marked to obtain the frequency modulation and voltage regulation scene interaction logic anchor point set. The set of frequency modulation and voltage regulation scenario interaction logic anchor points is transformed into primitives. The interaction logic corresponding to each anchor point is broken down into the smallest unit that can be matched repeatedly. Each unit corresponds to the feature description of a type of interaction action. All the smallest units are sorted together to obtain a dedicated sparse dictionary primitive library. The exclusive sparse dictionary primitive library is hierarchically mapped. The hierarchical structure of the primitive library is set according to the distribution hierarchy of real-time running data, so that the global coverage level primitive corresponds to the global data distribution, and the single data level primitive corresponds to the single data detail. After the hierarchical mapping is completed, the initial exclusive sparse dictionary is obtained. The initial dedicated sparse dictionary is bound one-to-one with the statistical distribution attributes of the real-time running data set, so that each dictionary level corresponds to a feature of a statistical distribution attribute, and each dictionary primitive corresponds to a detailed description of a statistical distribution attribute. After the binding operation is completed, the bound dedicated sparse dictionary is obtained. The bound dedicated sparse dictionary is applied to match real-time running data. The description content and hierarchical association of dictionary primitives are adjusted according to the matching results until the matching coverage of the dictionary and real-time running data reaches the scenario requirements, thus obtaining a dedicated sparse dictionary adapted to the frequency and voltage regulation scenario of virtual power plant and large power grid interaction.

3. The method as described in claim 2, characterized in that, The process involves applying the bound, dedicated sparse dictionary to match real-time operational data, adjusting the representation content and hierarchical association of dictionary primitives based on the matching results, until the matching coverage between the dictionary and the real-time operational data meets the scenario requirements. This yields a dedicated sparse dictionary adapted to the frequency and voltage regulation scenario of virtual power plants interacting with large power grids, including: A dedicated sparse dictionary adapted to the frequency and voltage regulation scenarios of virtual power plants and large power grids is matched element by element with the real-time operation data set. The data matching status and coverage of each dictionary primitive are recorded, and all the recorded contents are arranged to obtain the dictionary primitive matching degree description set. Gap identification is performed on the dictionary primitive matching degree description set to determine the dictionary primitives whose matching coverage does not meet the requirements. The real-time running data category and interaction scenario corresponding to each gap primitive are recorded. All gap information is collected in an orderly manner to obtain a primitive optimization requirement list. Detailed extraction is performed on the real-time running data set corresponding to the primitive optimization requirement list. All interaction data details in the scenario corresponding to the missing primitive are extracted, the extracted details are transformed into the smallest matching unit, and all supplementary units are sorted together to obtain the supplementary primitive set. The supplementary primitive set is embedded into a dedicated sparse dictionary adapted to the frequency and voltage regulation scenarios of virtual power plants and large power grids. It is then added to the level and position specified in the primitive optimization requirement list. The association path between the supplementary primitive and the original primitive is adjusted so that the supplementary primitive is integrated into the hierarchical structure of the dictionary, resulting in an expanded dedicated sparse dictionary. The expanded dedicated sparse dictionary is matched element by element again with the real-time running data set. The matching status and coverage of each dictionary primitive and supplementary primitive are recorded. All records are arranged to obtain the expanded dictionary matching degree description set. The expanded dictionary matching degree description set is judged to meet the standard, and it is confirmed that the matching coverage of all dictionary primitives meets the requirements. After all primitives meet the standard, the optimized exclusive sparse dictionary is obtained.

4. The method as described in claim 3, characterized in that, The expanded dictionary matching degree description set is then subjected to a compliance check to confirm that the matching coverage of all dictionary primitives meets the requirements. Once all primitives meet the requirements, an optimized dedicated sparse dictionary is obtained, including: The statistical distribution attributes corresponding to the optimized dedicated sparse dictionary are compared with the latest statistical distribution attributes of the real-time running data set, and the changes and ranges of each attribute are recorded. All comparison results are arranged to obtain a distribution attribute difference description set. Locate the dictionary primitives corresponding to the distribution attribute difference description set, determine the dictionary primitives directly associated with the changed attributes, record the hierarchical position and association path of each primitive to be adjusted, and sort all the primitives to be adjusted to obtain the dictionary primitive set to be adjusted. The latest statistical distribution attributes of the real-time running data set corresponding to the dictionary primitive set to be adjusted are extracted in detail. The interaction data details corresponding to the changed attributes are extracted, and the description of the primitive to be adjusted is replaced with the smallest unit form of the corresponding details to obtain the adjusted dictionary primitive set. The adjusted dictionary primitive set is used to replace the corresponding primitives to be adjusted in the optimized dedicated sparse dictionary. The association paths between the replaced primitives and other primitives are adjusted so that the replaced primitives are integrated into the hierarchical structure of the dictionary, resulting in a dynamically adjusted dedicated sparse dictionary. The dynamically adjusted dedicated sparse dictionary and the real-time running data set are matched element by element again. The matching status and coverage of each dictionary primitive are recorded. All the recorded contents are arranged to obtain the dynamically adjusted dictionary matching degree description set. The matching degree description set of the dynamically adjusted dictionary is judged to meet the requirements, and the matching coverage of all dictionary primitives is confirmed to meet the requirements, so as to obtain a special sparse dictionary adapted to the latest distribution attributes.

5. The method as described in claim 1, characterized in that, The process involves using the dedicated sparse dictionary as a processing benchmark to perform sparse representation operations on the real-time operating data set, obtaining a set of sparse components that have responsiveness to frequency modulation and voltage regulation coordinated control. Simultaneously, redundant data components unrelated to frequency modulation and voltage regulation coordinated control are filtered out from the real-time operating data set, including: The real-time running data set is matched one by one with the dedicated sparse dictionary that adapts to the latest distribution attributes. Each real-time running data is mapped to the dictionary primitive with the highest matching degree. The primitive correspondence and matching details of each data are recorded. All the records are arranged to obtain the data-primary association mapping table. Scenario matching is performed on the primitives in the data-primitive association mapping table, and the primitives are mapped to the anchor points in the frequency modulation and voltage regulation scenario interaction logic anchor point set. The anchor point category and interaction action corresponding to each data are marked, and all the marked contents are arranged to obtain the response directional data mark table. The response-oriented data labeling table is filtered to extract real-time operating data that is directly related to frequency and voltage regulation actions. All the filtered data are then systematically aggregated to obtain a subset of real-time operating data with response orientation. The real-time running data subset with response orientation is sparsely transformed, and each data item is represented as a combination of dictionary primitives. The transformed content retains only the part related to response orientation. All the transformed content is then collected in an ordered manner to obtain a sparse component set with response orientation. The response directional data labeling table is filtered in reverse to extract real-time operating data that is not related to frequency and voltage regulation actions. All the filtered data are then collected in an orderly manner to obtain a redundant real-time operating data subset. The redundant real-time running data subset is isolated and removed, completely eliminating it from the processing flow of the real-time running data set, thus completing the screening operation of redundant data components in the real-time running data set.

6. The method as described in claim 5, characterized in that, The sparse transformation of the real-time running data subset with response orientation involves mapping each data point to a combination of dictionary primitives, ensuring that the transformed content retains only the parts related to response orientation. All transformed content is then ordered and aggregated to obtain a set of sparse components with response orientation, including: The set of sparse components with responsive orientation is matched component by component with a dedicated sparse dictionary that adapts to the latest distribution attributes. The dictionary primitive combination and association path corresponding to each sparse component are recorded. All records are arranged to obtain a component-primary correspondence table. Scene mapping is performed on the primitive association paths in the component-primary correspondence table. The paths are mapped to the interaction paths in the frequency modulation and voltage regulation scene interaction logic anchor point set. The complete interaction action chain corresponding to each component is marked, and all marked contents are arranged to obtain the component interaction path description set. Frequency statistics are performed on the interaction action chains in the component interaction path description set. The number of action triggers and responses in each chain are recorded and converted into frequency descriptions of the interaction actions corresponding to the components. All statistical results are arranged to obtain the component interaction intensity description set. The frequency descriptions in the component interaction intensity description set are assigned accordingly, and the frequency description of each component is converted into its proportion description in the collaborative control. The proportion value of each component and the corresponding interaction path are recorded, and all the allocation results are arranged to obtain the component proportion allocation table. The sparse component set with response orientation is weighted, and the content of each component is bound to the corresponding proportion in the component proportion allocation table, so that the weight information becomes the inherent attribute of the component. All weighted components are sorted together to obtain the weighted sparse component set with response orientation. The weighted sparse component set with response orientation is subjected to range unification processing, and the proportion values ​​of all components are adjusted to values ​​within a unified range to ensure that the proportion comparison between components is clear and directly comparable. All adjusted components are then systematically aggregated to obtain the weighted optimized sparse component set with response orientation.

7. The method as described in claim 6, characterized in that, The weighted set of sparse components with responsiveness is then subjected to range unification processing, adjusting the proportion values ​​of all components to a uniform range to ensure clear and direct comparison of proportions between components. All adjusted components are then systematically aggregated to obtain a weighted and optimized set of sparse components with responsiveness, including: The weighted and optimized set of sparse components with response orientation is compared by primitive combination to identify sparse components with completely consistent primitive combinations. The number of members and the proportion of each repeated component group are recorded, and all identification results are arranged to obtain a list of repeated component identifications. The proportion of each group of duplicate components in the duplicate component identification list is merged, the proportion values ​​of all components in the group are added up, the primitive combination content of any component in the group is retained, and all merging schemes are arranged to obtain the duplicate component merging scheme. The weighted optimized sparse component set with response orientation is subjected to duplicate component merging. According to the rules in the duplicate component merging scheme, the duplicate components are replaced with single components after the cumulative proportion. All merged components are sorted together to obtain the merged sparse component set with response orientation. The interaction paths of the merged sparse component set with response direction are compared to identify sparse components with conflicting interaction paths. The interaction path content and proportion of each conflicting component are recorded, and all conflicting components are sorted together to obtain the interaction path conflict component set. The path verification is performed on the frequency modulation and voltage regulation scenario interaction logic anchor point set corresponding to the interaction path conflict component set. The correct interaction logic of each conflict path is determined, the adjustment direction and target content of each conflict component are recorded, and all adjustment schemes are arranged to obtain the conflict component adjustment scheme. The merged set of sparse components with responsiveness is adjusted for conflicting components. The interaction path content of the conflicting components is modified according to the rules in the conflicting component adjustment scheme. All the adjusted components are then orderly aggregated to obtain a deduplicated and optimized set of sparse components with responsiveness.

8. The method as described in claim 1, characterized in that, The inverse transformation operation is performed on the sparse component set with response directionality to reconstruct the key operating data required for the coordinated frequency and voltage regulation control of the virtual power plant and the large power grid. Error verification is performed on the key operating data, and abnormal data units with reconstruction deviations exceeding a preset range are removed to obtain the verified key operating data, including: The deduplication and optimization set of sparse components with response orientation is inversely transformed to restore the primitive combination corresponding to each sparse component to the original representation of the real-time running data. All the restored contents are collected in an orderly manner to obtain the initial reconstruction key running data. The initial key operational data of reconstruction is compared with the real-time operational data subset with responsiveness one by one, and the differences and locations of each data are recorded. All comparison results are arranged to obtain the reconstruction data difference description set. The set of reconstructed data differences is filtered, and the initial key operation data of reconstruction with differences exceeding the preset range is extracted. All the filtered data are then systematically aggregated to obtain a subset of the initial key operation data of reconstruction with deviations exceeding the range. The deduplication and optimization of the sparse components with responsiveness corresponding to the initial reconstruction key running data subset with deviations exceeding the range are located, the primitive combination content corresponding to the components is adjusted, and all adjusted components are orderly collected to obtain the corrected sparse component subset with responsiveness. The modified sparse component subset with response direction is inversely transformed to restore the adjusted primitive combination to the original representation of the real-time running data. All restored contents are then collected in an orderly manner to obtain the modified key running data. The corrected critical operating data is merged with the initial reconstructed critical operating data, replacing the original data with excessive deviations. All merged data is then systematically aggregated to obtain the critical operating data that has passed verification.

9. The method as described in claim 8, characterized in that, The process involves merging the corrected key operational data with the initial reconstructed key operational data, replacing the original data with excessive deviations, and then systematically aggregating all the merged data, including: The key operational data that passed the verification were correlated in time. Each data point was linked together according to the time correlation path of the real-time operational data set. The time position of each data point and its correlation with adjacent data were marked. All the correlation content was arranged to obtain the time correlation map of key operational data. A time-series consistency check is performed on the key operational data time-series correlation graph to determine the content where the correlation between adjacent data and the correlation between real-time operational data set are inconsistent. The time-series information and differences at each abnormal position are recorded, and all check results are arranged to obtain a time-series correlation anomaly description set. The time-series associated anomaly description set is located, and the key operational data that has passed the verification at the corresponding position is extracted. All the extracted data are then systematically aggregated to obtain a subset of key operational data for time-series anomalies. The time-series correlation path of the real-time running data set corresponding to the subset of time-series abnormal key running data is extracted in detail. The correct correlation and data content corresponding to the abnormal position are extracted. The correction direction and target content of each abnormal data are recorded. All correction schemes are arranged to obtain the time-series correction scheme. Time-series correction is performed on the critical operational data subset with time-series anomalies. The time-series association content and self-description of the data are modified according to the rules in the time-series correction scheme. All the corrected data are then collected in an orderly manner to obtain the time-series corrected critical operational data subset. The critical operational data subset after timing correction is merged with the critical operational data that has passed verification, and the original data with timing anomalies is replaced. All the merged data is then systematically aggregated to obtain critical operational data with consistent timing.

10. A control device, characterized in that, include: processor; And a memory, wherein the memory stores a computer program that, when run by the processor, causes the processor to perform the method as described in any one of claims 1 to 9.

Images