A multi-scene-oriented virtual power plant flexible resource dynamic image generation and adaptive updating method

Abstract

The present application relates to a kind of virtual power plant flexible resource dynamic image generation and adaptive updating method for multi-scenario demand, comprising obtaining the multi-source data of virtual power plant flexible resource;Initial resource image is generated;The scene weight corresponding to current business scenario is generated;Scene adaptation calculation is carried out to initial resource image and forms controllable potential boundary after prediction;Correlation rule mining and verification are carried out;When meeting the portrait update trigger condition, initial resource image, scene weight and controllable potential boundary are updated, and the updated resource image is output. Using the virtual power plant flexible resource dynamic image generation and adaptive updating method for multi-scenario demand designed by the present application, the problems of virtual power plant flexible resource image static, insufficient scene adaptation and dynamic updating not in time can be solved.

Description

Technical Field

[0001] This invention relates to the field of technology, specifically to a method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant to meet the needs of multiple scenarios. Background Technology

[0002] With the expansion of distributed renewable energy, energy storage systems, electric vehicles, controllable loads, and microgrids, virtual power plants are gradually becoming an important technological form for aggregating dispersed and flexible resources and participating in grid dispatch, demand response, and electricity market transactions. Different types of flexible resources differ in terms of output fluctuations, response speeds, durations, regulation costs, reliability, and user willingness to respond. Furthermore, the value of the same resource varies across different business scenarios, including peak shaving, frequency regulation, electricity trading, demand response, and industrial park energy management. Therefore, virtual power plants need to establish data models that reflect the static attributes, dynamic operating status, predictive capabilities, and scenario adaptability of resources.

[0003] Existing virtual power plant resource modeling methods often rely on equipment type, rated capacity, or a single operating curve as the basis for resource description. This makes it difficult to simultaneously reflect the resource's responsiveness, economic efficiency, reliability, and controllable potential boundaries under different business scenarios. While some solutions can classify resources or forecast loads, they lack a linkage mechanism between scenario requirements and profile fields, as well as an adaptive update process based on data quality, communication status, weather changes, electricity price changes, and response deviations. This causes resource profiles to lag behind actual operating conditions, thereby affecting the accuracy of resource selection, aggregation scheduling, and market pricing.

[0004] Therefore, existing technologies have shortcomings and need to be improved and developed. Summary of the Invention

[0005] The present invention provides a method for generating and adaptively updating a dynamic profile of flexible resources in a virtual power plant that is oriented towards multiple scenarios, in order to solve the problems of static flexible resource profiles of virtual power plants, insufficient scenario adaptation, and untimely dynamic updates in the prior art.

[0006] This invention provides a method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant to meet the needs of multiple scenarios. The method is executed by a virtual power plant resource profile generation platform and includes:

[0007] Acquire multi-source data of flexible resources in a virtual power plant, and perform quality evaluation, cleaning, and time alignment on the multi-source data;

[0008] An initial resource profile is generated based on resource type, grid connection point, and cleaned multi-source data.

[0009] Based on the current business scenario, the scenario requirement template is invoked to generate a scenario weight corresponding to the current business scenario;

[0010] Based on the scenario weights, scenario adaptation calculations are performed on the initial resource profile, and the short-term, medium-term, and long-term operating states of flexible resources are predicted to form a controllable potential boundary.

[0011] We perform association rule mining on the flexible resource operation parameters and verify the association rules through a knowledge graph.

[0012] When the profile update trigger condition is met, the initial resource profile, scene weight, and controllable potential boundary are updated, and the updated resource profile is output. The updated resource profile includes profile fields, profile fit, controllable potential boundary, resource relationship matrix, and scene tags.

[0013] Furthermore, the acquisition of multi-source data on the flexible resources of the virtual power plant includes:

[0014] Acquire equipment profile data, grid connection point data, rated capacity data, real-time power data, state of charge data, start-stop status data, historical response data, historical response deviation data, meteorological data, electricity price data, market clearing data, dispatch instruction data, user response willingness data, and communication status data for distributed power sources, energy storage systems, controllable loads, electric vehicles, and microgrids.

[0015] The multi-source data is aligned according to a unified timestamp, and the data quality value is calculated according to the following formula: In the formula, For the first A flexible resource at any time Data quality values; For flexible resource serial numbers; This refers to the moment of data acquisition. Integrity weight; Score for completeness; For accuracy weighting; Score for accuracy; Consistency weight; For consistency score; Timeliness is a weighting factor; The score is based on timeliness; among which, Completeness score, accuracy score, consistency score, and timeliness score are all normalized scores, ranging from 0 to 1; completeness weight, accuracy weight, consistency weight, and timeliness weight are all non-negative weights.

[0016] Furthermore, the quality assessment, cleaning, and time alignment of the multi-source data include:

[0017] When the data quality value is lower than the preset quality threshold, missing data is filled in, data that exceeds the equipment operation constraints is marked as abnormal, and high-frequency noise in the power curve is denoised. The equipment operation constraints include at least one of the following: rated power constraints, upper and lower limits of state of charge constraints, ramp rate constraints, start-stop state constraints, and grid connection point capacity constraints.

[0018] A set of candidate features for portraits is extracted from the cleaned multi-source data. The set of candidate features for portraits includes time-domain features, frequency-domain features, response features, and economic features.

[0019] The time-domain features include mean, variance, skewness, kurtosis, and autocorrelation; the frequency-domain features include dominant frequency components and periodic intensity; the response features include response delay, response bias, ramp rate, and duration; and the economic features include adjustment costs, market revenue, compensating price sensitivity, and default costs.

[0020] Furthermore, the process of generating an initial resource profile based on resource type, grid connection point, and cleaned multi-source data includes:

[0021] For each flexible resource, a profile field is established, which includes resource type, grid connection point, rated capacity, adjustable capacity, response time, duration, ramp rate, adjustment accuracy, adjustment cost, lifespan loss, response willingness, reliability, and suitable scenarios.

[0022] Convert the profile fields into standardized profile features using the following formula:

[0023] ;

[0024] In the formula, For the first A flexible resource at any time The A standardized profile feature; For the image field sequence number; For the first A flexible resource at any time The Each portrait field value; For the first The maximum value of each portrait field in the sample set; For the first The minimum value of each portrait field in the sample set; This is a set of positive profile fields, where a larger field value indicates stronger resource capabilities. This is a set of reverse profiling fields, where a larger field value indicates weaker resource capabilities; when At that time, The value is assigned to 1; the positive profile field set and the negative profile field set do not overlap; the profile field value is determined by at least one of the following: device file data, operation statistics data, response statistics data, cost statistics data, and user response willingness data.

[0025] Furthermore, the step of generating a scenario weight corresponding to the current business scenario by calling a scenario requirement template based on the current business scenario includes:

[0026] The current business scenarios are identified as one of the following: industrial high-energy-consuming regulation scenario, commercial and residential flexible regulation scenario, electricity market trading scenario, microgrid group coordination scenario, demand response scenario, park integrated energy management scenario, electricity market scenario, peak shaving scenario, and frequency regulation scenario.

[0027] The scenario weight of each profile field in the business scenario is calculated according to the following formula: In the formula, For business scenarios Next The scenario weight of each profile field; Number the business scenarios; For business scenarios Next Demand intensity value for each profile field; Sum the sequence number of the portrait field; Total number of image fields; For business scenarios Next The demand intensity value of each profile field; the demand intensity value is determined by a preset scenario demand template matrix, which is used to record the demand intensity mapping relationship between business scenarios and profile fields;

[0028] In the scenario demand template corresponding to peak shaving, the demand intensity values ​​for adjustable capacity and duration are higher than the demand intensity value for response time; in the scenario demand template corresponding to frequency regulation, the demand intensity values ​​for response time, ramp rate, and regulation accuracy are higher than the demand intensity value for duration; in the scenario demand template corresponding to electricity market trading, the demand intensity values ​​for regulation cost, market revenue, and reliability are higher than the demand intensity value for ramp rate.

[0029] Furthermore, the step of performing scene adaptation calculation on the initial resource profile based on the scene weights includes:

[0030] The profile suitability of flexible resources in the current business scenario is calculated using the following formula: In the formula, For the first A flexible resource at any time For business scenarios Image fit; Total number of image fields; For business scenarios Next The scenario weight of each profile field; For the first A flexible resource at any time The A standardized profile feature;

[0031] The profile fit is written into the resource profile, and the resource filtering order under the same business scenario is generated according to the profile fit. The resource filtering order is generated from large to small profile fit. The larger the profile fit, the more the corresponding flexible resource is adapted to the current business scenario.

[0032] Furthermore, the prediction of the short-term, medium-term, and long-term operating states of flexible resources to form a controllable potential boundary includes:

[0033] 0 to 4 hours is used as the short-term forecasting scale, 4 to 72 hours as the medium-term forecasting scale, and more than 72 hours as the long-term forecasting scale.

[0034] A gradient boosting tree model is used to generate short-term operating status predictions, a sequence-to-sequence long short-term memory network model is used to generate medium-term operating status predictions, and a time-series decomposition model is used to generate long-term operating status predictions. The prediction results for wind power resources and photovoltaic resources are corrected using a time-series prediction model with attention mechanism.

[0035] The controllable potential boundary is generated according to the following formula, where the controllable potential boundary is the range of active power adjustment, load shedding, or energy storage charging and discharging power provided by flexible resources within the prediction scale:

[0036] ;

[0037] In the formula, For the first A flexible resource at any time Prediction scale Significance level The controllable potential boundary below; For prediction scale; The significance level; For the first A flexible resource at any time Prediction scale The lower quantile predicted value; For the first A flexible resource at any time Prediction scale The high quantile predicted value.

[0038] Furthermore, the step of mining association rules for flexible resource operation parameters and validating these association rules using a knowledge graph includes:

[0039] The continuous operating parameters are discretized, and candidate association rules are generated using the FP-Growth algorithm.

[0040] The support and confidence of candidate association rules are calculated using the following formula:

[0041] ;

[0042] In the formula, Candidate association rules Support level; For candidate association rules Confidence level; This is the set of preceding runtime parameters; For the set of subsequent running parameters; For the running sample set; For a set of running parameters that includes the preceding item The set of running samples; For a set of runtime parameters including the following item The set of running samples; The number of samples in the running sample set; For a set of running parameters that includes the preceding item The number of samples; To simultaneously include the previous set of running parameters and subsequent set of running parameters The number of samples;

[0043] When the support is not less than the support threshold, the confidence is not less than the confidence threshold, and there is a device physical mechanism relationship, operation specification relationship, or resource location relationship in the knowledge graph that is consistent with the candidate association rule, the candidate association rule is written into the resource relationship matrix; otherwise, the candidate association rule is marked as a rule to be reviewed; the knowledge graph includes device entities, operation parameter entities, geographical location entities, business scenario entities, and operation specification entities;

[0044] The method of verifying association rules through knowledge graphs includes mapping the previous and subsequent sets of running parameters in the candidate association rules to entities, and retrieving the relationship paths between the corresponding entities in the knowledge graph.

[0045] Furthermore, the step of updating the initial resource profile, scene weight, and controllable potential boundary when the profile update trigger condition is met includes:

[0046] The image update trigger flag is calculated using the following formula: In the formula, For the first A flexible resource at any time The image update trigger flag; Used as a marker for scene transition events; Indicators of sudden weather events; This serves as a marker for events involving excessive electricity prices. For the first A flag indicating an out-of-limit response deviation for a flexible resource; For the first A flag indicating an abnormal communication status event for a flexible resource; The maximum value function is defined as follows: The scenario switching event flag is determined based on whether the current business scenario differs from the previous one; it is 0 if they are the same and 1 if they are different. The meteorological change event flag is determined based on whether the change in meteorological variables exceeds a preset meteorological change threshold; it is 0 if it does not exceed the threshold and 1 if it does. The electricity price exceeding the limit event flag is determined based on whether the electricity price exceeds a preset electricity price threshold; it is 0 if it does not exceed the threshold and 1 if it does. The response deviation exceeding the limit event flag is determined based on whether the deviation between the actual response amount and the command response amount exceeds a preset deviation threshold; it is 0 if it does not exceed the threshold and 1 if it does. The communication status abnormality event flag is determined based on communication abnormalities such as communication delay, data packet loss, or device offline; it is 0 when none of the communication abnormalities exist and 1 when any of the communication abnormalities exist.

[0047] when When, the portrait is updated; when At that time, maintain the resource profile from the previous moment.

[0048] Furthermore, it also includes:

[0049] The update coefficient is determined based on the data quality value and the communication status abnormal event flag: when the data quality value is not less than the preset quality threshold and the communication status abnormal event flag is 0, the update coefficient is determined as the first update coefficient; when the data quality value is less than the preset quality threshold or the communication status abnormal event flag is 1, the update coefficient is determined as the second update coefficient, which is less than the first update coefficient.

[0050] Update the portrait field value using the following formula: In the formula, For the first A flexible resource at any time The updated version Each portrait field value; For the first A flexible resource at any time The update coefficient, and ; For the first The flexible resource in the previous moment Each portrait field value; For the first A flexible resource at any time After cleaning Observations for each portrait field;

[0051] The updated resource profile includes profile fields, profile adaptability, controllable potential boundaries, resource relationship matrix, and scenario tags. The scenario tags include resource type tags, technical characteristic tags, performance level tags, and operational characteristic tags. The updated resource profile is output as a data object for resource filtering, test case generation, aggregation scheduling, and market quotation invocation.

[0052] Beneficial effects:

[0053] As can be seen from the above technical solutions, the present invention provides a method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant to meet the needs of multiple scenarios, which has the following beneficial effects:

[0054] 1. This application incorporates data such as equipment files, real-time power, state of charge, historical response, meteorological data, electricity prices, dispatch instructions, and communication status into a single data benchmark through multi-source data quality assessment, cleaning, and time alignment. Because the data sources for the flexible resources of a virtual power plant are diverse, with different sampling periods and data formats, directly performing profiling calculations could easily result in missing, abnormal, or delayed data being used as valid input, causing the profiling fields to deviate from the actual operating state. This application first evaluates the completeness, accuracy, consistency, and timeliness of the data, and then performs imputation, anomaly marking, and noise reduction processing based on the evaluation results, ensuring that the data entering the profiling model is comparable and traceable. Therefore, the subsequently generated profiling fields, fit, and controllable potential boundaries are all based on quality-controlled data, reducing the interference of multi-source heterogeneous data on resource evaluation results.

[0055] 2. This application converts fields such as resource type, grid connection point, rated capacity, adjustable capacity, response time, duration, ramp rate, regulation accuracy, regulation cost, lifetime loss, response willingness, and reliability into standardized profile features, enabling flexible resources with different physical attributes and dimensions to be compared in a unified feature space. Distributed photovoltaic systems, energy storage systems, controllable loads, electric vehicles, and microgrids operate on different mechanisms; directly comparing rated capacity or real-time power cannot reflect their actual contribution to specific business scenarios. This application standardizes both forward and reverse fields, converting fields where "larger values ​​indicate stronger capabilities" and "larger values ​​indicate weaker capabilities" into profile features in the same direction. This provides a unified data representation basis for resource selection, aggregation ranking, and scheduling decisions, reducing discrepancies in the evaluation of heterogeneous resources.

[0056] 3. This application establishes a scenario requirement template and generates scenario weights for profile fields based on the current business scenario, enabling the same resource profile to be reinterpreted as the business scenario changes. Peak shaving scenarios focus more on adjustable capacity and duration, frequency regulation scenarios focus more on response time, ramp rate, and adjustment accuracy, and market transaction scenarios focus more on adjustment costs, market benefits, and reliability. If fixed weights are used to evaluate resources, some resources may be incorrectly ranked in specific scenarios. This application transforms business requirements into field weights through scenario requirement templates and then calculates profile suitability based on the weights, allowing resource evaluation results to change with scenario switching. As a result, the platform can automatically filter more suitable resource combinations in different business scenarios, improving the consistency between resource calls and business objectives.

[0057] 4. This application generates controllable potential boundaries through multi-scale predictions at short, medium, and long term, enabling resource profiles to reflect not only the current state but also the range of adjustable capacity over a future period. Virtual power plant dispatch typically requires pre-determining resource availability; single-point predictions cannot reflect the uncertainties in wind and solar power output, load demand, and user response. This application utilizes different prediction models to handle short-term real-time fluctuations, medium-term cyclical patterns, and long-term trend changes, and forms controllable potential boundaries using quantile prediction, allowing the dispatch system to obtain upper and lower boundary information before allocating resources. Therefore, resource selection and market pricing are based not only on current profile fields but also on the future adjustable capacity range, reducing the risks associated with over-commitment, insufficient response, or dispatch deviations.

[0058] 5. This application introduces knowledge graph verification after association rule mining, ensuring that the operational parameter relationships discovered through data-driven discovery are constrained by equipment physical mechanisms, operating specifications, and resource location relationships. Simply relying on association rule algorithms may misjudge random correlations as stable patterns, especially when sample size is insufficient, scenarios change frequently, or external factors are strongly coupled. Incorrect rules may affect demand response strategies and resource aggregation judgments. This application first filters candidate rules using support and confidence, then performs entity mapping and relationship path verification using a knowledge graph. Rules that conform to equipment mechanisms and operating specifications are written into the resource relationship matrix, while rules lacking support or exhibiting conflicts are marked as rules requiring review. Thus, the resource relationship matrix combines statistical data and domain knowledge constraints, improving the interpretability of the profiling results.

[0059] 6. This application employs an event-driven profile update mechanism, triggering updates when scenarios change, weather changes occur, electricity prices exceed limits, response deviations exceed limits, or communication status anomalies occur. The fusion ratio of historical profile field values ​​and current cleaned observations is controlled by an update coefficient. If the profile is not updated for a long time, changes in resource status cannot be reflected in a timely manner; if the current observations are used entirely at each sampling moment, it is easily affected by noisy data and communication anomalies. This application determines whether an update is needed by using an update trigger flag and determines the update coefficient based on data quality values ​​and communication status anomaly event flags. When the data is reliable, the weight of the current observation is increased; when the data is unreliable, more historical profile information is retained. Therefore, the profile update process balances real-time performance and stability, making it more suitable for the continuous operation requirements of virtual power plants across multiple scenarios.

[0060] It should be understood that all combinations of the foregoing concepts and the additional concepts described in more detail below can be considered part of the inventive subject matter of this disclosure, provided that such concepts do not contradict each other.

[0061] The foregoing and other aspects, embodiments, and features of the teachings of the present invention will be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the invention, such as features and / or beneficial effects of exemplary embodiments, will become apparent from the following description or may be learned through practice of specific embodiments according to the teachings of the present invention. Attached Figure Description

[0062] The accompanying drawings are not drawn to scale. In the drawings, each identical or nearly identical component shown in the various figures may be denoted by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the invention will now be described by way of example and with reference to the accompanying drawings, wherein:

[0063] Figure 1This is a flowchart illustrating a method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant to meet the needs of multiple scenarios, as described in this application.

[0064] Figure 2 This is a flowchart of step S104 of a method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant for multi-scenario needs, as described in an embodiment of this application.

[0065] Figure 3 This is a flowchart of step S106 of a method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant for multi-scenario needs, as described in an embodiment of this application.

[0066] Figure 4 This is a flowchart of step S110 of a method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant for multi-scenario needs, as described in an embodiment of this application.

[0067] Figure 5 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art to which this invention pertains.

[0069] The terms "first," "second," and similar words used in the specification and claims of this patent application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, unless the context clearly indicates otherwise, the singular forms of "an," "a," or "the," etc., do not indicate a quantity limitation, but rather indicate the presence of at least one. Terms such as "comprising" or "including" mean that the element or object preceding "comprising" encompasses the features, integrals, steps, operations, elements, and / or components listed following "comprising" or "including," and do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or collections thereof. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; these relative positional relationships may change accordingly when the absolute position of the described object changes.

[0070] Existing virtual power plant resource modeling methods often rely on equipment type, rated capacity, or a single operating curve as the basis for resource description. This makes it difficult to simultaneously reflect the resource's responsiveness, economic efficiency, reliability, and controllable potential boundaries under different business scenarios. While some solutions can classify resources or forecast loads, they lack a linkage mechanism between scenario requirements and profile fields, as well as an adaptive update process based on data quality, communication status, weather changes, electricity price changes, and response deviations. This causes resource profiles to lag behind actual operating conditions, thereby affecting the accuracy of resource selection, aggregation scheduling, and market pricing.

[0071] In view of this, embodiments of the present invention provide a method for dynamic generation and adaptive updating of flexible resource profiles in virtual power plants for multi-scenario needs. This method is executed by a virtual power plant resource profile generation platform, which can be deployed on a virtual power plant operation and management platform, a cloud server, or an edge collaborative computing platform. This platform receives flexible resource operation data, external environment data, and business scenario data, and completes data preprocessing, profile generation, scenario adaptation, prediction boundary calculation, association rule verification, and profile updating within the same execution entity. (Refer to...) Figure 1 ,include:

[0072] Step S102: Obtain multi-source data of the flexible resources of the virtual power plant, and perform quality evaluation, cleaning and time alignment on the multi-source data.

[0073] Step S104: Generate an initial resource profile based on the resource type, grid connection point, and cleaned multi-source data.

[0074] Step S106: Based on the current business scenario, call the scenario requirement template to generate the scenario weight corresponding to the current business scenario.

[0075] Step S108: Perform scenario adaptation calculations on the initial resource profile based on scenario weights, and predict the short-term, medium-term, and long-term operating status of flexible resources to form a controllable potential boundary.

[0076] Step S110: Perform association rule mining on the flexible resource operation parameters and verify the association rules through a knowledge graph.

[0077] Step S112: When the profile update trigger condition is met, update the initial resource profile, scene weight and controllable potential boundary, and output the updated resource profile; the updated resource profile includes profile fields, profile fit, controllable potential boundary, resource relationship matrix and scene tags.

[0078] The virtual power plant resource profile generation platform first acquires multi-source data on flexible resources and performs quality evaluation, cleaning, and time alignment on the data to ensure that data from different sources, sampling periods, and formats can participate in calculations under the same time benchmark. Then, it generates an initial resource profile based on resource type, grid connection point, and the cleaned multi-source data. After determining the current business scenario, it calls the scenario requirement template to generate scenario weights and calculates the adaptation result of the initial resource profile to the current business scenario based on these weights. Simultaneously, it predicts the short-term, medium-term, and long-term operating states of flexible resources to form controllable potential boundaries. Next, it performs association rule mining on the operating parameters and uses a knowledge graph for rule validation. When update trigger conditions are met, it updates the profile fields, scenario weights, and controllable potential boundaries, outputting data objects for resource selection, scheduling, and pricing.

[0079] This application integrates multi-source data quality control, scenario requirement templates, profile adaptation calculation, multi-scale controllable potential boundaries, association rule knowledge verification, and event-driven updates into a unified profile generation mechanism. Traditional resource modeling typically describes resources using static parameters or single prediction curves, making it difficult to reconstruct resource value as business scenarios change. This application, through the mapping relationship from "scenario requirements to profile fields," enables the same resource to generate different adaptation results in different scenarios such as peak shaving, frequency adjustment, market transactions, and demand response; through controllable potential boundaries, the profile expresses future adjustment capabilities; and through knowledge graph verification, data rules are constrained by equipment mechanisms and operational specifications. The above combination forms a dynamic profile mechanism for multiple scenarios, which can be distinguished from single classification, single prediction, or static profile solutions, and can transform scattered and flexible resource data into computable, updatable, and callable resource profile objects. Because the profile generation process simultaneously considers data quality, basic resource attributes, business scenario requirements, future prediction boundaries, and rule credibility, the output results are no longer limited to the rated parameters of the equipment, but can reflect the true adjustable value of the resource in the current scenario. For virtual power plant operation platforms, this profile object can directly support resource screening, test case generation, aggregated scheduling, and market quotation, reducing the workload of manually judging resource suitability and reducing scheduling deviations caused by changes in resource status, scene switching, or communication anomalies.

[0080] In some embodiments, acquiring multi-source data on the flexible resources of a virtual power plant includes:

[0081] Acquire equipment profile data, grid connection point data, rated capacity data, real-time power data, state of charge data, start-stop status data, historical response data, historical response deviation data, meteorological data, electricity price data, market clearing data, dispatch instruction data, user response willingness data, and communication status data for distributed power sources, energy storage systems, controllable loads, electric vehicles, and microgrids.

[0082] Align multi-source data according to a unified timestamp and calculate data quality values ​​using the following formula: In the formula, For the first A flexible resource at any time Data quality values; For flexible resource serial numbers; This refers to the moment of data acquisition. Integrity weight; Score for completeness; For accuracy weighting; Score for accuracy; Consistency weight; For consistency score; Timeliness is a weighting factor; The score is based on timeliness; among which, Completeness score, accuracy score, consistency score, and timeliness score are all normalized scores, ranging from 0 to 1; completeness weight, accuracy weight, consistency weight, and timeliness weight are all non-negative weights.

[0083] The data foundation for virtual power plant profiling has been expanded from single-device operation data to a multi-dimensional data set covering equipment, operation, market, dispatch, users, and communication status. It's not simply a data aggregation; instead, a data credibility index is first established through quality evaluation. The platform acquires data on distributed power sources, energy storage systems, controllable loads, electric vehicles, and microgrids, including equipment profiles, grid connection points, rated capacity, real-time power, state of charge, start-up / shutdown status, historical response volume, historical response deviation, weather, electricity prices, market clearing, dispatch instructions, user response intentions, and communication status. This data is then aligned using a unified timestamp. After alignment, the platform calculates data quality values ​​based on four dimensions: completeness, accuracy, consistency, and timeliness, serving as the basis for subsequent data cleaning, anomaly labeling, and profiling updates.

[0084] The integrity score is used to characterize the proportion of valid data points within the sampling window to the total number of data points to be collected; the accuracy score is used to characterize the extent to which data values ​​fall within the rated operating constraints of the equipment; the consistency score is used to characterize the degree of data deviation from different acquisition channels for the same resource; and the timeliness score is used to characterize the delay in data arrival time relative to acquisition time.

[0085] The above design improves the reliability of input data for resource profiling. Virtual power plants aggregate a large number of resources and have numerous access channels, making data gaps, delays, anomalies, and inconsistencies from multiple sources common. Without quality assessment, erroneous data can directly impact resource profiling fields, fit, and controllable potential boundaries. By unifying timestamp alignment and calculating data quality values, data from different resources, data sources, and sampling periods can participate in modeling under the same caliber. This also provides a basis for determining cleaning and update coefficients, thereby improving the stability and usability of resource profiling.

[0086] In some embodiments, quality assessment, cleaning, and time alignment of multi-source data include:

[0087] When the data quality value is lower than the preset quality threshold, missing data is imputed, data exceeding equipment operation constraints is marked as anomalies, and high-frequency noise in the power curve is denoised. Equipment operation constraints include at least one of the following: rated power constraints, upper and lower limits of state of charge constraints, ramp rate constraints, start-stop state constraints, and grid connection point capacity constraints. For randomly missing data, interpolation results of similar resources at adjacent sampling times are used for imputed; for continuously missing data, historical data of the same period or time series predicted values ​​are used for imputed; for data exceeding equipment operation constraints, the original collected values ​​are retained and an anomaly marking field is added to prevent outliers from directly participating in the calculation of the profile field.

[0088] A set of candidate features for profiling is extracted from the cleaned multi-source data. The set of candidate features for profiling includes time-domain features, frequency-domain features, response features, and economic features.

[0089] Time-domain characteristics include mean, variance, skewness, kurtosis, and autocorrelation; frequency-domain characteristics include dominant frequency components and periodic intensity; response characteristics include response delay, response bias, ramp rate, and duration; and economic characteristics include adjustment costs, market returns, compensation price sensitivity, and default costs. Specifically, time-domain characteristics reflect the statistical distribution and temporal correlation of data, frequency-domain characteristics reflect periodic patterns, response characteristics reflect the resource's ability to execute control commands, and economic characteristics reflect the cost-benefit characteristics of resource participation in scheduling and market transactions.

[0090] This application goes beyond simply cleaning the raw data; it directly applies the cleaning results to profile feature engineering. For virtual power plant resources, the mean, variance, skewness, kurtosis, autocorrelation characteristics, dominant frequency component, periodicity, response delay, response deviation, ramp rate, duration, regulation cost, and market revenue indicators in the original power curve correspond to resource operational stability, periodicity, controllability, and economic efficiency, respectively. By grouping and extracting these features, the profile no longer only reflects the static attributes of the equipment but also expresses the dynamic behavior of resources under scheduling instructions, market prices, and changes in the external environment. This improves the business relevance of the profile fields: when selecting resources, virtual power plants need to know not only the resource capacity but also whether the resource is responsive, stable, cost-effective, and has a periodic availability pattern. By extracting time-domain, frequency-domain, response, and economic features from the cleaned data, the platform can provide richer data sources for subsequent standardized profile fields, enabling the profile to simultaneously serve operation scheduling and market decision-making, reducing biases caused by relying solely on equipment profile information for resource evaluation.

[0091] In some embodiments, reference is made to Figure 2 An initial resource profile is generated based on resource type, grid connection point, and cleaned multi-source data, including:

[0092] Step S1041: Establish profile fields for each flexible resource. The profile fields include resource type, grid connection point, rated capacity, adjustable capacity, response time, duration, ramp rate, adjustment accuracy, adjustment cost, lifespan loss, response willingness, reliability, and suitable scenarios.

[0093] Step S1042: Convert the portrait fields into standardized portrait features according to the following formula:

[0094] ;

[0095] In the formula, For the first A flexible resource at any time The A standardized profile feature; For the image field sequence number; For the first A flexible resource at any time The Each portrait field value; For the first The maximum value of each portrait field in the sample set; For the first The minimum value of each portrait field in the sample set; This is a set of positive profile fields. The larger the field value in the positive profile field set, the stronger the resource capability. This is a set of reverse profiling fields. Larger field values ​​in this set indicate weaker resource capabilities. At that time, The value is assigned to 1; the positive profile field set and the negative profile field set do not overlap; the profile field value is determined by at least one of the following: device file data, operation statistics data, response statistics data, cost statistics data, and user response willingness data.

[0096] Adjustable capacity, duration, ramp rate, adjustment accuracy, and reliability are considered positive profile fields; response time, adjustment cost, lifetime loss, and default cost are considered negative profile fields. For fields whose positive or negative attributes cannot be determined, they can be output as tag fields only and will not be included in the profile adaptation calculation.

[0097] For numerical fields that can participate in calculations, standardization is performed based on the positive or negative attributes of the fields, converting fields with different dimensions and different directional meanings into unified standardized profile features; when the maximum and minimum values ​​of a certain field in the sample set are equal, the standardized profile feature is assigned a value of 1 to avoid calculation interruption due to a denominator of 0, thus establishing unified profile fields and standardized expressions for heterogeneous and flexible resources.

[0098] Energy storage systems, electric vehicles, controllable loads, and distributed power sources have different physical mechanisms. Directly using raw field values ​​makes unified sorting and scenario adaptation impossible. This application processes data through forward and reverse field sets separately, ensuring that fields like adjustable capacity, duration, and ramp rate (where "larger is better") are expressed in the same evaluation direction as fields like response time, adjustment cost, and lifetime loss (where "smaller is better"). This approach allows resource profiling to retain the physical meaning of the fields while also enabling subsequent weighted calculations.

[0099] Therefore, it can solve the problem of incomparability of heterogeneous resource profile fields: through standardization, the platform can incorporate the capacity, speed, cost, reliability, and responsiveness of different resources into a unified feature space, providing a basis for scenario adaptability calculation. Since the positive and negative fields use different standardization directions, the evaluation results can reflect the true quality of resource capabilities, avoiding sorting errors caused by opposite field directions. Simultaneously, setting a fixed assignment rule when the maximum value equals the minimum value can improve the stability of the algorithm when there are few samples or when resource attributes are similar.

[0100] In some embodiments, reference is made to Figure 3 Based on the current business scenario, the scenario requirement template is invoked to generate a scenario weight corresponding to the current business scenario, including:

[0101] Step S1061: Identify the current business scenario as one of the following: industrial high-energy-consuming regulation scenario, commercial and residential flexible regulation scenario, electricity market trading scenario, microgrid group collaboration scenario, demand response scenario, park integrated energy management scenario, electricity market scenario, peak shaving scenario, and frequency regulation scenario.

[0102] Step S1062: Calculate the scenario weight of each profile field in the business scenario according to the following formula: In the formula, For business scenarios Next The scenario weight of each profile field; Number the business scenarios; For business scenarios Next Demand intensity value for each profile field; Sum the sequence number of the portrait field; Total number of image fields; For business scenarios Next The demand intensity value for each profile field is determined by a preset scenario demand template matrix. This matrix records the mapping relationship between the demand intensity of business scenarios and profile fields. The scenario demand template is stored as a two-dimensional matrix, with rows corresponding to business scenarios and columns corresponding to profile fields. Matrix elements represent the importance of the profile field within the corresponding business scenario. The demand intensity value can be a normalized value between 0 and 1; a larger value indicates a higher impact of the corresponding field on resource filtering, aggregation scheduling, or market pricing for that scenario.

[0103] In the scenario demand template corresponding to peak shaving, the demand intensity values ​​for adjustable capacity and duration are higher than the demand intensity value for response time; in the scenario demand template corresponding to frequency regulation, the demand intensity values ​​for response time, ramp rate, and regulation accuracy are higher than the demand intensity value for duration; in the scenario demand template corresponding to electricity market trading, the demand intensity values ​​for regulation cost, market revenue, and reliability are higher than the demand intensity value for ramp rate.

[0104] The current business scenario is identified as one of the following: industrial high-energy-consuming regulation, commercial and residential flexible regulation, electricity market trading, microgrid group coordination, demand response, integrated energy management of industrial parks, electricity market, peak shaving, and frequency regulation. The demand intensity values ​​of the corresponding profile fields are retrieved from a preset scenario demand template. By normalizing the calculations, the scenario weights of each profile field under the current business scenario are obtained. This allows the importance of each field in the current scenario to participate in subsequent profile adaptation calculations in numerical form, thereby constructing a quantitative mapping relationship between the business scenario and the profile fields.

[0105] Different scenarios prioritize different aspects of resource capabilities. For example, peak shaving scenarios emphasize adjustable capacity and duration, frequency regulation scenarios emphasize response time, ramp rate, and regulation accuracy, while market trading scenarios emphasize cost, revenue, and reliability. Traditional resource profiling typically uses fixed field weights, failing to reflect changes in business scenarios. This application transforms scenario differences into weight differences through scenario requirement templates, converting resource evaluation from a "fixed profile" to a "scenario-adaptive profile," thus improving the match between resource selection and business needs. When business scenarios change, the platform does not need to rebuild the entire resource model; it only needs to call the corresponding scenario requirement template and update the scenario weights to recalculate resource adaptability. Therefore, the same resource can obtain different ranking results in different business scenarios, consistent with the multi-scenario operation characteristics of virtual power plants. This mechanism reduces the arbitrariness of manual weight configuration and allows peak shaving, frequency regulation, market trading, and demand response operations to share the same profile foundation, improving the reusability of the profile model.

[0106] In some embodiments, scene adaptation calculation is performed on the initial resource profile based on scene weights, including:

[0107] The profile suitability of flexible resources in the current business scenario is calculated using the following formula: In the formula, For the first A flexible resource at any time For business scenarios Image fit; Total number of image fields; For business scenarios Next The scenario weight of each profile field; For the first A flexible resource at any time The A standardized profile feature.

[0108] Write the profile fit score into the resource profile, and generate the resource selection order under the same business scenario based on the profile fit score; the resource selection order is generated from large to small profile fit score, and the larger the profile fit score, the more the corresponding flexible resource is adapted to the current business scenario.

[0109] The scenario weights of each profile field in the business scenario are multiplied by the corresponding standardized profile features and summed to obtain the profile fit of flexible resources in the current business scenario. After the profile fit is written into the resource profile, the platform generates a resource screening order according to the profile fit of each resource in the same business scenario from largest to smallest, so that resources with higher fit will be given priority to enter the testing, aggregation scheduling or market quotation process.

[0110] Profile adaptation is used to homogenize and rank heterogeneous resources within the same business scenario. When different resources have different physical types, capacity units, and response mechanisms but have all been converted into standardized profile features, the platform performs weighted aggregation according to scenario weights, enabling energy storage, controllable loads, distributed power sources, and electric vehicles to participate in the screening under a unified evaluation scale.

[0111] The resource profile is transformed from descriptive data into computational metrics that directly drive resource selection. Standardized profile features reflect the resource's inherent capabilities, while scenario weights reflect key business needs. The combination of these two elements forms the profile fit. This fit is neither a single capacity metric nor a single cost metric, but rather a comprehensive evaluation result combining multiple capabilities according to scenario requirements. Therefore, in peak shaving, frequency regulation, or market transaction scenarios, the platform can generate different selection orders based on different resource value standards.

[0112] The above transformation improves the rationality of virtual power plant resource allocation. The order of resource selection directly affects the aggregated scheduling results and market bidding capabilities. If the selection criteria are singular, problems such as large capacity but slow response, low cost but poor reliability, and fast response but insufficient duration may occur. By sorting by profile suitability, the platform can comprehensively consider multiple profile fields and adjust the degree of influence of fields according to the current business scenario, thereby prioritizing resource combinations that better meet business objectives and improving scheduling execution efficiency and market participation quality.

[0113] In some embodiments, the short-, medium-, and long-term operating states of flexible resources are predicted to form a controllable potential boundary, including:

[0114] The forecasting scale is divided into three periods: 0 to 4 hours for short-term forecasting, 4 to 72 hours for medium-term forecasting, and over 72 hours for long-term forecasting. Short-term operational status forecasts are mainly used for frequency regulation, real-time scheduling, and intraday rolling corrections; medium-term operational status forecasts are mainly used for peak shaving, demand response, and day-ahead plan corrections; and long-term operational status forecasts are mainly used for resource availability assessment, capacity planning, and periodic corrections of scenario templates.

[0115] A gradient boosting tree model is used to generate short-term operational status predictions, a sequence-to-sequence long short-term memory network model is used to generate medium-term operational status predictions, and a time-series decomposition model is used to generate long-term operational status predictions. The prediction results for wind power resources and photovoltaic resources are corrected using a time-series prediction model with attention mechanism.

[0116] The controllable potential boundary is generated according to the following formula, which represents the range of active power adjustment, load reduction, or energy storage charging and discharging power provided by flexible resources within the forecast scale:

[0117] ;

[0118] In the formula, For the first A flexible resource at any time Prediction scale Significance level The controllable potential boundary below; For prediction scale; The significance level; For the first A flexible resource at any time Prediction scale The lower quantile predicted value; For the first A flexible resource at any time Prediction scale The high quantile predicted value.

[0119] This application employs different prediction models for different scales, forming controllable potential boundaries based on low-quantile and high-quantile predicted values ​​to represent the uncertain range of future resource adjustability. Operational status prediction results are incorporated into the resource profile, and future controllable potential is expressed as boundaries. Traditional resource profiles primarily rely on current status or historical statistical characteristics, failing to reflect the available range of resources within future scheduling cycles. This application uses different prediction models for short-term, medium-term, and long-term scales, adapting to different needs such as frequency regulation, intraday scheduling, day-ahead planning, and long-term resource assessment. By forming boundaries through quantile predictions, rather than simply outputting single-point predicted values, uncertainty can be incorporated into the profile, giving it risk representation capabilities.

[0120] Therefore, it can reduce the uncertainty risks in virtual power plant dispatching and market bidding. Wind power, solar power, load, and electric vehicle charging and discharging behaviors all fluctuate, and single-point forecasts can easily lead to over-utilization or underestimation of adjustable capacity. By using controllable potential boundaries, the platform can determine the upper and lower limits of a resource's future adjustable capacity during resource screening, reserve safety margins during dispatching, and reduce deviation risks during bidding. This boundary can also serve as an important field for updating the resource profile, expanding the resource profile from a static description to a dynamic data object oriented towards future operating cycles.

[0121] In some embodiments, reference is made to Figure 4 The association rules for flexible resource operation parameters are mined and validated using a knowledge graph, including:

[0122] Step S1101: Discretize the continuous operating parameters and generate candidate association rules using the FP-Growth algorithm.

[0123] Step S1102: Calculate the support and confidence of the candidate association rules according to the following formula:

[0124] ;

[0125] In the formula, For candidate association rules Support level; For candidate association rules Confidence level; This is the set of preceding runtime parameters; For the set of subsequent running parameters; For the running sample set; For a set of running parameters that includes the preceding item The set of running samples; For a set of runtime parameters including the following item The set of running samples; The number of samples in the running sample set; For a set of running parameters that includes the preceding item The number of samples; To simultaneously include the previous set of running parameters and subsequent set of running parameters The number of samples.

[0126] Step S1103: When the support is not less than the support threshold, the confidence is not less than the confidence threshold, and there is a device physical mechanism relationship, operation specification relationship or resource location relationship in the knowledge graph that is consistent with the candidate association rule, write the candidate association rule into the resource relationship matrix; otherwise, mark the candidate association rule as a rule to be reviewed; the knowledge graph includes device entities, operation parameter entities, geographical location entities, business scenario entities and operation specification entities.

[0127] Step S1104: Verifying association rules through a knowledge graph includes mapping the previous and subsequent sets of running parameters in the candidate association rules to entities, and retrieving the relationship paths between the corresponding entities in the knowledge graph.

[0128] When a candidate association rule indicates that a change in a certain environmental parameter will cause a change in resource power, the platform maps the environmental parameter, resource type, and power change result to entity nodes in the knowledge graph. If there are equipment physical mechanisms, operating specifications, or location relationships in the knowledge graph that support the change path, the candidate association rule is confirmed. If there are opposite relationships or no supporting path, the rule is marked as a rule to be reviewed.

[0129] First, the continuous operating parameters are discretized. Then, the FP-Growth algorithm is used to generate candidate association rules, and the support and confidence of the candidate association rules are calculated. When the support and confidence reach preset thresholds, the platform further verifies the candidate association rules using a knowledge graph. If the knowledge graph contains a device physical mechanism relationship, operating specification relationship, or resource location relationship consistent with the candidate association rule, the candidate association rule is added to the resource relationship matrix; otherwise, the candidate association rule is marked as a rule to be reviewed.

[0130] This application combines statistical associations obtained through data mining with domain knowledge from a knowledge graph. While relying solely on the FP-Growth algorithm can discover high-frequency co-occurrence relationships, it struggles to determine whether these relationships align with equipment mechanisms or operational specifications. This application utilizes a knowledge graph to perform entity mapping and relationship path verification on candidate rules, ensuring that the rules in the resource relationship matrix are supported by both sample statistics and domain knowledge. This process can distinguish between stable operational patterns, accidental correlations, and pseudo-associations caused by abnormal data. This improves the credibility and interpretability of relationship information in the resource profile. Virtual power plant scheduling requires not only understanding individual resources but also the relationships between different operating parameters, between different resources, and between resources and location conditions. Association rule mining can uncover hidden patterns, and knowledge graph verification can eliminate rules that do not conform to the underlying mechanisms. The rules ultimately written into the resource relationship matrix can be used for demand response strategy formulation, resource combination screening, and abnormal operation review, thereby enhancing the resource profile's ability to support complex scheduling scenarios.

[0131] In some embodiments, when the profile update trigger condition is met, the initial resource profile, scene weights, and controllable potential boundaries are updated, including:

[0132] The image update trigger flag is calculated using the following formula: In the formula, For the first A flexible resource at any time The image update trigger flag; Used as a marker for scene transition events; Indicators of sudden weather events; This serves as a marker for events involving excessive electricity prices. For the first A flag indicating an out-of-limit response deviation for a flexible resource; For the first A flag indicating an abnormal communication status event for a flexible resource; This is the function for finding the maximum value.

[0133] The scene switching event flag is determined based on whether the current business scene is different from the previous business scene; it takes a value of 0 when they are the same and a value of 1 when they are different. The meteorological change event flag is determined based on whether the change in meteorological variables exceeds a preset meteorological change threshold; it takes a value of 0 when it does not exceed the threshold and a value of 1 when it does. The electricity price exceeding the limit event flag is determined based on whether the electricity price exceeds a preset electricity price threshold; it takes a value of 0 when it does not exceed the threshold and a value of 1 when it does. The response deviation exceeding the limit event flag is determined based on whether the deviation between the actual response amount and the command response amount exceeds a preset deviation threshold; it takes a value of 0 when it does not exceed the threshold and a value of 1 when it does. The communication status abnormal event flag is determined based on the communication abnormal status of communication delay, data packet loss, or device offline; it takes a value of 0 when none of the communication abnormal statuses exist and a value of 1 when any of the communication abnormal statuses exist.

[0134] when When, the portrait is updated; when At that time, maintain the resource profile from the previous moment.

[0135] Profile updates are triggered using an event-driven approach; an update is activated whenever any event flag is set to 1. This method avoids frequent recalculation of profiles when resource conditions are stable, and also prevents resource selection bias caused by using the previous profile when scene, electricity price, weather, or communication status changes.

[0136] The platform detects scene switching events, sudden weather events, electricity price exceeding limits events, response deviation exceeding limits events, and communication status abnormal events, and sets the flag of each event to 0 or 1 according to preset rules. When any event occurs, the platform sets the profile update trigger flag to 1 by taking the maximum value function and starts the profile update; when no events occur, the profile update trigger flag is 0, and the platform maintains the resource profile of the previous moment.

[0137] This event-driven approach directly links profile updates to changes in business scenarios and operational status, transforming profile updates from fixed-period updates to multi-event triggered updates. Profile changes for flexible resources in a virtual power plant are often caused by events such as scenario switching, weather changes, electricity price fluctuations, response deviations, and communication anomalies. If updates are only performed at fixed intervals, the old profile may still be used after critical events occur; if updates are performed at every sampling moment, it will increase the computational burden and amplify the impact of noise. This solution constructs unified trigger flags through event flags, giving profile updates clear triggering criteria and covering multiple dimensions such as scenario, environment, market, execution effect, and communication status.

[0138] This design improves the timeliness and computational efficiency of profile updates. When changes occur in business scenarios, electricity prices, weather conditions, or resource response effects, the platform can promptly trigger profile updates, ensuring that profile fields, scenario weights, and controllable potential boundaries reflect the latest operational status. When resource status is stable and no triggering events occur, the platform maintains the profile from the previous moment, reducing unnecessary redundant calculations. This approach is suitable for virtual power plant scenarios with massive and flexible resource access, and is beneficial for maintaining the stable operation of the profile system under conditions of limited computing resources and high data traffic.

[0139] In some embodiments, it also includes:

[0140] The update coefficient is determined based on the data quality value and the communication status abnormal event flag: when the data quality value is not less than the preset quality threshold and the communication status abnormal event flag is 0, the update coefficient is determined as the first update coefficient; when the data quality value is less than the preset quality threshold or the communication status abnormal event flag is 1, the update coefficient is determined as the second update coefficient, which is less than the first update coefficient.

[0141] The first update coefficient enhances the impact of current cleaned observations on the profile field and is suitable for situations with high data quality and normal communication. The second update coefficient increases the retention rate of historical profile field values ​​and is suitable for situations with low data quality, significant communication latency, or offline devices. Through these processes, the profile update process can remain stable even when data reliability changes.

[0142] Update the portrait field value using the following formula: In the formula, For the first A flexible resource at any time The updated version Each portrait field value; For the first A flexible resource at any time The update coefficient, and ; For the first The flexible resource in the previous moment Each portrait field value; For the first A flexible resource at any time After cleaning Each portrait field observation value.

[0143] The updated resource profile includes profile fields, profile adaptability, controllable potential boundaries, resource relationship matrix, and scenario tags. Scenario tags include resource type tags, technical characteristic tags, performance level tags, and operational characteristic tags. The updated resource profile is output as a data object for resource filtering, test case generation, aggregation scheduling, and market quotation invocation.

[0144] By limiting the adaptive updates of profile field values ​​and the data object structure of the updated profile, the platform determines the update coefficient based on data quality values ​​and communication status anomaly event flags. It then uses a weighted fusion of the previous profile field values ​​and the currently cleaned profile field observations to obtain the updated profile field values. The updated resource profile includes profile fields, profile fit, controllable potential boundaries, a resource relationship matrix, and scenario tags. Scenario tags include resource type tags, technical characteristic tags, performance level tags, and operational characteristic tags. The platform outputs the updated resource profile as a data object for resource filtering, test case generation, aggregation scheduling, and market quotation invocation.

[0145] The update process doesn't simply replace historical profiles with current observations. Instead, it determines update coefficients based on data quality and communication status, merging historical profiles and current observations according to their reliability. When data quality is high and communication is normal, current observations better reflect resource status, allowing for a larger update coefficient. Conversely, when data quality is low or communication is abnormal, current observations may be unreliable, necessitating a smaller update coefficient to retain more historical profile information. This mechanism enables the profile update process to be adaptive, preventing sudden changes in profiles due to data anomalies.

[0146] This design improves the continuity, stability, and reusability of resource profiles. During virtual power plant operation, communication delays, equipment offline, data gaps, and abnormal sampling are difficult to completely avoid. Directly updating the profile with abnormal observations could lead to unreasonable fluctuations in resource selection order and scheduling strategies. This application controls the fusion ratio of historical and current values ​​through update coefficients, ensuring the profile updates with changes in the actual state while maintaining smoothness even when data quality is insufficient. The updated profile is further output in the form of standard data objects, facilitating direct access by testing, scheduling, and pricing modules, thus increasing the reusability of the profile results within the business system.

[0147] Another embodiment of the present invention provides a flexible resource dynamic profile generation and adaptive update system for virtual power plants that meets the needs of multiple scenarios. This system is built on a virtual power plant resource profile generation platform and includes:

[0148] The acquisition module is used to acquire multi-source data of the flexible resources of the virtual power plant, and to perform quality evaluation, cleaning and time alignment on the multi-source data.

[0149] The first generation module is used to generate an initial resource profile based on the resource type, grid connection point, and cleaned multi-source data.

[0150] The second generation module is used to call the scenario requirement template according to the current business scenario and generate the scenario weight corresponding to the current business scenario.

[0151] The calculation module is used to perform scene adaptation calculations on the initial resource profile based on the scene weights, and to predict the short-term, medium-term and long-term operating status of flexible resources to form a controllable potential boundary.

[0152] The mining module is used to mine association rules for flexible resource operation parameters and verify the association rules through a knowledge graph.

[0153] The update module is used to update the initial resource profile, scene weights, and controllable potential boundaries when the profile update trigger conditions are met, and output the updated resource profile. The updated resource profile includes profile fields, profile fit, controllable potential boundaries, resource relationship matrix, and scene tags.

[0154] It should be noted that although several units or sub-units of the device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0155] Based on the same inventive concept as the above method embodiments, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it enables the electronic device to implement the control method described in the above embodiments.

[0156] In one embodiment, the electronic device may be a server, and in this embodiment, the structure of the electronic device may be as follows: Figure 5 As shown, it includes a memory, a communication module, and one or more processors.

[0157] Memory is used to store computer programs executed by the processor. Memory can be mainly divided into a program storage area and a data storage area. The program storage area can store the operating system and programs required to run instant messaging functions, etc.; the data storage area can store various instant messaging information and operation instruction sets, etc.

[0158] Memory can be volatile memory, such as random access memory (RAM); memory can also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or memory can be any other medium capable of carrying or storing a desired computer program having the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory can be a combination of the above-mentioned types of memory.

[0159] A processor may include one or more central processing units (CPUs) or digital processing units, etc. A processor is used to implement the aforementioned data processing methods when a computer program stored in memory is invoked.

[0160] The communication module is used to communicate with terminal devices and other servers.

[0161] This application embodiment does not limit the specific connection medium between the above-described memory, communication module, and processor. This application embodiment... Figure 5 The memory and processor are connected via a bus, and the bus is in... Figure 5 The connections between other components are illustrated with arrows and are for illustrative purposes only, not as limiting information. Buses can be categorized as address buses, data buses, control buses, etc. For ease of description, Figure 5 The text uses only one arrow to describe it, but does not indicate that there is only one bus or one type of bus.

[0162] Based on the same inventive concept as the above-described method embodiments, embodiments of the present invention also provide a computer-readable storage medium for storing a computer program. When the computer program is run on a computer, it enables an electronic device to implement the control methods described in the above embodiments. The computer-readable storage medium can be a readable signal medium or a readable storage medium. A readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0163] Based on the same inventive concept as the above-described method embodiments, embodiments of the present invention also provide a computer program product. The computer program product includes a computer program that, when run on an electronic device, causes the electronic device to perform the steps of the control methods described above according to various exemplary embodiments of this application. The program product may take the form of any combination of one or more readable media. These computer program commands can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the commands executed by the processor of the computer or other programmable data processing device generate a process for implementing... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0164] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Those skilled in the art can make various modifications and refinements without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention shall be determined by the claims.

Claims

1. A method for generating and adaptively updating flexible resource dynamic profiles of virtual power plants to meet the needs of multiple scenarios, executed by a virtual power plant resource profile generation platform, characterized in that... include: Acquire multi-source data of flexible resources in a virtual power plant, and perform quality evaluation, cleaning, and time alignment on the multi-source data; An initial resource profile is generated based on resource type, grid connection point, and cleaned multi-source data. This includes: establishing profile fields for each flexible resource, whereby the profile fields include resource type, grid connection point, rated capacity, adjustable capacity, response time, duration, ramp rate, adjustment accuracy, adjustment cost, lifetime loss, response willingness, reliability, and suitable scenarios; and converting the profile fields into standardized profile features according to the following formula: In the formula, For the first A flexible resource at any time The A standardized profile feature; For the image field sequence number; For the first A flexible resource at any time The Each portrait field value; For the first The maximum value of each portrait field in the sample set; For the first The minimum value of each portrait field in the sample set; This is a set of positive profile fields, where a larger field value indicates stronger resource capabilities. This is a set of reverse profiling fields, where a larger field value indicates weaker resource capabilities; when At that time, The value is assigned to 1; the positive profile field set and the negative profile field set do not overlap; the profile field value is determined by at least one of the following: device file data, operation statistics data, response statistics data, cost statistics data, and user response willingness data; Based on the current business scenario, the scenario requirement template is invoked to generate a scenario weight corresponding to the current business scenario; Based on the scenario weights, scenario adaptation calculations are performed on the initial resource profile, and the short-term, medium-term, and long-term operating states of flexible resources are predicted to form a controllable potential boundary. Association rule mining is performed on flexible resource operation parameters, and the association rules are validated using a knowledge graph. This includes: discretizing continuous operation parameters and generating candidate association rules using the FP-Growth algorithm; and calculating the support and confidence of the candidate association rules according to the following formula: In the formula, Candidate association rules Support level; Candidate association rules Confidence level; This is the set of preceding runtime parameters; For the set of subsequent running parameters; For the running sample set; For a set of running parameters that includes the preceding item The set of running samples; For a set of runtime parameters including the following item The set of running samples; The number of samples in the running sample set; For a set of running parameters that includes the preceding item The number of samples; To simultaneously include the previous set of running parameters and subsequent set of running parameters The number of samples; when the support is not less than the support threshold, the confidence is not less than the confidence threshold, and there is a device physical mechanism relationship, operation specification relationship, or resource location relationship in the knowledge graph that is consistent with the candidate association rule, the candidate association rule is written into the resource relationship matrix; otherwise, the candidate association rule is marked as a rule to be reviewed; the knowledge graph includes device entities, operation parameter entities, geographical location entities, business scenario entities, and operation specification entities; the verification of association rules through the knowledge graph includes mapping the previous and subsequent operation parameter sets in the candidate association rule to entities, and retrieving the relationship path between the corresponding entities in the knowledge graph; when the profile update trigger condition is met, the initial resource profile, scene weight, and controllable potential boundary are updated, and the updated resource profile is output, including: calculating the profile update trigger flag according to the following formula: In the formula, For the first A flexible resource at any time The image update trigger flag; Used as a marker for scene transition events; Indicators of sudden weather events; This serves as a marker for events involving excessive electricity prices. For the first A flag indicating an out-of-limit response deviation for a flexible resource; For the first A flag indicating an abnormal communication status event for a flexible resource; The function is to take the maximum value; the scenario switching event flag is determined based on whether the current business scenario is different from the previous business scenario, with a value of 0 when they are the same and a value of 1 when they are different; the meteorological change event flag is determined based on whether the change in meteorological variables exceeds a preset meteorological change threshold, with a value of 0 when it does not exceed the threshold and a value of 1 when it does; the electricity price exceeding the limit event flag is determined based on whether the electricity price exceeds a preset electricity price threshold, with a value of 0 when it does not exceed the threshold and a value of 1 when it does; the response deviation exceeding the limit event flag is determined based on whether the deviation between the actual response amount and the command response amount exceeds a preset deviation threshold, with a value of 0 when it does not exceed the threshold and a value of 1 when it does; the communication status abnormal event flag is determined based on the communication abnormal status of communication delay, data packet loss, or device offline, with a value of 0 when none of the communication abnormal statuses exist and a value of 1 when any of the communication abnormal statuses exist; when When, the portrait is updated; when At the same time, the resource profile of the previous moment is maintained; the updated resource profile includes profile fields, profile adaptability, controllable potential boundary, resource relationship matrix and scene tags.

2. The method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant oriented towards multi-scenario needs, as described in claim 1, is characterized in that... The multi-source data for acquiring flexible resources of the virtual power plant includes: Acquire equipment profile data, grid connection point data, rated capacity data, real-time power data, state of charge data, start-stop status data, historical response data, historical response deviation data, meteorological data, electricity price data, market clearing data, dispatch instruction data, user response willingness data, and communication status data for distributed power sources, energy storage systems, controllable loads, electric vehicles, and microgrids. The multi-source data is aligned according to a unified timestamp, and the data quality value is calculated according to the following formula: In the formula, For the first A flexible resource at any time Data quality values; For flexible resource serial numbers; This refers to the moment of data acquisition. Integrity weight; Score for completeness; For accuracy weighting; Score for accuracy; Consistency weight; For consistency score; Timeliness is a weighting factor; The score is based on timeliness; among which, Completeness score, accuracy score, consistency score, and timeliness score are all normalized scores, ranging from 0 to 1; completeness weight, accuracy weight, consistency weight, and timeliness weight are all non-negative weights.

3. The method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant oriented towards multi-scenario needs, as described in claim 2, is characterized in that... The process of quality assessment, cleaning, and time alignment of the multi-source data includes: When the data quality value is lower than the preset quality threshold, missing data is filled in, data that exceeds the equipment operation constraints is marked as abnormal, and high-frequency noise in the power curve is denoised. The equipment operation constraints include at least one of the following: rated power constraints, upper and lower limits of state of charge constraints, ramp rate constraints, start-stop state constraints, and grid connection point capacity constraints. A set of candidate features for portraits is extracted from the cleaned multi-source data. The set of candidate features for portraits includes time-domain features, frequency-domain features, response features, and economic features. The time-domain features include mean, variance, skewness, kurtosis, and autocorrelation; the frequency-domain features include dominant frequency components and periodic intensity; the response features include response delay, response bias, ramp rate, and duration; and the economic features include adjustment costs, market revenue, compensating price sensitivity, and default costs.

4. The method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant oriented towards multi-scenario needs, as described in claim 3, is characterized in that... The step of calling the scenario requirement template according to the current business scenario and generating the scenario weight corresponding to the current business scenario includes: The current business scenarios are identified as one of the following: industrial high-energy-consuming regulation scenario, commercial and residential flexible regulation scenario, electricity market trading scenario, microgrid group coordination scenario, demand response scenario, park integrated energy management scenario, electricity market scenario, peak shaving scenario, and frequency regulation scenario. The scenario weight of each profile field in the business scenario is calculated according to the following formula: In the formula, For business scenarios Next The scenario weight of each profile field; Number the business scenarios; For business scenarios Next Demand intensity value for each profile field; Sum the sequence number of the portrait field; Total number of image fields; For business scenarios Next The demand intensity value of each profile field; the demand intensity value is determined by a preset scenario demand template matrix, which is used to record the demand intensity mapping relationship between business scenarios and profile fields; In the scenario demand template corresponding to peak shaving, the demand intensity values ​​for adjustable capacity and duration are higher than the demand intensity value for response time; in the scenario demand template corresponding to frequency regulation, the demand intensity values ​​for response time, ramp rate, and regulation accuracy are higher than the demand intensity value for duration; in the scenario demand template corresponding to electricity market trading, the demand intensity values ​​for regulation cost, market revenue, and reliability are higher than the demand intensity value for ramp rate.

5. The method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant oriented towards multi-scenario needs, as described in claim 4, is characterized in that... The step of performing scene adaptation calculation on the initial resource profile based on the scene weights includes: The profile suitability of flexible resources in the current business scenario is calculated using the following formula: In the formula, For the first A flexible resource at any time For business scenarios Image fit; Total number of image fields; For business scenarios Next The scenario weight of each profile field; For the first A flexible resource at any time The A standardized profile feature; The profile fit is written into the resource profile, and the resource filtering order under the same business scenario is generated according to the profile fit. The resource filtering order is generated from large to small profile fit. The larger the profile fit, the more the corresponding flexible resource is adapted to the current business scenario.

6. The method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant oriented towards multi-scenario needs, as described in claim 1, is characterized in that... The prediction of the short-term, medium-term, and long-term operating states of flexible resources to form a controllable potential boundary includes: 0 to 4 hours is used as the short-term forecasting scale, 4 to 72 hours as the medium-term forecasting scale, and more than 72 hours as the long-term forecasting scale. A gradient boosting tree model is used to generate short-term operating status predictions, a sequence-to-sequence long short-term memory network model is used to generate medium-term operating status predictions, and a time-series decomposition model is used to generate long-term operating status predictions. The prediction results for wind power resources and photovoltaic resources are corrected using a time-series prediction model with attention mechanism. The controllable potential boundary is generated according to the following formula, where the controllable potential boundary is the range of active power adjustment, load shedding, or energy storage charging and discharging power provided by flexible resources within the prediction scale: ； In the formula, For the first A flexible resource at any time Prediction scale Significance level The controllable potential boundary below; For prediction scale; The significance level; For the first A flexible resource at any time Prediction scale The lower quantile predicted value; For the first A flexible resource at any time Prediction scale The high quantile predicted value.

7. The method for generating and adaptively updating a flexible resource dynamic profile of a virtual power plant oriented towards multi-scenario needs, as described in claim 1, is characterized in that... Also includes: The update coefficient is determined based on the data quality value and the communication status abnormal event flag: when the data quality value is not less than the preset quality threshold and the communication status abnormal event flag is 0, the update coefficient is determined as the first update coefficient. When the data quality value is less than the preset quality threshold or the communication status abnormal event flag is 1, the update coefficient is determined to be a second update coefficient that is less than the first update coefficient. Update the profile field value using the following formula: In the formula, For the first A flexible resource at any time The updated version Each portrait field value; For the first A flexible resource at any time The update coefficient, and ; For the first The flexible resource in the previous moment Each portrait field value; For the first A flexible resource at any time After cleaning Observations for each portrait field; The updated resource profile includes profile fields, profile adaptability, controllable potential boundaries, resource relationship matrix, and scenario tags. The scenario tags include resource type tags, technical characteristic tags, performance level tags, and operational characteristic tags. The updated resource profile is output as a data object for resource filtering, test case generation, aggregation scheduling, and market quotation invocation.

Images