Virtual power plant optimization control system based on source network load storage cooperation

Abstract

The invention relates to the technical field of virtual power plants, and discloses a virtual power plant optimization control system based on source-grid-load-storage collaboration, which comprehensively acquires and preprocesses wind and light output, load, weather and power grid operation data through a multi-source data acquisition module, provides accurate original data for error analysis, and improves the accuracy of the system. The prediction error space-time correlation analysis module mines space correlation and time self-correlation characteristics of errors and generates an error scene set containing correlation characteristics, the source load scene library module combines with the error scene set to update and form a dynamic scene library containing risks, and the robust optimization scheduling module carries out robust optimization scheduling on the basis of the dynamic scene library. A scheduling scheme is solved by taking economy and robustness as targets, and a real-time scheduling execution module is linked with a system monitoring feedback module, so that error perceptibility, scene dynamic updating and scheduling adaptive adjustment are integrally realized, the ability of a virtual power plant to deal with uncertainty is improved, the economy and power grid stability are guaranteed, and the method is suitable for large-scale popularization and application. And the reliability and competitiveness of system operation are enhanced.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant technology, and in particular to an optimized control system for virtual power plants based on source-grid-load-storage coordination. Background Technology

[0002] Existing virtual power plant control systems collect historical meteorological and load data to construct a database of source-load curve scenarios for typical days, which serves as the basis for predicting wind and solar power output and electricity demand. With maximizing economic efficiency or minimizing cost as the primary objective function, and considering equipment physical constraints and grid security constraints, optimization algorithms are used to calculate the scheduling plans for energy storage, distributed power sources, and other units over a future period. The entire system operates within an open-loop framework of "prediction-optimization-execution," and the accuracy of its decisions highly depends on the precision of the initial prediction data, assuming that future operating scenarios closely match historical typical scenarios.

[0003] However, the core problem with the aforementioned existing technologies lies in the fact that their prediction models severely neglect the spatial and temporal correlation of prediction errors. Spatial correlations include situations where a sudden storm can simultaneously affect all photovoltaic power plants in a region, causing a sharp drop in all photovoltaic output within a short period. Such spatial correlations amplify and compound errors, causing significant impact on the system. Temporal correlations arise because prediction errors are not random. If the prediction at the current moment is too high, the prediction at the next moment is also likely to be too high. This temporal autocorrelation leads to error accumulation, causing the optimized scheduling scheme to completely deviate from its optimal path after several time periods.

[0004] The existing method and system for dynamic optimization scheduling of virtual power plants with source-grid-load-storage coordination (announcement number CN117332963A) attempts to address the stochasticity of renewable energy generation and provide typical data for day-ahead optimization scheduling by constructing a source-load curve scenario library based on external factors (such as season and weather). However, its scenario library is generated based on historical clustering, essentially transforming uncertainty into several definite typical scenarios for optimization. This averaging process smooths out the spatiotemporal correlation characteristics inherent in extreme events and prediction errors, failing to capture and address real risks such as simultaneous sharp drops in output from multiple sites or persistently high prediction errors.

[0005] Furthermore, an existing optimization method for a virtual power plant under source-grid-load-storage coordination (announcement number CN118944156A) employs an improved Zebra Algorithm to determine the optimal capacity configuration and operation scheme for each unit within the system, aiming at system economy and load deficit rate. However, during the optimization process, the wind and solar power output data used by this algorithm comes from the output of an ideal mathematical model and does not embed a set of prediction error scenarios with spatiotemporal correlation. Therefore, the optimal solution obtained by this algorithm is based on deterministic ideal conditions and fails to test the robustness of the scheme under the impact of correlation errors, making it very vulnerable in the real world. Summary of the Invention

[0006] The technical problem to be solved by this invention is that the existing technology has the disadvantage of being unable to capture and cope with the real risks such as the simultaneous sharp drop in output of multiple stations or the continuous high prediction error. To this end, we propose a virtual power plant optimization control system based on source-grid-load-storage coordination.

[0007] To achieve the above objectives, this application adopts the following technical solution: a virtual power plant optimization control system based on source-grid-load-storage coordination, comprising:

[0008] The multi-source data acquisition module is used to collect wind and solar power output data, load data, meteorological data, and power grid operation data in the virtual power plant coverage area, and stores the data after preprocessing.

[0009] The prediction error spatiotemporal correlation analysis module is used to mine the spatial correlation characteristics and temporal autocorrelation of prediction errors and generate a set of error scenarios containing correlation characteristics.

[0010] The Source Load Scenario Library module is used to update historical typical scenarios by combining the error scenario set to form a dynamic scenario library containing risks.

[0011] The robust optimization scheduling module is used to call dynamic scenario library data and solve scheduling schemes for energy storage and distributed power sources with the goals of economy and robustness, combined with equipment and grid constraints.

[0012] The real-time scheduling and execution module is used to execute the scheduling scheme and provide feedback on the execution data.

[0013] The system monitoring and feedback module is used to monitor the operating status. If it deviates from the plan, it will link with the robust optimization scheduling module to make adjustments.

[0014] The technical effects and advantages of this invention are as follows:

[0015] This invention comprehensively collects and preprocesses wind and solar power output, load, meteorological, and power grid operation data through a multi-source data acquisition module, providing accurate raw data for error analysis and solving the problem of weak analytical foundation caused by single data or insufficient preprocessing in traditional methods. The predictive error spatiotemporal correlation analysis module mines the spatial correlation and temporal autocorrelation characteristics of errors, generating an error scenario set containing correlation characteristics, eliminating the traditional deficiency of ignoring the spatiotemporal correlation of errors. The source-load scenario library module combines the error scenario set with updates to form a dynamic scenario library containing risks, breaking through the limitations of static historical typical scenarios. The robust optimization scheduling module, based on the dynamic scenario library, solves scheduling schemes with economic efficiency and robustness as objectives, avoiding the vulnerability of deterministic optimization under error impact. The real-time scheduling execution module and the system monitoring feedback module work together to ensure dynamic adjustment of the scheme, solving the problem of decision deviation under the open-loop framework. Overall, it achieves error perception, dynamic scenario updates, and adaptive scheduling adjustment, which not only improves the virtual power plant's ability to cope with uncertainty but also ensures economic efficiency and power grid stability, enhancing the reliability and competitiveness of system operation. Attached Figure Description

[0016] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts:

[0017] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0018] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0019] Reference Figure 1 As shown, the present invention provides a technical solution: a virtual power plant optimization control system based on source-grid-load-storage coordination, comprising:

[0020] The multi-source data acquisition module collects wind and solar power output data, load data, meteorological data, and grid operation data from the virtual power plant coverage area. After preprocessing, the data is stored and used as raw data for the prediction error spatiotemporal correlation analysis module. This module mines the spatial correlation and temporal autocorrelation characteristics of prediction errors, generating an error scenario set containing these correlation characteristics. This set is used to update the source-load scenario library module. The source-load scenario library module updates historical typical scenarios based on the error scenario set, forming a dynamic scenario library with risks. This module provides input data for the robust optimization scheduling module. The robust optimization scheduling module calls upon data from the dynamic scenario library, aiming for economy and robustness, and considering equipment and grid constraints to solve for scheduling schemes for energy storage and distributed power sources. It outputs the scheduling scheme to the real-time scheduling execution module. The real-time scheduling execution module executes the scheduling scheme and provides feedback on the execution data. The system monitoring feedback module monitors the operating status and, if deviations from the scheme occur, coordinates with the robust optimization scheduling module for adjustments.

[0021] Furthermore, as a preferred embodiment, the prediction error spatiotemporal correlation analysis module specifically includes:

[0022] The spatial correlation modeling unit is used to analyze the geographical correlation of wind and solar power stations and load nodes within a region, identify equipment clusters that are synchronously disturbed, quantify the spatial coupling degree of errors within the cluster, and establish a spatial correlation model. The temporal autocorrelation analysis unit is used to calculate the autocorrelation coefficient of errors at different time periods based on historical predictions and actual data, identify the temporal pattern of persistent error bias, and construct a temporal autocorrelation model. The correlation error scenario generation unit is used to combine the spatial correlation model and the temporal autocorrelation model to generate multiple sets of error scenarios containing spatiotemporal correlation characteristics, and label the probability of occurrence and the degree of impact of the scenarios.

[0023] In this preferred embodiment, the method is used to analyze the geographical correlation between wind and solar power plants and load nodes within a region, identify equipment clusters experiencing synchronization disturbances, quantify the spatial coupling degree of errors within the clusters, and establish a spatial correlation model. Specifically, the following operations are performed:

[0024] Historical forecast and actual operating data of all wind and solar power stations and load nodes within the region are collected. The forecast error of each device in each time period is calculated, and the forecast error is the difference between the actual and predicted values. Simultaneously, the latitude and longitude coordinates of the devices are acquired, and a spatial distribution heatmap of the error within the region is generated based on a spatial interpolation algorithm to initially identify areas with high error correlation. For devices in areas with high error correlation in the heatmap, the covariance of the error sequences of any two devices is calculated. The covariance reflects the synchronicity of error changes. A spatial error covariance function is constructed by combining the geographical distance between devices, and the covariance value is transformed into spatial correlation strength. A density clustering algorithm is used, with spatial correlation strength as the clustering basis, and the cluster radius and maximum density are set. With a small sample size, devices with spatial correlation strength exceeding the clustering threshold are grouped into the same cluster to identify synchronously disturbed device clusters. For each cluster, the covariance matrix of errors of all devices within the cluster is calculated, and principal eigenvalues ​​are extracted through matrix eigenvalue decomposition. The principal eigenvalues ​​reflect the overall correlation level of cluster errors. A spatial coupling degree calculation model is constructed by combining the geographical distribution range of the cluster, with the geographical distribution range being the radius of the largest circumscribed circle, to quantify the spatial coupling degree of errors within the cluster. Finally, based on the cluster division results, spatial error covariance matrix, principal eigenvalues, and coupling degree data, a spatial correlation model that includes error statistical correlation and geographical distribution characteristics is established, providing a statistical basis for the subsequent generation of correlation error scenarios.

[0025] Formula for spatial error covariance function:

[0026] ;

[0027] : Spatial error covariance between the i-th and j-th devices; the larger the value, the stronger the synchronization of device errors. : The length of the time series of historical data; : Time index of the time series; : The prediction error of the i-th device at time t; The prediction error of the j-th device at time t; : The average historical prediction error of the i-th device; : The average historical prediction error of the j-th device; The straight-line distance between the i-th device and the j-th device; Spatial attenuation scale parameter.

[0028] Formula for intra-cluster error space coupling: ;

[0029] Error space coupling degree of the k-th device cluster; : The principal eigenvalue of the k-th cluster error covariance matrix; The m-th eigenvalue of the k-th cluster error covariance matrix; : The number of devices in the k-th cluster; The radius of the largest circumcircle of the geographic distribution of the k-th cluster; The radius of the outermost circle of the virtual power plant's coverage area.

[0030] The specific processing procedure for the above formula is as follows:

[0031] Error data preprocessing: Import historical forecast data and actual data from the equipment, and calculate the error for each time period. , and solve , Import the device's latitude and longitude coordinates and calculate... and Spatial error covariance calculation: Substitute the error data and geographic parameters into... Formula: Calculate the spatial error covariance of all device pairs to generate a covariance matrix; Device cluster identification: Set DBSCAN clustering parameters, with the cluster radius based on λ, and the minimum sample size adjusted according to device density. Clustering is performed based on the association criteria, and a list of devices for each cluster is output. and Spatial coupling degree calculation: Perform eigenvalue decomposition on the covariance matrix of each cluster to obtain... and Substitute Formulas calculate cluster coupling degree; spatial association model output: integrated cluster partitioning results, covariance matrix, and principal eigenvalues. Coupling degree and geographical parameters Generate a spatial association model.

[0032] Specifically, by collecting historical predictions and actual data with high temporal granularity to calculate equipment prediction errors, and combining this with Kriging interpolation to generate an error spatial distribution heatmap, highly correlated areas are initially identified, solving the problem of traditional methods that make it difficult to intuitively locate error spatial clusters. Through the spatial error covariance function, error synchronization is coupled with geographical distance to calculate correlation strength, overcoming the one-sidedness of relying solely on distance or single error statistics. The DBSCAN algorithm, based on correlation strength clustering, accurately identifies clusters of synchronously disturbed equipment, avoiding the subjectivity of manual cluster division. Through the formula for spatial coupling degree of errors within a cluster, combined with the principal eigenvalues ​​of the covariance matrix and the geographical distribution range to quantify the coupling degree, a quantitative characterization of error spatial correlation is achieved. The overall constructed spatial correlation model reflects both the statistical correlation characteristics of errors and integrates geographical distribution features, providing a precise tool for capturing the risk of regional error superposition caused by storms, significantly improving the ability to perceive and characterize spatial correlation errors.

[0033] In this preferred embodiment, the method is used to calculate the autocorrelation coefficient of errors in different time periods based on historical forecasts and actual data, identify the temporal pattern of error persistence bias, and construct a time autocorrelation model. Specifically, the following operations are performed:

[0034] Historical prediction error time series data from the acquisition equipment were collected, with time granularities including 15 minutes, 1 hour, and 1 day, covering at least three complete seasons. The error series was divided into three levels according to time scale: short-term (15 minutes - 2 hours), medium-term (2 - 24 hours), and long-term (1 - 7 days). For each time scale, the autocorrelation coefficients of different lag orders were calculated using the sliding window method, with the window size dynamically adjusted according to the scale: 4 time periods for the short-term, 12 time periods for the medium-term, and 7 time periods for the long-term, resulting in a multi-scale autocorrelation matrix. Trend tests (such as the Mann-Kendall test) were used to identify persistent trends in the error series. The bias period is defined as five or more consecutive periods where the error deviates in the same direction, with the deviation amplitude gradually increasing or remaining stable. Feature parameters such as the bias start time, duration, and deviation growth rate are extracted. Based on the multi-scale autocorrelation coefficient and the bias feature parameters, a comprehensive autocorrelation index with time scale weights is constructed to quantify the error correlation strength of different lag periods. Finally, the multi-scale autocorrelation matrix, bias feature parameters, and comprehensive autocorrelation index are integrated to construct a time autocorrelation model that includes short-term fluctuation correlation, medium-term trend continuation, and long-term cycle recurrence, providing a basis for generating the time correlation characteristics of error scenarios.

[0035] The specific formula for the comprehensive autocorrelation index is as follows: ;

[0036] The composite time autocorrelation index when the lag order is τ; Lag order Time-scale index (1=short-term, 2=medium-term, 3=long-term); The weighting coefficients for the s-th time scale (satisfying) =1, with 0.5 for short-term, 0.3 for medium-term, and 0.2 for long-term (adjustable based on the significance of the error). : The autocorrelation coefficient with lag τ at the s-th time scale; Time decay coefficient; : The actual time interval corresponding to the lag order τ at the s-th time scale (e.g., short-term τ=1 corresponds to 0.25h, medium-term τ=1 corresponds to 1h).

[0037] The specific processing procedure for the above formula is as follows:

[0038] Data stratification: Historical error time series were resampled at 15-minute, 1-hour, and 1-day granularities to obtain error series at three scales; Autocorrelation coefficient calculation: For each scale series, the autocorrelation coefficient was calculated using... Calculate the autocorrelation coefficients for lags τ = 1, 2, ..., 10. Let be the error at time t on the s-scale. The mean of the s-scale error. (Sequence length); Bias feature extraction: Detect consecutive error segments in the same direction using a sliding window, and record the start time and duration of the bias. and deviation growth rate; composite index calculation: , , , Substitute into the formula and calculate. Model construction: Integrating autocorrelation coefficients, bias characteristic parameters, and... The matrix generates a time autocorrelation model.

[0039] Specifically, by collecting error sequences with multiple time granularities and covering multiple seasons, and stratifying them into short-term, medium-term, and long-term segments, and combining this with a dynamic sliding window to calculate the autocorrelation coefficient, the problem of traditional single-time-scale analysis's inability to capture error correlations across different time periods is solved. The Mann-Kendall test is used to identify periods of persistent bias and extract characteristic parameters, overcoming the neglect of the persistent same-direction deviation pattern of errors. By comprehensively formulating the autocorrelation index and integrating multi-scale autocorrelation coefficients with time decay characteristics, the strength of error correlations across different lag periods is quantified, avoiding the one-sidedness of single-scale analysis. The constructed time autocorrelation model covers short-term fluctuations, medium-term trends, and long-term cyclical correlations, accurately characterizing the time autocorrelation of errors, such as the pattern that a current high prediction is likely to be high at the next moment, providing a quantitative tool for capturing the risk of accumulated errors and significantly improving the ability to perceive time-related errors.

[0040] In this preferred embodiment, the method combines a spatial correlation model and a temporal autocorrelation model to generate multiple sets of error scenarios containing spatiotemporal correlation characteristics, and labels the probability of occurrence and the degree of impact of each scenario. Specifically, the following operations are performed:

[0041] The spatial correlation model extracts equipment cluster partitioning results, spatial coupling degree data, and correlation matrix to determine the spatial synchronization characteristics of equipment errors within each cluster. These characteristics represent the coupling strength of unidirectional fluctuations in equipment errors within the cluster. Simultaneously, the model extracts multi-scale autocorrelation indices, bias characteristic parameters, and a comprehensive autocorrelation matrix to clarify the temporal correlation patterns of errors at different lag times. These temporal correlation patterns represent short-term fluctuation persistence and medium-to-long-term trend persistence. An improved Monte Carlo simulation method is employed, introducing spatial and temporal constraints when generating the initial error scenario. Spatial constraints allow for the setting of synchronization fluctuation coefficients for equipment errors within the same cluster based on spatial coupling degree. Temporal constraints allow for the setting of synchronization fluctuation coefficients for errors in adjacent time periods based on comprehensive correlation matrix. The autocorrelation index sets association weights to ensure that the scenarios simultaneously reflect spatial cluster synchronicity and temporal bias persistence. The generated massive number of scenarios are filtered and optimized. High-risk scenarios are retained by calculating the similarity between the scenarios and historical extreme events. Similarity can be a scenario where regional error drops are caused by storms. K-means clustering algorithm is used to merge similar scenarios to reduce the size of the scenario set. For each set of filtered scenarios, the probability of scenario occurrence is calculated by combining the frequency of similar spatiotemporal correlation errors in historical data. The impact of the scenario is quantified by simulating the impact of the error scenario on the virtual power plant dispatch scheme. The impact can be expressed as the resulting power deficit or economic loss. This process completes the generation and labeling of multiple sets of error scenarios with spatiotemporal correlation characteristics.

[0042] Specifically, by extracting data such as cluster partitioning and coupling degree from the spatial correlation model, the spatial synchronization characteristics of errors are clarified. By extracting multi-scale exponents and bias parameters from the temporal autocorrelation model, the temporal correlation patterns are clarified. Then, an improved Monte Carlo simulation is used to introduce spatial and temporal constraints to generate initial scenarios. High-risk scenarios are retained through historical extreme event similarity calculation, and the scale is reduced by K-means clustering. Finally, the probability and impact of labeled scenarios are simulated and labeled by combining historical frequency and scheduling impact. The generated error scenarios can simultaneously reflect spatial cluster synchronization and temporal bias persistence. They include high-risk scenarios such as the sudden drop in regional errors caused by storms, and the scenario set is made efficient and usable through simplification. The labeled probability and impact provide accurate risk basis for subsequent robust optimization scheduling, enabling the scheduling scheme to adapt to spatiotemporal correlation error impacts in advance, greatly improving the virtual power plant's ability to cope with uncertainties, and ensuring operational stability and economy.

[0043] In this preferred embodiment, the prediction error spatiotemporal correlation analysis module further includes:

[0044] The error propagation simulation unit simulates the propagation path of correlated errors in each stage of the power generation, grid, load, and storage system, analyzes the impact of error superposition on the accuracy of wind and solar power output forecasting and load forecasting results, and outputs an error propagation impact report. The scenario selection and optimization unit, based on the error propagation impact report, selects high-risk correlated error scenarios that significantly impact the system, eliminates redundant scenarios, and optimizes the scale and effectiveness of the scenario set. The error propagation simulation unit simulates the propagation path of correlated errors in each stage of the power generation, grid, load, and storage system and analyzes the superposition impact, outputting a propagation report. The scenario selection and optimization unit then selects high-risk scenarios, eliminates redundancies, and optimizes the scenario set based on the report. Ultimately, this accurately locates the impact points of high-risk errors, simplifies the scenario set to improve effectiveness, and provides accurate risk input for subsequent robust scheduling.

[0045] Furthermore, as a preferred embodiment, the robust optimization scheduling module includes:

[0046] The system comprises several functional modules: a multi-objective function construction unit, which aims to minimize the operating cost of the virtual power plant as the economic objective and minimize the load deficit rate and grid-connected power fluctuation under associated error impacts as the robustness objectives; a dynamic constraint adjustment unit, which dynamically adjusts the energy storage charging and discharging power boundaries, the distributed power output upper limit, and the grid interconnection power constraints based on the output fluctuation range and load deviation range of the error scenario set; a scheduling scheme solution unit, which calls the optimization algorithm to solve the energy storage and distributed power scheduling schemes under different error scenarios based on the multi-objective optimization function and dynamic constraints, and outputs the optimal robust scheduling scheme set; and an algorithm parameter adaptation unit, which adapts the algorithm parameters according to the complexity of the error scenario set and the field conditions. The number of scenarios is adjusted to optimize the algorithm's iteration count, search step size, and other parameters, improving the algorithm's solution efficiency and accuracy in scenarios with correlated errors. The robustness verification unit is used to substitute different correlated error scenarios into the optimal scheduling scheme set to verify whether the scheme's operating indicators meet the requirements under error impact. Schemes with insufficient robustness are eliminated, and the optimal scheme is retained. The generated optimal robust scheduling scheme set can dynamically balance economy and robustness, avoiding both weak error resistance caused by solely pursuing cost and resource waste caused by overemphasizing robustness. The schemes verified in multiple scenarios can effectively cope with correlated error impacts, significantly reduce load deficits and power fluctuations, ensure grid stability, control operating costs, and significantly improve the operational reliability and competitiveness of virtual power plants under uncertainty.

[0047] In this preferred embodiment, an optimization algorithm is invoked to solve energy storage and distributed power scheduling schemes under different error scenarios based on a multi-objective optimization function and dynamic constraints, and to output the optimal robust scheduling scheme set. Specifically, the following operations are performed:

[0048] A mixed-integer programming algorithm is used to transform the multi-objective optimization function into a single-objective weighted function. The weights in the single-objective weighted function are determined by the analytic hierarchy process (AHP), with an economic weight of 40% and a robustness weight of 60%. The energy storage charging and discharging state is set as an integer variable, while the energy storage charging and discharging power and distributed power output are set as continuous variables. In the constraint construction phase, dynamic constraints are transformed into piecewise constraint equations. Dynamic constraints are parameters that change with the error scenario. Piecewise constraint equations include the energy storage charging state limit constraint equation under high-risk scenarios and the distributed power output fluctuation constraint equation under medium- and low-risk scenarios. The algorithm employs a scenario-based collaborative iteration strategy. The first iteration solves for common constraints across all error scenarios, such as basic grid safety constraints, obtaining a global basic solution. The second iteration groups the solutions by scenario type and adjusts the basic solutions based on the personalized dynamic constraints for each scenario group. For example, the backup capacity constraint in the wind and solar power drop scenario yields the optimized solutions for each group. The third iteration performs cross-scenario verification on the optimized solutions for each group, calculating the objective function deviation of the same solution in different scenario groups. If the deviation is less than a preset threshold, the solution is retained; if the deviation exceeds the threshold, the iteration returns to the second round to readjust the constraint parameters. After the iteration converges, all retained solutions are sorted by multi-objectives, and the solutions with no obvious shortcomings in terms of economy and robustness are selected to form the optimal robust scheduling scheme set. At the same time, the scenario adaptation range corresponding to each scheme is output, such as scheme A adapting to high-risk wind and solar power error scenarios and scheme B adapting to medium- and low-risk load deviation scenarios.

[0049] Specifically, by invoking a mixed-integer programming algorithm, the multi-objective optimization function is transformed into a single-objective weighted function with weights determined by the analytic hierarchy process (AHP), distinguishing between energy storage charging / discharging states and power, and distributed power generation output. Dynamic constraints that change with the scenario are converted into piecewise constraint equations, and a three-round scenario-based collaborative iteration is employed. After convergence, the solution without bottlenecks is selected and the scenario adaptation range is output. This ensures that the generated set of optimal robust scheduling schemes accurately balances economy and robustness, avoiding bottlenecks caused by solely pursuing cost or robustness. Each scheme has a clearly defined applicable scenario, specifically addressing different situations such as high-risk wind and solar power errors and medium-to-low-risk load deviations. Cross-validation ensures the stability of the schemes across multiple scenarios, significantly improving the virtual power plant's ability to cope with correlated errors and guaranteeing grid safety and operational efficiency.

[0050] Furthermore, as a preferred embodiment, the real-time scheduling execution module includes:

[0051] The system comprises several functional units: a data receiving unit, a dispatch command correction unit, and an execution log unit. The data includes real-time data on the actual operation of energy storage, distributed power sources, and load nodes, such as energy storage state of charge, power output, and load consumption. The dispatch command correction unit compares the actual operating data with the dispatch plan data. If the deviation is within acceptable limits, the original command is maintained. If the deviation exceeds the limits, the dispatch command is initially corrected based on real-time error data from the prediction error spatiotemporal correlation analysis module. The command issuance and execution unit issues the corrected dispatch command or the original command to each execution unit, ensuring that energy storage charging and discharging and distributed power output are executed according to the command, and simultaneously recording the command execution status. The emergency response unit activates the emergency dispatch mode when correlation errors cause system power shortages or excesses to exceed thresholds, prioritizing the use of backup energy storage resources and adjustable load resources to quickly balance system power. The execution log generation unit records the content of dispatch commands, actual execution data, deviations, and correction measures for each time period, forming an execution log to provide data support for subsequent optimization. This ensures real-time response to dispatch deviations, rapid power balancing, and the implementation of dispatch plans, enhancing the system's flexibility and stability in dealing with correlation errors.

[0052] Furthermore, as a preferred embodiment, the system monitoring and feedback module includes:

[0053] The system comprises several modules: Operational Status Monitoring Unit (AS / RS) for real-time monitoring of the virtual power plant's operational status, including wind and solar power output prediction deviations, energy storage operating parameters, grid node voltage and frequency, and actual load consumption versus prediction deviations; Deviation Cause Analysis Unit (DSO) for analyzing whether deviations between operational data and the dispatch plan are caused by associated errors, and identifying the spatiotemporal correlation type and impact range of the errors, based on scenario data from the prediction error spatiotemporal correlation analysis module; Feedback Adjustment Trigger Unit (PATI) for triggering a feedback signal to the robust optimization dispatch module if the deviation is caused by associated errors and exceeds tolerance, initiating a re-optimization process for the dispatch plan; Robustness Index Evaluation Unit (BOC) for periodically evaluating the robustness indicators of the system's dispatch plan based on historical operational data and associated error scenarios, including the frequency of plan adjustments under error impacts, load guarantee rate, and economic loss rate, generating a robustness evaluation report; and Index Optimization Suggestion Unit (IOS) for identifying system optimization suggestions based on the robustness evaluation report. The robustness bottleneck provides optimization suggestions for scenario optimization in the prediction error spatiotemporal correlation analysis module and objective function adjustment in the robust optimization scheduling module; the anomaly alarm unit generates tiered alarm information and pushes it to management personnel according to preset methods when the operation status monitoring unit detects power grid frequency anomalies, energy storage failures, or power imbalance risks caused by correlation errors; the historical data archiving unit archives and stores monitored operation data, deviation analysis results, robustness assessment reports, and alarm records to form a historical database, providing historical data support for model optimization in the prediction error spatiotemporal correlation analysis module and scenario updates in the source-load scenario library module, monitoring the status of each link of the virtual power plant in real time, accurately analyzing whether deviations are correlation errors and their impact range, and triggering re-optimization of the scheduling scheme when the tolerance is exceeded; the robustness is regularly evaluated and optimization suggestions are provided, tiered alarms are triggered when anomalies occur, and archived data supports model and scenario updates, ensuring the system can stably cope with correlation errors and improving operational reliability.

[0054] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A virtual power plant optimization control system based on source network load storage coordination, characterized in that, The application relates to a virtual power plant robust optimization scheduling system. The application comprises the following steps: A multi-source data acquisition module is used to acquire wind and light output data, load data, meteorological data and power grid operation data of a virtual power plant coverage area, and the data is stored after being preprocessed; A prediction error space-time correlation analysis module is used to mine the space correlation characteristics and time autocorrelation of prediction errors, and generate an error scene set containing correlation characteristics; A source and load scene library module is used to update historical typical scenes in combination with the error scene set, and form a dynamic scene library containing risks; A robust optimization scheduling module is used to call dynamic scene library data, and solve a scheduling scheme of energy storage and distributed power supply in combination with equipment and power grid constraints, aiming at economy and robustness; A real-time scheduling execution module is used to execute the scheduling scheme and feed back execution data; 2.The source network load storage coordination based virtual power plant optimization control system according to claim 1, characterized in that, A system monitoring feedback module is used to monitor the operation state, and if the state deviates from the scheme, the robust optimization scheduling module is linked to adjust. The prediction error space-time correlation analysis module comprises the following steps: A space correlation modeling unit is used to analyze the geographical correlation of wind and light power stations and load nodes in the region, identify a synchronous disturbed device cluster, quantify the space coupling degree of errors in the cluster, and establish a space correlation model; A time autocorrelation analysis unit is used to calculate autocorrelation coefficients of errors in different time periods based on historical prediction and actual data, identify the time law of error continuous bias, and build a time autocorrelation model; 3.The source-grid-load-storage collaborative based virtual power plant optimization control system of claim 2, wherein, An associated error scene generation unit is used to generate multiple error scenes containing space-time correlation characteristics in combination with the space correlation model and the time autocorrelation model, and label the scene occurrence probability and influence degree. The space correlation modeling unit is used to analyze the geographical correlation of wind and light power stations and load nodes in the region, identify a synchronous disturbed device cluster, quantify the space coupling degree of errors in the cluster, and establish a space correlation model, and the following operations are performed: Historical prediction data and actual operation data of all wind and light power stations and load nodes in the region are collected, and the prediction error of each device in each time period is calculated; The latitude and longitude coordinates of the devices are obtained, and an error space distribution heat map in the region is generated based on a spatial interpolation algorithm, and the error high correlation region is preliminarily identified; For the devices in the error high correlation region in the heat map, the covariance of error sequences of any two devices is calculated, a space error covariance function is constructed in combination with the geographical distance between the devices, and the covariance value is converted into a space correlation strength; ； : spatial error covariance between the ith device and the jth device, the larger the value, the stronger the device error synchronism; : time series length of historical data; : time index of time series; : prediction error of the ith device at time t; : prediction error of the jth device at time t; : mean of historical prediction error of the ith device; : mean of historical prediction error of the jth device; : straight-line distance between the ith device and the jth device; : spatial attenuation scale parameter; The space error covariance function formula is as follows: A density clustering algorithm is adopted, the space correlation strength is taken as the clustering basis, the clustering radius and the minimum sample number are set, the devices with a space correlation strength higher than a clustering threshold value are divided into the same cluster, and the identification of the synchronous disturbed device cluster is realized; For each cluster, the covariance matrix of errors of all devices in the cluster is calculated, the main eigenvalue is extracted through matrix eigenvalue decomposition, a space coupling degree calculation model is constructed in combination with the geographical distribution range of the cluster, and the space coupling degree of errors in the cluster is quantified; ； : error space coupling degree of the kth device cluster; : the principal eigenvalue of the kth cluster error covariance matrix; : the mth eigenvalue of the kth cluster error covariance matrix; : the number of devices within the kth cluster; : the maximum circumscribed circle radius of the geographical distribution of the kth cluster; : the maximum circumscribed circle radius of the virtual power plant coverage area; The error space coupling degree formula in the cluster is as follows: Finally, the space correlation model containing error statistical correlation and geographical distribution characteristics is established based on the cluster division result, the space error covariance matrix, the main eigenvalue and the coupling degree data.

4. The source-grid-load-storage collaborative based virtual power plant optimization control system according to claim 2 or 3, characterized in that, The method comprises the following steps of: Collecting historical prediction error time series data of the equipment, dividing the error sequence into short-term, medium-term and long-term three levels according to the time scale; for each time scale, the autocorrelation coefficients of different lag orders are calculated by using the sliding window method, and a multi-scale autocorrelation matrix is obtained; Identifying the time period of the persistent bias in the error sequence through trend test and extracting the bias characteristic parameters; Based on the multi-scale autocorrelation coefficients and the bias characteristic parameters, a comprehensive autocorrelation index is constructed by fusing the time scale weight, so as to quantify the error correlation strength in different lag periods; The comprehensive autocorrelation index is specifically calculated according to the following formula: ; : integrated time autocorrelation index with lag order τ; : lag order, : time scale index; : weight coefficient of s-th time scale; : autocorrelation coefficient with lag order τ in s-th time scale; : time decay coefficient; : actual time interval corresponding to lag order τ in s-th time scale; Finally, the multi-scale autocorrelation matrix, the bias characteristic parameters and the comprehensive autocorrelation index are integrated to construct a time autocorrelation model containing short-term fluctuation correlation, medium-term trend continuation and long-term cycle recurrence.

5. The source-grid-load-storage collaborative based virtual power plant optimization control system according to claim 4, wherein, The method comprises the following steps of: Extracting the device cluster division result, spatial coupling degree data and correlation matrix in the spatial correlation model to determine the spatial synchronization characteristics of the error of the devices in each cluster; At the same time, the multi-scale autocorrelation index, the bias characteristic parameters and the comprehensive autocorrelation matrix in the time autocorrelation model are extracted to determine the time correlation law of the error in different lag periods; An improved Monte Carlo simulation method is used to introduce spatial constraints and time constraints when generating the initial error scenario, so as to ensure that the scenario reflects the cluster synchronization in space and the bias persistence in time at the same time; The generated massive scenarios are screened and optimized, the high-risk scenarios are reserved by calculating the similarity between the scenarios and the historical extreme events, and the K-means clustering algorithm is used to merge similar scenarios to reduce the size of the scenario set; For each scenario after screening, the occurrence probability of the scenario is calculated by combining the occurrence frequency of the same type of space-time correlation error in the historical data, the impact degree of the scenario is quantified by simulating the impact of the error scenario on the virtual power plant scheduling scheme, and the generation and labeling of the multiple error scenarios with space-time correlation characteristics are completed.

6. The source-grid-load-storage collaborative based virtual power plant optimization control system of claim 1, wherein, The prediction error space-time correlation analysis module further comprises: An error propagation simulation unit is configured to simulate the propagation path of the associated error in the source, network, load and storage links, analyze the influence degree of error superposition on the prediction accuracy of wind and light output and the load prediction result, and output an error propagation influence report; A scenario screening and optimization unit is configured to screen high-risk associated error scenarios with significant impact on the system based on the error propagation influence report, eliminate redundant scenarios, and optimize the size and effectiveness of the scenario set. 7.The source-grid-load-storage collaborative based virtual power plant optimization control system of claim 1, wherein, The robust optimization scheduling module comprises: A multi-objective function construction unit is configured to set a target weight dynamic adjustment rule, construct a multi-objective optimization function, and take the minimum virtual power plant operation cost as an economic target, and take the minimum load shortage rate and the minimum grid-connected power fluctuation under the impact of associated errors as robustness targets. A dynamic constraint adjustment unit is configured to combine the output fluctuation range and the load deviation range in the error scenario set to dynamically adjust the energy storage charge and discharge power boundary, the distributed power output upper limit and the power grid tie-line power constraint to form a dynamic constraint condition; A scheduling scheme solving unit is configured to call an optimization algorithm, solve the energy storage and distributed power scheduling scheme under different error scenarios based on a multi-objective optimization function and the dynamic constraint condition, and output an optimal robust scheduling scheme set. 8.The source-grid-load-storage collaborative based virtual power plant optimization control system of claim 7, wherein, The method for calling the optimization algorithm, solving the energy storage and distributed power scheduling scheme under different error scenarios based on the multi-objective optimization function and the dynamic constraint condition, and outputting the optimal robust scheduling scheme set specifically performs the following operations: The mixed integer programming algorithm is called to convert the multi-objective optimization function into a single objective weighted function, set the energy storage charge and discharge state as an integer variable, and set the energy storage charge and discharge power and the distributed power output as continuous variables; In the constraint condition construction stage, the dynamic constraint condition is converted into a segmented constraint equation; In the algorithm solving stage, a scene coordination iteration strategy is adopted, the first round of iteration is performed on the common constraint of all error scenarios to obtain a global basic solution; The second round of iteration is performed on the basis of the scene type grouping, the personalized dynamic constraint of each group of scenes is adjusted to obtain a grouped optimization solution; The third round of iteration is performed on the grouped optimization solution to verify the cross-scene, calculate the objective function deviation of the same solution under different scene groups, and if the deviation is less than a preset threshold, the solution is retained. If the deviation exceeds the threshold, the second round of adjustment of the constraint parameter is returned. After the iteration converges, all retained solutions are subjected to multi-objective sorting, and a solution without obvious short board in economy and robustness is selected to form an optimal robust scheduling scheme set, and the scene adaptation range corresponding to each scheme is output. 9.The source-grid-load-storage collaborative based virtual power plant optimization control system of claim 1, wherein, The real-time scheduling execution module includes: An execution data receiving unit is configured to receive the actual operation data of the energy storage, the distributed power and the load node in real time; A scheduling instruction correction unit is configured to compare the actual operation data with the scheduling scheme data, maintain the original instruction if the deviation is within the allowable range, and preliminarily correct the scheduling instruction in combination with the real-time error data of the prediction error space-time correlation analysis module if the deviation exceeds the range; An instruction issuing execution unit is configured to issue the corrected scheduling instruction or the original instruction to each execution unit to ensure that the energy storage charge and discharge and the distributed power output are executed according to the instruction, and to record the instruction execution situation synchronously; An emergency response unit is configured to start an emergency scheduling mode when the associated error causes the system power shortage or surplus to exceed the threshold, and to preferentially call the standby energy storage resource and the adjustable load resource to quickly balance the system power; An execution log generation unit is configured to record the scheduling instruction content, the actual execution data, the deviation and the correction measures of each period to form an execution log.

10. The source-grid-load-storage collaborative based virtual power plant optimization control system of claim 1, wherein, The system monitoring feedback module includes: An operation state monitoring unit is configured to monitor the operation state of each link of the virtual power plant in real time; A deviation reason analysis unit is configured to analyze whether the deviation is caused by the associated error and identify the space-time correlation type and the influence range of the error in combination with the scene data of the prediction error space-time correlation analysis module when the operation data deviates from the scheduling scheme. The feedback adjustment triggering unit is configured to trigger a feedback signal to the robust optimization scheduling module to start a scheduling scheme re-optimization process if the deviation is caused by the associated error and exceeds the tolerance; The robustness index evaluation unit is configured to periodically evaluate the robustness index of the system scheduling scheme based on historical operation data and associated error scenarios, and generate a robustness evaluation report; The index optimization suggestion unit is configured to identify the system robustness short board according to the robustness evaluation report, and provide optimization suggestions for the scenario optimization of the prediction error spatio-temporal correlation analysis module and the target function adjustment of the robust optimization scheduling module; The abnormal alarm unit is configured to generate hierarchical alarm information when the operation state monitoring unit finds that the power grid frequency is abnormal, the energy storage fails or the power imbalance risk caused by the associated error, and push the alarm information to the management personnel in a preset manner; The historical data archiving unit is configured to archive and store the monitored operation data, deviation analysis results, robustness evaluation report and alarm records to form a historical database.

Images