Distributed collaborative optimization scheduling method for virtual power plant

Abstract

The invention relates to the technical field of virtual power plant scheduling, and discloses a distributed collaborative optimization scheduling method for a virtual power plant. The method includes collecting an operating state data set of a target virtual power plant. And performing distributed collaborative model construction processing on the operation state data set to generate collaborative scheduling features covering power distribution balance degree, constraint matching closeness and interactive response sensitivity. And calling a pre-trained optimization scheduling model to carry out multi-target collaborative optimization processing on the collaborative scheduling features to obtain an optimization scheduling result and a key collaborative region identifier. And based on the association relationship between the load demand fluctuation sequence and the equipment adjustment capability, performing operation environment compensation correction processing on the optimization scheduling result, and generating a corrected result. And generating a virtual power plant scheduling strategy set including a power transfer path adjustment scheme and an energy storage equipment configuration processing scheme according to the key cooperative region identifier. According to the method, distributed energy resources are effectively integrated through multi-dimensional collaborative optimization and dynamic correction.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant scheduling technology, specifically a distributed collaborative optimization scheduling method for virtual power plants. Background Technology

[0002] With the continuous increase in the penetration rate of new energy sources, the large-scale integration of distributed power sources, energy storage devices, and diverse loads has profoundly changed the operation mode of the power system. As an important carrier for integrating distributed energy resources, virtual power plants achieve the coordinated management and optimized utilization of distributed energy by aggregating decentralized power generation units, energy storage systems, and controllable loads, becoming a key technical means to improve the flexibility and economy of the power system.

[0003] The current dispatch and operation of virtual power plants face numerous challenges. Distributed power output exhibits significant randomness and volatility, greatly influenced by factors such as weather conditions and geographical environment. For example, fluctuations in wind and solar power output can make it difficult to maintain power balance within the virtual power plant. Simultaneously, load demand is diverse and dynamic, with significant differences in electricity consumption patterns between industrial, commercial, and residential loads, further increasing the difficulty of dispatch.

[0004] Existing scheduling methods mostly employ centralized optimization strategies, relying on a central node to acquire and process global information. This approach suffers from problems such as information transmission delays, high computational complexity, and insufficient privacy protection when dealing with large-scale distributed resources. In a centralized architecture, each distributed unit needs to upload a large amount of operational data to the central node, increasing the communication burden and potentially causing deviations in scheduling command execution due to data transmission interruptions or delays. Furthermore, traditional scheduling models often focus on single-objective optimization, such as minimizing operating costs or network losses, making it difficult to consider multiple requirements such as power balance and device safety constraints, thus limiting the practicality of the scheduling results.

[0005] The insufficient coordination between the charging and discharging strategies of energy storage devices and distributed power sources and loads is also a prominent problem in current dispatching methods. Existing methods underutilize the regulation capabilities of energy storage devices and fail to fully consider the role of energy storage in smoothing power fluctuations and optimizing power allocation, resulting in the inability to fully realize the economic efficiency and reliability of energy storage resources. Furthermore, when the operating environment undergoes drastic changes, such as a sudden increase in load demand or a sharp drop in distributed power output, existing dispatching strategies lack effective compensation and correction mechanisms, failing to respond quickly to environmental changes and potentially leading to safety hazards such as voltage exceeding limits and frequency fluctuations.

[0006] The inadequate coordination mechanisms between different areas within a virtual power plant and the low accuracy in identifying key collaborative areas lead to irrational power transfer path planning and a lack of targeted equipment configuration. The varying resource characteristics and regulation capabilities of different areas mean that the inability to accurately identify key collaborative areas and formulate differentiated dispatch strategies will impact the overall operational efficiency and stability of the virtual power plant. These issues restrict the efficient integration and optimized utilization of distributed energy resources by virtual power plants, making it difficult to meet the requirements of flexible dispatch and safe operation in new power systems. Summary of the Invention

[0007] The purpose of this invention is to provide a distributed collaborative optimization scheduling method for virtual power plants to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides a distributed collaborative optimization scheduling method for virtual power plants, the method comprising:

[0009] Collect a set of operational status data of the target virtual power plant in the operating environment. The set of operational status data includes distributed power output sequence, load demand fluctuation sequence and energy storage device charging and discharging monitoring data.

[0010] The operating status data set is processed by a distributed collaborative model to generate the collaborative scheduling characteristics of the target virtual power plant. The collaborative scheduling characteristics include power allocation balance, constraint matching tightness, and interaction response sensitivity.

[0011] The pre-trained optimization scheduling model is invoked to perform multi-objective collaborative optimization processing on the collaborative scheduling features, generating the optimized scheduling results of the target virtual power plant and key collaborative area identifiers;

[0012] The optimized scheduling result is subjected to operational environment compensation and correction processing to generate a corrected optimized scheduling result. The operational environment compensation and correction processing is based on the correlation between the load demand fluctuation sequence and the equipment adjustment capability.

[0013] A set of virtual power plant scheduling strategies is generated based on the key collaborative area identifier. The set of virtual power plant scheduling strategies includes power transfer path adjustment schemes and energy storage device configuration processing schemes.

[0014] Preferably, the step of performing distributed collaborative model construction processing on the operating status data set to generate the collaborative scheduling characteristics of the target virtual power plant includes:

[0015] The distributed power output sequence is divided into multiple output subsequences according to a time window, and each output subsequence corresponds to a scheduling optimization cycle.

[0016] For each of the aforementioned output subsequences, the following processing is performed:

[0017] The distributed collaborative topology of the target virtual power plant is constructed based on the charging and discharging monitoring data of the energy storage device. The distributed collaborative topology includes the spatial distribution data of the power output field, the load demand field, and the energy storage regulation field.

[0018] The distributed cooperative topology is coupled with the output subsequence for analysis to generate the cooperative model construction result for the current time window. The cooperative model construction result includes the spatial distribution matrix of power allocation components, constraint matching components and interaction response components.

[0019] The results of the collaborative model construction over multiple consecutive time windows are accumulated and superimposed to calculate the power allocation balance, constraint matching tightness, and interaction response sensitivity; wherein,

[0020] The power distribution balance is the maximum deviation rate of the power distribution components between the power source and the load.

[0021] The constraint matching tightness is the integral of the constraint matching component in the direction of the device operating limit.

[0022] The interaction response sensitivity is the ratio of the variance to the mean of the interaction response components within a predetermined time interval.

[0023] Preferably, the step of coupling the distributed cooperative topology structure with the output sub-sequence for analysis and processing to generate the cooperative model construction result for the current time window includes:

[0024] Based on the correspondence between the power output field and the power output components in the power output subsequence, a power output-demand mapping equation is established, and the first distribution function of the power allocation component is obtained by solving the power output-demand mapping equation.

[0025] Based on the correlation characteristics between the load demand field and equipment operation constraints, a constraint matching calculation model is constructed, which includes dynamic correction parameters for equipment adjustment rate and response delay.

[0026] By combining the spatial change rate of the energy storage regulation field with the device interaction frequency, an interactive response iterative calculation process is established, which includes a feedback correction mechanism for power increment and response increment.

[0027] The outputs of the first distribution function, the constraint matching calculation model, and the interactive response iterative calculation process are spatially interpolated and fused to generate distributed collaborative model data containing power allocation, constraint matching, and interactive response components.

[0028] Preferably, the step of invoking a pre-trained optimized scheduling model to perform multi-objective collaborative optimization processing on the collaborative scheduling features, generating optimized scheduling results and key collaborative area identifiers for the target virtual power plant, includes:

[0029] The power distribution balance is input into the first feature analysis layer of the optimized scheduling model, and the distribution coordinates of the uneven power distribution region and the power adjustment amplitude change curve are determined by the power deviation factor calculation module.

[0030] The constraint matching tightness is input into the second feature analysis layer of the optimization scheduling model to perform cumulative calculation of equipment operation constraints, and generate the constraint over-limit probability and adjustment rate prediction value of the equipment operation boundary.

[0031] The interaction response sensitivity is input into the third feature analysis layer of the optimized scheduling model, and the device response delay and interaction frequency evolution data of the interaction response surface are calculated based on the device interaction degradation model.

[0032] By integrating the power adjustment amplitude change curve, the constraint over-limit probability, and the equipment response delay, a comprehensive coordination index for the target virtual power plant is generated. The optimized scheduling result is determined based on the comparison result between the comprehensive coordination index and the preset scheduling threshold.

[0033] Based on the spatial superposition of the distribution coordinates, the predicted adjustment rate, and the interaction frequency evolution data, the geometric locations of regions with uneven power distribution, constraint overrun paths, and high-risk interaction delay regions are identified.

[0034] Preferably, the step of performing runtime environment compensation and correction processing on the optimized scheduling result to generate a corrected optimized scheduling result includes:

[0035] Extract the extreme load values ​​and load change frequencies from the load demand fluctuation sequence, and calculate the dynamic adjustment amount of the equipment regulation capacity as the load changes.

[0036] Based on the dynamic adjustment amount, load fluctuation compensation calculation is performed on the power distribution balance to generate a corrected power distribution balance.

[0037] Based on the correlation between the load change frequency and the equipment response characteristics, the constraint matching tightness is subjected to response delay correction processing to generate the corrected constraint matching tightness.

[0038] Based on the equipment regulation efficiency change data under extreme load, the interactive response sensitivity is adjusted to adapt to equipment performance, and a corrected interactive response sensitivity is generated.

[0039] The corrected power allocation balance, constraint matching tightness, and interaction response sensitivity are input into the optimized scheduling model for recalculation, generating optimized scheduling results after compensating for environmental factors.

[0040] Preferably, the step of calculating the load fluctuation compensation for the power distribution balance based on the dynamic adjustment amount to generate the corrected power distribution balance includes:

[0041] Obtain the initial regulation capacity and dynamic adjustment amount of the target virtual power plant under the baseline load, and establish a regulation capacity-load correlation function;

[0042] The power allocation increment is calculated based on the regulation capacity-load correlation function, whereby the power allocation increment is the product of the load change and the regulation capacity change.

[0043] The power allocation increment is superimposed on the calculation process of the power allocation balance to generate a power allocation balance correction value that includes the impact of load fluctuations;

[0044] The power distribution balance correction value is subjected to response delay effect compensation processing, which is based on the product factor of the equipment response delay curve and the load holding time.

[0045] Preferably, generating a set of virtual power plant scheduling strategies based on the key collaborative area identifier includes:

[0046] For the identified areas of uneven power distribution, the optimal power transfer path is calculated, which is achieved by adjusting the output distribution ratio of adjacent distributed power sources.

[0047] Based on the identifier of the constraint-exceeding path, an energy storage device configuration processing scheme is constructed, which includes the selection of the energy storage device access area and the optimized configuration of energy storage capacity parameters.

[0048] Based on the identification of the high-risk areas of interaction delay, a demand response adjustment strategy is generated. The demand response adjustment strategy dynamically adjusts the adjustment frequency and adjustment range according to the predicted interaction delay value.

[0049] The optimal power transfer path, the energy storage device configuration processing scheme, and the demand response adjustment strategy are prioritized to generate a set of scheduling strategies that include execution timing and implementation parameters.

[0050] Preferably, the energy storage device configuration processing scheme includes:

[0051] Extract the geometric features of the constraint-crossing path, and calculate the path radius of curvature and the crossing direction angle;

[0052] The access density of energy storage devices is selected based on the radius of curvature, and the access density is inversely proportional to the radius of curvature.

[0053] The power output direction of the energy storage device is adjusted based on the over-limit direction angle so that the output direction forms a predetermined angle with the constraint over-limit direction.

[0054] The response time of the energy storage equipment is dynamically adjusted based on the equipment regulation efficiency test data to ensure that the regulation power is below the critical value of the equipment's rated capacity.

[0055] Generate a configuration table containing configuration parameters such as access density, output direction, and response time.

[0056] Preferably, the method further includes:

[0057] The actual power deviation and constraint over-limit length of the target virtual power plant are collected within a preset verification period.

[0058] The actual power deviation and the predicted power adjustment amplitude are analyzed to generate a first error correction coefficient.

[0059] The constraint over-limit length and the predicted adjustment rate are compared in the time domain to generate a second error correction coefficient.

[0060] The weight parameters of the optimized scheduling model are adjusted according to the first error correction coefficient and the second error correction coefficient to generate the optimized scheduling model.

[0061] The optimized scheduling model is then applied to subsequent batches of virtual power plant scheduling optimization tasks.

[0062] Preferably, the set of operational status data of the target virtual power plant under the operating environment includes:

[0063] Outlier detection processing is performed on the distributed power output sequence, and abnormal output points exceeding three times the standard deviation of the historical mean are identified and removed by sliding window statistical method;

[0064] The load demand fluctuation sequence is subjected to trend decomposition processing to separate the periodic fluctuation component and the random fluctuation component.

[0065] The charging and discharging monitoring data of the energy storage device are time-aligned, and the data acquisition timestamps of each device are calibrated based on the synchronous clock.

[0066] Generate a preprocessed runtime status dataset that includes outlier removal, trend decomposition, and time alignment.

[0067] Compared with the prior art, the beneficial effects of the present invention are:

[0068] The distributed collaborative optimization scheduling method for virtual power plants provided by this invention collects distributed power output sequences, load demand fluctuation sequences, and energy storage device charging and discharging monitoring data of the target virtual power plant under its operating environment. This provides comprehensive and accurate basic information for subsequent collaborative scheduling. These data cover key parameters of virtual power plant operation, accurately reflecting the output characteristics of distributed energy sources, dynamic load changes, and the operating status of energy storage devices, thus avoiding scheduling deviations caused by missing or inaccurate information.

[0069] A distributed collaborative model was constructed and processed on the operational status data set to generate collaborative scheduling features such as power allocation balance, constraint matching tightness, and interaction response sensitivity, thereby achieving a quantitative description of the collaborative relationships within the virtual power plant. This feature extraction method breaks through the dependence on single parameters in traditional scheduling, characterizing the collaborative operation status between distributed resources from multiple dimensions. This allows the scheduling model to more comprehensively grasp the operational characteristics of the virtual power plant, laying a more scientific analytical foundation for subsequent optimization.

[0070] By invoking a pre-trained optimization scheduling model to perform multi-objective collaborative optimization on the cooperative scheduling features, the optimized scheduling results and key collaborative region identifiers are generated, overcoming the limitations of traditional single-objective optimization. Multi-objective collaborative optimization can simultaneously consider multiple requirements such as power balance, equipment constraints, and operational economy, ensuring that the scheduling results improve overall operational efficiency while meeting various constraints. The generation of key collaborative region identifiers provides clear direction for subsequent development of targeted scheduling strategies, helping to improve the accuracy and effectiveness of scheduling strategies.

[0071] Based on the correlation between load demand fluctuation sequences and equipment regulation capabilities, the optimized scheduling results are compensated and corrected for the operating environment, enhancing the adaptability of the scheduling results to dynamic operating environments. When load demand fluctuates drastically or the output of distributed power sources changes abruptly, this correction mechanism can quickly adjust scheduling instructions to avoid power imbalances or equipment overloads caused by environmental changes, ensuring the stable operation of the virtual power plant under complex operating conditions.

[0072] Based on the identification of key collaborative regions, a set of scheduling strategies is generated, including power transfer path adjustment schemes and energy storage device configuration processing schemes, enabling refined management of virtual power plant resources. By developing differentiated power transfer paths and energy storage configuration schemes tailored to the resource characteristics and collaborative needs of different regions, the regulation potential of each region can be fully utilized, improving the utilization efficiency of resources within the virtual power plant. Simultaneously, this regional collaborative scheduling strategy reduces unnecessary power transmission losses, lowers equipment operating pressure, and extends equipment lifespan. Attached Figure Description

[0073] Figure 1This is a schematic diagram illustrating the working principle of the virtual power plant distributed collaborative optimization scheduling method described in this invention.

[0074] Figure 2 A flowchart for load fluctuation compensation calculation;

[0075] Figure 3 The flowchart for optimizing the scheduling model verification and correction;

[0076] Figure 4 A flowchart for collecting runtime status data sets. Detailed Implementation

[0077] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] Please see Figures 1-4 The present invention provides a distributed collaborative optimization scheduling method for virtual power plants, the method comprising:

[0079] Collect a set of operational status data of the target virtual power plant in the operating environment. The set of operational status data includes distributed power output sequence, load demand fluctuation sequence and energy storage device charging and discharging monitoring data.

[0080] The operating status data set is processed by a distributed collaborative model to generate the collaborative scheduling characteristics of the target virtual power plant. The collaborative scheduling characteristics include power allocation balance, constraint matching tightness, and interaction response sensitivity.

[0081] The pre-trained optimization scheduling model is invoked to perform multi-objective collaborative optimization processing on the collaborative scheduling features, generating the optimized scheduling results of the target virtual power plant and key collaborative area identifiers;

[0082] The optimized scheduling result is subjected to operational environment compensation and correction processing to generate a corrected optimized scheduling result. The operational environment compensation and correction processing is based on the correlation between the load demand fluctuation sequence and the equipment adjustment capability.

[0083] A set of virtual power plant scheduling strategies is generated based on the key collaborative area identifier. The set of virtual power plant scheduling strategies includes power transfer path adjustment schemes and energy storage device configuration processing schemes.

[0084] Example 1:

[0085] When constructing a distributed collaborative model for the operational status data set to generate the collaborative scheduling characteristics of the target virtual power plant, the distributed power generation output sequence must first be divided into multiple output sub-sequences according to time windows. Each output sub-sequence corresponds to a scheduling optimization cycle. The division of time windows can be flexibly set according to the actual operating rhythm and scheduling requirements of the virtual power plant. For example, based on the periodic characteristics of intraday load fluctuations, the time window can be set to 15 minutes, 30 minutes, or 60 minutes. If the load in the virtual power plant's coverage area exhibits significantly different fluctuation patterns during the day and night, a dynamic time window division method can be adopted. A smaller time window, such as 15 minutes, can be used during the daytime period when load fluctuations are severe to improve scheduling accuracy; a larger time window, such as 60 minutes, can be used during the nighttime period when the load is relatively stable to reduce computational load. After the division is completed, each output sub-sequence will fully record the output changes of distributed power sources within the corresponding scheduling optimization cycle, including the real-time output values ​​and trends of different types of distributed power sources (such as photovoltaic power generation, wind power generation, small gas turbines, etc.).

[0086] For each output subsequence, a series of processing steps are required. A distributed collaborative topology for the target virtual power plant is constructed based on the charging and discharging monitoring data of the energy storage devices. This data includes real-time charging and discharging power, remaining capacity, number of charging and discharging cycles, and health status of each energy storage device. When constructing the distributed collaborative topology, this data must be combined with the geospatial information of the virtual power plant to clarify the spatial distribution of the power output field, load demand field, and energy storage regulation field. The power output field needs to be marked with the specific locations of all distributed power sources, including the installation areas of photovoltaic power plants, the distribution range of wind farms, and the connection points of small generator sets, and record the spatial distribution of parameters such as maximum output, minimum output, and current output of each power source. The load demand field needs to be divided into different load areas, such as residential load areas, commercial load areas, and industrial load areas, each corresponding to different load characteristics, and the load demand and load density of each area at different times must be marked. For energy storage regulation sites, it is necessary to clarify the installation location, capacity, charging and discharging efficiency, and other parameters of each energy storage device, as well as the regulation range that each energy storage device can cover, that is, which power output sites and load demand sites it can provide charging and discharging support for.

[0087] The distributed collaborative topology and output subsequences are coupled and analyzed to generate the collaborative model construction results for the current time window. The coupled analysis requires establishing the correlation between the topology and the output subsequences, analyzing how the power output of the power generation field is transmitted to the load demand field through the lines in the topology within the scheduling cycle corresponding to the current output subsequence, and how the energy storage regulation field plays a regulatory role in this process. During the analysis, the spatial distribution matrices of power allocation components, constraint matching components, and interaction response components need to be calculated. The spatial distribution matrix of the power allocation component describes the specific values ​​of power output allocated to load demand at different spatial locations of the virtual power plant. For example, the power allocation component of a certain region may reflect the proportion of photovoltaic power generation obtained in that region to the total demand, while another region may reflect the allocation of wind power generation. The spatial distribution matrix of the constraint matching component focuses on the degree of matching between the operating constraints of each device and the actual operating state, such as the matching relationship between the actual transmission power of a transmission line and the maximum transmission capacity of that line, or the matching relationship between the charging and discharging power of a certain energy storage device and the charging and discharging limit of that device. The values ​​in the matrix reflect the tightness of this matching. The spatial distribution matrix of the interaction response components is used to describe the interaction response between different devices. For example, when the output of a distributed power source changes, the spatial distribution of the response speed and response amplitude of the corresponding energy storage device or other power source is shown. The values ​​in the matrix reflect the intensity of this interaction response.

[0088] The results of collaborative model construction across multiple consecutive time windows are cumulatively superimposed to calculate power allocation balance, constraint matching tightness, and interaction response sensitivity. This cumulative superimposition process integrates the spatial distribution matrices of multiple adjacent time windows, analyzing the changing trends and cumulative effects of each component over time. Calculating power allocation balance requires comparing the power allocation differences between power sources and loads within different time windows to identify the maximum deviation rate of the power allocation component between power sources and loads. For example, within a continuous time window, when the power output of one region is significantly greater than the load demand, while the power output of another region is significantly less than the load demand, the power allocation deviation rate between these two regions is large, and the maximum value of this deviation rate represents the power allocation balance. Calculating constraint matching tightness requires integrating the constraint matching component along the direction of the equipment's operating limits. The integration interval is multiple consecutive time windows. The integration result reflects the cumulative change of the constraint matching component along the direction of the equipment's operating limits during this period. For example, for a certain energy storage device, the constraint matching component value is higher when its charging and discharging power is close to the maximum limit. Integration yields the cumulative degree to which the device approaches the limit over a continuous time period. The calculation of interactive response sensitivity requires first determining a predetermined time interval, calculating the variance and mean of the interactive response components within that interval, and then dividing the variance by the mean to obtain the ratio. This ratio reflects the degree of fluctuation of the interactive response components during this period relative to the average level. A larger ratio indicates poorer stability of the interactive response, while a smaller ratio indicates more stable interactive response. Through the above cumulative superposition and calculation process, three key indicators that reflect the characteristics of virtual power plant collaborative dispatch are finally obtained.

[0089] Example 2:

[0090] When coupling the distributed collaborative topology with the output subsequences for analysis to generate the collaborative model construction results for the current time window, it is necessary to establish a power output-demand mapping equation based on the correspondence between the power output field and the power output components in the output subsequence. The power output field contains information such as the location, output capacity, and output characteristics of various distributed power sources, while the output subsequence records the actual output changes of these power sources within a specific time window. Establishing the correspondence requires clarifying the spatial affiliation of each power output component; that is, the photovoltaic output at a certain moment corresponds to the spatial region where the photovoltaic power station is located, and the wind power output corresponds to the spatial range of the wind farm. Through this correspondence, the mapping equation can describe how the power output meets the load demand of different regions spatially. The first distribution function obtained after solving this equation can show the law of power distribution component variation with spatial location. For example, in areas with concentrated commercial loads, the power distribution component may show a higher value, while in areas with sparse loads, the value is lower. This distribution law directly reflects the spatial matching state between power output and load demand.

[0091] Based on the correlation between load demand and equipment operating constraints, a constraint matching calculation model is constructed. Changes in load demand directly affect the operating state of equipment. For example, when the load demand in a certain area suddenly increases, the transmission power of the corresponding transmission line will rise, possibly approaching the line's maximum transmission capacity. At this time, the correlation between equipment operating constraints and load demand changes. The dynamic correction parameters for equipment adjustment rate and response delay included in the model need to be set according to this correlation. The equipment adjustment rate parameter reflects the speed at which the equipment adjusts to load changes, such as the speed at which energy storage equipment goes from standby to full-power discharge. The response delay parameter reflects the time difference between receiving an adjustment command and actually executing the adjustment action. Dynamic correction means that these parameters are not fixed values, but will automatically adjust with changes in load demand. For example, when the load fluctuates rapidly, the adjustment rate parameter may increase to speed up the response, while the response delay parameter may be corrected according to the current operating state of the equipment (such as whether it is operating at full load).

[0092] An iterative calculation process for interactive response is established by combining the spatial change rate of the energy storage regulation field and the interaction frequency of the devices. The spatial change rate of the energy storage regulation field describes the spatial variation of the regulation capability of energy storage devices. For example, a region with a large number of energy storage devices may have a stronger regulation capability, while adjacent regions with fewer devices may have a weaker regulation capability; this spatial difference forms the change rate. The interaction frequency of the devices refers to the frequency of information exchange and coordinated actions between different devices (such as between distributed power sources and energy storage devices, or between energy storage devices and loads). During periods of large load fluctuations, the interaction frequency is usually higher. The iterative calculation process requires repeated calculations until the results stabilize. In each iteration, the power increment (the amount of power to be adjusted) is first calculated based on the current spatial change rate and interaction frequency. Then, the response increment (the degree of response required by the device to achieve this power adjustment) is calculated based on the power increment. A feedback correction mechanism plays a role in this process. If the response increment does not perfectly match the power increment, the difference is used as feedback information for the next iteration. Through multiple iterations, the deviation is gradually reduced, making the calculation results more consistent with actual operating conditions.

[0093] The outputs of the first distribution function, the constraint matching calculation model, and the interactive response iterative calculation process are spatially interpolated and fused to generate distributed collaborative model data containing power allocation, constraint matching, and interactive response components. The purpose of spatial interpolation and fusion is to integrate outputs from different sources into a unified model data set. These outputs may differ in spatial coverage or data density; for example, the output of the first distribution function may have denser data points in some areas, while the output of the constraint matching calculation model may be more detailed in others. Spatial interpolation fills in data gaps, ensuring the continuity and completeness of the model data across the entire virtual power plant. The weights of each output must be considered during interpolation. For instance, the first distribution function may have a higher weight near the power output field, while the constraint matching calculation model may have a greater weight in areas with concentrated equipment. The fused distributed collaborative model data comprehensively reflects the overall power allocation, equipment constraint matching, and equipment interactive response of the virtual power plant within the current time window, providing a foundation for subsequent collaborative scheduling feature calculations.

[0094] Example 3:

[0095] When the pre-trained optimization scheduling model is invoked to perform multi-objective collaborative optimization processing on the collaborative scheduling features, and the optimized scheduling results and key collaborative area identifiers for the target virtual power plant are generated, the power distribution balance is input into the first feature analysis layer of the optimization scheduling model. The power deviation factor calculation module determines the distribution coordinates of the uneven power distribution areas and the power adjustment amplitude variation curve. The power deviation factor calculation module analyzes the power distribution balance data point by point, and filters out areas where the power distribution differs significantly from the expected balance state based on the internally set deviation judgment criteria. The distribution coordinates of these areas are marked using the spatial coordinate system of the virtual power plant, accurate to the specific geographical partition or equipment cluster range. The power adjustment amplitude variation curve calculates the power adjustment amplitude that needs to be adjusted at each time node within the scheduling cycle based on the power gap or surplus of these areas. The fluctuations of the curve reflect the changes in the adjustment intensity at different times. For example, the adjustment amplitude may show an upward trend during peak load periods, while it is relatively flat during stable load periods.

[0096] The constraint matching tightness is input into the second feature analysis layer of the optimized scheduling model to perform cumulative calculation of equipment operation constraints, generating the constraint overrun probability and regulation rate prediction value at the equipment operation boundary. The cumulative calculation of equipment operation constraints continuously tracks and accumulates the constraint matching tightness value at the equipment operation boundary, statistically analyzing the ratio of the number of times the equipment operation state approaches or exceeds the constraint boundary to the total duration over a past period, thus obtaining the constraint overrun probability. The regulation rate prediction value is based on the changing trend of constraint matching tightness, combined with the equipment's historical regulation data, to infer the regulation speed that the equipment can achieve in future scheduling cycles. For example, if the constraint matching tightness of a transmission line continues to increase, its regulation rate prediction value may decrease accordingly, reflecting the decline in the line's regulation capacity when approaching full load.

[0097] The interaction response sensitivity is input into the third feature analysis layer of the optimized scheduling model. Based on the device interaction degradation model, the device response delay and interaction frequency evolution data of the interaction response surface are calculated. The device interaction degradation model considers the natural performance degradation of devices during long-term operation, as well as the impact of frequent interactions on device response speed. By analyzing the fluctuations in interaction response sensitivity, the time required for the device to complete its response after receiving an adjustment command is calculated, i.e., the device response delay. The interaction frequency evolution data records the changing trend of the number of interactions between devices per unit time. For example, during periods of large fluctuations in distributed power output, the interaction frequency may show periodic peaks, while remaining at a low level during periods of stable output.

[0098] By integrating the power adjustment amplitude variation curve, constraint over-limit probability, and equipment response delay, a comprehensive coordination index for the target virtual power plant is generated. The optimized scheduling result is determined based on the comparison between the comprehensive coordination index and a preset scheduling threshold. The calculation of the comprehensive coordination index integrates the three parameters according to certain rules. For example, different weights are assigned to the peak value of the power adjustment amplitude variation curve, the magnitude of the constraint over-limit probability, and the duration of the equipment response delay, and then a comprehensive value is obtained through weighted summation. The preset scheduling threshold is a critical value set based on the virtual power plant's safety operation standards and economic operation objectives. When the comprehensive coordination index is lower than this threshold, the corresponding scheduling scheme is the optimized scheduling result.

[0099] Based on the spatial overlay results of distribution coordinates, predicted regulation rates, and interaction frequency evolution data, the geometric locations of regions with uneven power distribution, constraint-overriding paths, and high-risk interaction delay regions are identified. Spatial overlay processing maps these data to the same spatial coordinate system, clearly displaying the specific range and interrelationships of each key collaborative region through layer overlay. For example, regions with uneven power distribution may partially overlap with high-risk interaction delay regions; these are distinguished using different geometric shapes and colors to intuitively identify regions requiring priority processing.

[0100] When generating a virtual power plant dispatch strategy set based on key collaborative area identifiers, the optimal power transfer path is calculated for the identifiers of areas with uneven power distribution. This optimal power transfer path is achieved by adjusting the output allocation ratio of adjacent distributed power sources. During the calculation, the current output of distributed power sources surrounding the unevenly distributed power distribution area, the carrying capacity of transmission lines, and the distribution of load demand are analyzed to find a path that minimizes transmission losses while meeting load demands. The output allocation ratio of adjacent distributed power sources is adjusted; for example, increasing the output of a photovoltaic power station while reducing the output of a nearby wind farm, so that power can flow along the optimal path to the power-deficient area.

[0101] Based on the identification of constraint-exceeding paths, a configuration and processing scheme for energy storage devices is constructed. This scheme includes the selection of energy storage device access areas and the optimized configuration of energy storage capacity parameters. The geometric features of the constraint-exceeding paths are extracted, and the path radius of curvature and the excess direction angle are calculated. The path radius of curvature reflects the curvature of the constraint-exceeding path; a smaller radius of curvature indicates a sharper turn and greater obstruction to power transmission. The excess direction angle indicates the direction of power transmission when an excess occurs. The access density of energy storage devices is selected based on the radius of curvature. Access density is inversely proportional to the radius of curvature; that is, more energy storage devices are deployed on sections with smaller radii of curvature to enhance local regulation capabilities. The power output direction of the energy storage devices is adjusted based on the excess direction angle, so that the output direction forms a predetermined angle with the constraint-exceeding direction. For example, when the excess direction is clockwise, the output direction of the energy storage device may be set counterclockwise to offset some of the excess power. The response time of the energy storage devices is dynamically adjusted based on the device regulation efficiency test data to ensure that the regulated power is below the device's rated capacity critical value. A configuration table containing configuration parameters such as access density, output direction, and response time is generated to provide specific basis for the installation and commissioning of energy storage devices.

[0102] Based on the identification of high-risk areas for interaction delay, a demand response adjustment strategy is generated. This strategy dynamically adjusts the adjustment frequency and magnitude according to the predicted interaction delay value. When the predicted interaction delay value is large, the adjustment frequency is reduced to decrease the number of device interactions, while the magnitude of each adjustment is increased to avoid frequent interactions exacerbating the delay. When the predicted interaction delay value is small, the adjustment frequency is appropriately increased and the adjustment magnitude is reduced to achieve more precise power regulation.

[0103] The optimal power transfer path, energy storage device configuration scheme, and demand response adjustment strategy are prioritized to generate a set of scheduling strategies, including execution timing and implementation parameters. Prioritization considers the implementation difficulty, impact on the stability of the virtual power plant, and post-implementation effects of each strategy. For example, energy storage device configuration schemes that can quickly alleviate constraint exceedances may be given higher priority and executed first. The execution timing specifies the start and end times of each strategy, while the implementation parameters include specific operational values ​​during strategy execution, such as the specific power transfer values ​​and the charging and discharging power of the energy storage devices.

[0104] Example 4:

[0105] To generate corrected optimized scheduling results by performing operational environment compensation processing, it is necessary to first extract extreme load values ​​and load change frequencies from the load demand fluctuation sequence, and then calculate the dynamic adjustment amount of equipment regulation capacity as the load changes. Extreme load values ​​refer to the maximum and minimum values ​​in the load demand fluctuation sequence that exceed the normal load range; these values ​​often have a significant impact on the equipment's regulation capacity. Load change frequency is the number of times load demand changes significantly per unit time. For example, in a day, a commercial load area may experience two sudden load changes during the morning opening and afternoon closing periods, leading to an increased load change frequency. The calculation of the dynamic adjustment amount of equipment regulation capacity needs to consider the equipment's technical parameters, such as maximum regulation power and regulation accuracy, as well as the difference between extreme load values ​​and normal load values, and the level of load change frequency, to determine the deviation between the actual achievable regulation capacity and the rated regulation capacity under different load conditions.

[0106] The power distribution balance is calculated based on load fluctuation compensation using dynamic adjustment, generating a corrected power distribution balance. The initial regulation capacity and dynamic adjustment of the target virtual power plant under the baseline load are obtained, and a regulation capacity-load correlation function is established. The baseline load is typically set as the average load value of the virtual power plant on a typical operating day, and the initial regulation capacity is the stable output power of the equipment under this baseline load. The regulation capacity-load correlation function reflects the actual regulation capacity of the equipment corresponding to different load values. For example, when the load is higher than the baseline load, the regulation capacity output by the function may decrease as the load increases; when the load is lower than the baseline load, the regulation capacity may show a different trend. The power distribution increment is calculated based on this function, and its calculation formula is as follows:

[0107] ΔP=ΔL×ΔC

[0108] Where ΔP represents the power allocation increment, ΔL represents the load change (i.e., the difference between the current load value and the baseline load value), and ΔC represents the change in regulation capacity (i.e., the dynamic adjustment). Adding the power allocation increment to the calculation of power allocation balance corrects the original balance value, which did not consider the impact of load fluctuations, generating a corrected power allocation balance value that includes the impact of load fluctuations. Then, this corrected value undergoes response delay compensation. The equipment response delay curve describes the time change from receiving a regulation command to the actual regulation effect under different load levels. Load holding time refers to the duration of a certain extreme load value. The product factor of these two factors quantifies the cumulative impact of response delay on power allocation. Including this factor in the compensation calculation makes the corrected power allocation balance more closely reflect the regulation lag in actual equipment operation.

[0109] Based on the correlation between load change frequency and equipment response characteristics, a response delay correction process is applied to the constraint matching tightness to generate a corrected constraint matching tightness. Equipment response characteristics include response speed and settling time. Different load change frequencies cause these characteristics to change. For example, when the load change frequency is too high, the equipment may experience a decrease in response speed and an increase in settling time due to frequent adjustments. Analyzing this correlation requires collecting equipment operating data under different load change frequencies to clarify the changing patterns of response characteristics. The correction process involves finding the corresponding change value of response characteristics based on the current load change frequency and adjusting the constraint matching tightness accordingly. For example, when the load change frequency exceeds a certain threshold, the decrease in equipment response speed will cause the actual constraint matching tightness to be lower than the calculated value, in which case the constraint matching tightness needs to be corrected downwards.

[0110] Based on data on changes in equipment regulation efficiency under extreme loads, the interaction response sensitivity is adjusted to adapt to equipment performance, resulting in a corrected interaction response sensitivity. Extreme loads may cause equipment to operate under off-design conditions, altering regulation efficiency. For example, when a battery discharges under extreme high loads, its energy conversion efficiency decreases, affecting its interaction response speed with distributed power sources. Data on changes in equipment regulation efficiency is obtained through long-term monitoring of equipment operating parameters under extreme load conditions, including the rate of efficiency decline and recovery time. During the adjustment process, the calculation parameters for the interaction response sensitivity are corrected based on this data. For instance, when extreme loads cause a 30% decrease in regulation efficiency, the calculation coefficient for the interaction response sensitivity is multiplied by 0.7, ensuring that the corrected interaction response sensitivity reflects the actual interaction capability of the equipment under extreme loads.

[0111] The revised power allocation balance, constraint matching tightness, and interaction response sensitivity are input into the optimized scheduling model for recalculation. Based on these updated collaborative scheduling characteristics, the model re-performs multi-objective collaborative optimization, ultimately generating optimized scheduling results after compensating for environmental factors. This result comprehensively considers environmental factors such as load fluctuations, equipment response delays, and performance changes under extreme loads, enabling more accurate guidance for the actual scheduling and operation of virtual power plants.

[0112] Example 5:

[0113] Within a preset verification period, the actual power deviation and constraint over-limit length of the target virtual power plant are collected. The preset verification period needs to be set in conjunction with the virtual power plant's operating cycle and dispatch frequency. For example, for a virtual power plant that is dispatched on a daily basis, the verification period can be set to one week to cover the operation on different working days and rest days; for scenarios with higher dispatch frequency, the verification period can be shortened to three days or less. Within the verification period, monitoring equipment deployed at each key node of the virtual power plant records the actual power deviation data in real time, that is, the difference between the actual power output of each area and the predicted power output in the optimized dispatch results. At the same time, the constraint over-limit length is recorded, that is, the duration for which transmission lines, energy storage devices, etc. exceed their rated operating limits during operation, as well as the specific time period when the over-limit occurs.

[0114] The actual power deviation and the predicted power adjustment amplitude are analyzed to generate the first error correction coefficient. This deviation analysis requires comparing the actual power deviation and the predicted power adjustment amplitude time-by-time within the validation period, calculating the absolute and relative differences for each time period, and statistically analyzing the distribution of these differences, such as the frequency of differences clustering within a certain range and the conditions under which the maximum difference occurs. Based on these analysis results, the value of the first error correction coefficient is determined. When the actual power deviation is generally greater than the predicted power adjustment amplitude, the coefficient may be greater than 1 to enhance the model's predictive power of the power adjustment amplitude; when the actual deviation is less than the predicted value, the coefficient may be less than 1 to avoid over-adjustment.

[0115] The constraint over-limit length and the predicted adjustment rate are compared in the time domain to generate a second error correction coefficient. This time-domain comparison aligns the constraint over-limit length and the predicted adjustment rate on the time axis, analyzing the degree of match between the predicted adjustment rate and the actual required adjustment rate during the over-limit period. For example, if the constraint over-limit length is long in a certain period but the predicted adjustment rate is high, it indicates that the model's assessment of the adjustment capability for that period is overly optimistic, and the corresponding prediction parameters need to be reduced. Based on this time-domain deviation, the second error correction coefficient is determined. Its value must reflect the magnitude of the deviation between the predicted adjustment rate and the actual over-limit situation, and it is used to correct the calculation modules related to the adjustment rate in the model.

[0116] The weight parameters of the optimized scheduling model are adjusted based on the first and second error correction coefficients to generate the optimized scheduling model. The optimized scheduling model includes multiple weight parameters used to balance different objectives. For example, in the calculation of power distribution balance and constraint matching tightness, their respective weights determine the model's emphasis on each. During the adjustment process, the first error correction coefficient is applied to the weight parameters related to power adjustment, enabling the model to more accurately reflect the actual power deviation in subsequent calculations; the second error correction coefficient is applied to the weight parameters related to the regulation rate, optimizing the model's ability to predict constraint exceedances. Through this adjustment, the model's output better reflects the actual operating state of the virtual power plant.

[0117] The optimized scheduling model is applied to subsequent batches of virtual power plant scheduling optimization tasks, enabling the updated model to play a role in the new scheduling cycle and continuously improving the adaptability and accuracy of scheduling optimization.

[0118] When collecting operational status data of the target virtual power plant under operating conditions, outlier detection processing is performed on the distributed power generation output sequence. Anomalies exceeding three times the standard deviation of the historical mean are identified and removed using a sliding window statistical method. The sliding window method requires setting an appropriate window size, such as one hour, to segment the distributed power generation output sequence. The historical mean and standard deviation within each window are calculated. Each output data point within the window is compared with the mean. If the difference between a data point and the mean exceeds three times the standard deviation, it is identified as an outlier output point and removed. Examples include sudden high power values ​​from photovoltaic power generation at night or abnormal output values ​​from wind power generation during windless periods.

[0119] Trend decomposition is performed on the load demand fluctuation sequence to separate periodic and stochastic fluctuation components. A time series decomposition algorithm is used to break down the load demand fluctuation sequence into multiple components. The periodic fluctuation component reflects the regular changes in load with a fixed cycle (such as daily or weekly cycles), for example, the peak fluctuations in residential load during the morning, noon, and evening. The stochastic fluctuation component includes irregular load changes caused by sudden factors (such as sudden weather changes or large-scale events). This decomposition allows for a clearer understanding of the different causes of load fluctuations.

[0120] Time alignment processing is performed on the charging and discharging monitoring data of energy storage devices, calibrating the data acquisition timestamps of each device based on a synchronized clock. Energy storage devices within a virtual power plant may come from different manufacturers, and the clocks of their data acquisition systems may deviate, causing monitoring data at the same moment to be marked with different times. Time alignment processing uses a unified synchronized clock signal, such as a GPS clock, to calibrate the acquisition timestamps of each device, ensuring that the charging and discharging data of all energy storage devices at the same point in time accurately correspond. For example, calibrating the 10:00:05 data of one energy storage device to 10:00:00 consistent with other devices ensures data consistency across the time dimension.

[0121] A preprocessed operational status dataset is generated, which includes outlier removal, trend decomposition, and time alignment. This dataset integrates processed distributed power output sequences, load demand fluctuation sequences, and energy storage device charging and discharging monitoring data, providing reliable data input for subsequent distributed collaborative model construction and optimized scheduling, and ensuring the smooth implementation of the entire scheduling method.

[0122] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0123] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A distributed collaborative optimization scheduling method for virtual power plants, characterized in that, The method includes: Collect a set of operational status data of the target virtual power plant in the operating environment. The set of operational status data includes distributed power output sequence, load demand fluctuation sequence and energy storage device charging and discharging monitoring data. The operating status data set is processed by a distributed collaborative model to generate the collaborative scheduling characteristics of the target virtual power plant. The collaborative scheduling characteristics include power allocation balance, constraint matching tightness, and interaction response sensitivity. The pre-trained optimization scheduling model is invoked to perform multi-objective collaborative optimization processing on the collaborative scheduling features, generating the optimized scheduling results of the target virtual power plant and key collaborative area identifiers; The optimized scheduling result is subjected to operational environment compensation and correction processing to generate a corrected optimized scheduling result. The operational environment compensation and correction processing is based on the correlation between the load demand fluctuation sequence and the equipment adjustment capability. A set of virtual power plant scheduling strategies is generated based on the key collaborative area identifier. The set of virtual power plant scheduling strategies includes power transfer path adjustment schemes and energy storage device configuration processing schemes.

2. The distributed collaborative optimization scheduling method for virtual power plants according to claim 1, characterized in that, The process of constructing a distributed collaborative model for the operational status data set to generate the collaborative scheduling characteristics of the target virtual power plant includes: The distributed power output sequence is divided into multiple output subsequences according to a time window, and each output subsequence corresponds to a scheduling optimization cycle. For each of the aforementioned output subsequences, the following processing is performed: The distributed collaborative topology of the target virtual power plant is constructed based on the charging and discharging monitoring data of the energy storage device. The distributed collaborative topology includes the spatial distribution data of the power output field, the load demand field, and the energy storage regulation field. The distributed cooperative topology is coupled with the output subsequence for analysis to generate the cooperative model construction result for the current time window. The cooperative model construction result includes the spatial distribution matrix of power allocation components, constraint matching components and interaction response components. The results of the collaborative model construction over multiple consecutive time windows are accumulated and superimposed to calculate the power allocation balance, constraint matching tightness, and interaction response sensitivity; wherein, The power distribution balance is the maximum deviation rate of the power distribution components between the power source and the load. The constraint matching tightness is the integral of the constraint matching component in the direction of the device operating limit. The interaction response sensitivity is the ratio of the variance to the mean of the interaction response components within a predetermined time interval.

3. The distributed collaborative optimization scheduling method for virtual power plants according to claim 2, characterized in that, The step of coupling the distributed cooperative topology with the output subsequence for analysis and processing to generate the cooperative model construction result for the current time window includes: Based on the correspondence between the power output field and the power output components in the power output subsequence, a power output-demand mapping equation is established, and the first distribution function of the power allocation component is obtained by solving the power output-demand mapping equation. Based on the correlation characteristics between the load demand field and equipment operation constraints, a constraint matching calculation model is constructed, which includes dynamic correction parameters for equipment adjustment rate and response delay. By combining the spatial change rate of the energy storage regulation field with the device interaction frequency, an interactive response iterative calculation process is established, which includes a feedback correction mechanism for power increment and response increment. The outputs of the first distribution function, the constraint matching calculation model, and the interactive response iterative calculation process are spatially interpolated and fused to generate distributed collaborative model data containing power allocation, constraint matching, and interactive response components.

4. The distributed collaborative optimization scheduling method for virtual power plants according to claim 1, characterized in that, The pre-trained optimization scheduling model is invoked to perform multi-objective collaborative optimization processing on the collaborative scheduling features, generating the optimized scheduling results and key collaborative area identifiers for the target virtual power plant, including: The power distribution balance is input into the first feature analysis layer of the optimized scheduling model, and the distribution coordinates of the uneven power distribution region and the power adjustment amplitude change curve are determined by the power deviation factor calculation module. The constraint matching tightness is input into the second feature analysis layer of the optimization scheduling model to perform cumulative calculation of equipment operation constraints, and generate the constraint over-limit probability and adjustment rate prediction value of the equipment operation boundary. The interaction response sensitivity is input into the third feature analysis layer of the optimized scheduling model, and the device response delay and interaction frequency evolution data of the interaction response surface are calculated based on the device interaction degradation model. By integrating the power adjustment amplitude change curve, the constraint over-limit probability, and the equipment response delay, a comprehensive coordination index for the target virtual power plant is generated. The optimized scheduling result is determined based on the comparison result between the comprehensive coordination index and the preset scheduling threshold. Based on the spatial superposition of the distribution coordinates, the predicted adjustment rate, and the interaction frequency evolution data, the geometric locations of regions with uneven power distribution, constraint overrun paths, and high-risk interaction delay regions are identified.

5. The distributed collaborative optimization scheduling method for virtual power plants according to claim 1, characterized in that, The step of performing runtime environment compensation and correction processing on the optimized scheduling result to generate a corrected optimized scheduling result includes: Extract the extreme load values ​​and load change frequencies from the load demand fluctuation sequence, and calculate the dynamic adjustment amount of the equipment regulation capacity as the load changes. Based on the dynamic adjustment amount, load fluctuation compensation calculation is performed on the power distribution balance to generate a corrected power distribution balance. Based on the correlation between the load change frequency and the equipment response characteristics, the constraint matching tightness is subjected to response delay correction processing to generate the corrected constraint matching tightness. Based on the equipment regulation efficiency change data under extreme load, the interactive response sensitivity is adjusted to adapt to equipment performance, and a corrected interactive response sensitivity is generated. The corrected power allocation balance, constraint matching tightness, and interaction response sensitivity are input into the optimized scheduling model for recalculation, generating optimized scheduling results after compensating for environmental factors.

6. The distributed collaborative optimization scheduling method for virtual power plants according to claim 5, characterized in that, The step of calculating load fluctuation compensation for the power distribution balance based on the dynamic adjustment amount to generate a corrected power distribution balance includes: Obtain the initial regulation capacity and dynamic adjustment amount of the target virtual power plant under the baseline load, and establish a regulation capacity-load correlation function; The power allocation increment is calculated based on the regulation capacity-load correlation function, whereby the power allocation increment is the product of the load change and the regulation capacity change. The power allocation increment is superimposed on the calculation process of the power allocation balance to generate a power allocation balance correction value that includes the impact of load fluctuations; The power distribution balance correction value is subjected to response delay effect compensation processing, which is based on the product factor of the equipment response delay curve and the load holding time.

7. The distributed collaborative optimization scheduling method for virtual power plants according to claim 4, characterized in that, The step of generating a set of virtual power plant scheduling strategies based on the key collaborative area identifier includes: For the identified areas of uneven power distribution, the optimal power transfer path is calculated, which is achieved by adjusting the output distribution ratio of adjacent distributed power sources. Based on the identifier of the constraint-exceeding path, an energy storage device configuration processing scheme is constructed, which includes the selection of the energy storage device access area and the optimized configuration of energy storage capacity parameters. Based on the identification of the high-risk areas of interaction delay, a demand response adjustment strategy is generated. The demand response adjustment strategy dynamically adjusts the adjustment frequency and adjustment range according to the predicted interaction delay value. The optimal power transfer path, the energy storage device configuration processing scheme, and the demand response adjustment strategy are prioritized to generate a set of scheduling strategies that include execution timing and implementation parameters.

8. The distributed collaborative optimization scheduling method for virtual power plants according to claim 7, characterized in that, The energy storage device configuration and processing scheme includes: Extract the geometric features of the constraint-crossing path, and calculate the path radius of curvature and the crossing direction angle; The access density of energy storage devices is selected based on the radius of curvature, and the access density is inversely proportional to the radius of curvature. The power output direction of the energy storage device is adjusted based on the over-limit direction angle so that the output direction forms a predetermined angle with the constraint over-limit direction. The response time of the energy storage equipment is dynamically adjusted based on the equipment regulation efficiency test data to ensure that the regulation power is below the critical value of the equipment's rated capacity. Generate a configuration table containing configuration parameters such as access density, output direction, and response time.

9. The distributed collaborative optimization scheduling method for virtual power plants according to claim 1, characterized in that, The method further includes: The actual power deviation and constraint over-limit length of the target virtual power plant are collected within a preset verification period. The actual power deviation and the predicted power adjustment amplitude are analyzed to generate a first error correction coefficient. The constraint over-limit length and the predicted adjustment rate are compared in the time domain to generate a second error correction coefficient. The weight parameters of the optimized scheduling model are adjusted according to the first error correction coefficient and the second error correction coefficient to generate the optimized scheduling model. The optimized scheduling model is then applied to subsequent batches of virtual power plant scheduling optimization tasks.

10. The distributed collaborative optimization scheduling method for virtual power plants according to claim 1, characterized in that, The collection of operational status data of the target virtual power plant under the operating environment includes: Outlier detection processing is performed on the distributed power output sequence, and abnormal output points exceeding three times the standard deviation of the historical mean are identified and removed by sliding window statistical method; The load demand fluctuation sequence is subjected to trend decomposition processing to separate the periodic fluctuation component and the random fluctuation component. The charging and discharging monitoring data of the energy storage device are time-aligned, and the data acquisition timestamps of each device are calibrated based on the synchronous clock. Generate a preprocessed runtime status dataset that includes outlier removal, trend decomposition, and time alignment.

Images