Resource group comprehensive optimization method and system of virtual power plant

Abstract

The invention discloses a resource group comprehensive optimization method and system of a virtual power plant, and relates to the technical field of virtual power plants. The resource group comprehensive optimization method and system for the virtual power plant comprises the following steps: preprocessing operation monitoring data and historical operation statistical data; generating a bottleneck suppression allocation mark, generating a single equipment list and outputting a boundary according to a bottleneck amplification distortion judgment result; boundary contraction is carried out according to equivalent constraint contraction strength analysis, and an apportionment suppression constraint injection and output plan is solved; and performing rolling correction triggering, monomer allocation instruction set reallocation, resource group reconstruction triggering and resource replacement according to a coupling amplification pushing correction evaluation result. The problem that in a heterogeneous distributed resource group, temperature, health degree or gradeability and other tiny performance degradation of single equipment can be coupled and amplified in the group, so that unified output and gradient hypothesis of an aggregation model are distorted, and a virtual power plant scheduling plan cannot be met at an actual execution end is solved.

Description

Technical Field

[0001] This invention relates to the field of virtual power plant technology, specifically to a method and system for comprehensive optimization of resource groups in a virtual power plant. Background Technology

[0002] Virtual power plants, as a new organizational form that participates in the electricity market and grid dispatch in an aggregated manner, typically integrate heterogeneous resources such as distributed photovoltaic, wind power, energy storage, electric vehicle charging and discharging, and controllable loads through a communication and control platform. This unified command issuance aims to achieve goals such as aggregated output planning, meeting power change rate constraints, maintaining power quality at grid connection points, and participating in ancillary services. In current engineering practices, virtual power plants typically achieve online access to distributed resources through communication networks and edge acquisition devices. Based on the platform, they uniformly manage information such as resource status, power capacity, response speed, and available time periods, and integrate the aggregated adjustable capacity as a whole to participate in grid-side planning submissions, dispatch command execution, and ancillary service responses. Simultaneously, virtual power plants need to adapt to grid operation constraints and electricity market rules, forming a closed-loop operating system from resource access, capacity assessment, planning to execution tracking. This improves the observability, controllability, and tradability of distributed resources, supporting the safe and stable operation of the grid and the transformation of the energy structure.

[0003] For example, the invention patent with announcement number CN120474111B discloses a method for optimizing the resource combination of a virtual power plant. This method includes dividing the energy transmission line buried in underground space into equal sections and dividing the line sections according to four orientations: facing the ground, left side wall, underground, and right side wall. Based on historical detection feature data, the correlation coefficient between ambient temperature and line energy consumption is statistically analyzed, and the abnormal ambient temperature value is predicted by combining the relationship between the current temperature and the section height, thus obtaining the predicted value of temperature loss energy consumption. At the same time, image detection is performed on the line section set to extract cracked sections, and the crack length and crack direction are statistically analyzed. The predicted energy consumption under the influence of crack length and crack inclination is predicted respectively. Then, the predicted value of temperature loss energy consumption and the predicted value of crack-related energy consumption are integrated to schedule the original energy allocation and output the actual energy allocation.

[0004] For example, the invention patent with publication number CN118659468A discloses a virtual power plant resource scheduling optimization method based on edge computing. The edge server obtains the original data from the user end and completes cleaning, repair and standardization preprocessing. It extracts load characteristics, power generation characteristics and energy storage system characteristics to construct a demand prediction model to obtain the predicted value of power resource demand. Based on the predicted value of power resource demand and equipment constraints, a power resource scheduling optimization model is established with the goal of minimizing the operating cost of the virtual power plant. The scheduling strategy is obtained by solving the problem using the alternating direction multiplier method. The scheduling tasks are prioritized and the scheduling strategy is adjusted accordingly to form the optimal scheduling strategy. Finally, the optimal scheduling strategy is uploaded to the cloud control center platform for execution to realize power resource scheduling.

[0005] Existing virtual power plants, when aggregating and scheduling resource groups, often use uniform output capacity and gradient constraints as the modeling basis. This makes it difficult to reflect subtle performance degradations in heterogeneous individual devices during actual operation, such as temperature changes, decreased health, or reduced ramp-up capabilities. Because these minute deviations are amplified by intra-group coupling, the equivalent output capacity and regulation capability of the resource group gradually deviate from the model assumptions. This leads to problems such as unreachability, accumulated deviations, and boundary violations at the execution end of the scheduling plan. Existing methods lack real-time quantification and dynamic adjustment mechanisms for such bottleneck changes, making it difficult to maintain the consistency and controllability of resource group scheduling.

[0006] Therefore, in order to address the above problems, there is an urgent need for a comprehensive optimization method and system for virtual power plant resource groups. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a comprehensive optimization method and system for resource groups in virtual power plants. It solves the problem that in heterogeneous distributed resource groups, minor performance degradation of individual devices, such as temperature, health, or ramp-up capability, can be amplified by intra-group coupling, causing distortion of the unified output and gradient assumptions of the aggregation model, resulting in the virtual power plant scheduling plan failing to meet actual execution requirements.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a comprehensive optimization method for resource groups in a virtual power plant, comprising: S1, collecting operation monitoring data and obtaining historical operation statistics; preprocessing the operation monitoring data and historical operation statistics; S2, determining bottleneck amplification distortion in the operation monitoring data and historical operation statistics, and generating bottleneck suppression allocation markers, generating individual equipment lists, and outputting boundary values ​​based on the bottleneck amplification distortion determination results; S3, performing equivalent constraint contraction strength analysis based on the bottleneck amplification distortion determination results, and performing boundary contraction, allocation suppression constraint injection, and output plan solution based on the equivalent constraint contraction strength analysis; S4, performing coupled amplification push correction evaluation on the operation monitoring data, and triggering rolling correction, reallocating individual allocation instruction sets, triggering resource group reconstruction, and replacing resources based on the coupled amplification push correction evaluation results.

[0009] Further, the specific process of collecting operational monitoring data and obtaining historical operational statistics is as follows: Collect operational monitoring data, including: energy storage battery temperature data, energy storage battery voltage data, energy storage battery current data, allowable power change rate, grid connection point voltage data, grid connection point current data, the total number of individual devices within the resource group, and the step sliding window; Obtain historical operational statistics and create a historical operational statistics database, including: historical energy storage battery temperature data, historical energy storage battery voltage data, historical energy storage battery current data, historical active power data, and historical allowable power; The allowable power change rate is calculated according to the executable power change rate in the grid connection specifications; the step sliding window is determined according to the segmented time window rules constructed based on the sampling period; and the historical allowable power is statistically obtained based on the allowable power change rate in the grid connection specifications within the historical operational interval.

[0010] Furthermore, the specific process for preprocessing the operation monitoring data and historical operation statistics data is as follows: A piecewise linear trend decomposition algorithm with first-order difference constraints is used to extract the time-varying baseline and isolate abnormal abrupt changes in the allowable power change rate; a phase-synchronized power frequency cycle alignment and robust RMS value estimation algorithm is used to perform cycle alignment and robust statistics on the grid connection point voltage and current data; a timestamp-consistent multi-source sequence resampling and missing segment hold-up interpolation algorithm is used to perform unified sampling interval alignment and short-term missing segment completion on the operation monitoring data and historical operation statistics data; and a distribution standardization and linear normalization algorithm is used to standardize and normalize the operation monitoring data and historical operation statistics data.

[0011] Furthermore, the specific process for determining bottleneck amplification distortion in operation monitoring data and historical operation statistics is as follows: Obtain operation monitoring data and historical operation statistics; for energy storage battery temperature data, use the sliding window median and MAD robust normalization algorithm to obtain the temperature deviation intensity; establish a second-order equivalent circuit model for energy storage battery voltage data, energy storage battery current data, and energy storage battery temperature data, and use the extended Kalman filter online state estimation algorithm and the sliding window quantile benchmark deviation to obtain the health state deviation intensity; for energy storage battery voltage data and energy storage battery current data, use the recursive least squares parameter identification algorithm with forgetting factor to estimate the equivalent internal resistance sequence, and use the sliding window median and MAD robust normalization algorithm to obtain the equivalent internal resistance deviation intensity; for real-time active power data, use the sliding window first-order polynomial least squares slope estimation algorithm to obtain the actual power change rate value, and after subtracting it from the allowable power change rate boundary, use the quantile scale normalization method to obtain the power change rate deviation intensity; use the Euclidean norm of the above four deviation intensities as the bottleneck change amount of a single device; and calculate the voltage and current data at the grid connection point using power calculations. Real-time active power data is used, and a cross-spectral coherence function estimation algorithm is employed to construct the individual coherence matrix. The intra-group coupling amplification intensity is obtained through spectral radius normalization. Historical operational statistics are used to calculate the historical bottleneck changes of individual devices. Tail samples are extracted using the peak-to-threshold method, and the tail extremum is obtained as the historical maximum bottleneck change value through maximum likelihood estimation. The complement of the intra-group coupling amplification intensity is calculated and its reciprocal is taken to obtain the coupling amplification reciprocal factor. For each individual device, the natural logarithm of the bottleneck change is calculated to obtain the individual bottleneck logarithmic amplitude. The calculation of the single... The absolute value of the ratio of the logarithmic amplitude of the bottleneck to the maximum historical change value is used to obtain the normalized amplitude of the individual bottleneck. The bottleneck sensitivity exponent is raised to the power of the normalized amplitude of the individual bottleneck to obtain the individual sensitivity amplitude term. The negative value of the ratio of the logarithmic amplitude of the individual bottleneck to the bottleneck mitigation coefficient is used to perform an exponential function operation to obtain the individual mitigation attenuation term. The individual bottleneck mitigation terms within the total number of individual devices in the resource group are summed to obtain the resource group bottleneck mitigation sum. The product of the coupling amplification reciprocal factor, the individual mitigation attenuation term, and the resource group bottleneck mitigation sum is calculated to obtain the bottleneck amplification distortion judgment value.

[0012] Furthermore, the specific process of generating bottleneck suppression allocation markers, generating individual equipment lists, and outputting boundaries based on the bottleneck amplification distortion judgment results is as follows: Real-time comparison of the bottleneck amplification distortion judgment value and the bottleneck amplification distortion judgment threshold. The bottleneck amplification distortion judgment threshold includes a primary judgment threshold and a secondary judgment threshold: When the bottleneck amplification distortion judgment value is less than the secondary judgment threshold, the output resource group equivalent output boundary and the resource group equivalent power change rate boundary are normal boundaries, and the individual unit's executable margin, response uncertainty scale, and coupling amplification strength are output; when the bottleneck amplification distortion judgment value is greater than or equal to the secondary judgment threshold and less than the primary judgment threshold, the resource group equivalent power change rate boundary is... The system performs tightening processing and selects the top m individual devices as the tail-end dominant devices based on the bottleneck change of individual devices. After selecting the number of clusters using the Bayesian information criterion, it obtains the cluster with the largest mean and determines the number of tail-end dominant devices. It issues ramp-up limiting instructions and discharge current upper limit constraint instructions to the top m individual devices and outputs the bottleneck suppression allocation mark and the list of ranked individual devices. When the bottleneck amplification distortion judgment value is greater than or equal to the first-level judgment threshold, it performs protective load reduction on the tail-end dominant devices and outputs the resource group reconstruction trigger mark and the list of recommended isolated individual devices. It also outputs the equivalent output boundary of the resource group before reconstruction, the equivalent power change rate boundary of the resource group, and the coupling amplification strength within the group.

[0013] Furthermore, the specific process of performing equivalent constraint shrinkage strength analysis based on the bottleneck amplification distortion judgment result is as follows: calculate the ratio of the bottleneck amplification distortion judgment value to the sum of the bottleneck amplification distortion judgment values ​​to obtain the bottleneck saturation factor; calculate the ratio of the bottleneck change of each individual unit within the range of the number of dominant units at the tail to the maximum historical bottleneck change value and perform a power operation on the bottleneck sensitivity index to obtain the sensitive amplification amount of the individual unit; add one to the ratio of the bottleneck change of the individual unit to the bottleneck slow-release coefficient to obtain the slow-release inhibition factor of the individual unit; calculate the ratio of the sensitive amplification amount of the individual unit to the slow-release inhibition factor of the individual unit and sum it based on the number of dominant units at the tail to obtain the aggregation amount of the tail bottleneck; calculate the ratio of the aggregation amount of the tail bottleneck to the number of dominant units at the tail to obtain the average amount of the tail bottleneck; calculate the product of the bottleneck saturation factor, the inverse coupling amplification factor, and the average amount of the tail bottleneck to obtain the bottleneck-perceived equivalent constraint shrinkage value.

[0014] Furthermore, the specific process of boundary contraction, allocation suppression constraint injection, and output plan solution based on equivalent constraint contraction strength analysis is as follows: Real-time comparison of equivalent constraint contraction values ​​and equivalent constraint contraction thresholds, including first-level and second-level contraction thresholds: When the equivalent constraint contraction value is less than the second-level contraction threshold, the equivalent output boundary and equivalent power change rate boundary of the resource group are used as optimization constraint boundaries. Under the condition that the bottleneck suppression allocation marker is empty, mixed integer programming is used to handle discrete decision-making of resource availability status. The continuous output allocation within the resource group is solved using the distributed alternating direction multiplier method, outputting the resource group output plan and individual allocation instruction set; when the equivalent constraint contraction value is greater than or equal to the second-level contraction threshold and less than the first-level contraction threshold... The system employs a comprehensive optimization mode based on shrinkage equivalent constraints. It performs shrinkage on the equivalent output boundary and the equivalent power change rate boundary of the resource group. The bottleneck suppression allocation marker and the tail-dominant unit list are used as allocation constraint inputs. The output is the resource group output plan and unit allocation instruction set under shrinkage constraints. When the equivalent constraint shrinkage value is greater than or equal to the first-level shrinkage threshold, the optimization objective is switched to prioritize the power satisfaction of the tie line plan and the system-level power balance satisfaction. The resource group reconfiguration trigger marker and the proposed isolated unit equipment list are used as solution mode inputs. Allocation solution is performed only on the set after removing the proposed isolated unit equipment list. The output is the resource group output plan, unit allocation instruction set, proposed isolated unit equipment list, and reconfiguration coordination marker with feasibility region priority.

[0015] Furthermore, the specific process of coupling amplification and push correction assessment of operation monitoring data is as follows: Real-time active power data is obtained from the grid connection point voltage and current data through power calculation; the planned active power value is read from the output resource group output plan; the equivalent output margin of the resource group is obtained by subtracting the planned active power value from the upper bound of the contracted equivalent output of the resource group and the lower bound of the planned active power value from the lower bound of the contracted equivalent output of the resource group; the difference between the real-time active power data and the planned active power value is calculated and its absolute value is taken to obtain the execution deviation amplitude; the ratio of the execution deviation amplitude to the resource group equivalent output margin plus the numerical stability factor is calculated to obtain the margin execution deviation term; the equivalent constraint contraction value of the bottleneck perception is calculated to obtain the constraint contraction amplification term; the product of the coupling amplification reciprocal factor, the margin execution deviation term, and the constraint contraction amplification term is calculated to obtain the bottleneck closed-loop execution correction judgment value.

[0016] Furthermore, the specific process of triggering rolling correction, reallocating individual allocation instruction sets, triggering resource group reconstruction, and replacing resources based on the coupling amplification push correction evaluation results is as follows: Real-time comparison of correction judgment values ​​and correction judgment thresholds, including primary and secondary judgment thresholds: When the correction judgment value is less than the secondary judgment threshold, the resource group output plan and individual allocation instruction set remain unchanged, a daily execution database is created, and the execution deviation magnitude and normal execution status are output and archived to the daily execution database; When the correction judgment value is greater than or equal to the secondary judgment threshold and less than the primary judgment threshold, rolling execution correction is triggered: a rolling correction trigger flag and a correction direction flag are output, and the next rolling… The periodic resource group output plan is corrected in the same direction according to the execution deviation magnitude, and a corrected resource group output plan is generated; the unit allocation instruction set is redistributed according to the bottleneck suppression allocation flag and the tail dominant unit list, and a corrected unit allocation instruction set is generated; when the correction judgment value is greater than or equal to the first-level judgment threshold, the resource group reconstruction trigger flag is output and the isolation action in the isolated unit equipment list is executed; the fast resource replacement rule is executed on the isolated resource group member set to introduce available unit equipment and form a reconstructed resource group member set; the update input of the reconstructed resource group equivalent output boundary and the resource group equivalent power change rate boundary is output and written back; and the starting point of the reconstructed scheduling trajectory is output.

[0017] The second aspect of this invention provides a resource group integrated optimization system for a virtual power plant, including a data acquisition and preprocessing module for acquiring operation monitoring data and obtaining historical operation statistics; preprocessing the operation monitoring data and historical operation statistics; a bottleneck identification and impact factor quantification module for determining bottleneck amplification distortion in the operation monitoring data and historical operation statistics, and generating bottleneck suppression allocation markers, generating individual equipment lists, and outputting boundary values ​​based on the bottleneck amplification distortion determination results; a resource group equivalent constraint modeling and integrated optimization module for performing equivalent constraint contraction strength analysis based on the bottleneck amplification distortion determination results, and performing boundary contraction, allocation suppression constraint injection, and output plan solution based on the equivalent constraint contraction strength analysis; and a rolling execution correction and resource group reconstruction module for performing coupled amplification and push correction evaluation on the operation monitoring data, and triggering rolling correction, reallocating individual allocation instruction sets, triggering resource group reconstruction, and replacing resources based on the coupled amplification and push correction evaluation results.

[0018] The present invention has the following beneficial effects: (1) This invention constructs a bottleneck identification and quantification criterion for resource groups, and explicitly incorporates the impact of slight performance degradation of individual devices on the adjustment capability of resource groups into the scheduling decision link, thereby achieving the effect of early identification and early intervention of hidden bottleneck risks, effectively solving the problem in the prior art that slight anomalies do not trigger alarms but continuously weaken the feasible domain and are difficult to detect in time.

[0019] (2) This invention establishes a bottleneck-aware resource group equivalent constraint modeling mechanism, which enables the resource group equivalent output boundary and equivalent power change rate boundary to dynamically and adaptively tighten with bottleneck risk, thereby achieving the effect of consistent optimization feasible domain and physical executable capability, effectively solving the problem that the scheduling plan cannot be satisfied at the execution end due to the distortion of the unified output and unified gradient assumptions in the prior art.

[0020] (3) This invention combines the rolling correction mechanism of the hierarchical optimization framework with the constraint-driven solution mode switching, so that the solution strategy of prioritizing economy or feasibility is automatically selected under different risk levels, thereby achieving stable and satisfactory results under the constraints of power grid interconnection plan and system-level power balance, effectively solving the problem of difficulty in stably switching control targets when the operating conditions change suddenly or the feasible domain shrinks in the prior art.

[0021] (4) This invention introduces suppression of allocation constraints and single-unit constraint injection strategies, so that comprehensive optimization can actively avoid excessive dependence on weak equipment at the allocation level and improve the structural rationality of task allocation, thereby realizing the adaptive redistribution effect of adjusting responsibilities within the resource group, effectively solving the problem of global solution structure distortion caused by local bottlenecks through coupling propagation in the prior art.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 This is a flowchart of a resource group comprehensive optimization method for a virtual power plant according to the present invention; Figure 2 This is a structural diagram of a resource group integrated optimization system for a virtual power plant according to the present invention; Figure 3 The Pearson correlation thermogram of the parameters and shrinkage values ​​of this invention; Figure 4 Stacked bar charts showing the contribution factors decomposed for different scenarios of the present invention; Figure 5 This is a flowchart of the real-time correction and reconstruction decision-making process of the present invention. Detailed Implementation

[0024] 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.

[0025] Please see Figures 1-5This invention provides a technical solution: a comprehensive optimization method for resource groups in a virtual power plant, comprising the following steps: S1, collecting operation monitoring data and obtaining historical operation statistics; preprocessing the operation monitoring data and historical operation statistics; S2, determining bottleneck amplification distortion in the operation monitoring data and historical operation statistics, and generating bottleneck suppression allocation markers, generating individual equipment lists, and outputting boundaries based on the bottleneck amplification distortion determination results; S3, performing equivalent constraint contraction strength analysis based on the bottleneck amplification distortion determination results, and performing boundary contraction, allocation suppression constraint injection, and output plan solution based on the equivalent constraint contraction strength analysis; S4, performing coupled amplification push correction evaluation on the operation monitoring data, and triggering rolling correction, reallocating individual allocation instruction sets, triggering resource group reconstruction, and replacing resources based on the coupled amplification push correction evaluation results.

[0026] Specifically, the process of collecting operational monitoring data and obtaining historical operational statistics is as follows: Operational monitoring data includes: energy storage battery temperature data, energy storage battery voltage data, energy storage battery current data, allowable power change rate, grid connection point voltage data, grid connection point current data, the total number of individual devices within the resource group, and the step sliding window. The allowable power change rate is calculated based on the executable power change rate of the grid connection specifications applicable to the operating period to ensure that subsequent scheduling constraints remain consistent with grid-side safety requirements. The step sliding window is generated according to the segmented time window rules constructed based on the sampling period. The length of the step sliding window is determined based on the change point detection analysis of the current change rate sequence and short-time energy... The mutation characteristics were statistically determined by adjusting the window size based on the mutation frequency of the first-order difference in the current sequence and the degree of local energy accumulation. This allows the window to be shortened in areas of dense mutation and lengthened in areas of stability. The step sliding window ranges from 5 to 20 sampling points to obtain a time-series segmented structure consistent with real-time monitoring accuracy. The total number of individual devices within a resource group is dynamically counted based on the real-time access status to reflect changes in resource group members. The sampling period is determined based on the sampling frequency parameters of the monitoring device and the stability analysis of the operational monitoring data over a continuous time period. Through statistical evaluation of the rate of change and noise level of the monitoring signal, the sampling period falls within the operational adaptation range of 100 milliseconds to 1 second.

[0027] Historical operational statistics are acquired and a historical operational statistics database is created. The historical operational statistics include: historical energy storage battery temperature data, historical energy storage battery voltage data, historical energy storage battery current data, historical active power data, and historical allowable power. Among them, the historical allowable power is obtained by statistically analyzing the allowable power change rate corresponding to the applicable grid connection specifications within the historical operating range, so that the boundary conditions recorded in the historical operational statistics database can truly reflect the scheduling constraints of different operating ranges. The historical energy storage battery temperature data, historical energy storage battery voltage data, and historical energy storage battery current data are archived according to the original monitoring frequency to maintain the same time resolution as the operational monitoring data.

[0028] This implementation plan, through the systematic collection and structured organization of operational monitoring data and historical operational statistics, ensures that the resource group possesses a consistent, continuous, and quantifiable data foundation across both real-time operation and historical status timescales. This forms a complete input system supporting bottleneck amplification and distortion assessment, equivalent constraint contraction analysis, and subsequent comprehensive optimization solutions. Operational monitoring data, with a unified sampling period, possesses stable temporal resolution, accurately characterizing the dynamic changes in energy storage battery temperature, voltage, current, allowable power change rate, grid connection point voltage, and grid connection point current. Historical operational statistics constitute a reference benchmark for long-term operational behavior and boundary conditions, providing consistent support for the characteristic distribution of historical allowable power, historical energy storage battery temperature, historical energy storage battery voltage, and historical energy storage battery current. By constructing a historical operational statistics database, the full-cycle traceability of the resource group's operational status, power change boundaries, and individual device behavior can be ensured, enabling subsequent judgment processes to conduct quantitative analysis based on real operational constraints and historical distribution structures. The effect of this step is to form a complete data domain covering real-time operating status, historical boundary conditions, and resource group structure information, providing a unified, reliable, and verifiable data foundation for bottleneck identification, constraint contraction, output optimization, closed-loop correction, and resource group reconstruction.

[0029] Specifically, the preprocessing of operation monitoring data and historical operation statistics data is as follows: A piecewise linear trend decomposition algorithm with first-order difference constraints is used to extract the time-varying baseline and isolate abnormal abrupt changes in the allowable power change rate, forming a stable reference sequence that can be used to calculate the power change rate deviation intensity; a phase-synchronized power frequency cycle alignment and robust RMS estimation algorithm is used to perform cycle alignment and robust statistics on grid connection point voltage and current data, obtaining stable electrical parameters for subsequent power calculation and intra-group coupling amplification intensity calculation; a timestamp-consistent multi-source sequence resampling and missing segment hold-up interpolation algorithm is used to perform unified sampling interval alignment and short-time missing segment completion on the operation monitoring data and historical operation statistics data, ensuring that subsequent cross-correlation peak delay estimation and correction judgment value calculation can be directly called upon; and a distribution standardization and linear normalization algorithm is used to standardize and normalize the operation monitoring data and historical operation statistics data.

[0030] This implementation plan, through the systematic collection and structured organization of operational monitoring data and historical operational statistics, ensures that the resource group possesses a consistent, continuous, and quantifiable data foundation across both real-time operation and historical status timescales. This forms a complete input system supporting bottleneck amplification and distortion assessment, equivalent constraint contraction analysis, and subsequent comprehensive optimization solutions. Operational monitoring data, with a unified sampling period, possesses stable temporal resolution, accurately characterizing the dynamic changes in energy storage battery temperature, voltage, current, allowable power change rate, grid connection point voltage, and grid connection point current. Historical operational statistics constitute a reference benchmark for long-term operational behavior and boundary conditions, providing consistent support for the characteristic distribution of historical allowable power, historical energy storage battery temperature, historical energy storage battery voltage, and historical energy storage battery current. By constructing a historical operational statistics database, the full-cycle traceability of the resource group's operational status, power change boundaries, and individual device behavior can be ensured, enabling subsequent judgment processes to conduct quantitative analysis based on real operational constraints and historical distribution structures. The effect of this step is to form a complete data domain covering real-time operating status, historical boundary conditions, and resource group structure information, providing a unified, reliable, and verifiable data foundation for bottleneck identification, constraint contraction, output optimization, closed-loop correction, and resource group reconstruction.

[0031] Specifically, the process for determining bottleneck amplification distortion in operational monitoring data and historical operational statistics is as follows: First, acquire operational monitoring data and historical operational statistics. Second, for energy storage battery temperature data, use the sliding window midpoint and MAD robust normalization algorithm to obtain the temperature deviation intensity, which is used to capture mild anomalies such as slow temperature rise and local overheating. This ensures that real-time optimization meets the energy storage safety constraints and output constraints while avoiding assigning excessively rapid ramp-up tasks to heat-sensitive cells. Third, for energy storage battery voltage data, energy storage battery current data, and energy storage battery temperature data, establish a second-order equivalent circuit model, and use the extended Kalman filter online state estimation algorithm to output a health state sequence. The health status sequence is obtained by calculating the deviation from the sliding window quantile benchmark, which is used to transform the slow degradation of the health status into a quantifiable deviation intensity. For the energy storage battery voltage and current data, the equivalent internal resistance sequence is estimated using a recursive least squares parameter identification algorithm with a forgetting factor on identifiable current step segments. The equivalent internal resistance sequence is obtained using the sliding window mean and MAD robust normalization algorithm. Identifiable current step segments are: performing step sliding window difference detection on the energy storage battery current data, screening intervals where the mean current changes abruptly between adjacent windows and the magnitude of the change is significantly greater than the current fluctuation within the window, while requiring continuous and stable current before and after the change. The platform excludes slow ramp-up and jitter spikes; it uses a sliding window first-order polynomial least squares slope estimation algorithm to obtain the actual power change rate value from the real-time active power data, and then uses the quantile scaling normalization method after subtracting it from the allowable power change rate boundary to directly measure whether the ramp-up tracking capability of individual devices at the execution end is consistent with the ramp-up rate constraints in the constraint condition library; the Euclidean norm of the above four deviation intensities is used as the bottleneck change of individual devices; real-time active power data is obtained from grid connection point voltage data and grid connection point current data through power calculation; and a cross-spectral coherence function estimation algorithm (Hanning window piecewise averaging) is used to estimate the grid connection point voltage data and real-time active power data. Welch power spectral density estimation) constructs a single-unit coherence matrix and uses spectral radius normalization. The normalization factor is obtained by statistically analyzing the quantiles of the largest eigenvalue of the historical coherence matrix, ensuring that the largest eigenvalue falls between zero and less than one. The largest eigenvalue of the normalized coherence matrix is ​​taken as the intra-group coupling amplification strength, with a value range between 0 and 1. For historical energy storage battery temperature data, historical energy storage battery voltage data, historical energy storage battery current data, historical active power data, and historical allowable power change rate boundaries, the bottleneck change of historical single-unit equipment is calculated. The peak over-threshold method is used to extract tail samples, and the maximum likelihood estimation algorithm is used to obtain the tail extreme value as the maximum change value of the historical bottleneck.

[0032] The complement of the intra-group coupling amplification strength is calculated and its reciprocal is taken to obtain the coupling amplification reciprocal factor. When the intra-group coupling amplification strength approaches one, the coupling amplification reciprocal factor increases rapidly. This is used to characterize the risk of intra-group coupling amplifying small attenuations of individual units into distortions of the boundary between the equivalent output and equivalent power change rate of the resource group. For the bottleneck change of each individual device, the natural logarithm of the bottleneck change of the individual device is calculated. The natural logarithm mapping can compress extreme changes and highlight relative differences, allowing the bottleneck amplitudes of different devices to enter the nonlinear comparable space, thus obtaining the logarithmic amplitude of the individual bottleneck. The ratio of the logarithmic amplitude of the individual bottleneck to the maximum historical bottleneck change value is calculated and its absolute value is taken to obtain the normalized amplitude of the individual bottleneck. The absolute value processing can eliminate directional interference and reduce deviation. The quantity always maintains a positive definite attribute; the bottleneck sensitivity exponent is raised to the power of the normalized amplitude of the individual bottleneck to obtain the individual sensitivity amplitude term. This power operation can amplify and weight highly sensitive equipment, thereby forming a nonlinear focusing effect on the key bottleneck; the ratio of the logarithmic amplitude of the individual bottleneck to the bottleneck mitigation coefficient is negatively calculated and subjected to exponential function operation to obtain the individual mitigation attenuation term. The exponential attenuation structure reflects the dynamic characteristics of the bottleneck influence rapidly decaying with time and mitigation capacity; the individual bottleneck mitigation terms are summed over the total number of individual equipment within the resource group to obtain the resource group bottleneck mitigation sum; the product of the coupling amplification reciprocal factor, the individual mitigation attenuation term, and the resource group bottleneck mitigation sum is calculated to obtain the bottleneck amplification distortion judgment value. The specific calculation formula is as follows: ; In the formula, This represents the bottleneck amplification distortion judgment value, used to quantify the risk level of the amplification of the bottleneck change of a single device under the influence of the coupling amplification intensity within the group at the resource group level; This represents the total number of individual devices within a resource group, used to characterize the size of heterogeneous resource groups formed by dynamic clustering. This indicates the current calculation time, the online evaluation time point, and reflects the latest operating conditions; This represents the device number index within the resource group, used to align individual device constraints with individual device bottleneck changes one by one, ensuring that allocation and constraints can locate specific bottleneck devices rather than remaining at the aggregation layer; It represents the bottleneck change of a single device, used to characterize how slight performance degradation of a single device in terms of temperature, health status and ramping ability compresses the boundary of the rate of change of executable output and executable power. This indicates the intensity of intra-group coupling amplification. Yes, it is used to characterize the risk of intra-group coupling amplifying the small attenuation of a single unit into the equivalent output of the resource group and the boundary distortion of the equivalent power change rate. This represents the maximum change value of the historical bottleneck, used to establish differentiated historical extreme benchmarks for different individual devices, so that the resource group comprehensive optimization has a comparable scale when comparing the bottleneck degree of different individuals; The bottleneck sensitivity index is obtained by estimating the Hearst index using a detrended fluctuation analysis algorithm based on the historical changes in the bottleneck of individual equipment. The Hearst index ranges from 0 to 1. The bottleneck sensitivity index is obtained by linearly mapping the Hearst index to a closed interval of 1 to 4, with a value range of 1 to 4. The bottleneck mitigation coefficient is calculated by using the autocorrelation function sequence of the historical individual equipment bottleneck changes. The decay time constant of the autocorrelation function is fitted using the Levenberg-Marquardt exponential decay fitting algorithm to obtain the bottleneck mitigation coefficient, which is used to characterize the persistence and mitigation rate of the bottleneck effect. The value range is greater than 0.

[0033] This implementation scheme achieves unified quantification of key operational deviations, such as temperature deviation, health status deviation, equivalent internal resistance change, and power change rate deviation, by systematically determining bottleneck amplification distortion based on operational monitoring data and historical operational statistics. Based on this, an intra-group coupling amplification strength reflecting the intensity of interactions within the resource group is constructed, forming a coupling index system characterizing the overall stability of the resource group. This process integrates multi-source deviation characteristics into bottleneck changes for individual equipment and constructs differentiated scale benchmarks based on historical extreme conditions, ensuring that slight performance degradation of individual equipment is accurately represented within a unified evaluation framework. Subsequently, a nonlinear amplification and attenuation mechanism is established based on the bottleneck sensitivity index and bottleneck mitigation coefficient, quantifying the persistence, sensitivity, and mitigation capability of bottleneck evolution. Finally, this value, together with the inverse coupling amplification factor, forms the bottleneck amplification distortion determination value. This value reflects the degree to which the deviation of individual equipment is transmitted through intra-group coupling relationships and distorts the boundary between the equivalent output and equivalent power change rate of the resource group, providing stable, measurable, and interpretable risk input for subsequent bottleneck suppression, boundary contraction, output allocation, and resource group reconfiguration decisions.

[0034] Specifically, the process of generating bottleneck suppression allocation markers, generating individual equipment lists, and outputting boundaries based on the bottleneck amplification distortion judgment results is as follows: Real-time comparison of the bottleneck amplification distortion judgment value and the bottleneck amplification distortion judgment threshold. The bottleneck amplification distortion judgment threshold includes a primary judgment threshold and a secondary judgment threshold. When the bottleneck amplification distortion judgment value is less than the secondary judgment threshold, neither the individual unit performance degradation intensity nor the resource group coupling amplification intensity triggers a significant contraction of the executable boundary. The output resource group equivalent output boundary and the resource group equivalent power change rate boundary are conventional boundaries. The resource group equivalent output boundary is calculated by summing the maximum allowable discharge power boundary, the maximum allowable charging power boundary, and the real-time active power data of each individual unit within the resource group according to the same sign convention. Among them, the upper and lower bounds of the resource group are formed by summing the adjustable margins of each individual unit respectively. The equivalent power change rate boundary is obtained by summing the allowable power change rate boundaries of each individual unit within the resource group according to the upward and downward adjustment directions respectively. The individual executable margin and response uncertainty scale are combined. The degree and coupling amplification strength are passed to the resource group equivalent constraint modeling and comprehensive optimization module to calculate the individual unit power adjustment margin: the maximum allowable discharge power boundary minus the individual unit's real-time active power data; and the individual unit adjustment rate margin: the allowable power change rate boundary minus the estimated value of the individual unit's actual power change rate, which is obtained by slope estimation from the individual unit's real-time active power data. The tighter of the two margins is taken as the individual unit's executable margin. The response uncertainty scale is calculated from the deviation fluctuation level between the individual unit's actual executed active power sequence and the individual unit's target active power command sequence, which is used to quantify the randomness of the execution deviation. This allows the resource group equivalent constraint modeling and comprehensive optimization module to perform plan tracking and economic optimization according to the conventional feasible region. When the bottleneck amplification distortion judgment value is greater than or equal to the secondary judgment threshold and less than the primary judgment threshold, it indicates that the local bottleneck may have been amplified through coupling and affected the resource group equivalent gradient assumption. Therefore, a tightening process is applied to the boundary of the resource group equivalent power change rate: the upper bound of the resource group equivalent power change rate is limited to the lower bound of the quantile of the executable power change rate calculated from the individual executable margin. At each rolling scheduling time, based on the bottleneck change amount of individual devices, they are sorted from largest to smallest. The top m individual devices in the resource group are selected as the tail-leading individual devices, where m is adaptively clustered using a one-dimensional Gaussian mixture model. The results show that, after selecting the number of clusters using the Bayesian information criterion, the cluster with the largest mean is chosen. The number of samples in this cluster represents the number of the top few items, with a value ranging from 0 to K-1. Physical-level ramp-up limiting commands and discharge current upper limit constraints are issued to the top m individual devices. The power change rate limit is written into the control loop through the inverter controller to prevent the bottleneck device from experiencing excessively rapid power changes that could lead to execution failure. At the decision level, a bottleneck suppression allocation flag and a list of ranked individual devices are output as input flags and constraint information for the resource group equivalent constraint modeling and comprehensive optimization module, which are then used for subsequent allocation suppression modeling and comprehensive optimization solutions.

[0035] When the bottleneck amplification distortion judgment value is greater than or equal to the first-level judgment threshold, it indicates that the combined effect of the resource group coupling amplification strength and the individual unit performance degradation strength has caused a non-negligible distortion risk to the assumptions of unified output and unified power change rate. At the physical level, protective load reduction is performed on the tail-end dominant unit, including limiting the maximum allowable discharge power boundary, limiting the allowable power change rate boundary, and initiating a temperature control strategy. At the judgment level, the resource group reconstruction trigger flag and the list of recommended isolated individual equipment are output, along with the equivalent output boundary of the resource group before reconstruction, the equivalent power change rate boundary of the resource group, and the coupling amplification strength within the group. These serve as input information for the resource group equivalent constraint modeling and comprehensive optimization module and the rolling execution correction and resource group reconstruction module, and are used for subsequent resource group reconstruction, comprehensive optimization solution, and execution of closed-loop correction.

[0036] In this implementation scheme, by applying the bottleneck amplification distortion judgment value in the hierarchical judgment logic, the resource group can dynamically select three strategies—normal operation, contraction operation, and protection operation—based on the bottleneck evolution status of individual devices and the changing trend of intra-group coupling amplification intensity. This enables adaptive control of the resource group's equivalent output boundary and equivalent power change rate boundary. The judgment result can uniformly map the individual device performance deviation, intra-group coupling amplification effect, and the overall executability of the resource group. This allows the resource group to maintain a normal feasible domain under slight deviation, initiate boundary contraction and amortization suppression constraints during local bottleneck enhancement stages, and trigger protective load reduction and resource group reconstruction when bottleneck amplification reaches the distortion threshold. This effectively avoids plan infeasibility caused by the failure of the unified gradient assumption. This mechanism establishes a complete link from deviation identification, bottleneck location, boundary adjustment to individual device suppression and resource group structure adjustment, so that comprehensive optimization has clear feasible domain constraints and allocation strategy basis under different risk levels, thereby improving the executability of scheduling plans, resource group stability and operational security. It also constructs a dynamic boundary management system based on bottleneck amplification behavior, providing decision-driven constraint input for resource group comprehensive optimization, enabling resource groups to have self-adjustment capability and operational resilience in the context of heterogeneity, coupling and attenuation.

[0037] Specifically, the process of equivalent constraint contraction intensity analysis based on bottleneck amplification distortion judgment results is as follows: The ratio of the bottleneck amplification distortion judgment value to the sum of the two bottleneck amplification distortion judgment values ​​is calculated to obtain the bottleneck saturation factor. The bottleneck saturation factor is used to characterize the saturation degree of the bottleneck amplification distortion judgment value at a unit increment scale, enabling subsequent constraint contraction to adjust the contraction intensity according to the bottleneck growth rate. For each dominant unit within the range of tail-end numbers, the ratio of the individual bottleneck change to the historical maximum bottleneck change value is calculated, and the bottleneck sensitivity index is raised to the power of the value to obtain the individual sensitive amplification amount. This power mapping mechanism introduces nonlinear amplification capability for highly sensitive regions, allowing the most critical dominant unit in bottleneck evolution to receive priority response during constraint contraction. The ratio of the individual equipment bottleneck change to the bottleneck mitigation coefficient is then further... The increment operation yields the monomer sustained-release inhibition factor. Based on the dynamic decay characteristics of the bottleneck sustained-release coefficient, the inhibition weight of monomer devices with strong self-recovery capabilities is reduced during constraint contraction, thereby avoiding excessive conservative contraction of the resource group. The ratio of the monomer sensitive amplification amount to the monomer sustained-release inhibition factor is calculated and summed based on the number of tail-dominant monomers to obtain the tail-bottom bottleneck aggregation amount. The nonlinear aggregation of the tail region characterizes the bottleneck accumulation effect under the combined action of the tail-dominant monomers, enabling the resource group to accurately identify the collective behavior of the most unfavorable region. The ratio of the tail-bottom bottleneck aggregation amount to the number of tail-dominant monomers is calculated to obtain the tail-bottom average amount. The product of the bottleneck saturation factor, the inverse coupling amplification factor, and the tail-bottom average amount is calculated to obtain the bottleneck-perceived equivalent constraint contraction value. The specific calculation formula is as follows: ; In the formula, It represents the equivalent constraint contraction value, which is used to characterize the combined strength of the contraction required between the equivalent output boundary of the resource group and the equivalent power change rate boundary of the resource group; This represents the sorted individual device sequence number index, used to locate the dominant source of bottleneck propagation effects within heterogeneous resource groups; This indicates the current calculation time, adapting to changes in the feasible region caused by short-term forecast data. This represents the bottleneck amplification distortion judgment value, used to characterize the degree of distortion caused by the bottleneck change of a single device to the assumptions of the aggregation model under the intra-group coupling amplification intensity; It represents the intra-group coupling amplification strength, used to reflect the degree of linkage of resource groups formed by dynamic clustering under the constraints of electrical distance and complementary characteristics; This indicates the total number of individual devices within a resource group, used to adapt to changes in the number of members caused by dynamic clustering and dynamic reorganization of resource groups. This indicates the number of dominant monomers at the tail end, reducing the impact of excessively conservative contraction on the cost function and user comfort objectives; Indicates the first The bottleneck change of a single device; the first value after sorting all bottleneck changes of single devices in the resource group from largest to smallest. The bottleneck variation of individual devices enables the lower-level real-time optimization to take the individual devices that cause execution failure as the main targets of constraint contraction and amortization suppression. Indicates the first The maximum change value of the historical bottleneck is used to provide a historical extreme benchmark consistent with the characteristics of individual equipment, so that the dimensions and scales of the bottleneck changes of different individual equipment are comparable. Indicates the first A bottleneck sensitivity index is used to characterize the persistence and amplification sensitivity of the bottleneck change of the i-th individual device; Indicates the first A bottleneck mitigation coefficient is used to characterize the mitigation rate and duration of the bottleneck effect of the i-th individual device.

[0038] In this embodiment, Table 1 shows the calculation parameters and quantification results of the contraction value under different bottleneck scenarios. It records in detail the key quantification results of the bottleneck amplification judgment value, coupling amplification intensity, single bottleneck change, historical maximum bottleneck change value, bottleneck sensitivity index, bottleneck slow-release coefficient, single sensitive amplification amount, single slow-release inhibition factor, ratio of sensitive amplification amount to slow-release inhibition factor, tail bottleneck average amount, bottleneck saturation factor, coupling amplification reciprocal factor, and final equivalent constraint contraction value C(t) under three types of bottleneck scenarios: light bottleneck, medium bottleneck, and strong bottleneck. It is used to characterize the calculation process and index response law of the equivalent constraint contraction value under different bottleneck intensity conditions. Specifically: in the light bottleneck scenario, the bottleneck amplification judgment value is 0.20, the bottleneck saturation factor is 0.167, the average amount of the tail bottleneck is 0.12, and the final calculated C(t) is 0.024; in the medium bottleneck scenario, the bottleneck amplification judgment value is 0.60, the bottleneck saturation factor is 0.375, the average amount of the tail bottleneck is 0.18, and the final C(t) is 0.096; in the strong bottleneck scenario, the bottleneck amplification judgment value reaches 1.20, the bottleneck saturation factor increases to 0.545, the average amount of the tail bottleneck increases to 0.26, and the final C(t) is 0.259. Overall, the data shown in Table 1 fully presents the stepwise amplification trend of each parameter caused by bottleneck enhancement, providing a basic data basis for the design and optimization of the subsequent equivalent constraint contraction strategy.

[0039] Table 1. Calculation parameters and quantification results of shrinkage value under different bottleneck scenarios.

[0040] like Figure 3The diagram shows the Pearson correlation heatmap between the parameters of this invention and the shrinkage value. Combined with Table 1, it can be seen that different parameters have structural differences in their influence on the equivalent constrained shrinkage value C(t). Among them, the bottleneck saturation factor has the highest correlation with C(t), indicating that it plays a dominant role in the shrinkage value formation process; the inverse coupling amplification factor shows a significant positive correlation, indicating that the intra-group coupling strength has a sensitive influence on constrained shrinkage; the average amount of the tail bottleneck also has a prominent correlation with C(t), indicating that the dominant tail bottleneck monomer has a key contribution to the shrinkage calculation results; the monomer sensitive amplification amount and its ratio with the sustained-release inhibition factor also show a moderate correlation, reflecting that the performance degradation at the monomer level also plays an important role in C(t). Figure 4 The chart shown is a stacked bar chart of contribution factors for different scenarios. By standardizing and stacking the bottleneck saturation factor, the inverse coupling amplification factor, and the average amount of the tail bottleneck, it can be clearly seen that as the bottleneck intensifies from mild to severe, all contribution factors show a cumulative upward trend, consistent with the growth trajectory of the final C(t). Overall, Figure 3 and Figure 4 Together, these factors constitute the interpretability basis of the shrinkage value in this invention, providing clear quantitative support for parameter configuration optimization and bottleneck feature identification.

[0041] In this implementation scheme, by uniformly quantifying multi-dimensional parameters such as bottleneck amplification distortion judgment value, intra-group coupling amplification strength, tail-dominant single-unit performance decay characteristics, and historical extreme change values, the equivalent output boundary and equivalent power change rate boundary of the resource group can form a continuous, differentiable, and interpretable contraction response during the bottleneck evolution process. This accurately reflects the degree of influence of bottleneck accumulation, coupling amplification, and tail-dominant behavior on the equivalent executable domain, realizing the step-by-step contraction mapping and dynamic adaptation of the resource group from light bottleneck to strong bottleneck conditions. At the same time, through the comprehensive characterization of bottleneck saturation factor, coupling amplification reciprocal factor, and tail-bottom average quantity, the contraction value can directly describe the combined effect of bottleneck development speed, intra-group linkage strength, and tail-bottom structural risk on equivalent constraints. This ensures that the boundary contraction amplitude exhibits a stable, monotonic, and controllable change law with the strength of the bottleneck, thereby providing reliable quantitative input for subsequent generation of equivalent constraints, amortization suppression strategies, and comprehensive optimization solutions for the resource group. This allows constraint adjustment and scheduling execution to achieve a balance between safety, stability, and feasibility, and to continuously maintain the integrity of the feasible solution space.

[0042] Specifically, the process of boundary contraction, allocation of suppression constraints, and output plan solution based on equivalent constraint contraction strength analysis is as follows: Real-time comparison of equivalent constraint contraction values ​​and equivalent constraint contraction thresholds. Equivalent constraint contraction thresholds include first-level and second-level contraction thresholds. Lower-level real-time optimization needs to track the actual response capabilities of user-side adjustable resources, energy storage, and distributed power sources. Through hierarchical comparison of equivalent constraint contraction values, the equivalent output boundary and the equivalent power change rate boundary of the resource group can dynamically tighten with fluctuations in user-side load, disturbances in distributed power source output, and mild degradation of energy storage, avoiding amplification of tie-line planned power deviations and system-level power balance deviations due to distortion of the unified gradient assumption. When the equivalent constraint contraction value is less than the second-order contraction threshold, the equivalent output boundary and the equivalent power change rate boundary of the resource group are used as the optimization constraint boundary. On the grid side, the planned power tracking error of the tie line is within the dispatch allowable bandwidth, and the actual active power data returned by the resource group can meet the grid's response requirements for frequency regulation, peak shaving, and reserve in both the ramp-up and steady-state phases. There is no conflict between user-side comfort constraints and equipment capacity constraints. When the bottleneck suppression allocation marker is empty, the temperature deviation intensity, health status deviation intensity, and power change rate deviation intensity of each individual device do not form a dominant constraint on the adjustability of the resource group. Conventional allocation according to the cost function optimum is allowed, and it is not necessary to reduce the weight of specific user-side devices and specific energy storage units. In terms of solution strategy, mixed integer programming is used to handle the discrete decision of resource availability state. The mixed integer programming adopts augmented Lagrangian relaxation and branch and bound mechanism to solve for availability state, start-stop logic, and discrete switching variables. This approach enables the discrete feasible region to be uniformly processed within the continuous feasible region optimization framework. A mixed-integer decision space is constructed using the available state of individual devices, maximum adjustable capacity, and allowable power change rate boundaries. This allows for the joint solution of resource start-up / shutdown and adjustable intervals. A distributed alternating direction multiplier method is executed in parallel to solve the continuous output allocation within the resource group. This method divides the continuous output variables of the resource group into several local subproblems based on individual devices. Each subproblem independently minimizes its output, cost, and constraint functions. Augmented Lagrange coordination variables are alternately updated between local subproblems and group-level consistency constraints, achieving consistent convergence of power balance constraints. The global layer of the resource group is responsible for updating consensus variables and Lagrange multipliers, ensuring that the overall optimization converges towards the center of the feasible region after each iteration. This achieves rapid solution and real-time executability guarantees for large-scale resource groups in a distributed structure, outputting the resource group output plan and the transmission of individual allocation instruction sets.

[0043] When the equivalent constraint contraction value is greater than or equal to the secondary contraction threshold but less than the primary contraction threshold, some user-side resources or some energy storage units exhibit slight degradation that affects response without triggering alarms. A comprehensive optimization mode for contraction equivalent constraints is adopted: contraction is applied to the equivalent output boundary and the equivalent power change rate boundary of the resource group. The contraction magnitude is calculated by back-calculating the executable power change rate, which is the sum of the executable margins of the individual equipment sets corresponding to the number of tail-dominant units, within the group. The minimum contraction coefficient is taken such that the upper bound of the equivalent power change rate of the resource group after contraction does not exceed the executable power change rate, and the contraction result is formed. The executable power change rate of the tail-dominant unit under the current operating conditions is used as the upper bound of the true gradient of the resource group to prevent the rapid regulation capability promised by the virtual power plant to the grid from failing at the execution end. The bottleneck suppression allocation marker and the tail-dominant unit list are used as allocation constraint inputs, enabling comprehensive optimization at the allocation level for the tail-dominant unit list. By applying allocation suppression constraints to individual devices, the allocation of tasks with upward adjustments and rapid changes is reduced for bottleneck devices. Frequent adjustment tasks are allocated more to devices and user-side resources with smaller response uncertainty and more sufficient execution margin. This improves plan executability while meeting user comfort and production constraints. At the level of individual allocation variables, upper limits of bottleneck device allocation and upper limits of bottleneck device power change rate are added: for each device corresponding to the tail-dominant device list, the upper limit of its allocated planned power increment is limited to the function value of the device's execution margin, and the upper limit of its allocated planned power change rate is limited to the boundary of the device's allowed power change rate; for devices not corresponding to the tail-dominant device list, allocation is performed according to the normal boundary; at the same time, allocation consistency constraints are added at the resource group level to avoid scheduling instructions from over-relying on the tail-dominant device; the output is the resource group output plan and individual allocation instruction set obtained by solving the contraction constraints.

[0044] When the equivalent constraint contraction value is greater than or equal to the first-level contraction threshold, it represents a scenario of increased grid operation pressure. Continuing to prioritize the optimal cost function will lead to a continuous deviation in tie-line planned power and may trigger a violation of power balance constraints. Therefore, the optimization objective is switched to prioritizing tie-line planned power availability and system-level power balance availability. Resource group reconfiguration trigger flags and a list of recommended isolated unit devices are used as inputs to the solution mode. Optimization is performed only on the candidate set after the list of recommended isolated unit devices has been removed. The outputs are a feasible region-priority resource group output plan, unit allocation instruction set, list of recommended isolated unit devices, and reconfiguration coordination flag transmission.

[0045] In this implementation scheme, by comparing the equivalent constraint contraction value with the equivalent constraint contraction threshold in real time, the equivalent output boundary and the equivalent power change rate boundary of the resource group can form a continuous, measurable, and verifiable hierarchical contraction response under different operating conditions, thereby ensuring the accuracy of the scheduling plan's matching with the actual executable capability. Under mild contraction conditions, the normal feasible region of the resource group is maintained, allowing allocation and tracking to operate stably within the complete adjustable range. Under moderate contraction conditions, the limitation of the tail-dominant individual on the resource group's response capability is weakened by the injection of allocation suppression constraints, enabling the scheduling strategy to operate within a reasonable range. A dynamic balance is achieved between execution margin, response uncertainty scale, and power change rate constraint, improving the stability and controllability of plan execution. Under intensity contraction conditions, resource group reconfiguration and isolation mechanisms ensure that the power satisfaction of tie line plans and the system-level power balance satisfaction remain feasible under the tightest constraints, thereby avoiding further spread of boundary distortion caused by individual performance degradation and intra-group coupling amplification. Real-time adaptation of scheduling feasible domain, structured adjustment of allocation strategy, and dynamic optimization of resource group configuration are realized, so that scheduling results maintain consistent operational characteristics in terms of security, executability, and robustness.

[0046] Specifically, the process of coupling amplification and push-and-correction assessment of operation monitoring data is as follows: Real-time active power data is obtained from grid connection point voltage and grid connection point current data through power calculation. The power calculation adopts a synchronously tuned phase alignment and robust RMS extraction mechanism to ensure that the real-time active power data can maintain numerical stability under power frequency disturbances and noise backgrounds. The planned active power value is read from the output resource group output plan and the consistency with the real-time monitoring sequence is ensured by aligning with the time synchronization stamp. The equivalent output margin of the resource group is obtained by taking the smaller of the upper bound of the contracted equivalent output of the resource group minus the planned active power value and the lower bound of the planned active power value minus the contracted equivalent output of the resource group. By constructing a symmetrical feasible region between the upper and lower bounds, the push-and-correction has a consistent response mechanism to deviations in different directions.

[0047] The difference between real-time active power data and planned active power value is calculated and its absolute value is taken to obtain the execution deviation amplitude. This is used to suppress judgment errors caused by instantaneous spikes and ensure that the deviation amplitude reflects the true execution offset trend. The ratio of the execution deviation amplitude to the resource group's equivalent output margin plus the numerical stability factor is calculated to obtain the margin execution deviation term. The execution deviation is mapped to the remaining feasible margin using a normalized scale, so that the deviation exhibits non-linear growth as it approaches the margin boundary, thereby revealing potential pushover risks in advance. The equivalent constraint contraction value plus bottleneck perception is calculated to obtain the constraint contraction amplification term. The bottleneck evolution state is introduced into the execution correction logic, so that the contraction intensity is transmitted to the execution layer, realizing closed-loop coupling between the constraint layer and the execution layer. The product of the coupling amplification reciprocal factor, the margin execution deviation term, and the constraint contraction amplification term is calculated to obtain the bottleneck closed-loop execution correction judgment value. The specific calculation formula is as follows: ; In the formula, This represents the correction judgment value, used to characterize the pushing intensity of the execution deviation on the feasible region under the background of coupling amplification and constraint contraction; It represents the coupling amplification strength within the group, and describes the degree of linkage between individual devices within the resource group under the constraints of electrical distance and complementary characteristics; This represents the equivalent constraint contraction value, used to reflect the degree of tightness of the constraints after the equivalent output boundary and the equivalent power change rate boundary of the resource group have been contracted. It represents real-time active power data, corresponding to the actual active power injected into and absorbed by the virtual power plant at the grid connection point. It is a direct observation for the grid side to assess the availability of planned power from tie lines and the effectiveness of dispatch response. This indicates the planned active power value, and the output resource group output plan is read. This represents the equivalent output margin of a resource group, used to characterize the maximum allowable deviation in plan execution. The numerical stability factor is obtained by using the denominator minimum robustness constraint algorithm on the equivalent output margin of the resource group. The value ranges from 0.005 to 0.01.

[0048] In this implementation scheme, a coupled amplification and push-out correction assessment mechanism is constructed to dynamically quantify the deviation between the real-time active power data at the execution end and the planned active power value under the joint constraints of the equivalent output margin of the resource group and the equivalent constraint contraction value, thereby forming a closed-loop judgment capability for plan executability. This mechanism maps the coupling amplification strength, execution deviation amplitude, equivalent output margin of the resource group, and constraint contraction amplification effect into a unified correction judgment value, causing the execution deviation to exhibit a progressively amplifying trend as it approaches the feasible domain boundary, thus identifying potential output distortion and execution push-out risks in advance. Through a rolling correction and resource group reconfiguration strategy driven by the correction judgment value, the system can maintain plan stability during the mild deviation stage, perform consistent correction during the moderate deviation stage, and trigger protective reconfiguration during the severe deviation stage, achieving unified closed-loop linkage between the planning layer, constraint layer, and execution layer. This assessment mechanism significantly improves the plan executability assurance capability of the virtual power plant in complex coupled environments, enabling resource group output regulation to maintain stable, safe, and controllable operation performance when facing bottleneck evolution, user-side disturbances, and equipment dynamic degradation.

[0049] Specifically, the process of triggering rolling correction, reallocating individual amortized instruction sets, triggering resource group reconstruction, and replacing resources based on the evaluation results of coupling amplification and push correction is as follows: real-time comparison of correction judgment values ​​and correction judgment thresholds, whereby the correction judgment thresholds include primary judgment thresholds and secondary judgment thresholds: When the correction judgment value is less than the secondary judgment threshold, the resource group output plan and the individual allocation instruction set remain unchanged. A daily execution database is created, and the execution deviation amplitude and execution status normal mark are archived to the daily execution database. The daily execution data is sent back as the input for the next round of historical sequence update. The corresponding virtual power plant participates in the grid interconnection plan tracking and the operation status of regular frequency regulation and peak shaving. The actual active power data of the resource group can stably match the resource group output plan within the deviation band allowed by the grid dispatch. There is no continuous lag or saturation in the user-side load response and energy storage charging and discharging.

[0050] When the correction judgment value is greater than or equal to the secondary judgment threshold and less than the primary judgment threshold, rolling execution correction is triggered: the rolling correction trigger flag and correction direction flag are output, the resource group output plan of the next rolling cycle is corrected in the same direction according to the execution deviation magnitude, and the corrected resource group output plan is generated. The rolling cycle is determined by the cross-correlation peak delay estimation algorithm of the planned active power sequence and the actual active power sequence, and the obtained execution lag time is taken as the update step size of the rolling cycle. At the allocation level, the individual allocation instruction set is redistributed according to the bottleneck suppression allocation flag and the tail dominant individual list, and the corrected individual allocation instruction set is generated. The virtual power plant internal scheduling coordination prioritizes the allocation of the rapidly changing adjustment amount to the individual equipment with more sufficient execution margin and smaller response uncertainty scale, and reduces the adjustment tasks of the individual equipment corresponding to the tail dominant individual list to slow-changing or small-amplitude tasks, reducing the accumulation of resource group execution deviation caused by the bottleneck individual's tracking failure, and reducing the grid side's misjudgment of the virtual power plant's compensation capability. The corrected resource group output plan and the corrected individual allocation instruction set are issued for execution and a correction event log is generated.

[0051] When the correction judgment value is greater than or equal to the first-level judgment threshold, the resource group reconstruction trigger flag is output and the isolation action in the isolated single device list is executed; the fast resource replacement rule is executed on the isolated resource group member set to introduce available single devices and form a reconstructed resource group member set. In the candidate resource pool with similar electrical distance neighborhood or complementary characteristics, single devices with more stable available status are selected for replacement first, so as to maintain the original resource group's service capacity and cost function level to the power grid as much as possible, while reducing the disturbance of reconstruction to user-side comfort and production constraints; the updated inputs of the reconstructed resource group equivalent output boundary and the resource group equivalent power change rate boundary are output and written back to the next round of bottleneck identification and influence factor quantification module and resource group equivalent constraint modeling and comprehensive optimization module; the starting point of the reconstructed scheduling trajectory is output for subsequent rolling solution, and a reconstruction event log and a log comparing the feasible domain before and after reconstruction are generated.

[0052] like Figure 5The diagram illustrates the real-time correction and reconfiguration decision-making process. It details the real-time comparison of the correction judgment value and the dual-layer judgment threshold at the beginning of each rolling cycle. Based on the judgment result, the process is routed to one of three paths: normal mode, rolling correction mode, or reconfiguration mode. When the judgment result corresponds to the normal mode, the current plan remains unchanged, and daily execution archiving is completed. When the judgment result enters the rolling correction mode, the plan for the next cycle is directionally corrected based on the direction of the execution deviation. Simultaneously, the instruction sets of relevant individual devices are reconfigured at the allocation level, and correction events are recorded. When the judgment result meets the conditions for the reconfiguration mode, bottleneck unit isolation, resource replacement, and boundary updates are performed, and a new scheduling trajectory is generated for subsequent rolling solutions. Through the periodic closed-loop operation of this process, a hierarchical, stable, and traceable scheduling correction and resource group self-recovery mechanism can be formed under different deviation intensities.

[0053] In this implementation scheme, a hierarchical rolling correction and resource group reconfiguration mechanism based on correction judgment values ​​is established. This enables the virtual power plant to dynamically, interpretably, and directionally complete plan correction and resource group self-recovery when deviations occur at the execution end, bottleneck units continue to deteriorate, or intra-group coupling amplification effects intensify. During periods of minor deviation, it automatically maintains plan stability and archives daily execution information. During periods of moderate deviation, it performs directional rolling corrections to the next cycle's plan based on execution lag patterns and structurally redistributes the unit allocation instruction sets, allowing units with more sufficient execution margins to undertake the main regulation tasks, thereby effectively suppressing the continued deterioration of bottleneck units. During periods of severe deviation, reconfiguration triggers the isolation of bottleneck units, rapid replacement of available resources, and real-time updates of equivalent boundaries, ensuring that the resource group maintains feasible domain stability while meeting grid dispatch requirements. This hierarchical response system forms a continuous closed loop between the planning layer, constraint layer, and execution layer, improving the virtual power plant's dynamic executability, resistance to bottleneck propagation, and service continuity under disturbance conditions. It ensures that tie-line planned power tracking, system-level power balance, and user-side constraints remain stable and consistent in complex operating environments.

[0054] like Figure 2As shown, the second aspect of the present invention provides a resource group integrated optimization system for a virtual power plant, including a data acquisition and preprocessing module for acquiring operation monitoring data and obtaining historical operation statistics data; preprocessing the operation monitoring data and historical operation statistics data; a bottleneck identification and impact factor quantification module for determining bottleneck amplification distortion in the operation monitoring data and historical operation statistics data, and generating bottleneck suppression allocation markers, generating individual equipment lists, and outputting boundary values ​​based on the bottleneck amplification distortion determination results; a resource group equivalent constraint modeling and integrated optimization module for performing equivalent constraint contraction strength analysis based on the bottleneck amplification distortion determination results, and performing boundary contraction, allocation suppression constraint injection, and output plan solution based on the equivalent constraint contraction strength analysis; and a rolling execution correction and resource group reconstruction module for performing coupled amplification and push correction evaluation on the operation monitoring data, and triggering rolling correction, reallocating individual allocation instruction sets, triggering resource group reconstruction, and replacing resources based on the coupled amplification and push correction evaluation results.

[0055] In this implementation plan, a closed-loop control system is constructed, consisting of bottleneck identification and impact factor quantification, resource group equivalent constraints and comprehensive optimization, rolling execution correction and resource group reconstruction. This system enables the virtual power plant to promptly identify and quantify the coupling amplification effect of minor performance degradations such as temperature shifts, health degradation, or reduced ramping capacity within heterogeneous distributed resource groups. This avoids distortion of the unified output assumption and unified gradient assumption at the actual execution end. Through dynamic contraction of equivalent constraints, amortization suppression of tail-dominant units, directional rolling correction of execution deviations, and resource group reconstruction when necessary, the system effectively blocks the bottleneck propagation path. This allows the scheduling plan to continuously maintain a gradient structure consistent with the actual executability of individual equipment, significantly improving the plan executability, power point tracking stability, and system-level power balance capability of the virtual power plant under complex operating conditions.

[0056] 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.

[0057] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for comprehensive optimization of resource groups in a virtual power plant, characterized in that, Includes the following steps: S1, Collect operational monitoring data and obtain historical operational statistics; preprocess the operational monitoring data and historical operational statistics. S2 performs bottleneck amplification distortion judgment on operation monitoring data and historical operation statistics data, and generates bottleneck suppression allocation markers, individual equipment list and boundary output based on the bottleneck amplification distortion judgment results; S3, Based on the bottleneck amplification distortion judgment result, perform equivalent constraint shrinkage strength analysis, and solve the boundary shrinkage, amortization suppression constraint injection and output plan according to the equivalent constraint shrinkage strength analysis; S4 performs coupled amplification and push correction assessment on the operation monitoring data, and triggers rolling correction, reallocation of individual unit allocation instruction sets, resource group reconstruction and resource replacement based on the results of the coupled amplification and push correction assessment.

2. The resource group comprehensive optimization method for a virtual power plant according to claim 1, characterized in that: The specific process for collecting operational monitoring data and obtaining historical operational statistics is as follows: Collect operational monitoring data, including: energy storage battery temperature data, energy storage battery voltage data, energy storage battery current data, allowable power change rate, grid connection point voltage data, grid connection point current data, total number of individual devices in the resource group, and step sliding window; Obtain historical operation statistics data and create a historical operation statistics database. The historical operation statistics data include: historical energy storage battery temperature data, historical energy storage battery voltage data, historical energy storage battery current data, historical active power data, and historical allowable power. Specifically, the allowable power change rate is calculated according to the executable power change rate in the grid connection specification; the step sliding window is determined according to the segmented time window rule constructed according to the sampling period; and the historical allowable power is obtained by statistically analyzing the allowable power change rate in the grid connection specification within the historical operating interval.

3. The resource group comprehensive optimization method for a virtual power plant according to claim 1, characterized in that: The specific process for preprocessing the operation monitoring data and historical operation statistics data is as follows: A piecewise linear trend decomposition algorithm with first-order differential constraints is used to extract time-varying baselines and isolate anomalous changes in the allowable power change rate; a phase-synchronized power frequency cycle alignment and robust RMS estimation algorithm is used to perform cycle alignment and robust statistics on grid connection point voltage and current data. By using a timestamp-consistent multi-source sequence resampling and missing segment-preserving imputation algorithm, the operation monitoring data and historical operation statistics data are aligned with a unified sampling interval and short-term missing data are filled in. By using distribution standardization and linear normalization algorithms, the operation monitoring data and historical operation statistics data are standardized and normalized.

4. The resource group comprehensive optimization method for a virtual power plant according to claim 1, characterized in that: The specific process for determining bottleneck amplification distortion in operational monitoring data and historical operational statistics is as follows: The system acquires operational monitoring data and historical operational statistics. For the energy storage battery temperature data, it uses the sliding window quantile and MAD robust normalization algorithm to obtain the temperature deviation intensity. A second-order equivalent circuit model is established for the energy storage battery voltage, current, and temperature data. An extended Kalman filter online state estimation algorithm is used, and the health state deviation intensity is obtained through the sliding window quantile benchmark deviation. For the energy storage battery voltage and current data, a recursive least squares parameter identification algorithm with forgetting factor is used to estimate the equivalent internal resistance sequence, and the equivalent internal resistance deviation intensity is obtained through the sliding window quantile and MAD robust normalization algorithm. The actual power change rate value is obtained by using the sliding window first-order polynomial least squares slope estimation algorithm on the real-time active power data, and the deviation intensity of the power change rate is obtained by subtracting it from the allowable power change rate boundary and then using the quantile scale normalization method. The Euclidean norms of the above four deviation intensities are used as the bottleneck changes of individual equipment; real-time active power data are obtained from grid connection point voltage and grid connection point current data through power calculation, and a single-unit coherence matrix is ​​constructed using the cross-spectral coherence function estimation algorithm. The intra-group coupling amplification intensity is obtained by normalizing the spectral radius; historical bottleneck changes of individual equipment are calculated from historical operating statistics, tail samples are extracted using the peak threshold method, and the tail extreme value is obtained as the maximum historical bottleneck change value through the maximum likelihood estimation algorithm. Calculate the complement of the coupling amplification intensity within the group and take its reciprocal to obtain the coupling amplification reciprocal factor; calculate the natural logarithm of the bottleneck change for each individual device to obtain the individual bottleneck logarithmic amplitude; calculate the ratio of the individual bottleneck logarithmic amplitude to the historical maximum bottleneck change value and take its absolute value to obtain the individual bottleneck normalized amplitude; perform bottleneck sensitivity exponent operation on the individual bottleneck normalized amplitude to obtain the individual sensitivity amplitude term; calculate the ratio of the individual bottleneck logarithmic amplitude to the bottleneck mitigation coefficient, take its negative value, and perform exponential function operation to obtain the individual mitigation attenuation term; Summing the bottleneck mitigation items for each individual device within the resource group within the total number of individual devices yields the total bottleneck mitigation for the resource group. The bottleneck amplification distortion determination value is obtained by multiplying the inverse coupling amplification factor, the monomer slow-release attenuation term, and the sum of the bottleneck slow-release of the resource group.

5. The resource group comprehensive optimization method for a virtual power plant according to claim 1, characterized in that: The specific process of generating bottleneck suppression allocation markers, generating individual device lists, and outputting boundaries based on the bottleneck amplification distortion determination results is as follows: Real-time comparison of bottleneck amplification distortion judgment value and bottleneck amplification distortion judgment threshold. The bottleneck amplification distortion judgment threshold includes a primary judgment threshold and a secondary judgment threshold. When the bottleneck amplification distortion judgment value is less than the secondary judgment threshold, the output resource group equivalent output boundary and the resource group equivalent power change rate boundary are normal boundaries, and the output unit's executable margin, response uncertainty scale and coupling amplification strength are output. When the bottleneck amplification distortion judgment value is greater than or equal to the secondary judgment threshold and less than the primary judgment threshold, the equivalent power change rate boundary of the resource group is tightened, and the top m single devices are selected as the tail dominant single devices based on the bottleneck change amount of the single device. After selecting the number of clusters using the Bayesian information criterion, the cluster with the largest mean is obtained and the number of tail dominant single devices is determined. The ramp-up limiting command and the upper limit of discharge current constraint command are issued to the top m single devices, and the bottleneck suppression allocation mark and the list of ranked single devices are output. When the bottleneck amplification distortion judgment value is greater than or equal to the first-level judgment threshold, protective load reduction is performed on the tail-end dominant unit, and the resource group reconstruction trigger flag and the list of recommended isolated unit equipment are output. The equivalent output boundary of the resource group, the equivalent power change rate boundary of the resource group and the coupling amplification strength within the group are output before reconstruction.

6. The resource group comprehensive optimization method for a virtual power plant according to claim 1, characterized in that: The specific process of performing equivalent constraint shrinkage strength analysis based on the bottleneck amplification distortion determination result is as follows: The bottleneck saturation factor is obtained by calculating the ratio of the bottleneck amplification distortion judgment value to the sum of the bottleneck amplification distortion judgment values. For each dominant unit within the tail end range, the ratio of the bottleneck change to the historical maximum bottleneck change value is calculated, and the bottleneck sensitivity index is raised to the power of this ratio to obtain the unit sensitive amplification amount. The ratio of the bottleneck change to the bottleneck slow-release coefficient is incremented to obtain the unit slow-release inhibition factor. The ratio of the unit sensitive amplification amount to the unit slow-release inhibition factor is calculated and summed based on the tail end dominant unit quantity to obtain the tail end bottleneck aggregation amount. The ratio of the tail end bottleneck aggregation amount to the tail end dominant unit quantity is calculated to obtain the tail end bottleneck average amount. The bottleneck saturation factor, the inverse coupling amplification factor, and the tail end bottleneck average amount are multiplied to obtain the bottleneck perception equivalent constraint contraction value.

7. The resource group comprehensive optimization method for a virtual power plant according to claim 1, characterized in that: The specific process of boundary shrinkage, distributed suppression constraint injection, and output plan solution based on equivalent constraint shrinkage strength analysis is as follows: Real-time comparison of equivalent constraint shrinkage value and equivalent constraint shrinkage threshold, which includes primary shrinkage threshold and secondary shrinkage threshold: When the equivalent constraint contraction value is less than the second-level contraction threshold, the equivalent output boundary and the equivalent power change rate boundary of the resource group are used as the optimization constraint boundary. Under the state where the bottleneck suppression allocation mark is empty, mixed integer programming is used to process the discrete decision of the resource availability state. The continuous output allocation within the resource group is solved by the distributed alternating direction multiplier method, and the output of the resource group output plan and the individual allocation instruction set are output. When the equivalent constraint contraction value is greater than or equal to the second-level contraction threshold and less than the first-level contraction threshold, the contraction equivalent constraint comprehensive optimization mode is adopted to perform contraction on the resource group equivalent output boundary and the resource group equivalent power change rate boundary. The bottleneck suppression allocation mark and the tail dominant unit list are used as allocation constraint inputs, and the resource group output plan and unit allocation instruction set under contraction constraints are output. When the equivalent constraint contraction value is greater than or equal to the first-level contraction threshold, the optimization objective is switched to prioritize the power availability of the tie line plan and the system-level power balance availability. The resource group reconfiguration trigger flag and the list of recommended isolated unit equipment are used as inputs to the solution mode. The amortization solution is only performed on the set after the list of recommended isolated unit equipment is removed. The output is the feasible domain-priority resource group output plan, the unit amortization instruction set, the list of recommended isolated unit equipment, and the reconfiguration coordination flag.

8. The resource group comprehensive optimization method for a virtual power plant according to claim 1, characterized in that: The specific process for coupling, amplifying, pushing, correcting, and evaluating the operational monitoring data is as follows: Real-time active power data is obtained by calculating the voltage and current data at the grid connection point; the planned active power value is read from the output plan of the resource group; the equivalent output margin of the resource group is obtained by subtracting the planned active power value from the upper limit of the equivalent output of the contracted resource group and the lower limit of the equivalent output of the contracted resource group from the planned active power value. Calculate the difference between real-time active power data and planned active power value and take the absolute value to obtain the execution deviation amplitude; calculate the ratio of the execution deviation amplitude to the resource group equivalent output margin plus numerical stability factor to obtain the margin execution deviation term; calculate the equivalent constraint contraction value of bottleneck perception to obtain the constraint contraction amplification term; calculate the product of the coupling amplification reciprocal factor, the margin execution deviation term and the constraint contraction amplification term to obtain the bottleneck closed-loop execution correction judgment value.

9. The resource group comprehensive optimization method for a virtual power plant according to claim 1, characterized in that: The specific process of triggering rolling correction, reallocating individual amortized instruction sets, triggering resource group reconstruction, and replacing resources based on the evaluation results of coupling amplification push correction is as follows: Real-time comparison of correction judgment values ​​and correction judgment thresholds, including primary judgment thresholds and secondary judgment thresholds: When the correction judgment value is less than the secondary judgment threshold, keep the resource group output plan and the individual allocation instruction set unchanged, create a daily execution database, and output the execution deviation magnitude and the normal execution status mark and archive it to the daily execution database. When the correction judgment value is greater than or equal to the secondary judgment threshold and less than the primary judgment threshold, rolling correction is triggered: the rolling correction trigger flag and correction direction flag are output, the resource group output plan of the next rolling cycle is corrected in the same direction according to the execution deviation magnitude and the corrected resource group output plan is generated; the unit allocation instruction set is reallocated according to the bottleneck suppression allocation flag and the tail dominant unit list and the corrected unit allocation instruction set is generated. When the correction judgment value is greater than or equal to the first-level judgment threshold, output the resource group reconstruction trigger flag and execute the isolation action in the isolated single device list; execute the fast resource replacement rule on the isolated resource group member set to introduce available single devices and form the reconstructed resource group member set; output the update input of the reconstructed resource group equivalent output boundary and the resource group equivalent power change rate boundary and write it back; output the starting point of the reconstructed scheduling trajectory.

10. A resource group comprehensive optimization system for a virtual power plant, employing the resource group comprehensive optimization method for a virtual power plant according to any one of claims 1-9, comprising: The data acquisition and preprocessing module is used to collect operational monitoring data and obtain historical operational statistics. Preprocessing of operation monitoring data and historical operation statistics; The bottleneck identification and impact factor quantification module is used to determine the bottleneck amplification distortion of operation monitoring data and historical operation statistics, and to generate bottleneck suppression allocation markers, individual equipment lists and boundary outputs based on the bottleneck amplification distortion determination results. The resource group equivalent constraint modeling and comprehensive optimization module is used to perform equivalent constraint shrinkage strength analysis based on the bottleneck amplification distortion judgment results, and to perform boundary shrinkage, amortization suppression constraint injection and output plan solution based on the equivalent constraint shrinkage strength analysis. The rolling execution correction and resource group reconstruction module is used to perform coupled amplification and push correction assessment on operation monitoring data, and to trigger rolling correction, reallocate individual amortization instruction sets, trigger resource group reconstruction, and replace resources based on the results of the coupled amplification and push correction assessment.

Images