A multi-virtual power plant coordination configuration method based on sensitivity partitioning

Abstract

This invention discloses a multi-virtual power plant coordinated configuration method based on sensitivity partitioning. The method first constructs a multi-virtual power plant system model to calculate the power interaction sensitivity matrix, and then uses a clustering algorithm to dynamically partition the virtual power plants based on a comprehensive regulation potential index. Second, based on the partitioning results, a two-layer distributed coordinated optimization model is constructed: the upper-layer central coordinator optimization model generates power commands, and the lower-layer virtual power plant robust optimization model optimizes internal resources accordingly. Subsequently, a distributed iterative algorithm is used to solve the problem, and the intermediate solutions are constrained to obtain a preliminary solution. Finally, the operating status is verified: if the correction conditions are met, voltage correction is performed; otherwise, the system returns to repartitioning until convergence and the final solution is output. This invention solves the problems of low coordination efficiency due to fixed partitioning in existing scheduling methods, difficulty in balancing global economic efficiency and local uncertainty security, and the tendency for large-scale optimization calculations to diverge and lack dynamic adaptive capabilities.

Description

Technical Field

[0001] This invention relates to the field of new power distribution network optimization operation, and in particular to a method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning. Background Technology

[0002] Virtual power plants (VPPs), as the core carriers aggregating diverse resources such as distributed power sources, shared energy storage, and flexible loads, have become a key technology for bridging the gap in grid regulation capacity and supporting supply-demand balance due to their "observable, measurable, and controllable" characteristics. However, existing multi-VPP coordinated configuration and optimized operation technologies still have many key shortcomings, making it difficult to meet the economic, security, and flexibility requirements of high-proportion renewable energy grids.

[0003] With the deep integration of the digital economy and the new power system, diversified resources such as data centers, distributed energy storage, and flexible loads on the user side are being deployed on a large scale on the distribution network side, transforming the distribution network from the traditional "passive power reception" to "active regulation through source-load interaction." The power consumption / output characteristics of these resources have strong spatiotemporal differences. For example, the computing load of data centers fluctuates between day and night, and the output of distributed photovoltaics varies with sunlight, leading to frequent reconfiguration of the power flow distribution in the distribution network. Traditional power flow control methods based on fixed time periods and fixed paths, such as tie-line power setpoint control and static load transfer strategies, cannot adapt to dynamic power flow changes, increasing the risk of overload on distribution network lines. At the same time, the flexible adjustment potential of distributed resources has not been fully explored, and the contradiction between the economic efficiency and security of system operation is becoming increasingly prominent.

[0004] Faced with this challenge, simply increasing transmission capacity by expanding distribution network lines would not only result in high investment costs and long construction periods, but also fail to fundamentally solve the problem of dynamic power flow fluctuations, which contradicts the development concept of "efficiently utilizing existing resources" in distribution networks. In contrast, under the single resource autonomous control model, data centers, energy storage, and other entities adjust based solely on their local operating status, which can easily lead to insufficient power flow complementarity between regions. For example, when energy storage in one region is fully utilized and idle, another region may still need to curtail solar power due to insufficient energy storage capacity, resulting in resource waste.

[0005] Against this backdrop, collaborative power flow optimization of distribution networks with multiple resources has become a key direction for overcoming bottlenecks. However, these multiple resources belong to different stakeholders, such as data centers belonging to internet companies and energy storage belonging to third-party operators. There are privacy protection requirements for information such as operating data and cost parameters. Traditional centralized power flow optimization needs to obtain global information of all resources, which not only faces communication bandwidth pressure and the need to transmit massive amounts of operating data in real time, but also poses a risk of commercial information leakage. At the same time, the adjustment capabilities of multiple resources are time-varying. For example, the state of charge of energy storage changes with charging and discharging, and the migration of computing load in data centers has time constraints. Existing collaborative strategies mostly use static optimization models, which cannot dynamically match the real-time adjustment boundaries of resources. This leads to a disconnect between power flow optimization schemes and actual operating conditions, and repeated problems of line overload or resource idleness.

[0006] Therefore, developing a multi-virtual power plant coordination optimization method based on sensitivity partitioning, which can protect the information privacy of each resource subject while dynamically adapting the real-time adjustment capability of resources, and achieve the optimal power flow distribution of the distribution network with the lowest operating cost, is an urgent problem to be solved in the current technical field. Summary of the Invention

[0007] To address the aforementioned shortcomings of existing technologies, this invention provides a multi-virtual power plant coordinated configuration method based on sensitivity partitioning. By integrating dynamic partitioning technology that combines power interaction sensitivity and regulation potential, dual-layer distributed coordinated modeling, and an iterative correction mechanism with constraint processing, this invention solves the technical problems in existing virtual power plant scheduling, such as low coordination efficiency due to fixed partitioning, difficulty in balancing global economic efficiency and local uncertainty security, and the tendency for large-scale optimization calculations to diverge and lack of dynamic adaptive capabilities.

[0008] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0009] A method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning includes the following steps:

[0010] S1. Construct a multi-virtual power plant system model, calculate the power interaction sensitivity matrix between virtual power plants based on the multi-virtual power plant system model, process the power interaction sensitivity matrix and the comprehensive regulation potential index using a clustering algorithm, dynamically partition the multiple virtual power plants, and output the partitioning results.

[0011] S2. Based on the partitioning results, a two-layer distributed coordination optimization model is constructed, including an upper-layer central coordinator optimization model and a lower-layer virtual power plant robust optimization model. The upper-layer central coordinator optimization model generates power commands and optimizes internal resources according to the power commands through the lower-layer virtual power plant robust optimization model.

[0012] S3. The two-layer distributed coordination optimization model is solved by a distributed iterative algorithm, and the intermediate solutions of the iteration are constrained during the iteration process to obtain the preliminary scheduling scheme and iteration state information.

[0013] S4. Perform operation status verification based on the preliminary scheduling scheme and iteration status information. If the preset correction conditions are met, perform voltage correction operation; otherwise, return to step S1 until the convergence conditions are met and the final configuration scheme is output.

[0014] As a preferred embodiment, in step S1, the multi-virtual power plant system model includes:

[0015] The mathematical expression for the distributed power generation model is as follows:

[0016] ;

[0017] In the formula: Let be the active and reactive power outputs of the semi-controlled power supply i at time t, respectively. and Let be the minimum and maximum power factor angles of the semi-controlled power supply i, respectively. Let i be the maximum active power output that the semi-controlled power supply can provide at time t. Let be the probability that the event is true. For the confidence level of opportunity constraints, For a given reactive power quantile function under the given conditions For random variables The inverse function of the cumulative distribution function;

[0018] The mathematical expression for the shared energy storage model is as follows:

[0019] ;

[0020] In the formula: Virtual power plants The allocated energy storage capacity and power capacity, For virtual power plants The shared costs of energy storage that should be allocated This refers to the set of all virtual power plants participating in shared energy storage. Let S be any subset of the alliance that does not contain virtual power plant i, and let C(S) be the cost function of the alliance S.

[0021] The mathematical expression for the flexible load model is as follows:

[0022] ;

[0023] Where: Where: The computational load of virtual power plant i during time period t after migration. Virtual power plant before migration The computational load, For virtual power plants Towards Migration load, For virtual power plants Maximum load receiving capacity;

[0024] The mathematical expression for the network power flow model is as follows:

[0025] ;

[0026] In the formula: , and These represent the active power output of the nth thermal power unit at node m in the virtual power plant i, the predicted active power output of the wind turbine unit, and the base load power, respectively. They are nodes It is a collection of thermal power units, virtual power plants, and wind power. For nodes Voltage phase angle, Let be the voltage phase angle of node j adjacent to node m. For the line Reactance , Virtual power plants Middle node First The charging and discharging power of each energy storage unit.

[0027] As a preferred embodiment, in step S1, the power interaction sensitivity matrix is ​​derived from the Lagrange function;

[0028] The power interaction sensitivity matrix is ​​represented as follows:

[0029] ;

[0030] In the formula: For the power interaction sensitivity matrix, For the subproblem's Lagrangian function, These are the optimal dual variables corresponding to the equality constraints. Let be the optimal dual variable corresponding to the inequality constraint. Coupled decision parameters are issued by the upper-level coordinator and treated as known in the sub-problems;

[0031] The Lagrange function is expressed as:

[0032] ;

[0033] In the formula: Let be the objective function. For equality constraint functions, is the inequality constraint function.

[0034] As a preferred embodiment, in step S1, the specific processing steps of using a clustering algorithm to process the power interaction sensitivity matrix and the comprehensive regulation potential index, dynamically partitioning multiple virtual power plants, and outputting the partitioning results include:

[0035] S101. The power interaction sensitivity matrix is ​​normalized to obtain the normalized coupling degree matrix, which is expressed as:

[0036] ;

[0037] In the formula: This is the normalized coupling degree matrix; The magnitude of the impact on the operating cost or power balance state of virtual power plant j when the power decision variable of virtual power plant i changes by a unit;

[0038] S102. Based on the normalized coupling matrix and combined with the comprehensive regulation potential index of each virtual power plant, construct the sample feature vector of each virtual power plant.

[0039] S103. The elbow rule and the contour coefficient method are used to analyze the feature vector of the sample to determine the optimal number of clusters K;

[0040] S104. Initialize K cluster centers, calculate the Euclidean distance from each sample feature vector to each cluster center, divide the sample into the nearest cluster, and update each cluster center to the average value of all sample feature vectors in that cluster, until the cluster centers converge, and output the partitioning result.

[0041] As a preferred option, step S1 also includes a dynamic partition update mechanism, the specific processing steps of which include:

[0042] The system monitors its operating status in real time. When the preset dynamic update triggering conditions are met, the power interaction sensitivity matrix calculation and clustering algorithm are re-executed to update the partitioning results. After updating the partitioning results, the rationality of the new partitioning is verified. If the verification is successful, the new partitioning results are sent to the upper-level central coordinator optimization model. If the verification fails, the clustering parameters are adjusted and the clustering process is re-executed until the partitioning results meet the collaborative scheduling requirements.

[0043] The dynamic update triggering conditions include: the fluctuation range of new energy output exceeds a preset fluctuation threshold; the standard deviation of the comprehensive regulation potential index of virtual power plants in the region exceeds a preset complementarity threshold; the proportion of cross-regional power interaction to the total regional output exceeds a preset interference threshold; and a fixed time period trigger point is reached.

[0044] The verification of the rationality of the new partition specifically includes: calculating the collaborative efficiency index within the region and the interference index between regions; when the collaborative efficiency index within the region is greater than or equal to the efficiency threshold and the interference index between regions is less than or equal to the interference threshold, the verification is deemed successful.

[0045] As a preferred embodiment, in step S2, the specific processing procedure of the upper-level central coordinator optimization model includes:

[0046] With the goal of minimizing the total operating cost of the entire network, a power command including tie-line power reference values ​​and standby capacity configuration instructions is generated by combining the power interaction sensitivity matrix and partitioning results.

[0047] The objective function for the total operating cost of the entire network is expressed as:

[0048] ;

[0049] In the formula: To minimize the total network operating cost for the upper-level central coordinator, Let I be the time set, and let I be the set of IDs for all virtual power plants within the distribution network. This represents the power generation cost of virtual power plant i during time period t, including the power generation costs of thermal power units and distributed power sources. The shared energy storage usage cost of virtual power plant i during time period t. The calculation of load migration cost for virtual power plant i during time period t. The penalty cost for wind and solar curtailment of virtual power plant i during time period t;

[0050] The constraints of the objective function for the total operating cost of the entire network include:

[0051] Inter-regional power balance constraints:

[0052] ;

[0053] In the formula: For adjacent regions, The power transmitted by the inter-regional tie line of virtual power plant i during time period t;

[0054] Tether line transmission capacity constraints:

[0055] ;

[0056] In the formula: The minimum transmission power of tie line l, Here, l represents the maximum transmission power of tie line l, and l is the number of all tie lines between areas.

[0057] Constraints on the adequacy of network-wide backup capacity:

[0058]

[0059] In the formula: This represents the uplink reserve capacity of virtual power plant i during time period t. To meet the uplink backup requirements of the system during time period t. This represents the downlink reserve capacity of virtual power plant i during time period t. This is for the system's downlink backup requirements during time period t.

[0060] As a preferred embodiment, in step S2, the specific processing procedure of the robust optimization model of the lower-level virtual power plant includes:

[0061] Each virtual power plant receives tie-line power reference values ​​and reserve capacity configuration instructions from the upper-level central coordinator model, and constructs a two-stage robust optimization model with the objectives of minimizing the total day-ahead scheduling cost and minimizing the total intraday scheduling cost, in order to optimize the scheduling of internal controllable resources.

[0062] The objective function that minimizes the total day-ahead scheduling cost is:

[0063] ;

[0064] In the formula: Let i be the day-ahead dispatch total cost. Let i be the power generation cost of the thermal power unit of the virtual power plant during time period t. Let i be the start-up and shutdown cost of the thermal power unit in the virtual power plant during time period t. Reserved standby costs for the thermal power units of virtual power plant i during time period t. The charging and discharging cost of virtual power plant i during time period t. Reserved backup costs for virtual power plant i during time period t;

[0065] The objective function that minimizes the total cost of intraday scheduling is:

[0066] ;

[0067] In the formula: Let i be the total daily dispatch cost of virtual power plant i. The standby call cost for the thermal power unit of virtual power plant i during time period t. The load shedding penalty cost for virtual power plant i during time period t;

[0068] The constraints of the two-stage robust optimization model include:

[0069] Distributed power generation output constraints:

[0070] ;

[0071] In the formula: For the set of thermal power units within virtual power plant i, Let n be the minimum output of thermal power unit n. This represents the maximum output of thermal power unit n.

[0072] Energy storage charge / discharge power and state of charge constraints:

[0073] ;

[0074] ;

[0075] In the formula: The energy storage charging power of virtual power plant i during time period t. The maximum charge and discharge power of the energy storage of virtual power plant i. Let i be the energy storage discharge power of the virtual power plant during time period t. The state of charge of the energy storage of virtual power plant i during time period t. This is the minimum state of charge for energy storage. This represents the maximum state of charge of the energy storage.

[0076] Load migration time constraints:

[0077] ;

[0078] In the formula: This represents the actual delay time for the batch processing load. This is the maximum allowable delay time for batch processing load;

[0079] Node voltage and line power constraints:

[0080] ;

[0081] ;

[0082] In the formula: Let m be the actual voltage value of the i-th node at the m-th voltage monitoring point during time period t. and The lower and upper voltage limits of the m-th voltage monitoring point associated with the i-th node are respectively defined.

[0083] As a preferred embodiment, in step S3, the specific processing steps for solving the two-layer distributed coordinated optimization model using a distributed iterative algorithm include:

[0084] S301. Dynamically adjust the penalty parameter using an adaptive step size strategy; the adjustment formula for the penalty parameter is:

[0085] ;

[0086] In the formula: Let be the penalty parameter for the nth iteration. For the original residual, For dual residuals, All of these are parameter adjustments;

[0087] S302. The iterative intermediate solution is projected onto the feasible region of the virtual power plant using a feasible region projection mechanism; the feasible region is solved by vertex enumeration and satisfies the following active-reactive power regulation capability constraints:

[0088] ;

[0089] In the formula: For the feasible domain of the virtual power plant, and They are respectively the current active and reactive power outputs. and These are the power increments;

[0090] S303. Introduce a binary consistency variable for active and reactive power, and achieve consistent cost increment rates for each virtual power plant through iterative updates; the update rule is as follows:

[0091] ;

[0092] In the formula: Let be the active consistency variable after the (k+1)th iteration of the i-th VPP. Let be the reactive power consistency variable after the (k+1)th iteration of the i-th VPP. The coefficients are the state transition matrix coefficients. All are power deviations. All are consistency adjustment coefficients;

[0093] S304. Generate a preliminary scheduling scheme based on the projected iterative intermediate solution, and record the iterative residual as iterative state information.

[0094] As a preferred embodiment, step S4, the specific processing procedure for verifying the running status based on the preliminary scheduling scheme and iteration status information, includes:

[0095] S401. Monitor the operating status of each virtual power plant in real time to determine whether there is any power or voltage exceeding the limit.

[0096] S402. If there is a power limit violation, perform a power limit violation correction operation.

[0097] S403. If a voltage limit violation exists, perform a voltage limit violation correction operation.

[0098] S404. If the collaborative efficiency within the region decreases, calculate the difference in the incremental cost rate. When the difference in the incremental cost rate is greater than the preset growth rate threshold, repeat step S1.

[0099] S405. Check if the convergence condition is met. If it is, output the final configuration scheme.

[0100] As a preferred embodiment, step S402, the power over-limit correction operation, includes: adjusting the over-limit virtual power plant output to the boundary of the virtual power plant's feasible region according to the equal power factor principle, wherein the corrected active and reactive power outputs satisfy:

[0101] ;

[0102] In the formula: This represents the corrected active power output of the i-th VPP after exceeding its power limit during time period t. Let be the active power boundary value of the virtual power plant feasible region for the i-th VPP in time period t. Let be the rated power factor angle of the i-th VPP. The corrected reactive power output of the i-th VPP after exceeding the power limit in time period t;

[0103] In step S403, the voltage over-limit correction operation includes: calculating the node voltage change based on the voltage sensitivity model.

[0104] ;

[0105] In the formula: Let be the voltage change at node m. This represents the total number of VPPs within the distribution network. These are the active and reactive power adjustments for the i-th VPP, respectively. For the set of routes, These represent the resistance and reactance of line j, respectively. Let be the voltage amplitude at endpoint j of the line;

[0106] In step S404, the verification of whether the convergence condition is met includes: calculating the global maximum relative residual.

[0107] ;

[0108] In the formula: δ is the global maximum relative residual. These are the variable values ​​for the t-th and t-1-th iterations, respectively. These represent the active and reactive power outputs of VPP, respectively. For energy storage charging and discharging power, This refers to the power of the tie line.

[0109] Compared with the prior art, the present invention has the following technical effects:

[0110] 1. Compared with existing fixed partitioning strategies, this invention quantifies electrical coupling based on power interaction sensitivity matrix and reflects the differences in resource regulation capabilities of each entity by combining comprehensive regulation potential index. The above two are used as inputs to the clustering algorithm for dynamic partitioning, decomposing the large-scale global optimization problem into several medium-sized regional sub-problems. This effectively reduces the solution dimension and computational burden of subsequent optimization models for large-scale systems. It can also adaptively adjust the partitioning structure according to the real-time operating status of the power grid and source-load fluctuations, enhancing the system's adaptability to the uncertainty of high proportion of new energy sources.

[0111] 2. To address the shortcomings of existing solutions that focus on single resources, this invention constructs a two-layer distributed coordination and optimization model of "central coordination - regional autonomy." At the upper layer, a central coordinator generates power commands with the goal of maximizing global economy or absorption capacity, ensuring overall system optimization. At the lower layer, a robust optimization model is introduced to address the uncertainties in distributed power output and load forecasting, ensuring that the resources within the virtual power plant still meet safety constraints even in the worst-case scenario. This effectively balances global scheduling economy with local operational safety. Furthermore, this distributed architecture eliminates the need for lower-level virtual power plants to upload sensitive internal parameters such as energy storage SOC and detailed load curves; they only need to exchange power commands and boundary information. This achieves efficient collaborative scheduling while ensuring data privacy protection and independent operational rights for all parties involved.

[0112] 3. For high-dimensional nonlinear problems involving the joint optimization of multiple virtual power plants, this invention employs distributed iterative algorithms (such as ADMM and its improved versions) to decompose complex problems into parallel-computable subproblems, significantly reducing the computational burden of a single iteration. Simultaneously, a constraint handling mechanism (such as feasible region projection) is introduced during the iteration process, forcing each intermediate solution to fall within the physical feasible region of the virtual power plant, thus avoiding the intermediate solution exceeding the limit problem in traditional algorithms. Furthermore, this invention supports synchronous iterative optimization of active and reactive power variables, breaking the traditional decoupling mode of "active power first, reactive power later," and achieving coordinated control of voltage support and power balance, further improving power quality and system regulation accuracy.

[0113] 4. This invention implements a hierarchical response mechanism by rigorously verifying the operational status of the initial scheduling scheme: when local risks such as voltage exceeding limits are detected, voltage correction operations are prioritized to quickly eliminate hidden dangers; when conventional corrections cannot resolve structural blockages or significantly reduce coordination efficiency, a re-partitioning mechanism is triggered to return to step S1. This dynamic feedback strategy not only ensures that the globally optimal configuration scheme can always be obtained under complex and variable operating conditions, but also guarantees that the final output scheme strictly satisfies physical constraints such as voltage and power flow while achieving mathematical convergence, significantly improving the engineering practicality of the scheduling scheme and the safety and reliability of virtual power plants participating in grid scheduling. Attached Figure Description

[0114] To make the objectives, technical solutions, and advantages of the invention clearer, the invention will now be described in further detail with reference to the accompanying drawings, wherein:

[0115] Figure 1 This is a flowchart of a multi-virtual power plant coordinated configuration method based on sensitivity partitioning disclosed in this invention;

[0116] Figure 2 This is a flowchart illustrating the distributed resource aggregation and optimized scheduling process in an embodiment of the present invention.

[0117] Figure 3 This is a flowchart illustrating the specific process of monitoring and dynamically correcting the operating status in an embodiment of the present invention. Detailed Implementation

[0118] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0119] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to represent selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0120] It should be noted that similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the figures, or the orientation or positional relationship commonly used when the product is in use. They are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance. In addition, the terms "horizontal," "vertical," etc., do not indicate that the component is required to be absolutely horizontal or suspended, but can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted. In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0121] Example:

[0122] With the deepening of the construction of new power systems, the aggregation of massive distributed resources in the form of virtual power plants (VPPs) to participate in grid dispatch has become an inevitable trend. However, current multi-virtual power plant collaborative dispatch technology faces severe challenges: First, traditional dispatch modes are mostly based on fixed geographical partitions or administrative affiliations, ignoring the differences in the dynamic electrical coupling strength and regulation potential between virtual power plants. This leads to low coordination efficiency within partitions when source and load fluctuate drastically, and it is difficult to adapt to the uncertainties brought about by the high proportion of new energy access. Second, existing optimization methods often fail to balance global economy and local security. Centralized solutions face the "curse of dimensionality" and the risk of data privacy leakage, while conventional distributed algorithms are prone to generating intermediate solutions that violate physical constraints during iteration, leading to convergence difficulties or even divergence. Third, existing technical solutions lack effective dynamic correction mechanisms. Once the initial partitioning strategy fails or safety hazards such as voltage exceeding limits occur, it is often impossible to adaptively adjust the topology or perform precise correction, making it difficult to implement the dispatch scheme in actual engineering. Based on the aforementioned problems and shortcomings, this invention proposes a multi-virtual power plant coordinated configuration method based on sensitivity partitioning. Through dynamic partitioning clustering, a two-layer robust optimization architecture, an improved distributed solution algorithm, and a sensitivity feedback correction mechanism, it achieves efficient coordination of multi-virtual power plant resources. While protecting the privacy of information of each resource subject, it dynamically adapts the real-time adjustment capability of resources, thereby improving the economy, security, and operational flexibility of high-proportion renewable energy power grids.

[0123] Specifically, the multi-virtual power plant coordinated configuration method based on sensitivity partitioning proposed in this invention, such as... Figure 1 As shown, it includes the following steps:

[0124] S1. Construct a multi-virtual power plant system model, calculate the power interaction sensitivity matrix between virtual power plants based on the multi-virtual power plant system model, process the power interaction sensitivity matrix and the comprehensive regulation potential index using a clustering algorithm, dynamically partition the multiple virtual power plants, and output the partitioning results.

[0125] In this embodiment, the virtual power plant aggregates various distributed resources, such as distributed power sources and demand-side resources, within a certain area to form a virtual power plant (VPP). The power grid coordinates and controls the VPP, enabling it to participate in grid dispatching and market transactions, thereby enhancing the power system's flexible adjustment capabilities. The VPP-based operational control mode can provide the upper-level dispatch layer with the VPP's adjustment capability boundaries, while also effectively protecting the privacy of individual users within the VPP. Precise mathematical models are established for the different operational characteristics of distributed power sources, shared energy storage, and flexible data center loads, achieving comprehensive aggregation of diverse resources.

[0126] Distributed power generation model: A semi-controlled power generation model considering the uncertainty of wind and solar power output is constructed based on the Gaussian mixture module (GMM). The chance constraints are transformed into deterministic constraints through quantile transformation, facilitating model solution. Its mathematical expression is as follows:

[0127] ;

[0128] In the formula: Let be the active and reactive power outputs of the semi-controlled power supply i at time t, respectively. and Let be the minimum and maximum power factor angles of the semi-controlled power supply i, respectively. Let i be the maximum active power output that the semi-controlled power supply can provide at time t. Let be the probability that the event is true. For the confidence level of opportunity constraints, For a given reactive power quantile function under the given conditions For random variables The inverse function of the cumulative distribution function;

[0129] The mathematical expression for the shared energy storage model is as follows:

[0130] ;

[0131] In the formula: Virtual power plants The allocated energy storage capacity and power capacity, For virtual power plants The shared costs of energy storage that should be allocated This refers to the set of all virtual power plants participating in shared energy storage. Let S be any subset of the alliance that does not contain virtual power plant i, and let C(S) be the cost function of the alliance S.

[0132] The mathematical expression for the flexible load model is as follows:

[0133] ;

[0134] In the formula: The computational load of virtual power plant i during time period t after migration. Virtual power plant before migration The computational load, For virtual power plants Towards Migration load, For virtual power plants Maximum load receiving capacity;

[0135] Network power flow model: The DistFlow model is used to describe the topological constraints of the distribution network, clarifying the power balance of nodes and the transmission limitations of lines, and adapting to the radial topology characteristics of the distribution network. Its mathematical expression is as follows:

[0136] ;

[0137] In the formula: , and These represent the active power output of the nth thermal power unit at node m in the virtual power plant i, the predicted active power output of the wind turbine unit, and the base load power, respectively. They are nodes It is a collection of thermal power units, virtual power plants, and wind power. For nodes Voltage phase angle, Let be the voltage phase angle of node j adjacent to node m. For the line Reactance , Virtual power plants Middle node First The charging and discharging power of each energy storage unit.

[0138] In this embodiment, the power interaction sensitivity matrix between virtual power plants is derived based on the Lagrange function, quantifying the impact of changes in the operating status of different virtual power plants on other power plants and the entire network, providing a core basis for dynamic zoning; the power interaction sensitivity matrix is ​​expressed as:

[0139] ;

[0140] In the formula: For the power interaction sensitivity matrix, For the subproblem's Lagrangian function, These are the optimal dual variables corresponding to the equality constraints. Let be the optimal dual variable corresponding to the inequality constraint. Coupled decision parameters are issued by the upper-level coordinator and treated as known in the sub-problems;

[0141] The Lagrange function is expressed as:

[0142] ;

[0143] In the formula: Let be the objective function. For equality constraint functions, is the inequality constraint function.

[0144] In this embodiment, a clustering algorithm is used to dynamically partition the virtual power plant based on electrical distance and regulation potential indicators. The dynamic partitioning is achieved using the partial derivative of the Lagrangian function with respect to the power decision variables of the virtual power plant, with electrical coupling degree and regulation potential as dual indicators, and employing the K-means clustering algorithm. Specifically:

[0145] S101, the power interaction sensitivity matrix λ calculated in step S1 * Based on this, the degree of electrical coupling between virtual power plants is quantified. Sensitivity matrix λ * It is an n×n square matrix, where n is the total number of virtual power plants, and the element λ ij * λ represents the magnitude of the impact on the operating cost or power balance state of virtual power plant j when the power decision variable of virtual power plant i changes by a unit. ij * The larger the absolute value, the higher the electrical coupling between the two components, and the more significant the impact of power interaction on their respective operating states. To eliminate the influence of dimensions, the sensitivity matrix is ​​normalized to obtain the normalized coupling matrix, which is expressed as:

[0146] ;

[0147] In the formula: This is a normalized coupling degree matrix; the closer it is to 1, the tighter the coupling. The magnitude of the impact on the operating cost or power balance state of virtual power plant j when the power decision variable of virtual power plant i changes by a unit;

[0148] S102. Based on the normalized coupling matrix and combined with the comprehensive regulation potential index of each virtual power plant, construct the sample feature vector of each virtual power plant. The first n elements are the normalized coupling degree of the virtual power plant with all other virtual power plants, and the last element is the comprehensive regulation potential index.

[0149] S103. The elbow rule and the contour coefficient method are used to analyze the feature vector of the sample to determine the optimal number of clusters K;

[0150] In practice, the elbow rule uses the inflection point of the sum of squares within a cluster (WCSS) as a function of K as a candidate K value, and the silhouette coefficient method uses the maximum value of the silhouette coefficient (within the range of [-1, 1], the closer to 1, the better the clustering effect) to verify the optimal K value, so as to ensure that the number of clusters meets the requirements of collaborative scheduling and avoids the region division being too fine or too coarse.

[0151] S104. Initialize K cluster centers, calculate the Euclidean distance from each sample feature vector to each cluster center, divide the sample into the nearest cluster, and update each cluster center to the average value of all sample feature vectors in that cluster, until the cluster centers converge, and output the partitioning result.

[0152] The clustering objective function is:

[0153] ;

[0154] In the formula: K is the number of candidate clusters, i.e. the number of collaboratively regulated regions; Let VPP be the set of VPPs contained in the k-th cluster; Let be the sample feature vector of the i-th VPP; Let the center vector of the k-th cluster be denoted as 'k'. The total clustering error for the K regions;

[0155] The formula for updating each cluster center to the average of the feature vectors of all samples within that cluster is as follows:

[0156] ;

[0157] In the formula: It is the k-th cluster center in the (t+1)-th iteration; Let VPP be the number of regions in the k-th region during the t-th iteration; In the t-th iteration, the summation of the feature vectors of all samples in the k-th region.

[0158] Furthermore, to adapt to dynamic changes in system operating status, such as fluctuations in new energy output, changes in load levels, and updates to the SOC status of energy storage, a dynamic partitioning update mechanism is established to ensure that the partitioning strategy is always in an optimal adaptation state. Specifically, the system operating status is monitored in real time. When the preset dynamic update trigger conditions are met, the power interaction sensitivity matrix calculation and clustering algorithm are re-executed to update the partitioning results. After updating the partitioning results, the rationality of the new partitioning is verified. If the verification is successful, the new partitioning results are sent to the upper-level central coordinator optimization model. If the verification fails, the clustering parameters are adjusted and the clustering process is re-executed until the partitioning results meet the collaborative scheduling requirements.

[0159] The dynamic update triggering conditions include: the fluctuation range of new energy output exceeds a preset fluctuation threshold (10% in this embodiment); the standard deviation of the comprehensive regulation potential index of virtual power plants in the region exceeds a preset complementarity threshold (0.15 in this embodiment); the proportion of cross-regional power interaction to the total regional output exceeds a preset interference threshold (20% in this embodiment); and a fixed time period triggering point is reached (in this embodiment, the sensitivity matrix is ​​updated every hour, and the clustering optimization is performed every 4 hours).

[0160] The verification of the rationality of the new partition specifically includes: calculating the collaborative efficiency index within the region and the inter-regional interference index; when the collaborative efficiency index within the region is greater than or equal to an efficiency threshold, and the inter-regional interference index is less than or equal to an interference threshold, the verification is deemed successful. In this embodiment, the collaborative efficiency index within the region is 0.8, and the inter-regional interference index is 0.2.

[0161] Compared to existing fixed partitioning strategies, this embodiment quantifies electrical coupling based on a power interaction sensitivity matrix and combines a comprehensive regulation potential index to reflect the differences in resource regulation capabilities of each entity. By using both of these as inputs to a clustering algorithm for dynamic partitioning, the large-scale global optimization problem is decomposed into several medium-sized regional sub-problems. This effectively reduces the solution dimensionality and computational burden of subsequent optimization models for large-scale systems. Furthermore, it can adaptively adjust the partitioning structure according to the real-time operating status of the power grid and source-load fluctuations, enhancing the system's adaptability to the uncertainties of high-proportion renewable energy sources.

[0162] S2. Based on the partitioning results, a two-layer distributed coordination optimization model is constructed, including an upper-layer central coordinator optimization model and a lower-layer virtual power plant robust optimization model. The upper-layer central coordinator optimization model generates power commands and optimizes internal resources according to the power commands through the lower-layer virtual power plant robust optimization model.

[0163] In this embodiment, the specific processing steps of the upper-level central coordinator optimization model include:

[0164] With the goal of minimizing the total operating cost of the entire network, a power command including tie-line power reference values ​​and standby capacity configuration instructions is generated by combining the power interaction sensitivity matrix and partitioning results.

[0165] The objective function for the total operating cost of the entire network is expressed as:

[0166] ;

[0167] In the formula: To minimize the total network operating cost for the upper-level central coordinator, Let I be the time set, and let I be the set of IDs for all virtual power plants within the distribution network. This represents the power generation cost of virtual power plant i during time period t, including the power generation costs of thermal power units and distributed power sources. The shared energy storage usage cost of virtual power plant i during time period t. The calculation of load migration cost for virtual power plant i during time period t. The penalty cost for wind and solar curtailment of virtual power plant i during time period t;

[0168] The constraints of the objective function for the total operating cost of the entire network include:

[0169] Inter-regional power balance constraints:

[0170] ;

[0171] In the formula: For adjacent regions, The power transmitted by the inter-regional tie line of virtual power plant i during time period t;

[0172] Tether line transmission capacity constraints:

[0173] ;

[0174] In the formula: The minimum transmission power of tie line l, Here, l represents the maximum transmission power of tie line l, and l is the number of all tie lines between areas.

[0175] Constraints on the adequacy of network-wide backup capacity:

[0176]

[0177] In the formula: This represents the uplink reserve capacity of virtual power plant i during time period t. To meet the uplink backup requirements of the system during time period t. This represents the downlink reserve capacity of virtual power plant i during time period t. This is for the system's downlink backup requirements during time period t.

[0178] In this embodiment, the specific processing steps of the robust optimization model for the lower-level virtual power plant include:

[0179] Each virtual power plant receives tie-line power reference values ​​and reserve capacity configuration instructions from the upper-level central coordinator model, and constructs a two-stage robust optimization model with the objectives of minimizing the total day-ahead scheduling cost and minimizing the total intraday scheduling cost, in order to optimize the scheduling of internal controllable resources.

[0180] The objective function that minimizes the total day-ahead scheduling cost is:

[0181] ;

[0182] In the formula: Let i be the day-ahead dispatch total cost. Let i be the power generation cost of the thermal power unit of the virtual power plant during time period t. Let i be the start-up and shutdown cost of the thermal power unit in the virtual power plant during time period t. Reserved standby costs for the thermal power units of virtual power plant i during time period t. The charging and discharging cost of virtual power plant i during time period t. Reserved backup costs for virtual power plant i during time period t;

[0183] The objective function that minimizes the total cost of intraday scheduling is:

[0184] ;

[0185] In the formula: Let i be the total daily dispatch cost of virtual power plant i. The standby call cost for the thermal power unit of virtual power plant i during time period t. The load shedding penalty cost for virtual power plant i during time period t;

[0186] The constraints of the two-stage robust optimization model include:

[0187] Distributed power generation output constraints:

[0188] ;

[0189] In the formula: For the set of thermal power units within virtual power plant i, Let n be the minimum output of thermal power unit n. This represents the maximum output of thermal power unit n.

[0190] Energy storage charge / discharge power and state of charge constraints:

[0191] ;

[0192] ;

[0193] In the formula: The energy storage charging power of virtual power plant i during time period t. The maximum charge and discharge power of the energy storage of virtual power plant i. Let i be the energy storage discharge power of the virtual power plant during time period t. The state of charge of the energy storage of virtual power plant i during time period t. This is the minimum state of charge for energy storage. This represents the maximum state of charge of the energy storage.

[0194] Load migration time constraints:

[0195] ;

[0196] In the formula: This represents the actual delay time for the batch processing load. This is the maximum allowable delay time for batch processing load;

[0197] Node voltage and line power constraints:

[0198] ;

[0199] ;

[0200] In the formula: Let m be the actual voltage value of the i-th node at the m-th voltage monitoring point during time period t. and The lower and upper voltage limits of the m-th voltage monitoring point associated with the i-th node are respectively defined.

[0201] This embodiment addresses the shortcomings of existing solutions that focus on single resources by constructing a two-layer distributed coordination and optimization model of "central coordination-regional autonomy." At the upper layer, a central coordinator generates power commands with the goal of maximizing global economy or absorption capacity, ensuring overall system optimization. At the lower layer, a robust optimization model is introduced to address the uncertainties in distributed power output and load forecasting, ensuring that the resources within the virtual power plant still meet safety constraints even in the worst-case scenario. This effectively balances global scheduling economy with local operational safety. Furthermore, this distributed architecture eliminates the need for lower-level virtual power plants to upload sensitive internal parameters such as energy storage SOC and detailed load curves; they only need to exchange power commands and boundary information. This achieves efficient collaborative scheduling while ensuring data privacy protection and independent operational rights for all parties involved.

[0202] S3. The two-layer distributed coordination optimization model is solved by a distributed iterative algorithm, and the intermediate solutions of the iteration are constrained during the iteration process to obtain the preliminary scheduling scheme and iteration state information.

[0203] In this embodiment, the improved alternating direction multiplier method (ADMM), which incorporates a multivariate optimization strategy, is used to solve the two-level model in a distributed manner. Specifically, the steps include:

[0204] S301. Dynamically adjust the penalty parameters using an adaptive step size strategy;

[0205] In practice, the original residual and dual residual of the current iteration are first calculated; it is then determined whether the norm of the original residual or dual residual is greater than a preset threshold; if so, the penalty parameter is increased to accelerate convergence; if not, the penalty parameter is decreased to ensure solution accuracy; wherein, the adjustment formula for the penalty parameter is:

[0206] ;

[0207] In the formula: Let be the penalty parameter for the nth iteration. For the original residual, For dual residuals, All of these are parameter adjustments. When the residual is large, increasing the step size accelerates convergence; when the residual is small, decreasing the step size ensures solution accuracy.

[0208] S302. Use the feasible region projection mechanism to project the iterative intermediate solution to the feasible region of the virtual power plant.

[0209] In practice, the feasible region convex hull of the virtual power plant is constructed, and the set of vertices of the feasible region convex hull is determined by solving the optimization problem under different power factor angles.

[0210] The active and reactive power increments obtained from iterative calculations are mapped onto the convex hull of the feasible region, satisfying the following active-reactive power regulation capability constraint:

[0211] ;

[0212] In the formula: For the feasible domain of the virtual power plant, and They are respectively the current active and reactive power outputs. and These are the power increments;

[0213] S303. Introduce active and reactive dual consistency variables, and achieve consistent cost increment rates for each virtual power plant through iterative updates.

[0214] In practice, active power consistency variables and reactive power consistency variables are constructed separately.

[0215] Based on the weighted average of the consistency variables of neighboring nodes and the local power deviation, the consistency variables for the next time step are updated according to the following rules:

[0216] ;

[0217] In the formula: For the active power consistency variable after the (k+1)th iteration of the i-th VPP, it is used to achieve consistency of the incremental rate of active power cost across the entire network; For the reactive power consistency variable after the (k+1)th iteration of the i-th VPP, it is used to achieve consistency of the incremental rate of reactive power cost across the entire network. The coefficients of the state transition matrix are calculated from the elements of the communication topology Laplace matrix: , Let be the Laplace matrix elements from node i to node j; All are power deviations. All are consistency adjustment coefficients;

[0218] S304. Generate a preliminary scheduling scheme based on the projected iterative intermediate solution, and record the iterative residual as iterative state information.

[0219] This embodiment addresses high-dimensional nonlinear problems involving the joint optimization of multiple virtual power plants. The invention employs a distributed iterative algorithm (such as ADMM and its improved versions) to decompose the complex problem into parallel-computable subproblems, significantly reducing the computational burden of a single iteration. Simultaneously, a constraint handling mechanism (such as feasible region projection) is introduced during the iteration process, forcing each intermediate solution to fall within the physical feasible region of the virtual power plant. This avoids the intermediate solution exceeding the limit problem in traditional algorithms. Furthermore, the invention supports synchronous iterative optimization of active and reactive power variables, breaking the traditional decoupling mode of "active power first, reactive power later," and achieving coordinated control of voltage support and power balance, further improving power quality and system regulation accuracy.

[0220] S4. Perform operation status verification based on the preliminary scheduling scheme and iteration status information. If the preset correction conditions are met, perform voltage correction operation; otherwise, return to step S1 until the convergence conditions are met and the final configuration scheme is output.

[0221] In this embodiment, a dynamic correction mechanism is proposed. The output of the virtual power plant is adjusted to the power limit handling at the boundary of the feasible region according to the principle of equal power factor. The voltage limit correction of the binary consistency variable is corrected based on the voltage sensitivity model. When the cooperative efficiency in the region decreases, the sensitivity matrix is ​​recalculated for partition optimization. All corrections are based on the power interaction sensitivity matrix calculated in step S1.

[0222] The specific processing steps for this operational status verification include:

[0223] S401. Monitor the operating status of each virtual power plant in real time to determine whether there is any power or voltage exceeding the limit.

[0224] S402. If there is a power limit violation, perform a power limit violation correction operation.

[0225] In practice, the operating status of each virtual power plant is monitored in real time during the iteration process, including whether node voltage, line power, energy storage state of charge, and distributed power output exceed limits. When power exceeds limits, the output of the virtual power plant is adjusted to the boundary of the feasible domain of the virtual power plant according to the principle of equal power factor. The corrected active and reactive power outputs satisfy the following:

[0226] ;

[0227] In the formula: The corrected active power output of the i-th VPP after exceeding the power limit in time period t is located at the VFR boundary after adjustment, while keeping the power factor unchanged; The active power boundary value of the virtual power plant feasible region for the i-th VPP in time period t is obtained by the vertex enumeration method; Let be the rated power factor angle of the i-th VPP, typically taken as . (Lag) to ensure that the power factor remains unchanged after adjustment; The corrected reactive power output of the i-th VPP after exceeding the power limit in time period t;

[0228] S403. If a voltage limit violation exists, perform a voltage limit violation correction operation.

[0229] In specific implementation, the voltage over-limit correction operation includes: calculating the node voltage change based on the voltage sensitivity model.

[0230] ;

[0231] In the formula: The voltage change at node m is used to quantify the impact of VPP output adjustment on node voltage and determine whether the limit is exceeded. This represents the total number of VPPs within the distribution network, corresponding to the set of all VPPs involved in voltage regulation. These are the active and reactive power adjustment values ​​for the i-th VPP, respectively, which are obtained by correction using binary consistency variables and are used to improve node voltage over-limit. The set of lines is the intersection of the lines from node i to the power inflow node of the distribution network and the lines from node m to the power inflow node. These are the resistance and reactance of line j, respectively, taken from the distribution network topology parameters, reflecting the impact of line loss on voltage; The voltage amplitude at point j of the line is taken as the real-time operating value and used to calculate the voltage sensitivity coefficient.

[0232] S404. If the collaborative efficiency within the region decreases, calculate the difference in the incremental cost rate. When the difference in the incremental cost rate is greater than the preset growth rate threshold, repeat step S1.

[0233] In practice, when the difference in the rate of cost increment is greater than 5%, the sensitivity matrix is ​​recalculated and the partitioning strategy is adjusted.

[0234] S405. Check if the convergence condition is met. If it is, output the final configuration scheme.

[0235] In practice, when the iterative residual is less than the set threshold (ε=0.001) and there are no over-limit states, the iteration converges, and the day-ahead-intraday generation plan, shared energy storage dispatch strategy, calculated load migration scheme, tie-line power, and standby configuration scheme for each virtual power plant are output. The formula for calculating the iterative residual is as follows:

[0236] ;

[0237] In the formula: δ is the global maximum relative residual. These are the variable values ​​for the t-th and t-1-th iterations, respectively. These represent the active and reactive power outputs of VPP, respectively. For energy storage charging and discharging power, This refers to the power of the tie line.

[0238] This embodiment implements a hierarchical response mechanism by rigorously verifying the operational status of the initial scheduling scheme: when local risks such as voltage over-limit are detected, voltage correction operations are prioritized to quickly eliminate the hidden dangers; when conventional corrections cannot resolve structural blockages or significantly reduce coordination efficiency, a re-partitioning mechanism is triggered to return to step S1. This dynamic feedback strategy not only ensures that the globally optimal configuration scheme can always be obtained under complex and variable operating conditions, but also guarantees that the final output scheme strictly meets physical constraints such as voltage and power flow while achieving mathematical convergence, significantly improving the engineering practicality of the scheduling scheme and the safety and reliability of virtual power plants participating in grid scheduling.

[0239] like Figure 2 The diagram shown illustrates the specific flowchart of the distributed resource aggregation and optimized scheduling process in this embodiment. It can be seen that each virtual power plant aggregates diverse resources such as distributed power sources, shared energy storage, and flexible loads. Based on the aggregated resource model and network power flow model, the regulation capability and power interaction sensitivity of each VPP are calculated. According to the sensitivity matrix and regulation potential index, the VPPs are dynamically partitioned using a clustering algorithm to form multiple collaborative regulation regions. Within each region, the VPPs receive tie-line power and reserve capacity instructions from the upper-level central coordinator. Each VPP performs a two-stage robust optimization, optimizing internal generation, energy storage, load migration, and other resources, and outputs active / reactive power to the common coupling node of the distribution network. Through the above process, the aggregation, partitioning, and collaborative optimization of multiple virtual power plant resources are achieved, supporting the safe and economical operation of the distribution network.

[0240] like Figure 3 The diagram shows the specific flowchart of the operation status monitoring and dynamic correction process in this embodiment. It can be seen that during the distributed iterative solution process, the operating status of each virtual power plant, including node voltage, line power, energy storage status, and distributed power output, is monitored in real time. It determines whether a power limit violation has occurred; if so, the VPP output is adjusted to the feasible region boundary according to the equal power factor principle. It also determines whether a voltage limit violation has occurred; if so, the active-reactive binary consistency variable is corrected based on the voltage sensitivity model. Finally, it determines whether the regional coordination efficiency has decreased (cost increment difference > 5%); if so, the sensitivity matrix is ​​recalculated and the partitioning strategy is adjusted. A convergence judgment is performed: if the iteration residual is less than a set threshold and there are no limit violations, the optimized solution is output; otherwise, the iteration continues. Through the above closed-loop correction mechanism, the optimization process is ensured to converge quickly while meeting safety constraints, outputting a feasible and economical scheduling solution.

[0241] In summary, this embodiment discloses a multi-virtual power plant coordinated configuration method based on sensitivity partitioning, aiming to address the bottlenecks of existing technologies in dealing with source-load uncertainty, large-scale computational dimensionality, data privacy protection, and dynamic security verification. The method first constructs a multi-virtual power plant system model, innovatively integrating the power interaction sensitivity matrix and comprehensive regulation potential index. It then utilizes a clustering algorithm to achieve dynamic partitioning adapted to electrical topology and resource characteristics, improving coordination efficiency and reducing solution dimensionality from the source. Next, it constructs a two-layer distributed architecture of "upper-layer central coordination - lower-layer robust optimization," ensuring global economic efficiency while effectively mitigating wind and solar fluctuation risks and protecting the data privacy of each entity through robust strategies. In the solution phase, an improved distributed iterative algorithm is adopted, introducing an intermediate solution constraint processing mechanism (such as feasible region projection), significantly accelerating convergence and ensuring the physical feasibility of the iterative process. Finally, a closed-loop defense system is established, dynamically executing voltage fine-tuning or triggering partition re-optimization based on the operating status verification results. This embodiment achieves safe, efficient, and privacy-friendly collaborative scheduling of multiple virtual power plant systems under complex and variable operating conditions through full-chain technological innovation, providing strong theoretical support and technical means for the flexible interaction and stable operation of new power systems.

[0242] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit the technical solutions. Those skilled in the art should understand that any modifications or equivalent substitutions to the technical solutions of the present invention without departing from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning, characterized in that, Includes the following steps: S1. Construct a multi-virtual power plant system model, calculate the power interaction sensitivity matrix between virtual power plants based on the multi-virtual power plant system model, process the power interaction sensitivity matrix and the comprehensive regulation potential index using a clustering algorithm, dynamically partition the multiple virtual power plants, and output the partitioning results. S2. Based on the partitioning results, a two-layer distributed coordination optimization model is constructed, including an upper-layer central coordinator optimization model and a lower-layer virtual power plant robust optimization model. The upper-layer central coordinator optimization model generates power commands and optimizes internal resources according to the power commands through the lower-layer virtual power plant robust optimization model. S3. The two-layer distributed coordination optimization model is solved by a distributed iterative algorithm, and the intermediate solutions of the iteration are constrained during the iteration process to obtain the preliminary scheduling scheme and iteration state information. S4. Perform operation status verification based on the preliminary scheduling scheme and iterative status information. If the preset correction conditions are met, perform voltage correction operation. Otherwise, return to step S1 until the convergence condition is met and the final configuration scheme is output.

2. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 1, characterized in that, In step S1, the multi-virtual power plant system model includes: The mathematical expression for the distributed power generation model is as follows: ； In the formula: Let be the active and reactive power outputs of the semi-controlled power supply i at time t, respectively. and Let be the minimum and maximum power factor angles of the semi-controlled power supply i, respectively. Let i be the maximum active power output that the semi-controlled power supply can provide at time t. Let be the probability that the event is true. For the confidence level of opportunity constraints, For a given reactive power quantile function under the given conditions For random variables The inverse function of the cumulative distribution function; The mathematical expression for the shared energy storage model is as follows: ； In the formula: Virtual power plants The allocated energy storage capacity and power capacity, For virtual power plants The shared costs of energy storage that should be allocated This refers to the set of all virtual power plants participating in shared energy storage. Let S be any subset of the alliance that does not contain virtual power plant i, and let C(S) be the cost function of the alliance S. The mathematical expression for the flexible load model is as follows: ； In the formula: The computational load of virtual power plant i during time period t after migration. Virtual power plant before migration The computational load, For virtual power plants Towards Migration load, For virtual power plants Maximum load receiving capacity; The mathematical expression for the network power flow model is as follows: ； In the formula: , and These represent the active power output of the nth thermal power unit at node m in the virtual power plant i, the predicted active power output of the wind turbine unit, and the base load power, respectively. They are nodes It is a collection of thermal power units, virtual power plants, and wind power. For nodes Voltage phase angle, Let be the voltage phase angle of node j adjacent to node m. For the line Reactance , Virtual power plants Middle node First The charging and discharging power of each energy storage unit.

3. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 2, characterized in that, In step S1, the power interaction sensitivity matrix is ​​derived from the Lagrange function; The power interaction sensitivity matrix is ​​represented as follows: ； In the formula: For the power interaction sensitivity matrix, For the subproblem's Lagrangian function, These are the optimal dual variables corresponding to the equality constraints. Let be the optimal dual variable corresponding to the inequality constraint. Coupled decision parameters are issued by the upper-level coordinator and treated as known in the sub-problems; The Lagrange function is expressed as: ； In the formula: Let be the objective function. For equality constraint functions, is the inequality constraint function.

4. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 3, characterized in that, In step S1, the specific processing steps for using a clustering algorithm to process the power interaction sensitivity matrix and the comprehensive regulation potential index, dynamically partitioning multiple virtual power plants, and outputting the partitioning results include: S101. The power interaction sensitivity matrix is ​​normalized to obtain the normalized coupling degree matrix, which is expressed as: ； In the formula: This is the normalized coupling degree matrix; The magnitude of the impact on the operating cost or power balance state of virtual power plant j when the power decision variable of virtual power plant i changes by a unit; S102. Based on the normalized coupling matrix and combined with the comprehensive regulation potential index of each virtual power plant, construct the sample feature vector of each virtual power plant. S103. The elbow rule and the contour coefficient method are used to analyze the feature vector of the sample to determine the optimal number of clusters K; S104. Initialize K cluster centers, calculate the Euclidean distance from each sample feature vector to each cluster center, divide the sample into the nearest cluster, and update each cluster center to the average value of all sample feature vectors in that cluster, until the cluster centers converge, and output the partitioning result.

5. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 4, characterized in that, Step S1 also includes a dynamic partition update mechanism, the specific processing steps of which include: The system monitors its operating status in real time. When the preset dynamic update triggering conditions are met, the power interaction sensitivity matrix calculation and clustering algorithm are re-executed to update the partitioning results. After updating the partitioning results, the rationality of the new partitioning is verified. If the verification is successful, the new partitioning results are sent to the upper-level central coordinator optimization model. If the verification fails, the clustering parameters are adjusted and the clustering process is re-executed until the partitioning results meet the collaborative scheduling requirements. The dynamic update triggering conditions include: the fluctuation range of new energy output exceeds a preset fluctuation threshold; the standard deviation of the comprehensive regulation potential index of virtual power plants in the region exceeds a preset complementarity threshold; the proportion of cross-regional power interaction to the total regional output exceeds a preset interference threshold; and a fixed time period trigger point is reached. The verification of the rationality of the new partition specifically includes: calculating the collaborative efficiency index within the region and the interference index between regions; when the collaborative efficiency index within the region is greater than or equal to the efficiency threshold and the interference index between regions is less than or equal to the interference threshold, the verification is deemed successful.

6. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 5, characterized in that, In step S2, the specific processing procedure of the upper-level central coordinator optimization model includes: With the goal of minimizing the total operating cost of the entire network, a power command including tie-line power reference values ​​and standby capacity configuration instructions is generated by combining the power interaction sensitivity matrix and partitioning results. The objective function for the total operating cost of the entire network is expressed as: ； In the formula: To minimize the total network operating cost for the upper-level central coordinator, Let I be the time set, and let I be the set of IDs for all virtual power plants within the distribution network. This represents the power generation cost of virtual power plant i during time period t, including the power generation costs of thermal power units and distributed power sources. The shared energy storage usage cost of virtual power plant i during time period t. The calculation of load migration cost for virtual power plant i during time period t. The penalty cost for wind and solar curtailment of virtual power plant i during time period t; The constraints of the objective function for the total operating cost of the entire network include: Inter-regional power balance constraints: ； In the formula: For adjacent regions, The power transmitted by the inter-regional tie line of virtual power plant i during time period t; Tether line transmission capacity constraints: ； In the formula: The minimum transmission power of tie line l, Here, l represents the maximum transmission power of tie line l, and l is the number of all tie lines between areas. Constraints on the adequacy of network-wide backup capacity: In the formula: This represents the uplink reserve capacity of virtual power plant i during time period t. To meet the uplink backup requirements of the system during time period t. This represents the downlink reserve capacity of virtual power plant i during time period t. This is for the system's downlink backup requirements during time period t.

7. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 6, characterized in that, In step S2, the specific processing procedure of the robust optimization model of the lower-level virtual power plant includes: Each virtual power plant receives tie-line power reference values ​​and reserve capacity configuration instructions from the upper-level central coordinator model, and constructs a two-stage robust optimization model with the objectives of minimizing the total day-ahead scheduling cost and minimizing the total intraday scheduling cost, in order to optimize the scheduling of internal controllable resources. The objective function that minimizes the total day-ahead scheduling cost is: ； In the formula: Let i be the day-ahead dispatch total cost. Let i be the power generation cost of the thermal power unit of the virtual power plant during time period t. Let i be the start-up and shutdown cost of the thermal power unit in the virtual power plant during time period t. Reserved standby costs for the thermal power units of virtual power plant i during time period t. The charging and discharging cost of virtual power plant i during time period t. Reserved backup costs for virtual power plant i during time period t; The objective function that minimizes the total cost of intraday scheduling is: ； In the formula: Let i be the total daily dispatch cost of virtual power plant i. The standby call cost for the thermal power unit of virtual power plant i during time period t. The load shedding penalty cost for virtual power plant i during time period t; The constraints of the two-stage robust optimization model include: Distributed power generation output constraints: ； In the formula: For the set of thermal power units within virtual power plant i, Let n be the minimum output of thermal power unit n. This represents the maximum output of thermal power unit n. Energy storage charge / discharge power and state of charge constraints: ； ； In the formula: The energy storage charging power of virtual power plant i during time period t. The maximum charge and discharge power of the energy storage of virtual power plant i. Let i be the energy storage discharge power of the virtual power plant during time period t. The state of charge of the energy storage of virtual power plant i during time period t. This is the minimum state of charge for energy storage. This represents the maximum state of charge of the energy storage. Load migration time constraints: ； In the formula: This represents the actual delay time for the batch processing load. This is the maximum allowable delay time for batch processing load; Node voltage and line power constraints: ； ； In the formula: Let m be the actual voltage value of the i-th node at the m-th voltage monitoring point during time period t. and The lower and upper voltage limits of the m-th voltage monitoring point associated with the i-th node are respectively defined.

8. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 7, characterized in that, In step S3, the specific process of solving the two-layer distributed coordinated optimization model using a distributed iterative algorithm includes: S301. Dynamically adjust the penalty parameter using an adaptive step size strategy; the adjustment formula for the penalty parameter is: ； In the formula: Let be the penalty parameter for the nth iteration. For the original residual, For dual residuals, All of these are parameter adjustments; S302. The iterative intermediate solution is projected onto the feasible region of the virtual power plant using a feasible region projection mechanism; the feasible region is solved by vertex enumeration and satisfies the following active-reactive power regulation capability constraints: ； In the formula: For the feasible domain of the virtual power plant, and They are respectively the current active and reactive power outputs. and These are the power increments; S303. Introduce a binary consistency variable for active and reactive power, and achieve consistent cost increment rates for each virtual power plant through iterative updates; the update rule is: ； In the formula: Let be the active consistency variable after the (k+1)th iteration of the i-th VPP. Let be the reactive power consistency variable after the (k+1)th iteration of the i-th VPP. The coefficients are the state transition matrix coefficients. All are power deviations. All are consistency adjustment coefficients; S304. Generate a preliminary scheduling scheme based on the projected iterative intermediate solution, and record the iterative residual as iterative state information.

9. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 8, characterized in that, In step S4, the specific process of verifying the running status based on the preliminary scheduling scheme and iteration status information includes: S401. Monitor the operating status of each virtual power plant in real time to determine whether there is any power or voltage exceeding the limit. S402. If there is a power limit violation, perform a power limit violation correction operation. S403. If a voltage limit violation exists, perform a voltage limit violation correction operation. S404. If the collaborative efficiency within the region decreases, calculate the difference in the incremental cost rate. When the difference in the incremental cost rate is greater than the preset growth rate threshold, repeat step S1. S405. Check if the convergence condition is met. If it is, output the final configuration scheme.

10. The method for coordinated configuration of multiple virtual power plants based on sensitivity partitioning according to claim 9, characterized in that, In step S402, the power over-limit correction operation includes: adjusting the over-limit virtual power plant output to the boundary of the virtual power plant's feasible region according to the equal power factor principle, wherein the corrected active and reactive power outputs satisfy: ； In the formula: This represents the corrected active power output of the i-th VPP after exceeding its power limit during time period t. Let be the active power boundary value of the virtual power plant feasible region for the i-th VPP in time period t. Let be the rated power factor angle of the i-th VPP. The corrected reactive power output of the i-th VPP after exceeding the power limit in time period t; In step S403, the voltage over-limit correction operation includes: calculating the node voltage change based on the voltage sensitivity model. ； In the formula: Let be the voltage change at node m. This represents the total number of VPPs within the distribution network. These are the active and reactive power adjustments for the i-th VPP, respectively. For the set of routes, These represent the resistance and reactance of line j, respectively. Let be the voltage amplitude at endpoint j of the line; In step S404, the verification of whether the convergence condition is met includes: calculating the global maximum relative residual. ； In the formula: δ is the global maximum relative residual. These are the variable values ​​for the t-th and t-1-th iterations, respectively. These represent the active and reactive power outputs of VPP, respectively. For energy storage charging and discharging power, This refers to the power of the tie line.

Images