A multi-element reactive power resource regional coordination control method and system across voltage

Abstract

The present application relates to a kind of reactive resource collaborative control methods and systems across voltage, based on the linear mapping relationship of the variation of node voltage amplitude and phase angle and active, reactive power disturbance quantity of cross voltage alternating current flow equation of vertical multistage network is established, and construct comprehensive complex sensitivity matrix of whole network across voltage;Construction is based on the quality comprehensive evaluation function of partition of profile coefficient and modularity, and the best number is determined by optimization self-adaption, and leading node is established;Construction vertical cross voltage coordinated control double-layer optimization mathematical model, the upper layer global coordination layer model minimizes whole network total active network loss and discrete asset action penalty, and the lower layer local flexible autonomous layer model introduces adaptive affine decision rule minimizes voltage violation risk and photovoltaic active reduction penalty;Distributed parallel solution based on master-slave splitting theory is used, and reactive resource collaborative control scheme is obtained.The present application effectively realizes the decoupling linkage configuration of multiple resources under cross voltage scene while strictly guaranteeing topology privacy.

Description

Technical Field

[0001] This invention relates to the field of power grid reactive power dispatch and voltage control technology, and in particular to a method and system for coordinated control of multi-reactive power resource regions across voltage levels. Background Technology

[0002] With the deepening of the construction of new power systems, the large-scale and deep unordered integration of distributed photovoltaic (PV) systems supported by inverters is accelerating the evolution of the physical form and organizational structure of modern power grids towards a vertical multi-level network spanning transmission (110 kV and above main grid layer), distribution (10 kV medium voltage layer), and active terminal microgrids (380 V low voltage layer). Fully utilizing the four-quadrant flexible reactive power throughput capability of distributed PV inverters, combined with traditional voltage regulation equipment for voltage smoothing, has gradually become a recognized mainstream technical paradigm.

[0003] However, in real-world cross-voltage application scenarios characterized by strong heterogeneity among multiple stakeholders in the transmission, distribution, and microgrid systems, and with deeply intertwined uncertainties on both the source and load sides, existing reactive power voltage control methods, network partitioning architectures, and distributed solution mechanisms reveal the following significant shortcomings and technical bottlenecks in practical engineering applications: The cascading effect of voltage violations across voltage levels is difficult to eliminate: due to the tight electromagnetic coupling between transformers at each level of the main-distribution-micro network, a large-scale reverse active power flow is heavily injected into the medium and low voltage feeders at noon. The local "reverse voltage overflow effect" will propagate upwards along the cascading interfaces of multiple transformers in the longitudinal direction, causing the voltage at the boundary of the high-voltage main transmission network to synchronously exceed the limit. Existing technologies lack an algebraic interaction interface that can break down the physical barriers between levels and coordinate multi-level reactive resources across voltage levels.

[0004] Multi-temporal and spatial scale elements are difficult to coordinate and configure: Traditional discrete mechanical assets such as OLTC and SC operate on a minute-by-minute basis and are constrained by the number of operations and mechanical lifespan, making it completely unable to track the transient high-frequency fluctuations of photovoltaics on a second- or even millisecond-by-second basis. Continuous flexible resources (photovoltaic inverters) have high-speed continuous reactive power tracking characteristics, but the capacity of a single unit is limited by the circular domain boundary of instantaneous active power output. Existing technologies cannot organically cascade the normal operating base of discrete equipment with the transient adaptive absorption of flexible resources, resulting in severe fatigue wear of discrete equipment and frequent depletion of reactive power margin.

[0005] Traditional centralized architectures face the curse of dimensionality and topological privacy barriers: the number of control variables in distributed photovoltaic systems explodes, and centralized global optimization architectures bring intolerable communication latency, packet loss risks, and fatal computational bottlenecks. More importantly, the global collection of specific admittance matrices and load topology seriously violates the topological privacy and security specifications for heterogeneous multi-stakeholder networks. Existing distributed solution algorithms have slow convergence speeds when processing vertical multi-level networks and still require the exchange of some internal sensitive confidential information. Summary of the Invention

[0006] The purpose of this invention is to solve the problems of cross-voltage reactive power backflow and overvoltage violations caused by high-proportion distributed photovoltaic grid connection, and to provide a cross-voltage reactive power resource collaborative control method and system.

[0007] The objective of this invention can be achieved through the following technical solutions: As a first aspect of the present invention, a method for coordinated control of reactive power resources across voltage levels is provided, comprising: Based on the cross-voltage power flow equations of the entire network, a first-order Taylor expansion is performed to establish a linear mapping relationship between the changes in node voltage magnitude and phase angle and the active and reactive power disturbances. An active-reactive cross-coupling correction coefficient is introduced to construct a comprehensive complex sensitivity matrix across voltage ranges for the entire network; A quality evaluation function is constructed based on the contour coefficient and modularity, and the dominant node of boundary state interaction is established. A dual-layer optimization mathematical model for vertical cross-voltage coordinated regulation is constructed. The upper-layer global coordination layer model faces the high-voltage trunk transmission network and the global hierarchical boundary, while the lower-layer local flexible autonomous layer model is embedded in each horizontal spatial partition and runs independently and in parallel. An adaptive affine decision rule is introduced to correlate the reactive power output of the terminal inverter with the random disturbance of active power in real time. The optimal cooperative control scheme is obtained by using distributed parallel solution based on master-slave splitting theory.

[0008] As a preferred technical solution, for nodes within any voltage level The contour coefficient is calculated as follows: in, For nodes The average electrical distance between the node and all other nodes in the same cluster; For nodes The average electrical distance between the nearest neighboring node and all other nodes in the cluster.

[0009] As a preferred technical solution, the mathematical expression of the modularity is: These are the elements of the adjacency weight matrix after complex sensitivity truncation. and They are nodes and The sum of sensitivities; It is the sum of the total equivalent electrical edge weights of the entire network; and These are the partition labels to which the nodes belong; when hour This indicates that the two nodes belong to the same partition.

[0010] As a preferred technical solution, the comprehensive evaluation function for partition quality is constructed as follows: in, and These are adaptive weighting coefficients. (Through intervals) The internal process performs a one-dimensional exhaustive step-by-step optimization to achieve a comprehensive evaluation function. Reaching the global maximum value This is the optimal number of autonomous cluster partitions.

[0011] As a preferred technical solution, the adaptive weighting coefficient is adjusted according to preset conditions, which include: Calculate the expected value (mean) and standard deviation of the silhouette coefficient sequence and modularity sequence over the entire optimization interval. Divide the standard deviation of each index by its corresponding mean to obtain the coefficient of variation of the silhouette coefficient. Coefficient of variation with modularity The weighting coefficients are obtained by normalizing the two coefficients of variation. As a preferred technical solution, the objective function of the upper-level global coordination layer model is: Total active power loss for the entire network; This is the equivalent penalty cost coefficient for a single tap changer operation of an on-load tap-changing transformer. For the ratio adjustment amount, It aggregates all OLTC data across the network; For critical transmission sections l The weight of the current fluctuation, The deviation of the trend is a factor.

[0012] As a preferred technical solution, the objective function of the lower-level local flexible autonomous layer model is: And it satisfies the spatiotemporal bound constraints at the safe confidence level: In the formula: The reference voltage rating for each node; The penalty coefficient for proactive merit reduction. Let be the active power reduction of the j-th distributed power source; It is a set of fuzzy probability measures constructed based on the statistical characteristics of historical uncertainty data. The symbol for conditional probability. Given the maximum confidence level (overvoltage tolerance probability).

[0013] As a preferred technical solution, the dynamic boundary penalty function term based on physical residuals introduced in the individual fitness functions of the master control layer and the sub-autonomous layer is expressed as follows: In the formula, and These are the boundary active and reactive power flows from the perspective of the main control layer; and These are the boundary equivalent active and reactive power flows calculated and uploaded by the sub-autonomous layer, respectively. and This represents the penalty coefficient for the dynamic boundary physical residual. Through the constraint of this physical penalty function, the distributed algorithm adaptively generates a guiding vector during the iteration process, driving the boundary algebraic difference between the master and slave sides to rapidly approach zero.

[0014] As a preferred technical solution, the distributed parallel solution and closed-loop iteration based on the master-slave splitting theory are as follows: The main control layer solves the main network optimization model and sends the updated boundary node voltage magnitude command vertically to each sub-autonomous layer. Each sub-autonomous layer solves the lower-level adaptive affine absorption model in parallel and feeds back the latest boundary equivalent active and reactive power flow to the main control layer. Furthermore, a dynamic boundary penalty function term based on physical residuals is introduced into the individual fitness function for iterative solution.

[0015] As a second aspect of the present invention, a cross-voltage reactive power resource cooperative control system is provided, comprising: The integrated complex sensitivity matrix construction module is used to perform a first-order Taylor expansion based on the cross-voltage AC power flow equation and introduce active-reactive cross-coupling correction coefficients to construct the network-wide cross-voltage integrated complex sensitivity matrix. The horizontal network autonomous decoupling partitioning module is used to construct a comprehensive evaluation function for partitioning quality based on the cross-voltage electrical distance index, by introducing the profile coefficient and modularity, and to implement decoupling partitioning and establish the dominant node using an improved k-medoids clustering algorithm; The vertical cross-voltage coordinated control dual-layer optimization module includes an upper global coordination layer sub-module and a lower local flexible autonomous layer sub-module, which is used to construct a mathematical model for the linkage optimization of discrete assets and flexible resources in a vertical multi-level space. The distributed master-slave split parallel solution module is used to alternately iterate between the master control layer and the sub-autonomous layer. It introduces a dynamic boundary penalty function term based on physical residuals to drive the boundary state residuals to approach zero and output the globally optimal cooperative control strategy.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention effectively improves the problems of voltage violations at the end of reverse power flow across voltage, topological privacy barrier locking among multiple stakeholders, and centralized computing power bottleneck of high-dimensional distributed variables caused by the interweaving of random uncertainties on both sides of the source and load sides in the master-distributor-micro vertical multi-level network. By deriving the cross-voltage complex sensitivity matrix of the fusion cross-coupling coefficient, constructing a two-layer coordinated control optimization mathematical model of vertical spatial hierarchical decoupling, and implementing a master-slave split distributed decoupling solution mechanism based on the soft penalty term of cross-section physical residuals, this invention effectively improves the problems of voltage violation at the end of reverse power flow across voltage, topological privacy barrier locking among multiple stakeholders, and centralized computing power bottleneck of high-dimensional distributed variables caused by the interweaving of random uncertainties on both sides of the source and load sides in the master-distributor-micro vertical multi-level network. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a cross-voltage reactive resource collaborative control method according to an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram comparing the simulated voltage results of a conventional voltage regulation method and the method used in this invention, which is an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram comparing the voltage optimization results under over-limit conditions on the high-voltage side of a cross-voltage system.

[0020] Figure 4 for Figure 3 The result is a simulation system structure diagram. Detailed Implementation

[0021] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0022] Example 1 A flowchart illustrating a cross-voltage reactive power resource collaborative control method provided in this embodiment of the invention is shown below. Figure 1 As shown, the main steps include: S101. Based on the first-order Taylor expansion of the cross-voltage AC power flow equations of the vertical multi-level network between the transmission network, distribution network and active terminal microgrid, a linearized mapping relationship is established between the voltage amplitude and phase angle changes of the cross-voltage nodes covering the main grid layer, medium voltage layer and terminal low voltage layer and the active and reactive power disturbances. The global cross-voltage level voltage-reactive power sensitivity matrix and active power-voltage sensitivity matrix are derived and extracted.

[0023] S102. To comprehensively characterize the cross-interference and coupling relationship between heavy photovoltaic reverse active power flow injection and inverter reactive power throughput on voltage fluctuations under the high impedance ratio characteristics of low-voltage end feeders, an active-reactive cross-coupling correction coefficient is introduced to characterize the high impedance ratio characteristics of low-voltage end feeders. This achieves unified cross-voltage modeling of main grid voltage regulation equipment operation, distribution network power flow interaction, and microgrid load fluctuations, synthesizing a comprehensive complex sensitivity matrix for the entire network. Specific elements in the comprehensive complex sensitivity matrix are extracted to characterize the control sensitivity coefficients that represent the voltage changes at one node caused by reactive power disturbances at another node. Through a logarithmic mapping relationship, these coefficients are transformed into electrical distance indicators that quantify the electromagnetic tightness between nodes at each level. The higher the control sensitivity between nodes, the closer the corresponding electrical distance is to zero.

[0024] S103. A partition quality evaluation function is constructed by introducing contour coefficient and modularity. By performing one-dimensional exhaustive step-by-step optimization within the interval, the optimal number of autonomous cluster partitions is adaptively determined. An improved k-medoids clustering algorithm is used to implement the horizontal autonomous decoupling partitioning of the entire network topology in the spatial dimension, and to complete the spatial projection dimensionality reduction of high-dimensional control variables. The central node in each partition with the highest sensitivity to cross-voltage disturbance control of the entire network and the smallest sum of electrical distances to other nodes is established as the dominant node responsible for the boundary state interaction of multi-level voltage levels.

[0025] S104. Based on horizontal zoning, a vertical cross-voltage dual-layer optimization control model is constructed. The upper-layer global coordination layer, oriented towards the backbone network and the main distribution boundary, operates on an hourly / minute-level scale. Its objective is to minimize the total active power loss of the entire network and the penalty for the operation of traditional voltage regulating equipment. It optimizes and solves discrete variables such as transformer tap positions and the number of capacitor banks switched on and off, and generates the optimal dynamic active power allowable range boundary for distributed photovoltaic power as a rigid limit for vertical cascading. The lower-layer local flexible autonomous layer operates on a second-level scale. Its objective is to minimize local voltage deviation and photovoltaic absorption penalty. It introduces adaptive affine decision rules to locally correlate inverter reactive power output with active power disturbances, and combines distributed bar chance constraints to strictly ensure that the terminal voltage does not exceed the limit under a given safety confidence level.

[0026] S105. A distributed parallel solution algorithm based on master-slave splitting theory is introduced. The algorithm performs cross-sectional state splitting at the vertical virtual boundary, completely locking the impedance topology, load, and equipment limits of each entity locally. In the upper-level optimization, the voltage and power imbalance on the longitudinal interactive cross-section is extracted and reconstructed into a dynamic boundary physical residual soft penalty term, which is embedded into the individual fitness function, driving the boundary algebraic difference to approach zero. During execution, the upper-level master control layer solves the main network model and issues boundary voltage commands. Each sub-autonomous layer uses this as a boundary to solve its local affine absorption model in parallel and feeds back the boundary power flow. It uses a genetic algorithm with selection, crossover, and mutation operators for alternating iterative searches until the boundary state residual between two adjacent iterations is less than a preset high-precision threshold, at which point the algorithm terminates and outputs the globally optimal strategy.

[0027] Furthermore, an active-reactive cross-coupling correction coefficient is introduced to characterize the high impedance ratio of low-voltage end feeders, enabling unified cross-voltage modeling of main grid voltage regulation equipment operation, distribution network power flow interaction, and microgrid load fluctuations, thus constructing a comprehensive cross-voltage sensitivity matrix for the entire network: Based on the matrix elements in the comprehensive complex sensitivity matrix that can simultaneously characterize low-voltage terminal disturbances and high-voltage main grid responses, a logarithmic mapping relationship is used to define any two nodes at different voltage levels. and Electrical distance indicators between voltage ranges This is used to accurately quantify the tightness of electromagnetic control between nodes in a multi-level voltage network. To clarify the control responsibility boundaries of heterogeneous entities while maintaining strong cross-voltage correlation, a comprehensive evaluation function for partitioned quality is constructed by introducing the profile coefficient and modularity as dual performance indicators. For nodes within any voltage level... Its profile coefficient The calculation is as follows: in, For nodes The average electrical distance between the node and all other nodes in the same cluster; For nodes Average electrical distance to all nodes within the nearest neighboring cluster. Average profile coefficient of the entire network. The closer the value is to 1, the closer the electrical distance between nodes within the voltage partition, and the clearer the geometric boundaries of the cluster division.

[0028] Furthermore, the mathematical expression for modularity is: in, These are the elements of the adjacency weight matrix after complex sensitivity truncation. and They are nodes and The sum of sensitivities; It is the sum of the total equivalent electrical edge weights of the entire network; and These are the partition labels to which the nodes belong; when hour 0 indicates that the two nodes belong to the same partition, otherwise it is 0.

[0029] Furthermore, the comprehensive evaluation function for zoning quality is constructed as follows: in, and These are adaptive weighting coefficients. (Through intervals) The internal process performs a one-dimensional exhaustive step-by-step optimization to achieve a comprehensive evaluation function. Reaching the global maximum value This is the optimal number of autonomous cluster partitions: Utilizing the improved The medoids clustering algorithm implements horizontal autonomous decoupling partitioning of the entire network topology in the spatial dimension, automatically optimizes the number of autonomous cluster partitions and boundary assignments, completes spatial projection dimensionality reduction of high-dimensional control variables, and establishes the central node in each partition with the highest sensitivity to cross-voltage disturbance control of the entire network and the smallest sum of electrical distances to other nodes as the dominant node responsible for the boundary state interaction of multi-level voltage levels.

[0030] Furthermore, the adaptive weighting coefficients are adjusted according to preset conditions, which include: Calculate the expected value (mean) and standard deviation of the silhouette coefficient sequence and modularity sequence over the entire optimization interval. Divide the standard deviation of each index by its corresponding mean to obtain the coefficient of variation of the silhouette coefficient. Coefficient of variation with modularity The weighting coefficients are obtained by normalizing the two coefficients of variation. .

[0031] Furthermore, a two-level optimization mathematical model for vertical cross-voltage coordinated regulation was constructed: Based on horizontal partitioning, a two-layer multi-objective optimization mathematical model with vertical spatial hierarchical decoupling is constructed to achieve cross-voltage cascading linkage and complementary advantages between high-voltage slow-scale discrete mechanical assets and medium- and low-voltage fast-scale continuous flexible resources in vertical multi-level space: For high-voltage backbone transmission networks and global hierarchical boundaries, with the optimization objectives of minimizing total active power loss, mechanical wear penalty for tap changer (OLTC) tap position adjustment, and mechanical wear penalty for switching operations of parallel capacitor banks (SC), an upper-level global coordination layer model is constructed: In the formula: Total active power loss for the entire network; This is the equivalent penalty cost coefficient for a single tap changer operation of an on-load tap-changing transformer. For the ratio adjustment amount, It aggregates all OLTC data across the network; For critical transmission sections l The weight of the current fluctuation, The deviation of the trend is a contributing factor; , , They are nodes i Active power injection, reactive power injection, and voltage amplitude. , , These are the mutual conductance and mutual susceptance of the admittance matrix, respectively, and the phase angle difference between the nodal voltages. and Let represent the branch current and its upper limit of allowable capacity. The optimization variables of the upper-level global coordination layer model can be expressed as: Among them, discrete variables and Each is executed locally by the main network operator, while the continuous interval control variables... It then serves as the core interactive operator for vertical collaboration and is passed down to the lower-level cluster autonomous layer.

[0032] The lower-level local flexible autonomous layer model is embedded within each horizontal spatial partition and operates independently and in parallel, running on a second- or millisecond-level scale within the local circuits of medium and low voltage (distribution networks and microgrids). The lower layer receives the cross-voltage dynamic active power allowable range boundary issued by the upper layer. As a rigid limit constraint, an adaptive affine decision rule (ADR) is introduced to correlate the reactive power output of the end inverter with random active power disturbances in real time. Its objective function aims to minimize the risk of voltage violations at the end node and the active photovoltaic reduction penalty, thereby strictly suppressing the cross-voltage reverse power flow overvoltage cascading violations caused by end active power backfeed under a given safety confidence level. In the formula: The reference voltage rating for each node; The penalty coefficient for proactive merit reduction. For the first j Active power reduction of a distributed power source; It is a set of fuzzy probability measures constructed based on the statistical characteristics of historical uncertainty data. The symbol for conditional probability. Given the maximum confidence level (overvoltage tolerance probability).

[0033] Furthermore, this invention proposes a distributed parallel solution and closed-loop iterative method based on master-slave splitting theory: The master control layer will calculate the equivalent boundary power of each vertical boundary section. , Treating the boundary conditions as known, independently solve the main network optimization model, and update the boundary node voltage magnitude commands. The message is sent vertically down to each sub-autonomous level. Each sub-autonomous level receives it. Subsequently, a rigid bound is used as the local relaxed node to independently and in parallel solve the lower-level adaptive affine absorption model, without leaking the internal topology and inverter capacity limits. The latest boundary equivalent active and reactive power flow obtained from the solution is then fed back to the master control layer. In the individual fitness functions of the master control layer and the sub-autonomous layer, a dynamic boundary penalty function term based on physical residuals is introduced. This drives the boundary algebraic difference between the master and slave sides to rapidly approach zero: In the formula, and These are the boundary active and reactive power flows from the perspective of the main control layer; and These are the boundary equivalent active and reactive power flows calculated and uploaded by the sub-autonomous layer, respectively. and This represents the penalty coefficient for the dynamic boundary physical residual. Through the constraint of this physical penalty function, the distributed algorithm adaptively generates a guiding vector during the iteration process, driving the boundary algebraic difference between the master and slave sides to rapidly approach zero.

[0034] The system exchanges the aforementioned cross-sectional state variables between the main control layer and the dominant nodes in each region, continuously alternating and iterating in a loop. It uses the selection, crossover, and mutation operators of a genetic algorithm to perform distributed closed-loop optimization until the maximum absolute error of the cross-voltage boundary state residuals between two adjacent iterations and the deviation of the multi-objective function of the entire network meet the set convergence accuracy criterion threshold. The iteration then terminates, outputting the globally optimal cross-voltage multi-resource region collaborative control strategy and issuing it for execution.

[0035] For example, the voltages of each node obtained by simulation using the method of the present invention in the improved 14-node system are shown in Figure 2, and the output of each reactive resource is shown in Table 1.

[0036] Table 1. Output of various reactive resources According to simulation results, compared with traditional voltage regulation methods, the method proposed in this invention effectively reduces the possibility of voltage exceeding the limit at each node in a cross-voltage system. Figure 3 In order to be in Figure 4 The voltage optimization of the present invention under the condition of high voltage side voltage exceeding the limit is shown in the 220-110-10kV system. According to the simulation results, the method proposed in this invention also improves the problem of excessive network voltage caused by high voltage side voltage exceeding the limit.

[0037] Example 2 As another aspect of the present invention, this embodiment also provides a cross-voltage reactive power resource collaborative control system, comprising: The integrated complex sensitivity matrix construction module is used to perform a first-order Taylor expansion based on the cross-voltage AC power flow equation and introduce active-reactive cross-coupling correction coefficients to construct the network-wide cross-voltage integrated complex sensitivity matrix. The horizontal network autonomous decoupling partitioning module is used to construct a comprehensive evaluation function for partitioning quality based on the cross-voltage electrical distance index, by introducing the profile coefficient and modularity, and to implement decoupling partitioning and establish the dominant node using an improved k-medoids clustering algorithm; The vertical cross-voltage coordinated control dual-layer optimization module includes an upper global coordination layer sub-module and a lower local flexible autonomous layer sub-module, which is used to construct a mathematical model for the linkage optimization of discrete assets and flexible resources in a vertical multi-level space. The distributed master-slave split parallel solution module is used to alternately iterate between the master control layer and the sub-autonomous layer. It introduces a dynamic boundary penalty function term based on physical residuals to drive the boundary state residuals to approach zero and output the globally optimal cooperative control strategy.

[0038] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0039] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for coordinated control of reactive power resources across voltage levels, characterized in that, include: Based on the cross-voltage power flow equations of the entire network, a first-order Taylor expansion is performed to establish a linear mapping relationship between the changes in node voltage magnitude and phase angle and the active and reactive power disturbances. An active-reactive cross-coupling correction coefficient is introduced to construct a comprehensive complex sensitivity matrix across voltage ranges for the entire network; A quality evaluation function is constructed based on the contour coefficient and modularity, and the dominant node of boundary state interaction is established. A dual-layer optimization mathematical model for vertical cross-voltage coordinated regulation is constructed. The upper-layer global coordination layer model faces the high-voltage trunk transmission network and the global hierarchical boundary, while the lower-layer local flexible autonomous layer model is embedded in each horizontal spatial partition and runs independently and in parallel. An adaptive affine decision rule is introduced to correlate the reactive power output of the terminal inverter with the random disturbance of active power in real time. The optimal cooperative control scheme is obtained by performing distributed parallel solution based on the master-slave split theory.

2. The method for coordinated control of reactive power resources across voltage levels according to claim 1, characterized in that, For nodes within any voltage level The contour coefficient is calculated as follows: in, For nodes The average electrical distance between the node and all other nodes in the same cluster; For nodes The average electrical distance between the nearest neighboring node and all other nodes in the cluster.

3. The method for coordinated control of reactive power resources across voltage levels according to claim 1, characterized in that, The mathematical expression for the modularity is: These are the elements of the adjacency weight matrix after complex sensitivity truncation. and They are nodes and The sum of sensitivities; It is the sum of the total equivalent electrical edge weights of the entire network; and These are the partition labels to which the nodes belong; when hour This indicates that the two nodes belong to the same partition.

4. The method for coordinated control of reactive power resources across voltage levels according to claim 1, characterized in that, The quality evaluation function is constructed as follows: in, and For adaptive weighting coefficients; through intervals The internal process performs a one-dimensional exhaustive step-by-step optimization to achieve a comprehensive evaluation function. Reaching the global maximum value This is the optimal number of autonomous cluster partitions.

5. The cross-voltage reactive power resource coordinated control method according to claim 4, characterized in that, The adaptive weighting coefficients are adjusted according to preset conditions, which include: Calculate the expected value and standard deviation of the silhouette coefficient sequence and modularity sequence over the entire optimization interval. Divide the standard deviation of each index by its corresponding mean to obtain the coefficient of variation of the silhouette coefficient. Coefficient of variation with modularity The weighting coefficients are obtained by normalizing the two coefficients of variation. 。 6. The method for coordinated control of reactive power resources across voltage levels according to claim 1, characterized in that, The objective function of the upper-level global coordination layer model is: Total active power loss for the entire network; This is the equivalent penalty cost coefficient for a single tap changer operation of an on-load tap-changing transformer. For the ratio adjustment amount, It aggregates all OLTC data across the network; For critical transmission sections l The weight of the current fluctuation, The deviation of the trend is a factor.

7. The method for coordinated control of reactive power resources across voltage levels according to claim 1, characterized in that, The objective function of the lower-level local flexible autonomous layer model is: And it satisfies the spatiotemporal bound constraints at the safe confidence level: In the formula: The reference voltage rating for each node; The penalty coefficient for proactive merit reduction. Let be the active power reduction of the j-th distributed power source; It is a set of fuzzy probability measures constructed based on the statistical characteristics of historical uncertainty data. The symbol for conditional probability. Given the maximum confidence level.

8. The method for coordinated control of reactive resources across voltage levels according to claim 1, characterized in that, The distributed parallel solution based on the master-slave splitting theory is as follows: The main control layer solves the main network optimization model and sends the updated boundary node voltage magnitude command vertically to each sub-autonomous layer. Each sub-autonomous layer solves the lower-level adaptive affine absorption model in parallel and feeds back the latest boundary equivalent active and reactive power flow to the main control layer. Furthermore, a dynamic boundary penalty function term based on physical residuals is introduced into the individual fitness function for iterative solution.

9. The method for coordinated control of reactive power resources across voltage levels according to claim 8, characterized in that, In the individual fitness functions of the master layer and the sub-autonomous layer, the dynamic boundary penalty function term based on physical residuals is introduced as follows: In the formula, and These are the boundary active and reactive power flows from the perspective of the main control layer; and These are the boundary equivalent active and reactive power flows calculated and uploaded by the sub-autonomous layer, respectively. and This is the penalty coefficient for the physical residual at the dynamic boundary. Through the constraint of this physical penalty function, the distributed algorithm will adaptively generate a guiding vector during the iteration process, driving the boundary algebraic difference between the master and slave sides to rapidly approach zero.

10. A cross-voltage reactive power resource collaborative control system, characterized in that, include: The integrated complex sensitivity matrix construction module is used to perform a first-order Taylor expansion based on the cross-voltage AC power flow equation and introduce active-reactive cross-coupling correction coefficients to construct the network-wide cross-voltage integrated complex sensitivity matrix. The horizontal network autonomous decoupling partitioning module is used to construct a comprehensive evaluation function for partitioning quality based on the cross-voltage electrical distance index, by introducing the profile coefficient and modularity, and to implement decoupling partitioning and establish the dominant node using an improved k-medoids clustering algorithm; The vertical cross-voltage coordinated control dual-layer optimization module includes an upper global coordination layer sub-module and a lower local flexible autonomous layer sub-module, which is used to construct a mathematical model for the linkage optimization of discrete assets and flexible resources in a vertical multi-level space. The distributed master-slave split parallel solution module is used to alternately iterate between the master control layer and the sub-autonomous layer. It introduces a dynamic boundary penalty function term based on physical residuals to drive the boundary state residuals to approach zero and output the globally optimal cooperative control strategy.

Images