Power grid power flow risk control method based on topology analysis

Abstract

The invention discloses a power grid power flow risk control method based on topology analysis, and relates to the technical field of power flow risk control, and the method comprises the following steps: obtaining first data of a power grid in real time, and constructing a node admittance matrix; when a switch displacement signal or distributed power supply power mutation is detected, topology reconstruction is carried out based on a node admittance matrix, and a dynamic topological graph is generated; establishing a fault current characteristic matrix based on the dynamic topological graph in combination with the bidirectional power flow characteristics of the active power grid; establishing an adaptive risk early warning model based on the dynamic topological graph and the fault current characteristic matrix, and outputting the overload probability and the voltage out-of-limit risk index of each branch; and performing analysis according to the branch overload probability and the voltage out-of-limit risk index, and generating a multi-target optimization control instruction set. Through adaptive topology adjustment, load flow calculation and a risk early warning model, automatic power grid optimization control can be realized, and in combination with convergence verification of a virtual impedance observer, effectiveness of topology adjustment and control operation is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of power flow risk control, and more specifically, to a power grid power flow risk control method based on topological analysis. Background Art

[0002] With the gradual access of distributed energy sources (such as wind energy, solar energy, etc.) and energy storage systems to the power grid, traditional power grids are facing increasingly complex operation challenges. Especially in the actual operation of active power grids, due to multi-directional power flows and frequent topological changes, it is difficult to predict and control the power flow state and security risks of the power grid. How to efficiently and dynamically identify and control the power flow risks in the power grid has become a key technical problem for improving the security, stability and reliability of the power grid.

[0003] Traditional power grid risk control methods mainly rely on static power flow analysis and simple protection measures, such as traditional power flow calculation, switch operation and fault clearing mechanisms. However, with the increasing complexity of the power grid structure, these methods face the following main problems: Lag and inaccuracy in topological adjustment. The real-time monitoring and adjustment of the power grid topology still rely on manual settings and offline analysis, which results in insufficient response ability of the system to topological changes and inability to make effective adjustments quickly during the operation of the power grid, leading to the inability to resume normal operation in time when a fault occurs.

[0004] Slow fault identification and response. Most of the existing fault diagnosis methods rely on traditional single monitoring data and cannot comprehensively consider the influence of each node, branch and distributed energy source in the power grid on the fault current. Especially when multiple fault types appear alternately, traditional methods cannot accurately and timely identify the fault and formulate response measures.

[0005] Disconnection between static control and dynamic adjustment. At present, many power grid control methods make decisions based on static parameters and fixed thresholds, while the power grid state often changes at any time. Especially in the case of load fluctuations and changes in the output of distributed energy sources, fixed control strategies often cannot meet the real-time requirements and cannot flexibly respond to the changes in the power grid.

[0006] For example, a method, device, equipment and medium for one-key sequence control simulation test of a substation disclosed in the invention patent announcement with the announcement number of: CN112698584B. The method includes receiving a control instruction sent by a monitoring device; determining whether the control instruction meets a preset condition; if so, adjusting the state of the corresponding topological node in the power flow calculation topological graph according to the control instruction to obtain a new power flow calculation topological graph; determining the electrical property data of each topological node on the new power flow calculation topological graph; and sending the electrical property data to the monitoring device, so that through real-time power flow calculation in this application, the entire simulation test environment is closer to the actual operation environment of the substation, greatly improving the test verification effect, persuasiveness and accuracy of the one-key sequence control, and further making the relationship between the change of the switch position and the power flow voltage, current and power fully conform to the physical laws of power operation, and further making the simulation test more reliable and effectively improving the accuracy of commissioning verification.

[0007] For example, a main-distribution integrated power grid closed-loop control verification platform and method considering distributed new energy disclosed in the invention patent announcement with the announcement number of: CN118567332A. The platform obtains the CIME models of the main and distribution grids and splices them into a grid integrated model, and performs main and distribution grid power flow iterative calculation and grid frequency simulation based on the power flow section and the output of power plants and distributed new energy at the next moment. The output of power plants and distributed new energy is considered during the calculation process, and the injection amount of the power flow calculation is updated based on the execution result of the dispatching control instruction. The closed-loop control verification platform of the present invention realizes the laboratory closed-loop control verification of the second-level iterative interaction with the actual power grid dispatching control instruction, and automatically generates a control strategy evaluation report after the test ends. The present invention is used for laboratory detection and provides a true-type environment for the verification of the new energy control strategy of the power grid without interfering with the operation of the actual power grid dispatching control system.

[0008] In the above-disclosed technical solutions, there are at least the following technical problems: Traditional topological analysis relies on offline modeling and cannot track the dynamic changes of the power grid topology in real time (such as the access of distributed power sources, line switching and reclosing). The high-density access of large-scale distributed photovoltaics will cause frequent reconstruction of the power grid topology, and the update period of the existing model usually lags behind the actual changes, which may cause power flow calculation deviation. At the same time, in an active power grid, the fault current characteristics are significantly different from those of traditional passive networks, resulting in the failure of the topological identification method based on the current direction. For example, the reverse power supply of distributed power sources may misjudge the fault isolation strategy and increase the power outage area.

[0009] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0010] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide a power grid power flow risk control method based on topological analysis. By introducing a virtual impedance observer for convergence verification, it can determine in real time whether the topological adjustment is effective. If the topological adjustment is ineffective, the system will automatically trigger an active islanding control based on impedance trajectory prediction to solve the problem that the high-density access of large-scale distributed photovoltaics will cause frequent reconstruction of the power grid topology. The update period of the existing model usually lags behind the actual changes, which may cause power flow calculation deviation problems.

[0011] To achieve the above object, the present invention provides the following technical solutions: A power grid power flow risk control method based on topological analysis, comprising the following steps: obtaining first data of the power grid in real time and constructing a nodal admittance matrix; when a switch position change signal or a distributed power source power mutation is detected, performing topological reconstruction based on the nodal admittance matrix and generating a dynamic topological graph; establishing a fault current characteristic matrix based on the dynamic topological graph in combination with the bidirectional power flow characteristics of the active power grid; constructing an adaptive risk early warning model based on the dynamic topological graph and the fault current characteristic matrix, and outputting the overload probability of each branch and the voltage over-limit risk index; analyzing according to the branch overload probability and the voltage over-limit risk index, and generating a multi-objective optimization control instruction set.

[0012] In a preferred embodiment, when a switch position change signal or a distributed power source power mutation is detected, performing topological reconstruction based on the nodal admittance matrix and generating a dynamic topological graph specifically includes: monitoring the switch position signal and the distributed power source output power in real time and calculating the power change rate; when a switch state change or the power change rate is greater than a preset power change threshold is detected, determining it as a topological change event; performing rough topological adjustment based on the switch position change information, and performing fine-grained correction by real-time comparison of the voltage phase differences of adjacent nodes to generate a dynamic topological graph with confidence weight.

[0013] In a preferred embodiment, performing rough topological adjustment based on the switch position change information specifically includes: defining a set of switches to be adjusted, calculating the path weight based on the current change amount of each branch where the switch is located as the weight basis; according to the path weight and based on the Dijkstra algorithm, using the current power grid nodes as the graph vertices and the switch connection relationship as the graph edges, searching for the minimum weight operation path from the initial topology to the target topology; updating the switch state according to the search result, and synchronously correcting the corresponding branch admittance in the nodal admittance matrix to generate a preliminary topological structure.

[0014] In a preferred embodiment, the fine-grained correction by real-time comparison of the voltage phase differences between adjacent nodes is specifically as follows: Based on the PMU synchronized phasor data, the voltage phase angles of each node are obtained in real time, and the phase differences between adjacent nodes are calculated. If the phase difference between adjacent nodes is greater than the allowable deviation during normal operation, it is marked as a wrong connection. The determinant change rate of the nodal admittance matrix after perturbation of the wrong connection branch is calculated, and it is judged whether it is an unexpected topological change. For the connections that pass the verification, confidence weights are assigned according to the phase difference consistency and the determinant change rate.

[0015] In a preferred embodiment, the generation of the dynamic topological graph with confidence weights is specifically as follows: Based on the nodal admittance matrix, a dynamic topological graph is constructed. When the topology after coarse adjustment satisfies the minimization of the switch operation path weight and the determinant change rate after fine-grained correction is less than the preset determinant change rate threshold, the final dynamic topological graph is obtained; otherwise, it returns to the step of searching for the minimum-cost path again.

[0016] In a preferred embodiment, the establishment of the fault current characteristic matrix based on the dynamic topological graph combined with the bidirectional power flow characteristics of the active power grid is specifically as follows: Each node and branch of the power grid are analyzed to obtain the type and location of the fault, and based on the dynamic topological graph, the power grid topology is updated to determine the connection status of each node and the operation status of each branch. According to the change of the power grid topology, the nodal admittance matrix is updated. According to the bidirectional power flow characteristics of the active power grid, a bidirectional power flow model is established, considering the influence of distributed power generation sources and energy storage units on the bidirectional flow of current. Through the nodal admittance matrix and the dynamic topological graph, combined with the bidirectional power flow model, power flow calculation is carried out to obtain the current characteristics of each node and branch under fault conditions, and a fault current characteristic matrix is constructed.

[0017] In a preferred embodiment, the construction of the adaptive risk warning model based on the dynamic topological graph and the fault current characteristic matrix, and the output of the overload probability of each branch and the voltage over-limit risk index are specifically as follows: The dynamic topological graph is converted into a graph theory characteristic matrix; the second data of each adjacent node during the fault of each node is extracted and a node fault characteristic vector is formed; based on the topological feature encoding layer in the deep learning network, the graph theory characteristic matrix and the node fault characteristic vector are concatenated into a multi-dimensional input tensor, and feature learning is carried out based on the improved deep belief network architecture. Historical operation data is obtained to construct a training data set, and an adaptive risk warning model is obtained by using the training process of the deep belief network with unsupervised pre-training and supervised fine-tuning.

[0018] In a preferred embodiment, the feature learning based on the improved deep belief network architecture is specifically as follows: extracting the power grid topology structure features, power flow distribution features, and fault current features layer by layer based on the restricted Boltzmann machine to obtain an abstract representation of the fault features; calculating the overload probability of each branch based on the abstract representation of the fault features using a Softmax classifier, and calculating the voltage over-limit risk index through a linear regression layer.

[0019] Technical effects and advantages of the power grid power flow risk control method based on topological analysis of the present invention: 1. The present invention obtains the first data (SCADA data, PMU phasor data, smart meter data) of the power grid in real time and performs topological reconstruction based on the nodal admittance matrix. By using the dynamic topology map of the power grid and combining the fault current feature matrix and the deep learning model, it realizes real-time quantitative evaluation of the overload probability and voltage over-limit risk of each branch of the power grid. Through precise adjustment of the topology structure and a data-driven risk warning system, it can instantly conduct risk assessment and response when the topology of the power grid changes, effectively improving the stability and security of the power grid.

[0020] 2. The present invention solves the overload risk and voltage over-limit risk of the power grid through the particle swarm optimization algorithm and generates a control instruction set based on the optimization results. This strategy includes control means such as load transfer, voltage regulation, switch operation, and energy storage scheduling, which can automatically adjust the operating state of the power grid in case of emergency and reduce the overload risk and voltage over-limit risk. By combining the convergence verification of the virtual impedance observer, it ensures the effectiveness of the topology adjustment and control operations, thereby maximizing the operating stability and security of the power grid. Brief Description of the Drawings

[0021] Figure 1 It is a schematic flow chart of the power grid power flow risk control method based on topological analysis of the present invention. Detailed Embodiments

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] Embodiment 1, Figure 1 The power grid power flow risk control method based on topological analysis of the present invention is given, including the following steps: S1, obtaining the first data of the power grid in real time and constructing a nodal admittance matrix; The first data includes SCADA data and PMU phasor data; SCADA data: From the power grid monitoring system, which usually contains information such as voltage, power, load, current, etc. of each node. The SCADA system is the core of power grid monitoring, through which the real-time operation status of the entire power grid can be obtained.

[0024] PMU phasor data: The phasor measurement unit provides synchronous phasor data, including voltage phasors and phase current phasors of each node. PMU data can provide high-precision real-time power grid status data, especially suitable for dynamic state monitoring.

[0025] The real-time acquisition of the first power grid data and the construction of the nodal admittance matrix are specifically as follows: The intelligent control system is used to real-time acquire the first power grid data and perform data cleaning; According to the topological structure of the power grid, all nodes and branches are determined; Based on the first data after data cleaning, the self-admittance and mutual admittance of the nodes are calculated respectively; Based on the calculated self-admittance and mutual admittance, the elements of the nodal admittance matrix are filled to obtain the nodal admittance matrix.

[0026] The self-admittance of the node is obtained by calculating the sum of the admittances of all branches connected to the node.

[0027] The mutual admittance is obtained by calculating the negative value of the admittance of the branch connecting any two nodes.

[0028] S2. When a switch position change signal or a sudden change in distributed power generation power is detected, topology reconstruction is performed based on the nodal admittance matrix and a dynamic topology graph is generated; When a switch position change signal or a sudden change in distributed power generation power is detected, topology reconstruction is performed based on the nodal admittance matrix and a dynamic topology graph is generated, specifically as follows; The switch position signal and the output power of the distributed power generation are monitored in real time and the power change rate is calculated; When a change in switch state or a power change rate greater than a preset power change threshold is detected, it is determined as a topology change event; Based on the switch position change information, topology rough adjustment is performed, and fine-grained correction is performed through real-time comparison of the voltage phase differences of adjacent nodes to generate a dynamic topology graph with confidence weight.

[0029] The topology rough adjustment based on the switch position change information is specifically as follows: A set of switches to be adjusted is defined, and the path weight is calculated based on the current change amount of each branch where the switch is located as the weight basis; According to the path weight and based on the Dijkstra algorithm, with the current power grid nodes as the graph vertices and the switch connection relationship as the graph edges, the minimum weight operation path from the initial topology to the target topology is searched; Update the switch status according to the search results, synchronously correct the corresponding branch admittance in the nodal admittance matrix, and generate a preliminary topological structure.

[0030] The change in the switch status is specifically as follows: The change in the switch status includes the change in the status of the tie switch and the sectionalizing switch.

[0031] The path weight is specifically as follows:

[0032] Among them, is the path weight, is the current change amount.

[0033] The fine-grained correction by real-time comparison of the voltage phase differences of adjacent nodes is specifically as follows: Based on the PMU synchronized phasor data, obtain the voltage phase angles of each node in real time, and calculate the phase difference between adjacent nodes; If the phase difference between adjacent nodes is greater than the allowable deviation for normal operation, mark it as a wrong connection; Calculate the determinant change rate of the nodal admittance matrix after perturbation of the wrong connection branch and the original matrix, and determine whether it is an unexpected topological change; For the connections that pass the verification, assign a confidence weight according to the phase difference consistency and the determinant change rate.

[0034] The determination of whether it is an unexpected topological change is specifically as follows: If the determinant change rate is greater than the preset determinant change rate threshold, determine it as an unexpected topological change and send a warning for manual confirmation.

[0035] The determinant change rate is specifically as follows:

[0036] The confidence weight is specifically as follows:

[0037] Among them, is the determinant change rate, is the determinant of the nodal admittance matrix after perturbation, is the determinant of the original nodal admittance matrix, is the confidence weight, is the phase difference between adjacent nodes.

[0038] The generation of a dynamic topological graph with confidence weights is specifically as follows: Based on the nodal admittance matrix, a dynamic topology graph is constructed. When the topology after coarse adjustment satisfies the minimization of the switch operation path weight and the change rate of the determinant after fine-grained correction is less than the preset determinant change rate threshold, the final dynamic topology graph is obtained; otherwise, return to the step of re-searching for the minimum-cost path.

[0039] The dynamic topology graph, where: Vertices: power grid nodes, marked with real-time voltage amplitude and confidence weight; Edges: weighted directed edges, with the weight being the branch admittance amplitude, and the direction being determined by the power flow direction (dynamically adjusted according to the output of distributed power sources in an active power grid), and incorrect connections are represented by dashed lines.

[0040] S3. Based on the dynamic topology graph and combined with the bidirectional power flow characteristics of the active power grid, establish a fault current characteristic matrix: The establishment of the fault current characteristic matrix based on the dynamic topology graph and combined with the bidirectional power flow characteristics of the active power grid is specifically as follows: Analyze each node and branch of the power grid to obtain the type and location of the fault, and based on the dynamic topology graph, update the power grid topology to determine the connection status of each node and the operation status of each branch; Update the nodal admittance matrix according to the change of the power grid topology; According to the bidirectional power flow characteristics of the active power grid, establish a bidirectional power flow model, considering the influence of distributed power sources and energy storage units on the bidirectional flow of current; Through the nodal admittance matrix and the dynamic topology graph, combined with the bidirectional power flow model, perform power flow calculations to obtain the current characteristics of each node and branch under fault conditions, and construct a fault current characteristic matrix.

[0041] The establishment of the bidirectional power flow model according to the bidirectional power flow characteristics of the active power grid, considering the influence of distributed power sources and energy storage units on the bidirectional flow of current, is specifically as follows: Define the current flow direction in each node and branch of the power grid, and consider the forward power flow and reverse power flow respectively; Calculate the current flow direction under fault conditions, and inversely affect it through the power flow calculation model to ensure that the influence of the forward and reverse power flows on the current is accurately reflected; Determine the influence of the reverse power flow provided by distributed power sources or energy storage units on the current path and current amplitude when a fault occurs.

[0042] The fault current characteristic matrix is specifically as follows: The amplitude of the fault current, representing the amplitudes of the forward power flow current and the reverse power flow current respectively; The current flow direction, clarifying whether the current is transmitted from the upstream to the load end or from the local power source to the power grid; The duration of the current, specifically describing the occurrence and disappearance duration of the fault current; Calculate the change of the current flow path, especially the adjustment of the current path under the influence of reverse power flow.

[0043] The types of the faults include single-phase grounding fault, three-phase grounding fault, short-circuit fault, and open-circuit fault.

[0044] S4. Construct an adaptive risk early-warning model based on the dynamic topology graph and the fault current feature matrix, and output the overload probability of each branch and the voltage over-limit risk index. The construction of the adaptive risk early-warning model based on the dynamic topology graph and the fault current feature matrix, and the output of the overload probability of each branch and the voltage over-limit risk index are specifically as follows: Convert the dynamic topology graph into a graph theory feature matrix, where the graph theory feature matrix includes an adjacency matrix, a node parameter vector, and a branch parameter vector. Extract the second data of each adjacent node when each node fails and form a node fault feature vector, where the second data includes current amplitude, phase difference, and flow direction feature. Based on the topology feature encoding layer in the deep learning network, splice the graph theory feature matrix and the node fault feature vector into a multi-dimensional input tensor, and perform feature learning based on the improved deep belief network architecture. Obtain historical operation data to construct a training data set, and adopt the training process of the deep belief network with unsupervised pre-training and supervised fine-tuning to obtain an adaptive risk early-warning model.

[0045] The obtaining of historical operation data to construct a training data set, and the adoption of the training process of the deep belief network with unsupervised pre-training and supervised fine-tuning to obtain an adaptive risk early-warning model are specifically as follows: Construct a training data set, including normal, abnormal, and fault scenarios, collect historical operation data, and perform data augmentation and normalization processing. Through the training process of unsupervised pre-training and supervised fine-tuning, layer by layer train the deep belief network, use the Adam optimizer to minimize the total loss function, and obtain a trained adaptive risk early-warning model. Based on the adaptive risk early-warning model, quantitatively evaluate the overload probability of each branch and the voltage over-limit risk index of the node, and visually display the risk distribution through the dynamic topology graph to highlight the high-risk branches and nodes.

[0046] The performing of feature learning based on the improved deep belief network architecture is specifically as follows: Based on the restricted Boltzmann machine, layer by layer extract the power grid topology structure features, power flow distribution features, and fault current features to obtain an abstract representation of the fault features. Based on the abstract representation of the fault features, calculate the overload probability of each branch according to the Softmax classifier, and calculate the voltage over-limit risk index through the linear regression layer.

[0047] S5. Analyze based on the branch overload probability and the voltage violation risk index, and generate a multi-objective optimization control instruction set.

[0048] The analysis based on the branch overload probability and the voltage violation risk index and generating a multi-objective optimization control instruction set is specifically as follows: Calculate the overload probability of each branch and the voltage violation risk index of each node respectively, and compare them with the preset first and second thresholds respectively; If both are greater than the thresholds, construct a multi-objective optimization control model with the goal of minimizing the branch overload probability and the voltage violation risk index and maximizing the power grid stability; Solve the multi-objective optimization control model based on the particle swarm optimization algorithm, and conduct convergence verification to obtain the result of multi-objective optimization; Generate a control instruction set based on the result of multi-objective optimization.

[0049] The conduct of convergence verification is specifically as follows: For the power grid after topology adjustment, calculate the trajectory of the virtual impedance through a virtual impedance observer and make a judgment; If the trajectory of the virtual impedance converges to the preset safety region (i.e., restricted within a certain preset impedance range) in the complex plane, it is determined that the topology adjustment is effective, the power grid operation is stable, and the topology reconstruction has been successfully completed; If the virtual impedance trajectory does not converge to the preset safety region (i.e., the impedance value exceeds the safety range), it is determined that the topology adjustment is invalid, and trigger the active islanding control based on the impedance trajectory prediction.

[0050] The control instruction set is specifically as follows: Prioritize adjusting the reactive power output of the new energy inverter closest to the fault point, sub-optimally adjust the combination state of the tie switches, verify the topology convergence in real time after execution of the control, and confirm that the control is effective when the node voltage variance is less than the preset voltage variance, otherwise trigger the secondary correction based on the impedance sensitivity.

[0051] Load transfer instruction: When the branch overload risk is relatively high, transfer part of the load to other branches to reduce the overload risk.

[0052] Voltage regulation instruction: At the nodes with relatively high voltage violation risk, adjust the output of the distributed power generation source, or enable voltage regulation equipment (such as voltage regulating transformers, reactive power compensation equipment, etc.).

[0053] Switch operation instruction: In case of emergency, execute switch operations (such as cutting off part of the load or power source) to maintain the system stability.

[0054] Energy storage scheduling instruction: When the system load is too heavy or the voltage exceeds the limit, the energy storage device is scheduled to supplement or absorb energy to alleviate the risk.

[0055] Power flow optimization instruction: Optimize the power flowing in the power grid to ensure system stability and avoid overload or voltage over-limit.

[0056] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to get a formula closest to the real situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0057] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0058] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0059] In addition, in each embodiment of this application, the functional modules can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0060] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0061] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A power grid flow risk control method based on topology analysis, characterized in that: The following steps are involved: Obtain the first data of the power grid in real time and build the node admittance matrix; When a switch position change signal or a sudden change in the power of a distributed power source is detected, the topology is reconstructed based on the node admittance matrix and a dynamic topology diagram is generated; Based on the dynamic topology diagram and the bidirectional power flow characteristics of the active power grid, the fault current characteristic matrix is ​​established: An adaptive risk warning model is built based on the dynamic topology diagram and fault current characteristic matrix to output the overload probability of each branch and the voltage over-limit risk index; An analysis is performed based on the branch overload probability and voltage over-limit risk index, and a multi-objective optimization control instruction set is generated.

2. The power grid flow risk control method based on topology analysis according to claim 1 is characterized in that: When a switch position change signal or a power mutation of a distributed power source is detected, topology reconstruction is performed based on the node admittance matrix and a dynamic topology diagram is generated, specifically: Monitor switch position signals and distributed power output power in real time and calculate power change rate; When a switch state change or a power change rate greater than a preset power change threshold is detected, it is determined as a topology change event; The topology is roughly adjusted based on the switch position information, and fine-grained correction is performed through real-time comparison of the voltage phase difference of adjacent nodes to generate a dynamic topology map with confidence weights.

3. The power grid flow risk control method based on topology analysis according to claim 2 is characterized in that: The topology rough adjustment based on the switch position change information is specifically as follows: Define the set of switches to be adjusted, and calculate the path weight based on the current change of each branch where the switch is located; Based on the Dijkstra algorithm according to the path weight, the current grid nodes are used as graph vertices and the switch connection relationships are used as graph edges to search for the minimum weight operation path from the initial topology to the target topology; The switch status is updated according to the search results, and the corresponding branch admittance in the node admittance matrix is ​​corrected synchronously to generate a preliminary topology structure.

4. The power grid flow risk control method based on topology analysis according to claim 3 is characterized in that: The fine-grained correction is performed by real-time comparison of the voltage phase differences of adjacent nodes, specifically: Based on PMU synchronized phasor data, the voltage phase angle of each node is obtained in real time, and the phase difference between adjacent nodes is calculated; If the phase difference between adjacent nodes is greater than the allowable deviation for normal operation, it is marked as an incorrect connection; Calculate the determinant change rate of the node admittance matrix after the perturbation of the wrongly connected branch and the original matrix, and determine whether it is an unexpected topology change; For connections that pass the verification, confidence weights are assigned based on the phase difference consistency and the determinant change rate.

5. The power grid flow risk control method based on topology analysis according to claim 4 is characterized in that: The generating of the dynamic topological graph with confidence weights is specifically as follows: Based on the node admittance matrix, a dynamic topology map is constructed. When the coarse-adjusted topology satisfies the minimization of the switch operation path weight and the determinant change rate after fine-grained correction is less than the preset determinant change rate threshold, the final dynamic topology map is obtained. Otherwise, the minimum cost path step is returned to be searched again.

6. The power grid flow risk control method based on topology analysis according to claim 5 is characterized in that: The fault current characteristic matrix is ​​established based on the dynamic topology diagram combined with the bidirectional power flow characteristics of the active power grid, specifically: Analyze each node and branch of the power grid to obtain the type and location of the fault, and update the power grid topology based on the dynamic topology map to determine the connection status of each node and the operating status of the branch; Update the node admittance matrix according to the changes in the power grid topology; According to the bidirectional power flow characteristics of the active power grid, a bidirectional power flow model is established to consider the bidirectional flow effects of distributed generation sources and energy storage units on current. Through the node admittance matrix and dynamic topology diagram, combined with the bidirectional power flow model, power flow calculation is performed to obtain the current characteristics of each node and branch under fault conditions, and construct the fault current characteristic matrix.

7. The power grid flow risk control method based on topology analysis according to claim 6 is characterized in that: The adaptive risk warning model is constructed based on the dynamic topology diagram and the fault current characteristic matrix to output the overload probability of each branch and the voltage over-limit risk index, specifically: Convert the dynamic topological graph into a graph feature matrix; Extracting the second data of each adjacent node when each node fails and forming a node failure feature vector; Based on the topological feature encoding layer in the deep learning network, the graph feature matrix and the node fault feature vector are concatenated into a multi-dimensional input tensor, and feature learning is performed based on the improved deep belief network architecture; Historical operation data is obtained to construct a training data set, and a deep belief network training process of unsupervised pre-training and supervised fine-tuning is adopted to obtain an adaptive risk warning model.

8. The power grid flow risk control method based on topology analysis according to claim 7 is characterized in that: The feature learning based on the improved deep belief network architecture is specifically as follows: Based on the restricted Boltzmann machine, the grid topology structure characteristics, power flow distribution characteristics and fault current characteristics are extracted layer by layer to obtain the abstract representation of fault characteristics. The overload probability of each branch is calculated based on the Softmax classifier according to the abstract representation of fault features, and the voltage over-limit risk index is calculated through the linear regression layer.

9. The power grid flow risk control method based on topology analysis according to claim 8 is characterized in that: The determinant change rate is specifically: The confidence weight is specifically: in, is the determinant rate of change, is the determinant of the node admittance matrix after perturbation, is the determinant of the original node admittance matrix, is the confidence weight, is the phase difference between adjacent nodes.

Images