Power grid dispatching error prevention method and system based on topological analysis

Abstract

The present invention discloses a power grid dispatching error prevention method and system based on topological analysis, which relates to the technical field of power grid dispatching error prevention, and includes the following steps: constructing a telecontrol signal credibility evaluation model based on telecontrol signal evaluation data to obtain high-credibility telecontrol signal data; performing time-series logic verification based on the high-credibility telecontrol signal data and multi-source data fusion to generate a power grid topological model; partitioning the power grid topological model based on spectral clustering and state-driven edge weight method, and performing boundary equivalent processing to obtain several topological modules; performing simulated dispatching operations on the several topological modules based on the parsing result of the dispatching order objectification, and performing path traversal based on the depth-first search method, and combining the reinforcement learning method to perform error prevention analysis on the traversed path. The power grid dispatching error prevention method of the present invention combines telecontrol signal credibility evaluation, dynamic topological partitioning and reinforcement learning algorithm, comprehensively improving the error prevention verification ability and intelligent level of the power grid dispatching system.

Description

Technical Field

[0001] The present invention relates to the technical field of power grid dispatching error prevention, and more specifically, to a power grid dispatching error prevention method and system based on topological analysis. Background Art

[0002] As the core support for the safe and stable operation of the power system, the correctness of dispatching operations in the power grid dispatching system is directly related to power supply reliability and system operation safety. Traditional power grid dispatching operations rely on manual experience and static power grid topological information. Especially in terms of preventing misoperations, it mainly relies on means such as rule bases, five-prevention systems, and process control for pre-event prevention and post-event accountability. When facing complex and changeable power grid operation environments, frequently adjusted wiring methods, and uncertainties in telemetry data, such methods show obvious lag and limited adaptability.

[0003] For example, a power grid analysis system and analysis method based on a topological diagram disclosed in the invention patent announcement with the publication number: CN119599285A includes: collecting power grid device information and topological connection relationship data, constructing a device status library and a power grid topological model; obtaining several node importance degrees and several edge connection tightness degrees according to the power grid topological model, and using a multi-scale strongly connected region finite division method to partition the power grid to obtain several power grid partitions; obtaining a comprehensive risk index of several power grid partitions according to several node importance degrees and several edge connection tightness degrees, and screening out power grid partitions exceeding a preset risk threshold to obtain risk partitions; collecting operation data of the risk partitions, and inputting the operation data into a fault prediction model constructed based on deep learning to obtain the predicted fault time and fault devices of the risk partitions. The present invention relates to the technical field of power grid risk analysis and solves the technical problems of low analysis efficiency and insufficient pertinence in existing power grid analysis methods.

[0004] For example, a method for local topology estimation of a power system based on substation measurement information disclosed in the invention patent announcement with the publication number: CN103413044B includes logically analyzing the telemetry and telecontrol data of the power system collected by the SCADA system, identifying suspicious telemetry and telecontrol data, and locating the substations to which the suspicious data belongs. Using the power and current measurement information collected in real time in a single or adjacent multiple substations, a linear programming model is established to perform real-time estimation of the local topological structure of the transmission network and correct the errors existing in the data. The method proposed by the present invention can accurately identify topological errors in substations through optimization principles, provide reliable input data for the state estimation program, and further reduce the interference of suspicious measurements and avoid local errors affecting global results.

[0005] Among the above-disclosed technical solutions, there are at least the following technical problems:

[0006] In the prior art, the processing of telecontrol signal data has problems such as low credibility and inability to perform dynamic verification. The current system usually defaults that the telecontrol signal data is true and credible, lacking a mechanism for quantitative evaluation and multi-source verification of telecontrol signals, which is likely to cause topological identification errors and then lead to scheduling decision-making mistakes. For example, phenomena such as telecontrol signal jitter, signal delay, or false reporting of device status often cannot be identified and corrected in a timely manner in the existing system;

[0007] Most traditional topological modeling methods are based on static graph models or manual input methods, and it is difficult to adapt to the characteristics of frequent changes in the power grid structure and dynamic evolution of the operating state. This method cannot reflect the conduction state and operating logic of power grid equipment in real time, severely restricting the intelligent level and error prevention ability of scheduling decision-making. Moreover, traditional path verification mainly relies on rule matching, and it is unable to effectively discover potential linkage risks in complex logic paths or the possibility of incorrect operations under boundary conditions, further reducing the effectiveness of error prevention verification of scheduling instructions.

[0008] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0009] In order to overcome the above defects of the prior art, embodiments of the present invention provide a power grid scheduling error prevention method and system based on topological analysis. By combining a power grid scheduling error prevention method of telecontrol signal credibility evaluation, dynamic topological partitioning, graph neural search, and reinforcement learning algorithm, it is possible to realize dynamic simulation, error prevention verification, and intelligent path discrimination of the entire process of scheduling instructions, comprehensively improving the error prevention verification ability and intelligent level of the power grid scheduling system.

[0010] To achieve the above object, the present invention provides the following technical solutions:

[0011] A power grid scheduling error prevention method based on topological analysis includes the following steps: constructing a telecontrol signal credibility evaluation model based on telecontrol signal evaluation data to obtain high-credibility telecontrol signal data; performing sequential logic verification based on the high-credibility telecontrol signal data and multi-source data fusion to generate a power grid topological model; partitioning the power grid topological model based on spectral clustering and state-driven edge weight method, and performing boundary equivalent processing to obtain several topological modules; performing simulated scheduling operations on the several topological modules based on the parsing result of the scheduling instruction objectification, and performing path traversal based on the depth-first search method, and combining the reinforcement learning method to perform error prevention analysis on the traversed path.

[0012] In a preferred embodiment, time-series logic verification is performed based on highly reliable telecontrol data and multi-source data fusion to generate a power grid topology model. Specifically: Obtain primary equipment, and establish an initial structure diagram based on the power connection relationship and physical location; Perform time-series logic verification on the highly reliable telecontrol data based on SOE and oscillogram data; According to the verified telecontrol data, obtain the device operation state, where the device operation state is obtained by parsing the telecontrol data based on the mapping relationship between the verified telecontrol data and the primary equipment; Dynamically update the initial structure diagram according to the device operation state and the graph traversal algorithm, and perform connected component analysis on the updated initial structure diagram to identify the main wiring form of the substation and generate a power grid topology model.

[0013] In a preferred embodiment, the time-series logic verification of the highly reliable telecontrol data based on SOE and oscillogram data is specifically as follows: Obtain the telecontrol change time and the telecontrol state change point in the telecontrol data; Match and compare the telecontrol change time with the event time series registered in the SOE, and judge whether the telecontrol change is triggered by a real operation; Correlate the telecontrol state change point to the digital quantity change trajectory in the oscillogram data, extract the waveform slope mutation points within a preset time window, and judge whether there is a corresponding change; Obtain the telecontrol data with real operation of the verified telecontrol change and corresponding change.

[0014] In a preferred embodiment, the power grid topology model is partitioned based on spectral clustering and state-driven edge weight method. Specifically: Obtain the first several eigenvectors of the Laplacian matrix based on the normalized Laplacian matrix to form a spectral embedding space. The normalized Laplacian matrix is constructed based on the weighted adjacency matrix, and the weighted adjacency matrix includes state perception edge weight, logical linkage strength weight value, and historical misoperation statistical value; Use spectral clustering to perform embedding mapping on the power grid topology based on the weighted adjacency matrix, project the nodes into the spectral embedding space, and then introduce k-Means to perform node clustering on them; Obtain the partition category of each node after clustering and perform boundary division of the power grid topology model to obtain several subgraphs.

[0015] In a preferred embodiment, boundary equivalent processing is performed to obtain a number of topological modules. Specifically: traverse the edges of the several subgraphs to identify boundary nodes and cross-region edges; classify and process the cross-region edges according to the switch states, retain the cross-region edges that need to be impedance-equivalent processed, and obtain the operation data of the cross-region edges; introduce equivalent nodes in the several subgraphs, and set the initial voltage values of the equivalent nodes according to the operation data; connect the boundary nodes in the subgraphs and the equivalent nodes through impedance elements to form virtual branches; detect the boundary line states, and obtain the voltage values of each boundary node in real time to dynamically correct the initial voltage values of the equivalent nodes; store each subgraph after boundary equivalent processing as a topological module, and perform voltage and power injection simulation based on the virtual branches and boundary nodes for comparison and verification to obtain a number of topological modules.

[0016] In a preferred embodiment, based on the dispatching instruction objectified parsing result, perform simulated dispatching operations on a number of topological modules, and perform path traversal based on the depth-first search method, and combine the reinforcement learning method to perform error prevention analysis on the traversed paths. Specifically: write dispatching instructions based on graphical mock dispatching instructions and text mock dispatching instructions; perform objectified parsing on the dispatching instructions, decompose them into standardized digital tasks, and parse them into steps with the smallest granularity of individual device operations based on error prevention rules; perform simulated binding on the objectified dispatching instructions and the corresponding topological modules, and perform simulated execution based on the steps with the smallest granularity to update the topological model in real time; obtain the operation graph structure based on the updated topological model, and the operation graph structure is connected by each device node and logical conduction relationship in the topological model; traverse the conduction paths of the device nodes in the operation graph structure through the depth-first search method, and perform error prevention analysis on the conduction paths based on the reinforcement learning method.

[0017] In a preferred embodiment, traverse the conduction paths of the device nodes in the operation graph structure through the depth-first search algorithm, and perform error prevention analysis on the conduction paths based on the reinforcement learning method. Specifically: obtain the device state data at the starting point of the depth-first search algorithm as the state in reinforcement learning; take all the sets of dispatching instructions that can be executed under the current starting point as the actions in reinforcement learning; train the reinforcement learning policy Agent based on the probability of selecting the corresponding action under the current state; construct a path screening reward function; construct a reinforcement learning state-action model according to the state, action reinforcement learning policy Agent, and path screening reward function; based on the reinforcement learning state-action model, predict the path screening values for all the conduction paths traversed by the depth-first search algorithm; based on the path screening values, perform path judgment and mark potential error prevention risks.

[0018] In a preferred embodiment, the time jitter entropy data, and the specific acquisition method is as follows:

[0019] Suppose that a remote signaling point has undergone several position changes within a preset time window, and a time interval vector is obtained by calculating the sequence of adjacent position change time intervals; the time interval vector is mapped into several intervals with a fixed width for binning operation, the number of intervals contained in each interval is counted, and it is normalized into a probability distribution to obtain a probability density; based on the probability density, Shannon entropy is calculated to obtain time jitter entropy data.

[0020] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0021] 1. By introducing a remote signaling credibility evaluation model, integrating multi-source data indicators such as time jitter entropy, state mutation rate, and multi-channel signal consistency rate, and combining SOE and recorded wave data for timing logic verification, dynamic credibility judgment and abnormal elimination of remote signaling data are realized, and the accuracy and stability of power grid equipment state recognition are improved from the source. This method effectively avoids the interference risks of false alarms, missed reports, or glitch signals of remote signaling to topology modeling and dispatching operations, ensures that the data basis for constructing a dynamic topology structure is true and reliable, and the logical link is closed and rigorous, significantly enhancing the anti-misoperation locking mechanism and system security protection ability of dispatching operations, and has high engineering practical value and foresight.

[0022] 2. By introducing state-driven edge weight definition and spectral clustering algorithm to realize the dynamic partition of the power grid topology model, it can accurately identify logical disconnections and connectivity adjustments caused by changes in the state of switch equipment, and avoid the risk of misidentification caused by static processing of topology division; at the same time, combined with the boundary equivalent processing mechanism, the electrical effect of cross-region edges is retained after subgraph division, ensuring that each topology module has electrical integrity and boundary response ability in anti-misoperation analysis. This method significantly improves the real-time performance, accuracy, and robustness of partition modeling, supports the rapidity and local response ability of anti-misoperation analysis in power grid dispatching, takes into account global consistency and local sensitivity, and has extremely high practical application value.

[0023] 3. Through the standardized mapping and intelligent verification of dispatching instructions from text / graphical input to device-level operation sequences by means of order object parsing and depth-first search path analysis assisted by reinforcement learning, an anti-misoperation analysis model integrating operation sequences, device topology states, and high-risk node perception is constructed. Compared with the traditional static rule verification method, this method can not only dynamically simulate the process of order execution and update the topology state in real time, but also use the reinforcement learning mechanism to intelligently screen high-risk paths, identify illegal connections and locking logic conflicts in the search space, significantly improving the comprehensiveness, real-time performance, and intelligence level of anti-misoperation identification, and strongly supporting the automatic verification of dispatching instructions and operation risk prediction. Description of the Drawings

[0024] Figure 1Schematic flow chart of the power grid dispatching error prevention method based on topological analysis provided by the embodiment of the present application.

[0025] Figure 2 Schematic structural diagram of the power grid dispatching error prevention system based on topological analysis provided by the embodiment of the present application. Detailed implementation manners

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying 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.

[0027] Embodiment 1 Figure 1 Schematic flow chart of the power grid dispatching error prevention method based on topological analysis provided by the embodiment of the present application, including the following steps:

[0028] S1. Build a remote signal credibility evaluation model based on the remote signal evaluation data, and obtain remote signal data with high credibility.

[0029] The remote signal credibility dynamic evaluation mechanism ensures that the state of each device entering the topological recognition link is logically credible and stable in data. The essential goal of this step is to solve the common problems of remote signal anomalies, state errors, loss or false signals in actual engineering, and ensure that the device states on which the dynamic topological graph is constructed are real, credible, and stable, so as to avoid the risk of incorrect topological recognition or misoperation caused by remote signal anomalies.

[0030] The construction of the remote signal credibility evaluation model based on the remote signal evaluation data to obtain remote signal data with high credibility is specifically as follows:

[0031] Obtain the remote signal evaluation data and build a remote signal credibility evaluation model. The remote signal evaluation data includes time jitter entropy data, state mutation data, and multi-channel signal consistency rate data;

[0032] Obtain the remote signal data, and perform credibility evaluation on the remote signal points of each remote signal data based on the remote signal credibility evaluation model. If the output of the remote signal credibility evaluation model is greater than the preset credibility, the remote signal point at this time is of high credibility, and the remote signal data with high credibility is selected.

[0033] Furthermore, the remote signal evaluation data includes time jitter entropy data, state mutation data, and multi-channel signal consistency rate data.

[0034] Among them, the time jitter entropy data measures the uncertainty of the time interval between the state changes of telecontrol signals, that is, whether the state changes are regular. If the interval between telecontrol signal state changes is very random and the entropy value is high, it indicates that the signal stability is poor and there may be jitter or interference. Obtaining the time jitter entropy data to analyze the credibility of telecontrol data has the following benefits for the construction of the power grid topology model and subsequent error prevention:

[0035] Fine-grained characterization of the timing characteristics of telecontrol signal states: By constructing a sequence of intervals between telecontrol signal state changes and calculating its entropy value, the following situations can be identified: Strong regularity and low entropy value: Mostly caused by manual operations, normal protection / control-induced state changes; chaotic intervals and high entropy value: May be caused by abnormal factors such as communication jitter, electromagnetic interference, and contact burrs. This enables the topology recognition algorithm to achieve: Signal credibility weighted modeling based on entropy values; Dynamically determining the influence range of abnormal telecontrol signals, and improving the topology recognition accuracy from the source.

[0036] Under the trend of the intelligent dispatching platform evolving towards spatio-temporal coordination, state self-adaptation, and credible fusion, the limitations of traditional static telecontrol signal judgment methods are gradually emerging. The introduction of time jitter entropy provides an effective means to characterize the stability of telecontrol signals from the perspective of "time series statistics".

[0037] The specific method for obtaining the time jitter entropy data is as follows:

[0038] Suppose a telecontrol point undergoes several state changes within a preset time window, and a time interval vector is obtained by calculating the sequence of time intervals between adjacent state changes;

[0039] The time interval vector is mapped into several intervals with a fixed width for binning operations. The number of intervals contained in each interval is counted and normalized to a probability distribution to obtain the probability density;

[0040] Based on the probability density, Shannon entropy is calculated to obtain the time jitter entropy data.

[0041] The specific calculation formula for the sequence of time intervals between adjacent state changes is as follows:

[0042]

[0043] The specific calculation formula for the probability distribution is as follows:

[0044]

[0045] The specific calculation formula for the time jitter entropy data is as follows:

[0046]

[0047] In the formula, is the time interval between two adjacent state changes, is the event interval for the th remote signaling change, is the event interval for the th remote signaling change, is the probability density for the th interval, is the total number of remote signaling changes within a preset time window, is the number of time intervals falling into the th interval, is the time jitter entropy data, is the number of bins.

[0048] Among them, the state mutation data measures the rapid mutation trend of the remote signaling state within a period of time window. Essentially, it detects whether the remote signaling appears abnormal frequent jitter. By obtaining the state mutation data to analyze the credibility of the remote signaling data, it has the following benefits for the construction of the power grid topology model and subsequent anti-error:

[0049] Improve the credibility screening mechanism of remote signaling data: Through the state mutation frequency index, a dynamic remote signaling quality assessment model can be established to automatically identify the remote signaling points with abnormal frequent changes, and then achieve: automatic elimination or weight reduction of abnormal remote signaling signals; improvement of the robustness of the state judgment of specific equipment (such as disconnecting switches, earthing switches, etc.); reduction of the "false closing, false opening" phenomenon caused by glitch data in topology recognition.

[0050] Enhance the steady-state recognition ability of the topology model: Since the topology modeling depends on the remote signaling state to determine the state of circuit breakers and disconnecting switches, frequent jitter signals are likely to cause false states in the modeling logic. After the introduction of state mutation data: it can assist in identifying state jumps caused by "non-real operations" and filter out "non-effective operations";

[0051] Improve the misjudgment interception ability: When the remote signaling has a glitch and misjudges the device's open / closed state, the anti-error locking logic may be bypassed. After the introduction of the state mutation index: set a logical delay and redundant confirmation for the jitter signal; avoid the remote signaling misleading the judgment of the locking conditions of the operation ticket and improve the ability to prevent misoperations.

[0052] The specific acquisition method of the state mutation data is as follows:

[0053] Count the number of state mutations of the device in the past several time periods to form a time distribution function;

[0054] Calculate the mutation rate by calculating the standard deviation in the distribution function and the mean of the mutation time based on a preset mutation rate formula;

[0055] Compare and analyze the absolute value of the mutation rate based on a preset maximum mutation rate to quantitatively obtain the state mutation data.

[0056] The mutation rate has the following specific calculation formula:

[0057]

[0058] The state mutation data has the following specific calculation formula:

[0059]

[0060] In the formula, is the mutation rate, is the total number of state mutations within the statistical period, is the switching function, is the mean value of the mutation time, is the standard deviation, is the state mutation data, is the maximum mutation rate.

[0061] Among them, the multi-channel signal consistency rate data is for telemetry points with multiple tele-signal sources or redundant acquisitions (such as double-ended acquisition of bus disconnectors, primary and standby communication links), and calculates the time consistency rate of the states of different tele-signal channels. By obtaining the multi-channel signal consistency rate data to analyze the credibility of telemetry data, the following benefits are achieved for the construction of the power grid topology model and subsequent anti-error:

[0062] Enhance the robustness of tele-signal source fusion: In devices with multiple tele-signal sources (such as bilateral position tele-signals of disconnectors, concurrent tele-signals of control boxes + substation terminals): The consistency rate index can quantify the timing and state consistency degree of different signal channels; when the consistency rate is too low, a "signal conflict" handling mechanism can be triggered to prevent incorrect construction of the topology state.

[0063] Improve the redundant criterion ability of topology state modeling: Topology modeling often relies on the combined criterion of "switch device state + adjacent connection state", and the multi-channel consistency rate can be used as an auxiliary input for determining the credibility of state judgment signals.

[0064] Improve the credibility of interlocking condition judgment: In the interlocking judgment of operation ticket execution, if a certain telemetry point has multiple sources, the consistency rate can be used for: dynamically locking the acquisition selection: preferentially using the channel information with a high consistency rate; if the consistency rate is lower than the threshold, the system can automatically determine that the tele-signal is "unreliable" and trigger the strengthening of anti-error interlocking; avoid incorrect interlocking judgment caused by abnormalities in a certain signal channel.

[0065] The method for specifically obtaining the multi-channel signal consistency rate data is as follows:

[0066] Obtain the tele-signal state values of several acquisition channels of the telemetry point;

[0067] Under the unified time axis, the multi-channel signal consistency rate data is calculated based on a preset multi-channel signal consistency rate formula.

[0068] The multi-channel signal consistency rate formula is specifically as follows:

[0069]

[0070] In the formula, is the multi-channel signal consistency rate data, is the length of the detection time axis, is the start time of the detection time axis, is the th acquisition channel's remote signaling status value, is the indicator function, which is 1 when the condition is met, otherwise 0.

[0071] Based on the time jitter entropy data, state mutation data, and multi-channel signal consistency rate data, a remote signaling credibility evaluation model is constructed based on LSTM (Long Short-Term Memory Network).

[0072] The remote signaling credibility evaluation model is specifically as follows:

[0073]

[0074] In the formula, is the remote signaling credibility, is the multi-channel signal consistency rate data, is the state mutation data, is the time jitter entropy data, , , are the weights respectively.

[0075] S2. Based on the remote signaling data with high credibility and the fusion of multi-source data, perform timing logic verification to generate a power grid topology model.

[0076] In this embodiment, by performing logical verification on the timing dependence between multiple signals, signal anomalies (such as overstepping tripping, remote signaling jitter) can be identified, which helps to construct a dynamically evolving topology model and achieve state-driven structure modeling. Moreover, through timing logic analysis, the following can be achieved: logical consistency verification between multiple types of remote signaling (such as closing position, opening position, anomaly, remote control flag) of the same device; verifying the integrity of the timing chain of "action → remote signaling → protection response" by combining SOE and waveform recording; providing high-precision event confirmation support for the anti-misoperation locking and auxiliary decision-making system.

[0077] The timing logic verification based on the remote signaling data with high credibility and the fusion of multi-source data to generate a power grid topology model, where the multi-source data includes SOE and waveform recording data, specifically:

[0078] Obtain power grid equipment data, and parse the power grid equipment data to obtain primary equipment in the power grid. The power grid equipment data includes GIS diagrams, a system diagram, and a real-time database. The primary equipment includes switchgear, busbars, disconnectors, voltage transformers, and current transformers;

[0079] Based on the modeling method of edges and nodes in graph theory, establish an initial structure diagram for the primary equipment based on the power connection relationship and physical location. The modeling method includes node-branch or node-equipment-node;

[0080] Perform time-sequence logic verification on highly reliable telecontrol signal data based on SOE and oscillogram data;

[0081] According to the mapping relationship between the verified telecontrol signal data and the primary equipment, parse the telecontrol signal data into the equipment operation status;

[0082] Dynamically update the initial structure diagram according to the equipment operation status based on the graph traversal algorithm, perform connected component analysis on the updated initial structure diagram, identify the main wiring form of the substation, and generate a power grid topology model. The main wiring form includes single busbar, double busbar, and double busbar with bypass.

[0083] Furthermore, in the process of constructing the power grid topology model, comparing the change events of the telecontrol signal status with the physical event records in auxiliary data sources such as SOE (Sequence of Events data), oscillogram, and PMU in terms of time-sequence logic plays a key role in judging the authenticity of the telecontrol signal data and improving the reliability of the status. This is used to verify whether the telecontrol signal status truly reflects the changes in equipment operations or physical states. And the power grid topology modeling highly depends on the telecontrol signal status, such as whether the circuit breaker is closed or the disconnector is open, etc.; if there are false alarms in the telecontrol signals, it will directly lead to problems such as misjudgment of the topology structure like "the circuit breaker is actually open but modeled as closed". This step can effectively identify the abnormality of the telecontrol signals, eliminate untrustworthy signals, and improve the correctness of the topology model.

[0084] The performing time-sequence logic verification on highly reliable telecontrol signal data based on SOE and oscillogram data is specifically as follows:

[0085] Obtain the telecontrol signal change time and the telecontrol signal status change point in the telecontrol signal data;

[0086] Match and compare the telecontrol signal change time with the event time-sequence registered in the SOE. If the time difference is within the preset standard time difference and the event types match, it is determined that the telecontrol signal change is triggered by a real operation. The event types include equipment type and operation type, and the operation types include closing, tripping, and remote control, etc.;

[0087] Correspond the change points of the remote signal status to the digital quantity change tracks in the oscillographic recording data, extract the waveform slope mutation points within a preset time window, and determine whether there are corresponding changes. If there are corresponding changes, it indicates that there is a digital quantity jump event in the oscillographic recording, supporting that the remote signal change is a real physical change.

[0088] Obtain the remote signal data with real operations after verification of the remote signal change and with corresponding changes.

[0089] The waveform slope mutation point, the specific calculation formula is as follows:

[0090]

[0091] If within the window there exists:

[0092] such that , it indicates that there are corresponding changes;

[0093] In the formula, is the waveform slope mutation factor, is the time series value of the analog quantity in the oscillographic recording system, is the detected analog quantity mutation time, is the time series matching tolerance window, is the remote signal change time, is the slope mutation threshold.

[0094] S3. Based on spectral clustering and state-driven edge weight method, partition the power grid topology model and perform boundary equivalent processing to obtain several topology modules.

[0095] Fixed topology or fixed edge weight only considers the physical connection relationship and ignores the device state changes, which cannot reflect the device state changes, resulting in misclassifying disconnected devices and locked devices into the same area during partition, and unable to capture the logical disconnection or misconnection caused by the state. By introducing state-driven edge weight, the edge weight can dynamically reflect the current operating state (such as whether it is conducting, whether it is locked).

[0096] The partition of the power grid topology model based on spectral clustering and state-driven edge weight method is specifically as follows:

[0097] Construct a weighted adjacency matrix and dynamically adjust the weighted adjacency matrix according to the device operating state. The weighted adjacency matrix includes state perception edge weight, logical linkage strength weight value, and historical misoperation statistical value;

[0098] Construct a degree matrix according to the weighted adjacency matrix and construct a normalized Laplacian matrix in combination with the weighted adjacency matrix;

[0099] Obtain the first several eigenvectors of the Laplacian matrix to form a spectral embedding space, and use spectral clustering to perform an embedding mapping on the power grid topology based on the weighted adjacency matrix. After projecting the nodes into the spectral embedding space, introduce k-Means to perform node clustering on them;

[0100] Obtain the partition category to which each node belongs and perform boundary division on the power grid topology model to obtain several subgraphs.

[0101] It should be noted that the state-aware edge weight reflects the real-time availability and physical state of the connection, avoiding misjudgment caused by false tripping / rejection of tele-signals; the logical linkage strength weight represents the strength of the operational coupling or protection linkage relationship logically existing between nodes, making the topology division more in line with the dispatching cognition; the historical misoperation statistical value represents the frequency / impact degree of abnormal or accident caused by misoperation during the dispatching and operation and maintenance processes between equipment pairs in history, which can be used as a reference for adjusting the division boundary, so as to enhance the identification ability of the division for high-risk areas.

[0102] The weighted adjacency matrix is specifically:

[0103]

[0104] The degree matrix is specifically:

[0105]

[0106] The Laplacian matrix is specifically:

[0107]

[0108] In the formula, is the weighted adjacency matrix, is the equipment and the equipment the edge weight value between them, is the equipment the total strength with other equipment, is the total number of equipment, is the equipment and the equipment the connection edge weight of them, is the Laplacian matrix, is the identity matrix, is the inverse square root of the degree matrix, is the normalization of the weighted adjacency matrix of.

[0109] It should be noted that the weighted adjacency matrix is an n×n matrix used to represent the connection strength between devices in the power grid topology; the device operating states include when two devices are both closed, one device is open and the other is closed, or when a lock exists and both devices are open; the logical linkage strength between combined devices refers to when the node device is a combined device (such as a disconnecting switch - circuit breaker, earthing switch - isolating switch), a high coupling value is assigned.

[0110] Furthermore, the boundary equivalence method is performed on several subgraphs obtained after the partition, ensuring the completeness of the whole-network anti-error analysis after the topological partition. Through automatic power grid partitioning and boundary equivalence, a small-scale calculation model is formed, and each partition is calculated independently, reducing the overall operation scale, improving the calculation efficiency, and meeting the real-time anti-error requirements.

[0111] The boundary equivalence process is carried out to obtain several topological modules, specifically:

[0112] Traverse the edges of the several subgraphs to identify the boundary nodes and cross-region edges;

[0113] Classify and process the cross-region edges according to the switch states, and retain the cross-region edges that need impedance equivalence processing. The classification processing includes that when the switch is closed, impedance equivalence processing is required, and its electrical connection effect is retained; when the switch is open, the connection is ignored and its electrical effect is not modeled in the module.

[0114] Based on the retained cross-region edges, obtain the operation data of the cross-region edges, where the operation data includes real-time equivalent resistance, the voltage amplitude and phase of the boundary nodes, and the power flow of the boundary branches.

[0115] Introduce equivalent nodes in the several subgraphs, and set the initial voltage value of the equivalent nodes according to the operation data. The voltage value is obtained by comparing the voltage amplitude of the boundary nodes in the operation data with a preset voltage.

[0116] Connect the boundary nodes in the subgraphs to the equivalent nodes through impedance elements to form virtual branches;

[0117] Real-time detect the boundary line state, and obtain the voltage value of each boundary node, and dynamically correct the initial voltage value of the equivalent nodes.

[0118] Store each subgraph after the boundary equivalence process as a topological module, perform voltage and power injection simulations based on the virtual branches and boundary nodes, and conduct comparative verification to obtain several topological modules.

[0119] It should be noted that the comparative verification is carried out through voltage and power injection simulations, comparing with the results of the global model before modularization to ensure that the relative errors of voltage and power meet the preset thresholds.

[0120] S4. Based on the parsed result of the dispatching order objectification for the dispatching instruction, perform simulated dispatching operations on several topological modules, traverse the paths based on the depth - first search method, and perform error - prevention analysis on the traversed paths by combining the reinforcement learning method. Specifically:

[0121] Generate a dispatching instruction by writing the dispatching order based on the graphical draft dispatching order and the text draft dispatching order. The graphical draft dispatching order is based on the whole - network graph model data. Users can quickly generate dispatching instructions by graphical clicking, which is applicable to types including single - device operation, main - transformer status switching, bus - bar switching operation, bypass - banding task, etc. The text draft dispatching order is based on natural language processing, identifies the tasks of the text - input dispatching order, and maps the identified tasks to the specified graph to facilitate the subsequent generation of objectified steps based on this text.

[0122] Perform objectification parsing on the dispatching instruction, decompose it into standardized digital tasks, and further parse it into steps with the smallest granularity of single - device operation based on the error - prevention rules. Specifically:

[0123] Objectification of switch - state conversion: Parse the device - state conversion of the switch into a sequence of device - state switches of the switch, the disconnectors on both sides, and the relevant earthing switches; Objectification of bus - bar state conversion: Parse the state conversion of the bus - bar into a sequence of state switches of the bus - bar and its affiliated devices; Objectification of main - transformer state conversion: Parse the state conversion of the main - transformer into a sequence of state switches of the main - transformer and its affiliated devices; Objectification of line - state conversion: Parse the line - state conversion into the sequence of device - state conversions of the line local and the relevant earthing switches; Objectification of bypass - band state conversion: Divide it into two types: bypass substituting for line and bypass substituting for switch. Parse the bypass - band task state into the sequence of state switches of the switch, disconnector, and 4 - disconnector from the bypass to the specified device (switch, line); Objectification of bus - bar switching conversion: Intelligently distinguish between hot switching and cold switching, and objectify the task into the operation sequence of the relevant disconnectors and the bus - tie interval according to the bus - bar rules; Objectification of main - transformer inversion: Objectify the state conversion of the main - transformer into the sequence of state conversions of the three - side switches and the devices of each side's bus - tie and sectionalizing intervals according to the state - conversion principle.

[0124] Based on the device connection relationship, match the corresponding topological modules in several topological modules for the objectified dispatching instruction of the dispatching order, perform simulated binding, and sequentially simulate and execute each objectified step with the smallest granularity, and update the topological model in real - time.

[0125] Based on the operation - graph structure after updating the topological model, the operation - graph structure takes the device nodes in the topological model as graph vertices and the conduction relationship as edges.

[0126] Traverse the conduction paths of the device nodes in the operation - graph structure by the depth - first search method, and perform error - prevention analysis on the conduction paths by combining the reinforcement learning method.

[0127] It should be noted that the equipment whose state changes from the disconnected state to the conducting state due to a certain operation in the dispatching order (such as closing the switch or closing the disconnecting switch). The state changes of these equipment will directly affect the connected paths in the electrical topology and are the objects that must be key detected in the anti-error logic judgment.

[0128] Furthermore, the traditional DFS (Depth-First Search Algorithm) explores all paths exhaustively, without intelligence in path selection. And in large-scale topologies, DFS will generate a large number of redundant paths, resulting in low efficiency and lack of focus; by introducing reinforcement learning to provide a policy selection mechanism, among multiple possible dispatching paths, an intelligent selection is made of a verification path with high risk to improve the efficiency and quality of anti-error analysis and assist the dispatcher in quickly judging whether there are potential risks in the instructions.

[0129] Traversing the conducting paths of equipment nodes in the operation graph structure through the depth-first search algorithm and performing anti-error analysis on the conducting paths based on the reinforcement learning method specifically includes:

[0130] Obtain the equipment state data at the starting point of the depth-first search algorithm as the state in reinforcement learning, where the equipment state data includes operation permissions, locking states, and adjacent node states;

[0131] Take the set of all schedulable instructions that can be executed under the current starting point as the actions in reinforcement learning, and the set of all schedulable instructions includes closing the disconnecting switch, opening the disconnecting switch, putting into protection, and withdrawing the remote control interlock;

[0132] Train the reinforcement learning policy Agent based on the probability of selecting the corresponding action under the current state;

[0133] Construct a path screening reward function, where the path screening reward function includes that all operations in the path conform to the anti-error logic, there are high-risk misoperations in the path (such as closing the switch under the condition of ineffective locking), the reliability of the equipment remote signal is abnormal in the path, and the operation violates the anti-error interlock logic;

[0134] Construct a reinforcement learning state-action model according to the state, action, reinforcement learning policy Agent, and path screening reward function;

[0135] Based on the reinforcement learning state-action model, predict the path screening value for all the conducting paths traversed by the depth-first search algorithm;

[0136] Based on the path screening value, perform path judgment. When the path screening value is lower than the preset path screening threshold, mark the action of the corresponding scheduling instruction as a potential anti-error risk.

[0137] The specific calculation formula of the path screening reward function is as follows:

[0138]

[0139] In the formula, is the path screening value, is the anti-misoperation logic compliance reward value, is the high-risk misoperation penalty value, is the equipment telemetry signal anomaly penalty value, is the penalty value for violating the anti-misoperation interlock logic, is the anti-misoperation logic compliance reward value coefficient, is the high-risk misoperation penalty coefficient, is the equipment telemetry signal anomaly value penalty coefficient, is the penalty coefficient for violating the anti-misoperation interlock logic.

[0140] Among them, .

[0141] It should be noted that the operation diagram structure needs to be encoded in vector form, such as adjacency matrix embedding encoding, the operation sequence can be encoded by RNN, and the states are combined into a high-dimensional feature vector for input to the RL network. The reinforcement learning Agent refers to an intelligent agent that executes decisions, interacts with the environment, and continuously optimizes its strategy according to the reward feedback in the reinforcement learning framework. Its core task is to learn a strategy that maximizes the long-term reward accumulated in a given environment.

[0142] Example 2, Figure 2 is the schematic structural diagram of the power grid dispatching anti-misoperation system based on topological analysis provided by the embodiment of the present application, including a telemetry signal credibility evaluation module, a topological model construction module, a topological model processing module, and an anti-misoperation analysis module, and there are connections between the modules:

[0143] The telemetry signal credibility evaluation module is used to construct a telemetry signal credibility evaluation model based on the telemetry signal evaluation data and obtain high-credibility telemetry signal data;

[0144] The topological model construction module is used to perform time-sequence logic verification based on the high-credibility telemetry signal data and multi-source data fusion and generate a power grid topological model;

[0145] The topological model processing module is used to partition the power grid topological model based on spectral clustering and state-driven edge weight method and perform boundary equivalent processing to obtain several topological modules;

[0146] The anti-misoperation analysis module is used to perform simulated dispatching operations on several topological modules based on the dispatching instruction order objectification parsing result, perform path traversal based on the depth-first search method, and perform anti-misoperation analysis on the traversed path in combination with the reinforcement learning method.

[0147] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula that is closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0148] 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.

[0149] 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 document can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians 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.

[0150] In addition, the functional modules in each embodiment of this application can be integrated into a processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0151] 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.

[0152] 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 dispatching error prevention method based on topological analysis, characterized in that It includes the following steps: Construct a remote signaling credibility evaluation model based on remote signaling evaluation data to obtain remote signaling data with high credibility; Perform time-series logic verification based on the remote signaling data with high credibility and multi-source data fusion to generate a power grid topology model; Partition the power grid topology model based on spectral clustering and state-driven edge weight method, and perform boundary equivalent processing to obtain several topology modules; Perform simulated dispatching operations on several topology modules based on the parsing results of dispatching order objectification, traverse the paths based on the depth-first search method, and perform misoperation prevention analysis on the traversed paths in combination with the reinforcement learning method; The misoperation prevention analysis of the traversed paths in combination with the reinforcement learning method is specifically as follows: Obtain the device status data at the starting point of the depth-first search algorithm as the state in reinforcement learning; Take all the dispatching instruction sets that can be executed under the current starting point as the actions in reinforcement learning, and train the reinforcement learning policy Agent; Construct a path screening reward function, and construct a reinforcement learning state-action model in combination with the state, action, and reinforcement learning policy Agent. The path screening reward function includes that all operations in the path conform to the misoperation prevention logic, there are high-risk misoperations in the path, the credibility of device remote signaling in the path is abnormal, and the operation violates the misoperation prevention locking logic; For all the conducting paths traversed by the depth-first search algorithm, predict the path screening value based on the reinforcement learning state-action model, make path judgments, and mark potential misoperation prevention risks.

2. The method for preventing misoperation in power grid dispatching based on topological analysis according to claim 1, wherein The time-series logic verification based on the remote signaling data with high credibility and multi-source data fusion to generate a power grid topology model is specifically as follows: Obtain primary equipment and establish an initial structure diagram based on the power connection relationship and physical location; Perform time-series logic verification on the remote signaling data with high credibility based on SOE and waveform recording data; According to the verified remote signaling data, obtain the device operation status, and the device operation status is obtained by parsing the remote signaling data based on the mapping relationship between the verified remote signaling data and the primary equipment; Dynamically update the initial structure diagram according to the device operation status and the graph traversal algorithm, and perform connected component analysis on the updated initial structure diagram to identify the main wiring form of the substation and generate a power grid topology model.

3. The power grid dispatching error prevention method based on topological analysis according to claim 2, wherein, The time-series logic verification of the remote signaling data with high credibility based on SOE and waveform recording data includes: judging the authenticity of remote signaling position change operations and judging the corresponding position changes of remote signaling state change points; The judgment of the authenticity of remote signaling position change operations is obtained by comparing and analyzing the remote signaling position change time in the remote signaling data with the event time series registered in the SOE; The judgment of the corresponding position change of the remote signaling state change point is obtained by obtaining the digital quantity position change trajectory corresponding to the remote signaling state change point in the remote signaling data and extracting the waveform slope mutation points within the preset time window for judgment.

4. The method for preventing misoperation in power grid dispatching based on topological analysis according to claim 1, characterized in that, The partition of the power grid topology model based on spectral clustering and state-driven edge weight method is specifically as follows: Obtain the first several eigenvectors of the Laplacian matrix based on the normalized Laplacian matrix to form a spectral embedding space. The normalized Laplacian matrix is constructed based on the weighted adjacency matrix, and the weighted adjacency matrix includes state-aware edge weights, logical linkage strength weights, and historical misoperation statistics values; Spectral clustering is used to perform embedding mapping on the power grid topology based on the weighted adjacency matrix. After projecting the nodes into the spectral embedding space, k-Means is introduced to perform node clustering on them; Obtain the partition category of each node after clustering and perform boundary division on the power grid topology model to obtain several subgraphs.

5. The method for preventing misoperation in power grid dispatching based on topological analysis according to claim 4, wherein, The boundary equivalent processing is performed to obtain several topological modules, specifically: Traverse the edges of several subgraphs to identify boundary nodes and cross-region edges; Classify and process the cross-region edges according to the switch state, retain the cross-region edges that need to be impedance-equivalent processed, and obtain the operation data of the cross-region edges; Introduce equivalent nodes in several subgraphs and set the initial voltage value of the equivalent nodes according to the operation data; Connect the boundary nodes and equivalent nodes in the subgraph through impedance elements to form virtual branches; Detect the boundary line state and obtain the voltage value of each boundary node in real time to dynamically correct the initial voltage value of the equivalent nodes; Store each subgraph after boundary equivalent processing as a topological module, perform voltage and power injection simulation based on virtual branches and boundary nodes, and conduct comparison and verification to obtain several topological modules.

6. The method for preventing misoperation in power grid dispatching based on topological analysis according to claim 1, characterized in that, The following is the specific process of performing simulated dispatching operations on several topological modules based on the parsing result of the dispatching order objectification and performing anti-error analysis on the traversed path based on the depth-first search method and the reinforcement learning method: Generate dispatching instructions based on graphical and text-based draft dispatching orders; Perform objectification parsing on the dispatching instructions, decompose them into standardized digital tasks, and parse them into steps with the smallest granularity of single device operation based on anti-error rules; Simulate and bind the objectified dispatching instructions with the corresponding topological modules, and perform simulated execution based on the steps with the smallest granularity to update the topological model in real time; Obtain the operation graph structure based on the updated topological model. The operation graph structure is connected by each device node and logical conduction relationship in the topological model; Traverse the conduction paths of device nodes in the operation graph structure through the depth-first search method, and perform anti-error analysis on the conduction paths based on the reinforcement learning method.

7. The method for preventing misoperation in power grid dispatching based on topological analysis according to claim 2, wherein The telemetry evaluation data includes time jitter entropy data, and the specific acquisition method is as follows: Assume that a telemetry point has changed several times within a preset time window, and obtain the time interval vector by calculating the adjacent change time interval sequence; Map the time interval vector into several intervals with a fixed width for binning operation, count the number of intervals contained in each interval, and normalize it to a probability distribution to obtain the probability density; Calculate the Shannon entropy based on the probability density to obtain the time jitter entropy data.

8. A system using the power grid dispatching error prevention method based on topological analysis according to any one of claims 1-7, characterized in that, It includes a telemetry credibility evaluation module, a topological model construction module, a topological model processing module, and an anti-error analysis module. There are connections between the modules: The telemetry credibility evaluation module is used to construct a telemetry credibility evaluation model based on the telemetry evaluation data and obtain high-credibility telemetry data; The topological model construction module is used to perform time-series logic verification based on the high-credibility telemetry data and multi-source data fusion to generate a power grid topological model; The topology model processing module is used to partition the power grid topology model based on spectral clustering and state-driven edge weight method, and perform boundary equivalent processing to obtain several topology modules; The anti-error analysis module is used to perform simulated dispatching operations on several topology modules based on the parsing results of dispatching instruction objectification, traverse the paths based on the depth-first search method, and perform anti-error analysis on the traversed paths in combination with the reinforcement learning method.

Images