An intelligent substation secondary protection switching-on and switching-off anti-misoperation rule base optimization method

Abstract

The application discloses a kind of secondary protection of intelligent substation is put off and prevents mistake rule base optimization method, it is related to relay protection mistake prevention technical field, including the following steps: based on multi-source data, establish the knowledge graph as the carrier of mistake prevention rule base;Based on the equipment attribute and topological association in the knowledge graph, corresponding simulation model parameters are mapped and configured, and a digital twin simulation instance simulating the dynamic behavior of secondary system is constructed;Receive secondary protection switching instruction and use knowledge graph to deduce risk operation sequence, and dynamically simulate and check risk operation sequence, generate test feedback information;Based on test feedback information, identify and optimize secondary protection switching mistake prevention rules in knowledge graph.The present application is used to solve the problem that existing secondary protection mistake prevention rule base lacks adaptive optimization capability.

Description

Technical Field

[0001] This invention relates to the field of relay protection error prevention technology, and more specifically, to a method for optimizing the rule base for enabling and disabling secondary protection in intelligent substations to prevent errors. Background Technology

[0002] The activation and deactivation of secondary protection systems in intelligent substations is a crucial aspect of ensuring the safe operation of the power grid. Traditional secondary protection error prevention rule libraries suffer from incomplete rule coverage and a lack of real-time adaptive capabilities.

[0003] The current secondary protection rule base has the following shortcomings: First, the performance verification methods of the existing rule base mostly use static or low-fidelity judgment methods, lacking statistical quantitative analysis of the robustness of rules under complex working conditions, and making it difficult to accurately simulate the impact of secondary network communication latency and extreme external working conditions on the reliability of rules; Second, the optimization of the existing rule base usually relies on expert experience or simple accident backtracking, making it difficult to systematically diagnose logical deficiencies or conflicts in the rules, and the accuracy of the rules cannot be guaranteed.

[0004] Therefore, there is an urgent need to propose an optimization method for the rule base of secondary protection activation / deactivation in intelligent substations that can take into account both dynamic robustness verification and closed-loop adaptive optimization. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an optimization method for the rule base of secondary protection activation / deactivation in intelligent substations. By integrating a closed-loop feedback mechanism of multi-source knowledge graph and digital twin simulation dynamic verification, the method solves the problems of incomplete rule coverage and lack of adaptive optimization capability in the existing rule base for preventing errors.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for optimizing the rule base for preventing malfunctions in the secondary protection system of an intelligent substation includes the following steps: establishing a knowledge graph based on multi-source data as the carrier of the rule base; mapping and configuring corresponding simulation model parameters based on the equipment attributes and topology associations in the knowledge graph to construct a digital twin simulation instance simulating the dynamic behavior of the secondary system; receiving secondary protection activation / deactivation commands and deriving risk operation sequences using the knowledge graph, performing dynamic simulation verification on the risk operation sequences, and generating test feedback information. The risk operation sequences include pre-verification based on knowledge graph parsing commands and real-time data, followed by reverse tracing to deduce potential faults and generating risk operation sequences; and identifying and optimizing the secondary protection activation / deactivation malfunction prevention rules in the knowledge graph based on the test feedback information.

[0008] In a preferred embodiment, the establishment of the knowledge graph as the carrier of the anti-misoperation rule base includes: parsing the virtual terminal connection relationship based on the SCD file, and transforming the device entities, topological associations and initial anti-misoperation rules into a knowledge graph structure that can perform semantic reasoning according to the preset ontology model.

[0009] In a preferred embodiment, the construction of a digital twin simulation instance simulating the dynamic behavior of a secondary system includes: mapping device attributes and topological associations in a knowledge graph to a preset simulation component library; dynamically correcting key parameters of simulation components using state evaluation technology based on multi-source data; synchronously using semantic rules in the knowledge graph as logical verification benchmarks in the simulation process; and constructing a digital twin simulation instance by combining the corrected key parameters.

[0010] In a preferred embodiment, the step of dynamically correcting the key parameters of the simulation element based on multi-source data and using state assessment technology includes: initializing the basic parameters of the simulation element based on the static configuration data of the substation; extracting the features of the dynamic operation data of the substation and generating state indicators through state assessment; obtaining correction coefficients based on the state indicators and preset correction thresholds; and using the correction coefficients to dynamically correct the basic parameters to complete the dynamic configuration of the key parameters.

[0011] In a preferred embodiment, the risk operation sequence specifically includes: parsing secondary protection activation / deactivation instructions into an operation semantic set based on the ontology structure of a knowledge graph; acquiring real-time operating data of the secondary system and pre-verifying the secondary protection activation / deactivation instructions by combining the operation semantic set and the semantic rules of the knowledge graph; based on the pre-verification results, inferring the potential consequences using the fault modes defined in the knowledge graph, and using a graph traversal algorithm to perform reverse tracing under the topological association structure constraints of the knowledge graph to obtain the risk operation sequence.

[0012] In a preferred embodiment, the step of inferring potential faults through reverse tracing includes: extracting node state sequences on the reverse tracing path and using a phase space reconstruction method to map the node state sequences into state point cloud trajectories in phase space; performing multi-scale topological filtering on the state point cloud trajectories and calculating the continuous coherence features during the filtering process to generate a continuous barcode image; filtering long barcodes in the continuous barcode image that represent unexpected topological coherence features, and mapping the state point cloud trajectories corresponding to the long barcodes into temporal operation combinations as the risk operation sequence.

[0013] In a preferred embodiment, the step of dynamically simulating and verifying the risk operation sequence to generate test feedback information includes: configuring the initial state of the digital twin simulation instance based on the timing information and state constraints of the risk operation sequence, and constructing a simulation environment that includes secondary network anomalies and extreme operating conditions as boundary conditions; driving the digital twin simulation instance to execute the risk operation sequence in the simulation environment; and collecting test feedback information generated during the execution of the risk operation sequence, wherein the test feedback information includes secondary system behavior data and rule verification results.

[0014] In a preferred embodiment, the step of identifying and updating the secondary protection activation / deactivation anti-misoperation rules in the knowledge graph based on test feedback information includes: verifying the secondary system behavior data based on semantic rules, identifying abnormal actions and constructing an abnormal event chain; classifying the abnormal event chain to obtain a defect pattern set to identify logical defects in the semantic rules; obtaining the safety margin difference between the actual operating parameters in the secondary system behavior data and the limiting parameters in the semantic rules based on the rule verification results; determining the minimum safety margin based on the safety margin difference, and obtaining the parameter optimization space of the limiting parameters by comparing the limiting parameters and the minimum safety margin; generating correction instructions based on the logical defects and the parameter optimization space, and adaptively optimizing the secondary protection activation / deactivation anti-misoperation rules in the knowledge graph.

[0015] In a preferred embodiment, a clustering analysis algorithm is used to classify the abnormal event chain to obtain a defect pattern set, and a statistical safety and reliability analysis algorithm is used to analyze the safety margin difference to determine the minimum safety margin.

[0016] An electronic device includes: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the method for optimizing the rule base for preventing malfunctions in the secondary protection system of an intelligent substation.

[0017] The technical effects and advantages of the present invention regarding the optimization method for the rule base of secondary protection activation / deactivation in intelligent substations:

[0018] 1. This invention establishes a multi-source fusion knowledge graph based on multi-source data and uses the knowledge graph to deduce risk operation sequences, thereby upgrading rule verification from passive response to active prediction. This enables dynamic and structured management of rule knowledge for substation secondary systems, ensuring the security and foresight of the rules.

[0019] 2. This invention achieves automated construction and configuration of simulation models by generating digital twin simulation instances based on knowledge graphs and dynamically simulating and verifying risk operation sequences, thereby quantifying the dynamic robustness of rules and improving the reliability of rule base verification.

[0020] 3. This invention identifies and updates the secondary protection deployment and deactivation prevention rules in the knowledge graph based on test feedback information, thereby achieving data-driven adaptive optimization of the rule base, ensuring the integrity and accuracy of the rules, and improving the efficiency of the rule base's full lifecycle management. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of a method for optimizing the rule base for preventing malfunctions in the secondary protection system of an intelligent substation, as provided in an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of clustering analysis of feature vectors of abnormal event chains in an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of parameter optimization space analysis in an embodiment of the present invention.

[0024] Figure 4 This is a structural block diagram of an exemplary electronic device provided for implementing embodiments of the present disclosure. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Example 1, Figure 1 This invention presents an optimization method for the rule base of secondary protection activation / deactivation error prevention in intelligent substations, comprising the following steps:

[0027] S1, A knowledge graph is built based on multi-source data to serve as the carrier of the error prevention rule base;

[0028] In this embodiment, establishing the knowledge graph that serves as the carrier of the anti-misoperation rule base includes: parsing the virtual terminal connection relationship based on the SCD file, and transforming the device entities, topological associations, and initial anti-misoperation rules into a knowledge graph structure capable of semantic reasoning according to a preset ontology model, as detailed below:

[0029] S101. Collect multi-source data from the substation, including static configuration data, dynamic operation data, five-prevention logic definition files, and historical fault records. The specific steps are as follows:

[0030] Static configuration data is obtained by reading the substation's configuration description file (SCD file), which includes the protection device settings, software version number, communication parameters of the GOOSE / SV network, and the substation's physical topology connection information.

[0031] Dynamic operation data is acquired in real time through the manufacturing message specification protocol and GOOSE subscription mechanism. Specifically, this includes the real-time remote signaling status of substation switch quantities, the real-time telemetry values ​​of analog quantities, the action change information in the event sequence record, and communication status parameters that reflect the performance of the secondary network, such as the real-time traffic of the GOOSE / SV network, the queue depth of the switch port, and the processor load rate.

[0032] Obtain historical fault records from historical fault case databases or expert fault knowledge bases;

[0033] At the same time, obtain the five-prevention logic definition file containing the anti-misoperation interlocking logic expressions of all primary and secondary equipment in the substation, such as the logic rule library exported by the substation five-prevention host or the electronic operation procedure text.

[0034] S102. A pre-defined model based on the IEC61850 standard and relay protection principles is as follows:

[0035] The preset ontology model is a collection of classes, attributes and relationships defined in the ontology construction tool using the network ontology language based on the IEC61850 standard and relay protection principles. It includes equipment entity classes, topology association classes, anti-misoperation rule classes and fault mode classes.

[0036] The device entity class parses the logical nodes and common data class definitions defined in the SCD file, mapping the device model and rated parameters to static attributes, and the communication services and status interfaces to dynamic attribute interfaces, which are used to describe the inherent characteristics and operating status of primary and secondary devices.

[0037] The topology association class includes physical electrical connections, signal logical connections, and logical dependencies between devices; the physical electrical connections are obtained based on the single-line diagram description in the SCD file; the signal logical connections are defined based on virtual terminal configurations and are used to describe the transmission paths of control signals and protection action signals between different secondary devices; the logical dependencies include the interval to which the device belongs, the voltage level, and the associated protection relationship between the protection device and the primary device;

[0038] The anti-misoperation rule class is defined based on the relay protection principle and the five-prevention logic definition file. It includes the action time limit characteristics, setting constraints, verification logic and interlocking conditions of the protection function block. The verification logic is specifically the numerical constraint relationship between the setting value of the protection device and the real-time telemetry value. The interlocking conditions are used to ensure the safety of switching operations. For example, the disconnecting switch can only be closed when the circuit breaker is in the open position and the grounding switch is in the open position.

[0039] The fault mode class is defined based on historical fault records and includes possible abnormal states of the system and their causal characteristics. The abnormal states include protection failure to operate, malfunction, communication interruption, and logic interlocking failure. The causal characteristics define the directed causal relationship edges between the "causing node - fault node - potential consequences" to support subsequent risk reverse deduction.

[0040] S103. Based on the SCD file, parse the virtual terminal connection relationship and map the multi-source data into a spectrum instance, as follows:

[0041] An XML parser is used, based on the substation configuration language file parsing standard, to parse the SCD file and extract the input signal description part, i.e., the "Inputs" tag, of all intelligent electronic devices (IED) nodes; the external signal index tags in the input signal description part are traversed, and the external signal index tags define the IED name, logical node, and data object reference of the signal sender; by parsing the external signal index tags, the mapping relationship between the signal sender and receiver is obtained, thereby establishing the virtual terminal logical connection;

[0042] The five-prevention logic definition file is parsed to extract the Boolean logic expressions and numerical constraint expressions of the interlocking conditions and verification logic; and the unstructured annotation fields in the SCD file are extracted using natural language processing or regular expression matching algorithms; based on the IEC61850 naming standard, the unique identifier and logical variable name of the device instance are extracted from the SCD file and combined with the unstructured annotation fields to establish a naming resolution mapping table, which associates the variables in the logical expressions with the instantiated device instance nodes in the knowledge graph;

[0043] The variables of Boolean logic expressions and numerical constraint expressions are mapped to device instance nodes in the knowledge graph through a named parsing mapping table; the interlocking conditions and verification logic in the error prevention rules are transformed into semantic rules according to the storage format of the ontology model; and the virtual terminal logic connections are mapped to topological association edges connecting device instance nodes to construct the secondary system signal flow topology.

[0044] Multi-source data is used as instance data and instantiated into nodes in the knowledge graph according to the types defined in the ontology structure, and node attributes are populated. To more intuitively illustrate the transformation relationship from multi-source data to knowledge graph nodes, some typical data from the substation bay layer are selected as examples to establish an instance table of IEC61850 and knowledge graph ontology mapping, as shown in Table 1:

[0045] Table 1

[0046]

[0047] S104. Store the instantiated data to complete the knowledge graph construction. The specific steps are as follows:

[0048] Multi-source data instances and transformed semantic rules are stored in the resource description framework triple format defined by the ontology model to form a structured data set of "subject-predicate-object"; where the subject and object correspond to device entities or logical states, and the predicate corresponds to topological connection relationships.

[0049] It should be noted that the establishment of traditional error prevention rule bases relies on static configuration data and human experience, lacking a unified semantic structure, which leads to the rules being out of touch with the real-time operating conditions of substations. This invention constructs a knowledge graph by integrating multi-source data, transforming error prevention rules into reasonable semantic rules, realizing the standardization, structured management and dynamic synchronization of error prevention rule knowledge, and solving the problem of difficult rule maintenance.

[0050] S2, Based on the device attributes and topology associations in the knowledge graph, map and configure the corresponding simulation model parameters to construct a digital twin simulation instance that simulates the dynamic behavior of the secondary system;

[0051] In this embodiment, S2 includes:

[0052] S201. Map the device attributes and topological relationships in the knowledge graph to the preset simulation component library. The specific steps are as follows:

[0053] The simulation component library is based on the general power system simulation standard and the functional block model defined by IEC61850, and includes standardized object classes of circuit breakers, disconnect switches, lines and protection function modules. A graph parser is used to read the topological relationship edges and device instance nodes in the knowledge graph, and according to the preset mapping rules and the simulation component library, the graph structure is converted into a topology description file that the simulation platform can recognize, such as a custom XML format.

[0054] The mapping rules are based on the semantic correspondence between the knowledge graph ontology structure and the simulation component library. They define one-to-one or many-to-one mapping relationships between ontology classes and object attributes in the knowledge graph and specific component models and input / output ports in the simulation component library. For example, the mapping rules stipulate that nodes in the knowledge graph labeled "circuit breaker" must be mapped to the preset "CB_Model" component in the simulation component library, and the "trip command" relationship edge in the graph must be mapped to the "Trip_Input" port of the "CB_Model" component. After the mapping is completed, the topology description file is loaded into the simulation platform to generate the topology model of the secondary protection system.

[0055] S202. Based on multi-source data, the key parameters of the simulation components are dynamically corrected using state evaluation technology. The specific steps are as follows:

[0056] Static configuration data is used as the basic parameters of the simulation components. The simulation component parameters are the physical settings and logical attributes of each component in the topology model, including protection settings, transformer ratio, software version number, and behavioral parameters including component action characteristics and communication link delay baseline.

[0057] Data feature extraction technology is used to filter, denoise, and normalize dynamic operating data to obtain operating feature data. The operating feature data includes the RMS value of real-time load current, voltage change rate, real-time throughput of GOOSE / SV messages, and average component temperature. Status assessment technology and a preset expert rule set are used to perform rule judgment and quantification on the operating feature data to generate status indicators.

[0058] The correction coefficient is obtained based on the status indicator and a preset correction threshold, which is based on historical fault data and relay protection principles. If the status indicator exceeds the correction threshold, the status indicator is input into a preset parameter correction lookup table to derive the corresponding correction coefficient. The parameter correction lookup table is constructed based on laboratory test data and contains a discrete mapping relationship between the status indicator range and the correction coefficient. For example, when the GOOSE network traffic load rate X is 50%, the corresponding communication delay correction coefficient is set to 1.1, as shown in Table 2.

[0059] Table 2

[0060]

[0061] Based on a preset synchronization strategy, the simulation element parameter values ​​reflecting the current real-time operating conditions are obtained by proportionally adjusting the behavioral parameters in the basic parameters using correction coefficients. The synchronization strategy includes a periodic strategy and an event-driven strategy: the periodic strategy refers to periodically checking parameter updates according to a set time interval, which is set based on the data refresh frequency of the substation background monitoring system and the computing load capacity of the simulation server, and can be selected as a synchronization period of 5 to 15 minutes; the event-driven strategy refers to immediately triggering a check when a status indicator changes abruptly and exceeds a preset alarm threshold, which is set based on the limit value specified in the IEC61850 standard. For example, when the GOOSE network communication delay is detected to exceed 3 milliseconds, it is determined that the threshold has been exceeded.

[0062] S203. Synchronize the semantic rules in the knowledge graph as the logical verification benchmark in the simulation process, and combine them with the corrected key parameters to construct a digital twin simulation instance. The specific steps are as follows:

[0063] The semantic rules are formal security constraints stored in the knowledge graph that can be used for reasoning. A rule compiler is used to compile the semantic rules into function libraries or scripting languages ​​that can be directly called and executed by the simulation kernel. The function libraries or scripting languages ​​serve as logical verification benchmarks during subsequent dynamic verification to determine whether the actual actions of the simulation components are compliant. The compilation process maps specific semantic primitives defined in the knowledge graph, such as interlocking as predicates and less than as attributes, to the corresponding procedural function calls and numerical comparison instructions in the simulation kernel based on a semantic mapping table preset by expert experience.

[0064] The topology model, simulation component parameters, and simulation verification benchmark are loaded, compiled, and initialized in the simulation kernel to obtain a digital twin simulation instance. The simulation instance is a runnable and interactive virtual test environment with high-fidelity behavioral simulation capabilities, capable of simulating the latency of GOOSE / SV network communication and the action timing of protection devices.

[0065] This step constructs a high-fidelity digital twin simulation instance through knowledge graph-driven automated mapping and semantic rule compilation and transformation. Based on multi-source data, parameters are dynamically configured to improve the accuracy and credibility of the simulation instance and ensure the effectiveness of subsequent dynamic verification benchmarks.

[0066] S3 receives secondary protection activation / deactivation commands, uses a knowledge graph to deduce risk operation sequences, performs dynamic simulation verification of risk operation sequences, and generates test feedback information.

[0067] In this embodiment, S3 includes:

[0068] S301. Based on the ontology structure of knowledge graphs, the secondary protection activation / deactivation instructions are parsed into a set of operational semantics. The specific steps are as follows:

[0069] Receive secondary protection activation / deactivation instructions input in natural language text or DSL format, wherein the secondary protection activation / deactivation instructions include actions, target devices, or functions;

[0070] An ontology-based semantic parser is used to transform secondary protection activation / deactivation commands into a set of operational semantics that can be recognized by a knowledge graph. The set of operational semantics includes the operation subject, operation relationship, and operation object.

[0071] S302. Obtain real-time operating data of the secondary system, and pre-verify the secondary protection activation / deactivation commands by combining the semantic rules of the operation semantic set and knowledge graph. The specific steps are as follows:

[0072] Acquire real-time operating data of the secondary system that reflects the current operating status of the system, and use the real-time operating data of the secondary system as a pre-verification context constraint. The real-time operating data of the secondary system includes the opening and closing status of the equipment and real-time current and voltage values.

[0073] Map the real-time operation data of the secondary system to the corresponding ontology nodes of the knowledge graph and activate the semantic rules related to the current working condition;

[0074] The operation semantic set is pre-verified based on the activated semantic rules to determine whether the secondary protection activation / deactivation command violates the most basic safety interlock constraint in the semantic rules; the pre-verification result is a binary judgment, including pass or fail.

[0075] S303. Based on the pre-verification results, the failure modes defined in the knowledge graph are used as potential consequences for deduction. A graph traversal algorithm is used to perform reverse tracing under the constraints of the topological association structure of the knowledge graph to obtain the risk operation sequence. The specific steps are as follows:

[0076] If the pre-verification result is a failure, a blocking signal is sent to prevent the operation of the secondary protection activation / deactivation command.

[0077] If the pre-verification result is passed, filter the fault mode nodes that are functionally or topologically associated with the object being operated. For example, functional association can be the protection range, and topological association can be the electrical connection.

[0078] Using the fault mode node as the starting point for reverse tracing and the topological association structure as the static constraint for reverse tracing, a breadth-first search algorithm is used to perform multi-path traversal along the reverse edges of the causal relationship chain in the knowledge graph. During the traversal, node state sequences are extracted from the raw data streams obtained from the process layer and station control layer buses at a preset sampling frequency (e.g., once every 1ms). The node state sequences contain discrete logic bits of secondary equipment (e.g., protection pressure plates, circuit breakers) with sampling timestamps, continuous communication characteristics (e.g., real-time network delay and jitter values ​​of GOOSE / SV messages), and continuous electrical characteristics (e.g., instantaneous voltage and current values ​​of associated primary equipment).

[0079] For the discrete logic bits (with values ​​of 0 or 1), a Gaussian kernel function is used for convolution smoothing to transform them into continuously changing logic signals. The standard deviation of the Gaussian kernel function is set to a range of 50ms to 100ms, and the convolution window width is set to 6 times the standard deviation. The absolute value of the deviation between the continuous electrical characteristics and the preset rated values ​​is calculated, and the absolute value of the deviation is normalized by Min-Max to the continuously changing logic signals, continuous communication characteristics, and the continuous electrical characteristics, mapping them to the [0,1] interval to eliminate dimensional differences.

[0080] The normalized logic signal, continuous communication feature, and absolute value of deviation are multiplied by preset weighting coefficients respectively, and then concatenated to obtain the state vector; the preset weighting coefficients of the normalized logic signal, continuous communication feature, and deviation value can be set to 1.0, 1.0, and 0.5 to adjust the contribution of different source data in calculating Euclidean distance.

[0081] Based on the preset time delay parameter and embedding dimension, multi-vector embedding is performed on the state vector to obtain the phase space reconstruction points, as shown in the following formula:

[0082] ,

[0083] in, Indicates at time Constructed phase space reconstruction points, It is a state vector; The time delay parameter can be set from 10ms to 50ms to preserve effective information about the dynamic evolution of the system. To determine the embedding dimension, the range is typically between 4 and 7 to ensure that the phase space can be fully expanded without overlap.

[0084] The phase space reconstruction points are all A high-dimensional vector of phase space, the set of phase space reconstruction points in the phase space is the state point cloud trajectory, and the geometry of the state point cloud trajectory can characterize the nonlinear evolution law of the secondary system under specific network fluctuations and electrical environments.

[0085] Each reconstructed point in the state point cloud trajectory is assigned a unique digital label, i.e., an index ID; the index ID has a one-to-one mapping relationship with the sampling timestamp recorded during the data acquisition phase, the state vector at the corresponding time, and the phase space reconstructed point, and an index-time series mapping table is constructed accordingly;

[0086] Calculate the weighted Euclidean distance between any two points in the state point cloud trajectory, and construct an N×N distance matrix using the Euclidean distance as matrix elements. The formula for calculating the weighted Euclidean distance is as follows:

[0087] ,

[0088] ,

[0089] ,,

[0090] ,

[0091] in, , , These represent the sampling times. and Between these, there are state deviations in logic signals, communication characteristics, and electrical biases; This is the index for the time delay component, with values ​​ranging from 0, 1, ..., m-1; , These represent the normalized logic signals at the corresponding sampling times; , These represent the normalized communication characteristics at the corresponding sampling times. , These represent the normalized absolute values ​​of the electrical deviation at the corresponding sampling times; The weighted Euclidean distance between two reconstructed points in phase space; , , The weighting coefficients corresponding to the normalized logic signal, continuous communication characteristics, and deviation value are respectively set to 1.0, 1.0, and 0.5.

[0092] Since the input data has been normalized to the [0,1] interval, the maximum possible diameter of the phase space is approximately Define the filter scale parameters. The filtering scale parameter varies from 0 to a preset maximum coverage radius, which is a preset proportion of the maximum possible diameter, ranging from 30% to 50%. For example, for a system with an embedding dimension of m=4, the maximum coverage radius can be set to 1.5. The range of the filtering scale parameter is discretized into K steps (e.g., K=100) to obtain a scale sequence. ;

[0093] The Vietoris-Rips complex construction algorithm is used for each scale step in the scale sequence. The corresponding simplicial complex is constructed according to the following rules: each phase space reconstruction point in the state point cloud trajectory is treated as a vertex to obtain a 0-simplicial complex; the Euclidean distance in the distance matrix is ​​retrieved to be no greater than the current scale step size. Given any two vertices of a set of n+1 vertices, connect them with an edge to form a 1-simulacra; if the distance between any two vertices in the set of n+1 vertices is no greater than the current scale step, then... Then the n+1 vertices form an n-simplex. For example, connecting 3 points in pairs forms a triangle, i.e., a 2-simplex; connecting 4 points in pairs forms a tetrahedron, i.e., a 3-simplex. With the scale step... As the length gradually increases, the number of point pairs satisfying the distance condition increases, the number of edges and high-dimensional surfaces increases continuously, and the simplex of the previous scale is always contained in the simplex of the next scale, forming a simplex sequence that shows a nested growth trend.

[0094] For each specific topological dimension m, a corresponding incidence matrix is ​​constructed. The topological dimension represents the geometric properties of the components in the simplex, where 0 represents vertices, 1 represents edges, and 2 represents triangles. Suppose the current simplex contains a n-simplexes and b (n-1)-simplexes. Then the boundary matrix is ​​a sparse matrix with dimension b×a. The elements of the boundary matrix are based on the inclusion relationship. If the i-th (n-1)-simplex is a face of the j-th n-simplex, then the matrix elements in the i-th row and j-th column are 1, otherwise they are 0. For example, when calculating 1-dimensional features, a boundary matrix is ​​constructed. The columns of the matrix represent each "edge", and the rows represent each "vertex". If "edge A" connects "vertex 1" and "vertex 2", then the column corresponding to "edge A" is filled with 1 in the rows corresponding to "vertex 1" and "vertex 2", and the rest are 0.

[0095] The boundary matrix is ​​transformed using Gaussian column elimination. Each column of the matrix is ​​traversed from left to right, and column transformation operations (i.e., XOR operations) are performed to transform the boundary matrix into an upper triangular matrix or a diagonal matrix.

[0096] Based on the boundary matrix after elimination, the Betti number of the m-th dimension homology group is calculated, which is the persistent homology feature. The Betti number is used to quantify the number of independent topological holes at the current filtering scale. The calculation formula is as follows:

[0097] ,

[0098] in, For Betty's number; The number of closed loops represents the total number of closed paths formed by connecting the beginning and end in the current simple complex network. Its value is obtained by counting the null dimension of the boundary matrix, that is, counting how many sets of edges can form a closed loop. The number of filled loops represents the number of loops in the closed loop that are actually the edges of higher-dimensional solid structures (such as triangular faces). Its value is obtained by counting the column space dimension of the next-dimensional boundary matrix, that is, counting how many loops have been filled with higher-dimensional complexes.

[0099] As the filtering scale parameter changes, the size of the boundary matrix dynamically changes, recording the lifecycle of each independent topological hole (i.e., the persistent homology feature). Specifically, as the filtering scale increases, new connections are formed between points, initially enclosing a new hollow closed loop, leading to an increase in the Betti number. The filtering scale parameter value at this point is recorded as the start time of the hole's lifecycle. As the filtering scale further increases, the points inside the hollow closed loop are connected (e.g., forming triangular faces), causing the closed loop to be filled, which in turn leads to a decrease in the Betti number. The filtering scale parameter value at this point is recorded as the end time of the hole's lifecycle.

[0100] The lifecycle of each hole is treated as an interval, and the set of intervals corresponding to all holes is plotted on a chart with the filter scale as the horizontal axis to generate the continuous barcode image.

[0101] Long barcodes representing unexpected topological coherence features are selected from the continuous barcode image, and the state point cloud trajectories corresponding to the long barcodes are mapped to temporal operation combinations as the risk operation sequence.

[0102] Calculate the standard deviation and arithmetic mean of all barcode lengths, and multiply the standard deviation by a preset sensitivity coefficient, the sensitivity coefficient being between 2.5 and 3.0; add the resulting product to the arithmetic mean to obtain the persistence threshold;

[0103] From the persistent barcode image, barcodes with a lifecycle length greater than the persistence threshold are selected as long barcodes;

[0104] For each long barcode corresponding to the m-th dimension homology class, the corresponding homology generator is extracted using the representative element optimization algorithm in standard homology algebra. Geometrically, the homology generator is represented as a minimal closed chain consisting of a specific set of vertices and their connection relationships in the phase space point cloud. By backtracking the column transformation records of the boundary matrix during the elimination process, the simplex set corresponding to the long barcode is identified, and the closed chain with the fewest vertices and the simplest path in the simplex set is selected as the representative feature of the homology class. All simplexes constituting the homology generator are traversed, and the set of index IDs of all vertices in the state point cloud trajectory is extracted.

[0105] Based on the set of index IDs, the corresponding sampling timestamp and the state vector containing logical bits, communication characteristics and electrical deviations under the corresponding timestamp are obtained by querying the index-time mapping table; the queried state vectors are linearly arranged according to the chronological order of the time axis, and the components in the vectors are denormalized to restore the original data in the node state sequence, thereby forming a combination of time-series operations.

[0106] The timing operation combination is the risk operation sequence, which is specifically an ordered set of discrete actions with strict timing constraints and continuous environmental conditions; for example, the risk operation sequence may be manifested as "at time t_1 the circuit breaker is in the open position, at the same time at time t_2 the network traffic suddenly increases, causing the GOOSE delay to reach 8ms, and then at time t_3 the disconnector switch closing action is performed".

[0107] It should be noted that traditional error prevention mechanisms rely on static interlocking rules for passive and single-step pre-verification when faced with operation commands, thus lacking the ability to predict potential multi-step chain operation risks. This invention obtains the operation semantic set of the command and withdrawal instructions and uses knowledge graphs to reverse trace and deduce the risk operation sequence, thereby realizing multi-step proactive risk prediction and enhancing the secondary system's ability to diagnose complex error operation risks and its foresight.

[0108] S304. Based on the timing information and state constraints of the risk operation sequence, configure the initial state of the digital twin simulation instance and construct a simulation environment that includes secondary network anomalies and extreme operating conditions as boundary conditions. The specific steps are as follows:

[0109] The real-time operating data of the secondary system is configured as the global initial state of all components in the digital twin simulation instance. The timing information includes the order and time of each operation in the risk operation sequence. The state constraints are the preconditions and postconditions required for the success of each operation. Based on the timing information and state constraints, it is verified whether the preconditions required for the first operation of the risk operation sequence meet the global initial state. If the preconditions do not meet the global initial state, that is, the real-time operating data of the secondary system is inconsistent with the preconditions, the state of the local components that do not meet the preconditions is forcibly configured based on the preconditions required by the risk sequence, thereby ensuring that the simulation conditions meet the test requirements.

[0110] The simulation environment is determined based on the initial state. The simulation environment includes background conditions such as external system load and power supply characteristics, and secondary network anomalies and extreme conditions are used as boundary conditions, such as GOOSE message delay, packet loss, near-zone short circuit fault or system low voltage. Secondary network anomalies and extreme conditions are frequent critical scenarios for the maloperation or failure to operate of secondary system protection in smart substations. The values ​​of the boundary conditions are set based on historical fault data and safety margin requirements.

[0111] S305. In the simulation environment, drive the digital twin simulation instance to execute the risk operation sequence, the specific steps of which are as follows:

[0112] The operations and events defined in the risk operation sequence are transformed into a command stream that the simulation kernel can recognize, such as closing a disconnect switch or injecting a short-circuit fault, and then input into the virtual I / O interface of the digital twin simulation instance for simulation according to the timing information of the risk operation sequence.

[0113] S306. Collect test feedback information generated during the execution of risk operation sequences. The test feedback information includes secondary system behavior data and rule verification results. The specific steps are as follows:

[0114] The data is collected in real time through the data acquisition service interface of the digital twin simulation kernel, and the test feedback information containing secondary system behavior data and rule verification results is output in a structured manner. The secondary system behavior data is generated in real time through the internal log module of the digital twin simulation engine, including network communication message latency and protection device action records generated during the simulation process. The rule verification results are obtained by continuously comparing the simulated secondary system behavior data with semantic rules based on the compiled and transformed semantic rule script. The verification results include pass or fail status, logical conflict identifier, and timing limit violation flag.

[0115] This step verifies the dynamic robustness of the error prevention rules through knowledge graph-driven scenario setting and high-fidelity behavior-level simulation methods, providing a comprehensive data foundation with accurate temporal information for rule adaptive optimization.

[0116] S4, based on test feedback information, identify and optimize the secondary protection activation / deactivation and anti-misoperation rules in the knowledge graph;

[0117] In this embodiment, S4 includes:

[0118] S401. Verify the behavior data of the secondary system based on semantic rules, identify abnormal actions, and construct an abnormal event chain. The specific steps are as follows:

[0119] Semantic rules are used as the baseline for judgment, and secondary system behavior data is used as the actual observed facts. By comparing semantic rules and secondary system behavior data, single abnormal actions in the secondary system behavior data that do not conform to the rule expectations are identified, such as protection failure or false activation.

[0120] A time-series analysis algorithm, specifically a sliding window correlation analysis algorithm, is employed. A preset time window is pushed forward from the point in time of the abnormal action. This preset length represents the maximum time span of the causal chain of secondary system events, including the maximum time limit for preventing false alarms, the maximum allowable delay in network communication, and the sum of signal jitter prevention time. This length can be set from 10 to 60 seconds. Within this time window, all displacement events and alarm events that are directly or indirectly connected to the device with the abnormal action in terms of physical topology or logical connection are selected. These displacement events and alarm events are then combined according to their chronological order of occurrence to form an abnormal event chain.

[0121] S402. Classify the abnormal event chain to obtain a defect pattern set in order to identify logical defects in semantic rules. The specific steps are as follows:

[0122] Extract the feature vector of the abnormal event chain and establish a preset action coding mapping table. The mapping table stores the one-to-one correspondence between action text description and numerical identifier. For example, "switch tripping" is mapped to 1 and "protection start" is mapped to 2. Based on the action coding mapping table, the discrete action types in the abnormal event chain are mapped to numerical codes, and the time interval between adjacent actions is used as a continuous numerical feature to construct a multi-dimensional feature vector containing numerical codes and continuous numerical features.

[0123] The multidimensional feature vectors are input into the K-Means clustering analysis algorithm, which automatically groups the data based on similarity to obtain a set of defect patterns representing common fault categories. The algorithm execution process specifically involves: iteratively calculating the distance between the multidimensional feature vector of each abnormal event chain and each cluster center point using the Euclidean distance formula; the smaller the Euclidean distance value, the higher the similarity between the two event chains in terms of action logic and timing; event chains with an Euclidean distance less than a preset clustering radius threshold are grouped into a single defect pattern set; the clustering radius threshold is set based on the normalized distribution of the feature vectors and is typically between 0.1 and 0.3.

[0124] To visually demonstrate the clustering results, 150 sets of abnormal event chains were selected for visualization. The results are as follows: Figure 2 As shown; Figure 2 The horizontal axis represents the time interval characteristics of adjacent actions, i.e., the normalized value, and the vertical axis represents the coding characteristics of the action type. The algorithm divides the discrete abnormal event chain into three typical defect modes: timing disorder caused by communication congestion, logical conflict of protection device, and human error. Each cluster center is clearly identifiable.

[0125] For each defect pattern set, the Apriori algorithm is used to traverse all historical databases containing normal and abnormal operations, count the number of times a specific event combination occurs, and calculate the frequency ratio and confidence level of the event combination in the historical database. The confidence level is the proportion of historical records containing preceding events where subsequent failure events also occur.

[0126] Event combinations whose frequency percentage exceeds a preset frequency threshold and whose confidence level exceeds a preset confidence threshold are identified as key preconditions leading to failure. The dynamic parameter characteristics accompanying these event combinations are considered behavioral characteristics directly related to the failure. The preset frequency threshold can be set to 5% to 10%. The preset confidence threshold is set according to the rule's confidence requirements, for example, 80%.

[0127] The key preconditions and behavioral features are logically compared with the semantic rules to determine the logical defects of the rules: if there are no rules in the rule base that contain key preconditions, it is judged as a logical deficiency; if the derivation result of the existing rules in the rule base contradicts the behavioral features, it is judged as a logical conflict.

[0128] S403. Based on the rule verification results, obtain the safety margin difference between the actual operating parameters in the secondary system behavior data and the constraint parameters in the semantic rules. The specific steps are as follows:

[0129] Based on the rule verification results, the associated secondary system behavior data that passed the verification are filtered out, and the actual operating parameters of the associated secondary system behavior data are obtained. The actual operating parameters are the real-time values ​​recorded in the secondary system behavior data, such as network communication message delay and the actual time of protection actions. The safety margin difference is obtained by subtracting the actual operating parameters from the limiting parameters in the semantic rules, so as to quantify the conservatism of the current rule parameters. The limiting parameters in the semantic rules refer to the values ​​that define the safety threshold in the semantic rules.

[0130] S404. Determine the minimum safety margin based on the safety margin difference, and obtain the parameter optimization space of the limiting parameters by comparing the limiting parameters with the minimum safety margin. The specific steps are as follows:

[0131] A statistical safety and reliability analysis algorithm, specifically the tolerance interval analysis method, is adopted to analyze the statistical distribution of the safety margin difference to determine the minimum safety margin, which is the minimum safety buffer value that the rules need to maintain under the premise of ensuring the reliability of the secondary system.

[0132] Obtain the extreme values ​​of the actual operating parameters, including the longest network latency and protection action time; the difference between the limiting parameter and the sum of the minimum safety margin and the extreme values ​​is taken as the maximum value at which the limiting parameter can be safely relaxed, i.e., the parameter optimization space of the limiting parameter; for example... Figure 3 As shown, Figure 3 The horizontal axis represents the time parameter value, and the vertical axis represents the probability density. The minimum safety margin is 20ms, and the parameter optimization space for the locking time limit is 70ms, which improves operational efficiency while ensuring safety.

[0133] S405. Generate correction instructions based on logical defects and parameter optimization space, and adaptively optimize the secondary protection activation / deactivation rules in the knowledge graph. The specific steps are as follows:

[0134] To address the deficiency in logical missing types, an INSERT template containing condition field placeholders and conclusion field placeholders is invoked. Key preconditions are filled into the condition field placeholders, and behavioral characteristics are filled into the conclusion field placeholders to generate new rule instructions.

[0135] To address the shortcomings of logical conflict types, a MODIFY template containing a deletion pattern domain and an insertion pattern domain is invoked; a unique identifier corresponding to the old rule node in the knowledge graph that contradicts the behavioral features is obtained, and the unique identifier is used to lock the erroneous rule node in the graph; the erroneous rule node is filled into the deletion pattern domain of the template to generate a deletion instruction for removing the old logical relationship; at the same time, the key preconditions and behavioral features are filled into the insertion pattern domain of the template as the logical start point and end point, respectively, to generate an insertion instruction for establishing a new logical relationship.

[0136] For the parameter optimization space, the UPDATE template containing object locators, attribute locators, and value update fields is invoked; the unique identifier of the device instance in the knowledge graph is used as an index to retrieve and locate the specific rule node and its corresponding constraint parameter attribute field in the graph, such as action time limit or set value threshold; the sum of the constraint parameter and the parameter optimization space is used as the optimization value to fill the value update field of the UPDATE template to generate a parameter update instruction;

[0137] Finally, based on the generated correction instructions, the SPARQL update interface of the knowledge graph is automatically invoked to perform the addition, deletion, and modification operations on the secondary protection activation and deactivation anti-misoperation rules; the corrected semantic rules take effect immediately, and this closed-loop process realizes the adaptive optimization of the anti-misoperation rule base.

[0138] It should be noted that traditional error prevention rule base optimization relies on expert experience or simple incident backtracking, making it difficult to systematically diagnose logical deficiencies or conflicts in the rules, and the accuracy of the rules cannot be guaranteed. This invention, based on test feedback information, uses time series analysis and cluster analysis algorithms to determine the functional defects of the rules and adopts statistical safety and reliability analysis algorithms to quantify the parameter optimization space, realizing data-driven adaptive optimization of the rule base and improving the efficiency of the rule base's full lifecycle management.

[0139] See Figure 4 An electronic device, comprising:

[0140] Memory, used to store computer programs;

[0141] A processor is used to implement the steps of the method for optimizing the rule base for secondary protection activation / deactivation and error prevention in intelligent substations when executing the computer program.

[0142] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0143] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0144] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0145] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0146] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0147] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent substation secondary protection switching anti-misoperation rule base optimization method, characterized in that, Includes the following steps: A knowledge graph is built based on multi-source data to serve as the carrier of the error prevention rule base. Based on the device attributes and topology associations in the knowledge graph, the corresponding simulation model parameters are mapped and configured to construct a digital twin simulation instance that simulates the dynamic behavior of the secondary system. The system receives secondary protection activation / deactivation commands and uses a knowledge graph to deduce risk operation sequences. It then performs dynamic simulation verification of the risk operation sequences through the simulation example, generating test feedback information. The risk operation sequences include commands parsed based on the knowledge graph and pre-verified using real-time data, followed by reverse tracing to deduce potential faults and generate risk operation sequences. The process of inferring potential faults through reverse tracing specifically includes: extracting node state sequences on the reverse tracing path and using a phase space reconstruction method to map the node state sequences into state point cloud trajectories in phase space; performing multi-scale topological filtering on the state point cloud trajectories and calculating the continuous coherence features during the filtering process to generate a continuous barcode image; selecting long barcodes in the continuous barcode image that represent unexpected topological coherence features, and mapping the state point cloud trajectories corresponding to the long barcodes into temporal operation combinations as the risk operation sequence; Based on test feedback, identify and optimize the secondary protection deployment and deactivation rules in the knowledge graph to prevent erroneous actions.

2. The method for optimizing the rule base for preventing malfunctions in the secondary protection system of an intelligent substation according to claim 1, characterized in that, The establishment of the knowledge graph, which serves as the carrier of the error prevention rule base, includes: Based on the SCD file, the virtual terminal connection relationship is parsed, and according to the preset ontology model, the device entities, topological associations and initial anti-misoperation rules are transformed into a knowledge graph structure that can perform semantic reasoning. 3.The method of claim 1, wherein, The digital twin simulation example for constructing the dynamic behavior of the secondary system includes: Map the device attributes and topological relationships in the knowledge graph to a pre-defined simulation component library; Based on multi-source data, the key parameters of the simulation components are dynamically corrected using state evaluation technology. Semantic rules from the knowledge graph are synchronized as logical verification benchmarks during the simulation process. Combined with the corrected key parameters, a digital twin simulation instance is constructed.

4. The method of claim 3, wherein the method further comprises: The method of dynamically correcting key parameters of simulation components based on multi-source data and state evaluation technology includes: Initialize the basic parameters of the simulation components based on the static configuration data of the substation; Extract features from the dynamic operation data of the substation and generate status indicators through status assessment; The correction coefficient is obtained based on the status index and the preset correction threshold. The correction coefficients are used to dynamically adjust the basic parameters, thereby completing the dynamic configuration of key parameters.

5. The method of claim 1, wherein the method further comprises: The risk operation sequence specifically includes: Based on the ontology structure of knowledge graphs, the secondary protection activation / deactivation instructions are parsed into a set of operational semantics; Obtain real-time operation data of the secondary system, and pre-verify the secondary protection activation / deactivation instructions by combining the semantic rules of the operation semantic set and knowledge graph; Based on the pre-verification results, the failure modes defined in the knowledge graph are used as potential consequences for deduction. Using a graph traversal algorithm, reverse tracing is performed under the constraints of the topological association structure of the knowledge graph to obtain the risk operation sequence.

6. The method of claim 1, wherein the method further comprises: The process of dynamically simulating and verifying the risk operation sequence to generate test feedback information includes: Based on the timing information and state constraints of the risk operation sequence, configure the initial state of the digital twin simulation instance and construct a simulation environment that includes secondary network anomalies and extreme working conditions as boundary conditions. In a simulation environment, the digital twin simulation instance is driven to execute a sequence of risky operations; The test feedback information generated during the execution of risky operation sequences is collected, and the test feedback information includes secondary system behavior data and rule verification results.

7. The method for optimizing the rule base for preventing malfunctions in the secondary protection system of an intelligent substation according to claim 6, characterized in that, The method of identifying and updating the secondary protection deployment and deactivation rules in the knowledge graph based on test feedback information includes: The behavior data of the secondary system is validated based on semantic rules to identify abnormal actions and construct an abnormal event chain. The abnormal event chain is classified to obtain a set of defect patterns in order to identify logical defects in semantic rules; Based on the rule verification results, obtain the safety margin difference between the actual operating parameters in the secondary system behavior data and the restriction parameters in the semantic rules; The minimum safety margin is determined based on the safety margin difference, and the parameter optimization space of the limiting parameters is obtained by comparing the limiting parameters and the minimum safety margin. Correction instructions are generated based on logical defects and parameter optimization space, and the secondary protection deployment and deactivation rules in the knowledge graph are adaptively optimized. 8.The method of claim 7, wherein, Clustering analysis algorithm is used to classify abnormal event chains to obtain a defect pattern set, and statistical safety and reliability analysis algorithm is used to analyze the safety margin difference to determine the minimum safety margin.

9. An electronic device, comprising: include: Memory, used to store computer programs; A processor is configured to implement the steps of the intelligent substation secondary protection activation / deactivation anti-misoperation rule base optimization method as described in any one of claims 1 to 8 when executing the computer program.

Images