Power distribution equipment fault self-healing control method and device based on sample compensation

Abstract

The present application relates to power distribution equipment fault self-healing control technical field, disclose a kind of power distribution equipment fault self-healing control method and device based on sample compensation.Acquire fault state and current topology version, form electrical character path;According to the action prefix and guard state identification sample cover gap of candidate self-healing control chain;Current electrical character path is mapped to historical anchor sample, and compensation sample chain with source mark and credibility is generated in the physical safety envelope formed by protection setting, equipment capacity, interlocking and power flow boundary;According to real sample support, compensation sample credibility and dependency relationship, target control chain is filtered, and power distribution equipment control instruction is output item by item;After each action, compare actual and expected state transition, correct sample credibility, and re-decision is executed instruction corresponding to failed dependent compensation sample.The present application makes compensation sample and current control state transition maintain verifiable correspondence, avoid mismatch sample in continuous self-healing action accumulates control deviation.

Description

Technical Field

[0001] This invention relates to the field of power distribution equipment fault self-healing control technology, and in particular to a power distribution equipment fault self-healing control method and device based on sample compensation. Background Technology

[0002] Distribution automation systems can collect voltage, current, protection actions, and switch positions using feeder terminals, protection devices, and distribution network master stations. After a fault occurs in a distribution line, they can isolate the faulty section and restore power to non-faulty power-loss areas. Historical fault records can also serve as a sample source for fault diagnosis and recovery control.

[0003] Existing fault self-healing solutions typically perform overall balancing of historical samples according to fault categories or directly use models to generate control chains. They cannot determine whether supplementary samples correspond to specific device state transitions under the current topology, which may lead to continuous self-healing actions using mismatched samples and accumulating control deviations. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a self-healing control method and apparatus for power distribution equipment faults based on sample compensation. This solves the problem that existing fault self-healing schemes cannot determine whether the supplementary sample corresponds to the specific equipment state transition under the current topology, which may lead to continuous self-healing actions using mismatched samples and accumulating control deviations.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A sample-compensated self-healing control method for power distribution equipment faults includes: acquiring power distribution equipment measurements, protection actions, switch states, and the current topology version at the fault trigger time to form the current fault state and the current electrical role path; generating candidate self-healing control chains containing fault isolation and power restoration actions based on the current topology, and matching the topology versioned action transfer sample library according to action prefixes and guard states to determine sample coverage gaps; for the sample coverage gaps, mapping the electrical role paths of historical anchor point samples to the current electrical role paths, and projecting the pre-action state, control action, and post-action state within the physical safety envelope to generate a compensation sample chain with source tags and credibility; determining the target self-healing control chain based on the support of real action transfer samples, the credibility of the compensation sample chain, and sample dependencies; outputting the control instructions of the target self-healing control chain item by item and comparing the feedback state transition of each control instruction with the corresponding sample, continuing execution when the guard conditions are met, reducing the credibility of the corresponding compensation sample when the conditions are not met, invalidating the unexecuted control instructions that depend on the sample, and re-determining the sample coverage gaps, until the self-healing control result and the updated sample library version are output.

[0006] Preferably, the topology-versioned action transfer sample includes a topology version identifier, electrical role path, pre-action state vector, single control action, guard condition, expected post-action state vector, source marker, credibility, and before-and-after sample links. The source marker is used to distinguish between real action transfer samples and compensated action transfer samples, and the before-and-after sample links are used to record the action sequence in the same self-healing control chain.

[0007] Preferably, the determination of the sample coverage gap includes: forming a gap signature from the device role sequence of the current electrical role path, the fault evidence interval, the current action prefix, and the guard status; retrieving the directly corresponding real action transfer sample set according to the gap signature; and generating a sample coverage gap descriptor when the real action transfer sample set is empty, the number of samples is lower than the support threshold, or the dispersion of the expected post-action state exceeds the dispersion threshold. The support threshold is determined based on the minimum number of repetitions of the same state transfer in the historical action log verified by the control results, and the dispersion threshold is determined based on the operating allowable bandwidth of the corresponding measurement.

[0008] Preferably, the electrical role path includes equipment type, power supply direction, electrical distance level of the equipment relative to the suspected faulty equipment, upstream and downstream relationship, control capability, and interconnection power supply margin. The mapping of the electrical role path includes first matching the fault side boundary and the power supply side boundary, then mapping the controllable switches in sequence according to the electrical distance level, and determining that the corresponding historical anchor point sample cannot be mapped when there is a conflict in the power supply direction or inconsistent control capability.

[0009] Preferably, the generation of the compensation sample chain includes: converting the continuous state quantities of the historical anchor point samples into per-unit state quantities based on the equipment rating and the current operating benchmark; replacing the equipment identifier and target state corresponding to the control action according to the mapping result of the electrical role path; projecting the mapped expected post-action state onto the physical safety envelope; discarding the corresponding mapping result when the projection changes the discrete target state of the control action or does not satisfy the equipment interlocking relationship; otherwise, determining the initial credibility and forming the compensation sample chain according to the verification status and role mapping integrity of the historical anchor point samples, wherein the physical safety envelope is jointly determined by the protection setting, equipment capacity, node voltage allowable range, branch current allowable range, power supply path constraints, and equipment interlocking relationship.

[0010] Preferably, the screening of the target self-healing control chain includes: eliminating candidate self-healing control chains whose state does not satisfy the physical security envelope after any action; sorting the remaining candidate self-healing control chains in the following order: larger recovery priority weighted load, higher proportion of real action transfer samples supporting the chain, higher minimum confidence of the dependent samples, fewer dependent compensation action transfer samples, and fewer switching operations; when the sorting results are still the same, the target self-healing control chain is determined according to the lexicographical order of the first device unique identifier in the candidate self-healing control chain.

[0011] Preferably, the comparison of the feedback state transition includes: after receiving confirmation of the controlled object's position and detecting that the sliding measurement window has reached a stable condition, forming a state vector after the actual action; checking the deviation between the discrete switch states, physical safety envelope determination results, and continuous state quantities in the state vector after the actual action and the expected state vector after the action; when the discrete switch states are consistent, the physical safety envelope determination is passed, and the deviation is within the range of the guard condition, increasing the credibility of the corresponding sample and writing this state transition into the real action transition sample set; when any condition is not met, the control command times out, or the current topology version changes during execution, freezing the unexecuted control command, reducing the credibility of the corresponding compensation sample, making the unexecuted control command that depends on the corresponding compensation sample invalid, and re-determining the sample coverage gap with the latest feedback state and the latest topology version.

[0012] A self-healing control device for power distribution equipment faults based on sample compensation includes: a current state construction module, used to form the current fault state and current electrical role path based on power distribution equipment measurements, protection actions, switch states, and the current topology version; a coverage gap identification module, used to generate candidate self-healing control chains and determine sample coverage gaps based on action prefixes, guard states, and a topology-versioned action transfer sample library; a sample compensation module, used to map historical anchor point samples to the current electrical role path and generate a compensated sample chain with source tags and credibility within the physical safety envelope; a control chain decision module, used to determine the target self-healing control chain based on the support of real action transfer samples, the credibility of the compensated sample chain, and sample dependencies; and a feedback execution module, used to output control commands item by item, update the sample credibility and sample dependencies based on the feedback state transfer of each control command, and output the self-healing control result and the updated sample library version.

[0013] Preferably, it also includes a sample version storage module, a power distribution automation interface, and a safety envelope configuration interface. The sample version storage module is used to save topology versioned action transfer samples, source tags, credibility change records, and links between previous and subsequent samples. The power distribution automation interface is used to receive the status of the power distribution terminal and protection device and output control commands to the controllable switch. The safety envelope configuration interface is used to read protection settings, equipment capacity, allowable operating range, power supply path constraints, and equipment interlocking relationships.

[0014] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described sample-compensation-based self-healing control method for power distribution equipment faults.

[0015] This invention provides a method and apparatus for self-healing control of power distribution equipment faults based on sample compensation. It has the following beneficial effects: 1. This invention locates sample coverage gaps by using the action prefix and guard state of the current control chain, maps historical anchor point samples to the current topology through electrical role paths, and invalidates mismatched samples and their dependent actions in the action-by-action feedback, so that the compensation samples maintain a verifiable correspondence with the current control state transition, and avoids the accumulation of control deviations by continuous self-healing actions.

[0016] 2. This invention uses protection settings, equipment capacity, node voltage, branch current, power supply path and equipment interlocking to form a physical safety envelope. Before the compensation sample enters the control chain decision, the mapping results that change the discrete target state or violate the safety boundary are eliminated, reducing the possibility of non-physical compensation samples affecting the control of power distribution equipment.

[0017] 3. This invention records the control dependency relationship between real samples and compensation samples using source marking, credibility, and linking of previous and subsequent samples, and incrementally updates the corresponding records after each action feedback, so that changes in sample quality can be transmitted along the control chain to actions that have not been executed, facilitating online correction and traceability during fault handling.

[0018] 4. This invention adopts a ranking rule based on recovery priority weighted load, real sample support ratio, minimum confidence, number of compensating sample dependencies, and number of switching operations to form a unique target self-healing control chain under the premise of satisfying the safety boundary, thereby avoiding execution ambiguity caused by candidate control chains with the same score. Attached Figure Description

[0019] Figure 1 A flowchart of the self-healing control method for power distribution equipment based on sample compensation provided by the present invention; Figure 2 This is a diagram of the topology versioned action transfer sample and compensation sample chain structure provided by the present invention; Figure 3The control action feedback verification and sample dependency failure relationship diagram provided for this invention; Figure 4 The diagram shows the structure of the self-healing control device for power distribution equipment based on sample compensation provided by the present invention. Detailed Implementation

[0020] The technical solutions in 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.

[0021] Example 1: This example provides a self-healing control method for power distribution equipment faults based on sample compensation, such as... Figure 1 As shown, the process includes current state construction S100, sample coverage gap identification S200, compensation sample chain generation S300, target self-healing control chain determination S400, and control feedback and sample update S500. Each stage shares the same topology version identifier, and the output is only received if the version identifier carried in the output of the previous stage is consistent with the current version of the distribution network model.

[0022] S100: Obtain the power distribution equipment measurements, protection actions, switch status and current topology version at the time of fault triggering, form the current fault status and the current electrical role path consisting of the power supply side, suspected faulty equipment and controllable switches, and organize the verified historical self-healing logs into topology versioned action transfer samples.

[0023] Preferably, such as Figure 2 As shown, the topology versioned action transfer sample is saved according to the corresponding data structure and consists of the following field groups.

[0024] The topology version identifier records the distribution network model version used when the sample was formed, and the electrical role path records the order of electrical roles of the power supply side, fault side boundary, sectionalizing switch and tie switch involved in the sample in the corresponding model.

[0025] The pre-action state vector records the switch position, protection status, node voltage, branch current, power flow direction, and communication validity marker before the control action is issued. The single control action records the unique identifier of the equipment, the target position, and the action type.

[0026] The guard condition record records the discrete conditions and continuous range of quantities that allow the action to be started and the next state to be entered after the action is completed. The expected post-action state vector record records the equipment state and measurement state that should be formed when the guard conditions are met.

[0027] Source markers distinguish between real action transfer samples formed by actual control feedback and compensated action transfer samples formed by topological mapping, and credibility records the effectiveness of the sample when it is adopted by subsequent control chains.

[0028] The preceding and following sample links store the identifiers of adjacent action transfer samples within the same self-healing control chain. An empty preceding link indicates the start of the control chain, while an empty following link indicates the end of the control chain or that no subsequent action has yet been formed.

[0029] Specifically, the distribution network master station receives data from the data acquisition and monitoring control system, the distribution automation system, and the protection information interface within the same fault event window. Input fields include at least the device's unique identifier, sampling time stamp, quality code, phase voltage or line voltage, phase current or zero-sequence current, protection action code, switch position, remote control execution status, and communication status.

[0030] Records from different interfaces are first mapped to device nodes in the distribution network model according to the unified device ledger. Records that cannot be located in the current device ledger are retained in the exception queue and do not participate in the current fault status calculation; when duplicate records appear for the same device, the record with the newest timestamp and valid quality code is used first.

[0031] The system uses the earliest valid timestamp of the protection action or the change of the output switch as the fault trigger time, and captures the stable operation window forward and the fault development window backward. The length of the two windows comes from the event analysis configuration of the distribution automation master station. The configuration value is used to ensure that the state before the protection action and the first round of switch changes can be covered, and is not written into the sample as a fixed product parameter.

[0032] The current topology version is read from the distribution network model service. The version content includes device nodes, connection edges, switch on / off status, power connection points, and communication paths. If a model is published during the reading process, the system discards the current snapshot and rereads it to prevent the same current fault state from being used in two different topology versions.

[0033] The system forms a set of suspected faulty devices based on the direction of protection action, fault current, or undervoltage propagation relationship. When the evidence of a fault is insufficient to identify a single device, the smallest continuous fault section is used as the suspected faulty device, and the controllable switches at both ends of the section are used as the fault boundary.

[0034] The electrical role path traces from each valid power node along the energized connection edge to the suspected faulty device, and then from the faulty side boundary to the tie switch with restoration capability. Objects in the path are represented by device roles rather than field names, enabling mapping between devices performing the same electrical function on different feeders.

[0035] Continuous quantities in the pre-action state vector are stored using the rated values ​​from the equipment ledger and the current operating baseline, along with their original and per-unit values. Discrete quantities are stored using explicitly enumerated values. Missing continuous quantities are marked with a missing flag, and missing switch positions or protection action states cause the corresponding control path to enter a state where automatic execution is prohibited.

[0036] The historical self-healing log is broken down into individual control commands. Each command uses the previous confirmed feedback as the pre-action state, the current command and its target device as a single control action, and the state after the feedback stabilizes as the post-action state. This avoids treating the entire power restoration scheme as merely an unverifiable sample label.

[0037] The system only marks historical segments with complete chronological sequence and verified control results as actual action transfer samples. Historical segments lacking location confirmation, invalid measurement quality, or untraceable model versions are stored in the verification set and are not included in the automatic control chain support statistics.

[0038] The current fault state, current electrical role path, and topology versioned action transfer sample library formed by S100 enable subsequent sample compensation to be carried out around the specific equipment state transfer, rather than simply increasing the number of samples according to the fault category. The output of S100 enters S200.

[0039] S200: Based on the current topology version, generate a candidate self-healing control chain consisting of fault isolation actions and power restoration actions for non-fault power loss areas. Match the current electrical role path and current fault state with the topology versioned action transfer sample library, and determine the sample coverage gap according to the action prefix and guard state of the candidate self-healing control chain.

[0040] Preferably, the sample coverage gap is determined according to the following sub-steps.

[0041] S210: The gap signature is formed by the device role sequence of the current electrical role path, the fault evidence interval, the current action prefix and the guard status.

[0042] S220: Retrieve the set of real action transfer samples that directly correspond to the gap signature. The retrieval results only include real samples that can be directly matched with the topological role and whose verifiable status is valid.

[0043] S230: Compare the number of real action transition sample sets with the support threshold, which is determined based on the minimum number of repetitions of the same state transition in the historical action log that has passed the control result verification.

[0044] S240: Calculate the degree of dispersion of the expected post-action state in the real action transfer sample set and compare it with the dispersion threshold formed by the allowable bandwidth of the corresponding measurement operation.

[0045] S250: When the real action transfer sample set is empty, the number is lower than the support threshold, or the dispersion exceeds the dispersion threshold, generate a sample coverage gap descriptor containing the gap signature, missing action location, and required state fields.

[0046] Specifically, candidate self-healing control chains are generated directly from the current topology snapshot. The system first identifies the disconnectable boundary switches on both sides of the suspected faulty section, then searches for power supply paths with available power and controllable tie switches along the outer edge of the power-loss area, and combines them into candidate control chains according to the timing sequence of isolation followed by power restoration.

[0047] Each action node in the candidate control chain carries an action prefix. The action prefix is ​​a sequence of device actions that have been planned to be executed from the start of the control chain to the current action. An empty prefix corresponds to the first isolated action, while a non-empty prefix is ​​used to distinguish the sample support for the same control action in different prior states.

[0048] The guard status consists of the current switch position, the fault-side undervoltage status, the protection reset status, the tie-side energized status, and the communication effective status. The system uses interval labels for continuous quantities and enumeration labels for discrete quantities, so that stable gap signatures can still be formed even when the measurement has allowable fluctuations.

[0049] The directly corresponding real samples require that the device role sequence be consistent, the fault evidence interval be compatible, the action prefix be consistent after role replacement, and the guard status fall within the same allowed interval. Historical records with the same fault type but different action prefixes cannot be included in the direct support set of the current action.

[0050] The system supports storing thresholds in the sample policy configuration and associating them with sample categories and control action types. During configuration initialization, the minimum number of repetitions is calculated from verified logs. This number can be recalculated after maintenance personnel implement protection policy changes. The system will not use unverified compensation samples to increase the number of times the threshold is met.

[0051] The degree of dispersion is calculated separately for each continuous field of the desired post-action state. If the distribution range of any key field exceeds the allowed bandwidth for that field, the system considers the sample set to lack consistent support for the post-action state, and establishes a coverage gap even if the number of samples reaches the support threshold.

[0052] When a measurement field is missing, the system first determines whether the field is a required field for the current control action. If a non-required field is missing, it is written to the gap descriptor and the S300 is used to supplement the constraint. If a required field is missing, the corresponding candidate control chain is marked as prohibited from automatic execution, and the missing safety input is not bypassed by generating a composite value.

[0053] In addition to the gap signature, the sample coverage gap descriptor also records the candidate control chain identifier, the current action sequence number, the direct real sample identifier, the missing state field, and the reason for the gap determination. This structure serves as the sole entry point for the S300 to select anchor samples and generate compensation sample chains.

[0054] By using S200 to pinpoint insufficient samples to specific state transitions within the control chain, it is possible to distinguish the evidence differences of the same fault category under different topologies, different action prefixes, and different guard states, thus preventing irrelevant sample expansion from entering the control decision. The output of S200 is then fed into S300.

[0055] S300: To address the sample coverage gap, the electrical role path of the historical anchor sample is mapped to the current electrical role path, and the pre-action state, control action, and post-action state of the historical anchor sample are projected into the physical safety envelope of the current power distribution equipment to generate a compensation sample chain with source labeling and credibility.

[0056] Preferably, the compensation sample chain is generated according to the following sub-steps.

[0057] S310: First, match the fault-side boundary and power-side boundary in the historical electrical role path and the current electrical role path, then compare the equipment type, power direction, electrical distance level relative to the faulty equipment, upstream and downstream relationship, control capability and interconnection power supply margin.

[0058] S320: Map controllable switches sequentially according to electrical distance levels. If there is a conflict in power direction or inconsistent control capability, it is determined that the historical anchor point sample cannot be mapped.

[0059] S330: Based on the equipment ratings and current operating baseline, convert the continuous state variables of historical anchor point samples into per-unit state variables.

[0060] S340: Replace the unique device identifier and target state corresponding to the control action based on the mapping result of the electrical role path, while keeping the action sequence and the link between the preceding and following samples unchanged.

[0061] S350: Projects the mapped post-action state onto a physical safety envelope formed by protection settings, equipment capacity, node voltage allowable range, branch current allowable range, power supply path constraints, and equipment interlocking relationships.

[0062] S360: When the projection changes the discrete target state of the control action or does not meet the device interlocking relationship, the mapping result is discarded; otherwise, the initial credibility is determined according to the verification status of the historical anchor point sample and the integrity of the role mapping, and a compensation sample chain is formed.

[0063] Specifically, historical anchor point samples are selected from real samples with the same action type as the current gap. The system first filters out samples with untraceable sources, invalid verification status, and broken links, and then sorts them from high to low according to the number of matches in the electrical role path field.

[0064] Fault-side boundary matching ensures that compensation actions apply to the same isolation position relationship, while power-side boundary matching ensures that the power supply direction of the power restoration action is compatible. If either type of boundary cannot be matched, the system terminates the mapping of the corresponding anchor point sample.

[0065] The electrical distance level between the equipment and the suspected faulty equipment is represented by the number of segmentable switches passed through, not by geographical distance. The upstream and downstream relationship is based on the current power direction; when bidirectional power flow exists, the system simultaneously reads the protection direction configuration and the real-time power direction, and if the two conflict, no mapping is established.

[0066] The control capability field must at least distinguish between remotely controllable switching, remote signaling only, local interlocking, and maintenance interlocking. If a historical action requires remote switching but the current role's equipment is in a remote signaling only or interlocked state, the mapping result will still be discarded, even if the equipment type and electrical distance are the same.

[0067] Per-unit conversion of continuous state quantities is based on the rated values ​​of each role's equipment, while retaining the current operating baseline. The system converts changes in voltage, current, and load rate from historical samples into relative changes before applying them to the state before the current action, avoiding the direct replication of absolute quantities from equipment with different capacities.

[0068] Control action mapping only replaces the unique identifier of the device, without changing the action semantics of opening, closing, or protection reset. When there is no device in the current path that matches the historical control role, the system does not shorten the control chain, nor does it automatically replace adjacent devices as control objects, but instead records the reason why mapping is not possible.

[0069] The protection settings and equipment interlocking relationships in the physical safety envelope are read from the safety envelope configuration interface. Equipment capacity is obtained from the equipment ledger, the allowable ranges of node voltage and branch current are obtained from the operating mode configuration, and power supply path constraints are obtained from the current topology verification. All configuration data includes the effective version and the time stamp for reading.

[0070] For each mapping action, the system performs static connectivity checks and power flow boundary checks on the current topology to obtain the allowable set of states after the desired action. When the desired state exceeds the allowable set, the projection is only allowed to tighten the range of continuous states, and it is not allowed to change the tripping state to the closing state or vice versa.

[0071] Equipment interlocking checks include prohibiting the closing of interconnecting circuits when a faulty section is not isolated, prohibiting the closing of corresponding switches when the grounding switch is in the closed position, and prohibiting remote control of maintenance-locked equipment. Interlocking conditions are derived from the current equipment configuration; any missing mandatory interlocking field will render the mapping results invalid.

[0072] The initial credibility is determined by the verification level of the anchor sample, the completeness of the role field matching, the coverage of the pre-action state, and the projection amplitude of the security envelope. The system writes each criterion into the credibility source record, and the initial credibility of the compensation sample must not be higher than the credibility of the corresponding historical anchor sample.

[0073] Multiple gaps in the same candidate control chain are linked together by preceding and following sample links to form a compensation sample chain. If the expected state after the preceding compensation action cannot satisfy the guard condition of the following compensation action, the system disconnects the link and ensures that the following action covers the gap state, preventing independently generated samples from being forcibly spliced ​​together.

[0074] The compensation sample chain output by S300 carries anchor sample identifiers, role mapping tables, physical security envelope versions, and credibility sources, making each compensation action traceable to the real state transition and subject to the current electrical constraints, and the output enters S400.

[0075] S400: Based on the support of real action transfer samples, the credibility of compensation sample chains, and sample dependencies, candidate self-healing control chains that satisfy the physical security envelope are selected to obtain the target self-healing control chain.

[0076] Preferably, the target self-healing control chain is screened according to the following sub-steps.

[0077] S410: Verify the expected post-action state of candidate self-healing control chains on an action-by-action basis, and eliminate candidate self-healing control chains whose post-action state does not satisfy the physical safety envelope.

[0078] S420: Compare the recovery priority weighted loads among the remaining candidate self-healing control chains, and prioritize retaining the candidate self-healing control chains with larger recovery priority weighted loads.

[0079] S430: When the recovery priority weighted load is the same, compare the support ratio of real action transfer samples and the lowest confidence of control chain dependent samples in turn.

[0080] S440: When the two items are still the same, compare the number of compensation action transfer sample dependencies and the number of switching operations in turn, and prioritize retaining the candidate self-healing control chain with fewer dependencies and fewer switching operations.

[0081] S450: When the above sorting results are still the same, the unique target self-healing control chain is determined according to the lexicographical order of the unique identifier of the first device in the candidate self-healing control chain.

[0082] Specifically, the control chain decision module establishes an action dependency table for each candidate self-healing control chain. Each row of the dependency table corresponds to a control action, recording the real sample identifier supporting the action, the compensation sample identifier, the guard condition, the expected post-action state, and the preceding action identifier.

[0083] The real action transfer sample support ratio is the ratio of the number of actions with direct real sample support to the total number of actions in the control chain. When an action receives support from multiple real samples simultaneously, it is still counted as a single supported action, and the sample count is stored separately in the support field.

[0084] The control chain relies on the lowest confidence level among all adopted samples to prevent a low-confidence compensation action from being masked by a large number of high-confidence actions. Control chains that do not use compensation samples participate in the comparison with the lowest confidence level in the real sample set.

[0085] The priority-weighted load recovery is calculated based on the load priority configured in the distribution network operation system and the current power outage load. The system only includes loads that can be recovered within the allowable range of current power supply capacity, node voltage, and branch current; the recoverable amount cannot be increased by lowering the sample safety conditions.

[0086] The physical safety envelope verification is performed sequentially according to the action order. After a previous action changes the topology, the subsequent action uses the updated temporary topology for calculation, and does not use the same power flow result from the control chain starting point to replace the verification result of the entire action sequence.

[0087] When a candidate control chain contains a communication-unreachable device, the decision module marks the corresponding action as unexecutable and removes the entire control chain. If all candidate control chains are removed, the system outputs the automatic self-healing interlocking result and the reason for the blockage, without generating alternative actions that lack device control capabilities.

[0088] The ranking process uses a defined lexicographical order for hierarchical comparison, avoiding the mixing of multiple indicators into a composite score of unknown origin. The unique equipment identifier is only used to eliminate ties after all technical indicators are identical, without altering the priority of safety and power restoration objectives.

[0089] Once the target self-healing control chain is determined, the decision module freezes the chain identifier, topology version, action dependency table, and security envelope version, and sends the first control action and its guard conditions to the feedback execution module. Actions that have not yet been executed cannot be issued in batches in advance.

[0090] S400 can prioritize the control chain with more real evidence, less reliance on compensation, and simpler operation among candidate schemes with the same security conditions, and form a unique output with complete tie rules. The target self-healing control chain of S400 enters S500.

[0091] S500: Outputs control commands to the power distribution equipment item by item according to the target self-healing control chain, and compares the feedback state transition with the sample that supports the control command after each control command is executed. Based on the comparison result, it continues to execute or disables the unexecuted control commands that depend on the mismatched sample, until the self-healing control result and the updated sample library version are output.

[0092] Preferably, the control feedback and sample update are performed according to the following sub-steps.

[0093] S510: Output a control command to the power distribution equipment corresponding to the current action, and record the command identifier, equipment unique identifier, target location, issuance time stamp and current topology version.

[0094] S520: After receiving confirmation of the controlled object's position and detecting that the sliding measurement window has reached a stable condition, a state vector is formed after the actual action.

[0095] S530: Verify the discrete switch states, physical safety envelope determination results, and deviations between the continuous state variables and the expected state vector after the actual action in the state vector.

[0096] S540: When the discrete switch states are consistent, the physical safety envelope judgment is passed, and the continuous state quantity deviation is within the guard condition limit, the credibility of the corresponding sample is increased, the current state transition is written into the real action transition sample set, and the next control action is entered.

[0097] S550: If any comparison condition is not met, reduce the confidence of the corresponding compensation sample, invalidate the unexecuted control instructions that depend on the corresponding compensation sample, and re-execute S200 with the latest feedback state.

[0098] S560: Freeze unexecuted control commands when the control command times out or the current topology version changes during execution, record abnormal feedback, and re-execute S200 with the latest feedback status and the latest topology version.

[0099] Specifically, the feedback execution module issues only one control command at a time. The command is sent to the corresponding feeder terminal or switch control unit through the distribution automation interface. The interface returns the acceptance status, execution status, and location change status, all three of which are associated with the same command identifier.

[0100] Position confirmation is jointly determined by remote signaling changes and control feedback. If the two are inconsistent, the system treats it as unconfirmed and waits for the next round of valid feedback; if the remote control timeout value configured for the corresponding equipment by the distribution automation system is exceeded, the current control action is judged as an execution anomaly.

[0101] The sliding measurement window begins acquiring data from the moment the position is confirmed. Stability requirements stipulate that the range of change in key continuous quantities within the window falls within the stability band given by the operational configuration, and that the protection action code and communication quality code remain valid. Transient measurements should not be misjudged as post-action states before the window has stabilized.

[0102] After the actual action, the state vector adopts the same field order and per-unit reference as the state vector before the action. The system first compares the discrete switch states item by item, then performs a physical safety envelope check, and finally compares the deviation of the continuous state quantities; if the previous layer fails, it does not continue to use the similarity of continuous quantities to cover up the discrete state error.

[0103] The continuous quantity limit in the guard condition is derived from the intersection of the expected range of the supporting samples and the current allowed bandwidth. An empty intersection indicates that the sample is incompatible with the current operating conditions, and the corresponding action is blocked before it is sent out, without allowing the action to pass through by amplifying the tolerance.

[0104] When a feedback state transition is successful, the system combines the pre-action state, the actual control action, and the post-action state into a new real action transition sample, and associates it with the current topology version and the original compensation sample identifier. The reliability of the compensation sample increases, but does not exceed the upper limit defined by the anchor point verification level.

[0105] When a feedback state transition fails, the system records the specific reasons for discrete state inconsistencies, safety envelope failures, or continuous state quantity deviations. The reliability of the corresponding compensation sample decreases, and upon reaching the failure condition, it is added to the failure set, no longer supporting automatic control under the current topology version.

[0106] Sample dependency failure relationship such as Figure 3 As shown, the system iterates backward along the action dependency table from the failure compensation sample, marking all unexecuted actions that directly or indirectly depend on the sample; actions that have been verified by feedback remain valid and their true records are not revoked due to the failure of subsequent samples.

[0107] After a failure due to an inaction, the system uses the state after the most recent actual action as the new current fault state, generates a new topology version based on the actual switch position, and returns to S200 to regenerate the candidate control chain and gap signature. The re-decision will not reuse the failed sample version.

[0108] If a switch change, protection re-operation, or distribution network diagram update is detected during execution that is not caused by this control chain, the system considers the topology version to have changed. The feedback execution module stops issuing subsequent instructions, merges the external changes into the latest state, and then re-enters coverage gap identification.

[0109] When all actions of the target self-healing control chain pass feedback verification, the system outputs the success result, the recovered area, the still isolated area, the action record, the change in sample confidence, and the updated sample library version; when all security candidate chains are unexecutable, the system outputs the locking result and the reason for the block.

[0110] By embedding the validity verification of the compensation sample into every real device action through S500, the dependency propagation can be blocked when the first mismatch occurs and the latest real feedback can be used to make new decisions, avoiding the accumulation of control deviations along the continuous self-healing control chain by the mismatch sample.

[0111] like Figure 4 As shown, this embodiment also provides a self-healing control device for power distribution equipment faults based on sample compensation. The device includes a current state construction module, a coverage gap identification module, a sample compensation module, a control chain decision module, and a feedback execution module. Each module is connected sequentially through a data interface with a topology version identifier.

[0112] The current state construction module connects the power distribution automation interface and the sample version storage module. It receives power distribution equipment measurements, protection actions, switch statuses and the current topology version, and outputs the current fault status, the current electrical role path and traceable real action transfer samples.

[0113] The coverage gap identification module is connected to the current state construction module. Based on the current topology, it generates candidate self-healing control chains, establishes action prefixes and guard states for each action, and outputs sample coverage gap descriptors corresponding to the action positions of the candidate control chains.

[0114] The sample compensation module connects the coverage gap identification module, the safety envelope configuration interface, and the sample version storage module. It completes the historical anchor point sample screening, electrical role mapping, state projection, and credibility initialization, and outputs the compensation sample chain and mapping basis.

[0115] The control chain decision module connects to the sample compensation module, reads the real sample support, the compensation sample confidence and the action dependency table, performs a deterministic ranking of candidate control chains that meet the physical security envelope, and outputs a unique target self-healing control chain.

[0116] The feedback execution module connects the control chain decision module and the power distribution automation interface. It sends instructions to the controllable switch in a single-action manner, receives position confirmation and measurement feedback, and updates the sample credibility, action dependency and sample library version.

[0117] The sample version storage module uses version locking for each read and write operation. After the control chain decision begins, the relevant sample version is read-only; after an action is completed, a new version is generated, which saves the difference record and retains the previous version for fault handling and traceability.

[0118] The safety envelope configuration interface reads protection settings, equipment capacity, permissible operating range, power supply path constraints, and equipment interlocking relationships according to the current topology version. If any mandatory configuration is missing, the interface outputs a "cannot be automatically executed" flag to the control chain decision module.

[0119] The power distribution automation interface uses the same unique device identification system for input feedback and output commands, and retains the command identifier and feedback timestamp. When communication is interrupted, the interface returns a clear unreachable status, and the feedback execution module freezes subsequent commands accordingly.

[0120] Example 2: This example illustrates the complete operation process using a distribution feeder containing sectionalizing switches and tie switches experiencing a section fault. This scenario is for explaining data flow and control logic only and does not represent deployment by a specific user, nor does it provide unverified quantitative results.

[0121] Before the fault occurs, the current state construction module reads the power nodes, trunk segments, branch connections, tie paths, and switch positions from the distribution network diagram and generates the current topology version. The sample version storage module provides previously verified real action transfer samples and their preceding and following links.

[0122] After the feeder terminal sends out protection actions and the output switch changes position, the system takes the earliest valid event as the fault trigger time, associates the voltage, current, protection direction and switch status of adjacent equipment, and forms the current fault status including the quality code.

[0123] The system identifies suspected fault sections based on the direction of fault current propagation and the range of voltage loss. It uses the controllable sectionalizing switches on both sides of the section as the fault side boundary and then forms the current electrical role path from the available power supply along the energized path and the interconnection path.

[0124] The coverage gap identification module generates two candidate control chains. The first control chain isolates the suspected faulty section and restores the non-faulty power-loss area via the first communication path. The second control chain uses the same isolation action but restores power via the second communication path.

[0125] The first isolation action of both control chains is supported by direct real samples. The tie-closing action of the first control chain lacks real samples under the same action prefix, while the tie-closing action of the second control chain has real samples, but the dispersion of the current distribution after its action exceeds the current operating allowable bandwidth.

[0126] The system establishes sample coverage gap descriptors for the two tie-off closing actions. The descriptor records the power supply side role, fault side boundary, isolation action prefix, tie-off switch closing guard, and missing or discrete post-action status fields.

[0127] The sample compensation module retrieves verified historical anchor point samples from other topology versions. The anchor point samples have the same fault-side boundary role, the same power direction, and the same tie-off closing control capability, but the field device identification and device capacity are different.

[0128] The system first maps the switches and lines in the anchor point sample to the current role path according to the electrical distance level, then converts the voltage, current and load rate into per-unit state quantities, and finally replaces the unique device identifier of the control action while maintaining the semantics of the closing action and the preceding and following links.

[0129] The safety envelope configuration interface provides current protection settings, interconnection equipment capacity, allowable ranges for node voltage and branch current, fault isolation conditions, and equipment interlocking relationships. The mapped desired state is then verified against the current topology power flow boundary to form a compensation sample chain.

[0130] One of the compensation paths would fail to meet the power supply path constraints if a connection was closed during projection, so the sample compensation module discarded the mapping result. The other compensation path did not change the discrete target state and satisfied the interlocking and operation boundaries, so it entered the action dependency table with an initial confidence level lower than that of the anchor point sample.

[0131] The control chain decision module first eliminates candidate control chains that cannot meet the physical security envelope, and then compares the recoverable priority-weighted load, the proportion of real sample support, the minimum confidence level, the number of compensating dependencies, and the number of switching operations to obtain a unique target self-healing control chain.

[0132] The feedback execution module issues the first fault-side boundary trip command. After receiving the position confirmation, the module waits for the key measurement window to stabilize, confirms that the fault section is de-energized, the boundary switch position is correct, and the protection status meets the guard conditions, and then transfers this status to the real sample set.

[0133] The system then issues a second boundary trip command and completes the checks on discrete states, safety envelopes, and continuous quantity deviations in the same manner. After the two isolation actions are successful, the action prefix of the tie-closing action is consistent with the preceding link of the compensation sample chain record.

[0134] After the connection closing command is executed, if the actual power flow direction, node voltage and branch current all fall within the range of the guard condition limit, the system will transfer and register the feedback state as a real sample, improve the reliability of the corresponding compensation sample and continue the remaining power restoration actions.

[0135] If the actual power flow direction is found to be inconsistent with the expected direction after the connection is closed, the feedback execution module immediately determines that the guard condition is not met, reduces the reliability of the corresponding compensation sample, and causes subsequent load recovery actions that depend on the compensation sample to fail.

[0136] The system retains the completed and verified isolation actions, forms a new current fault state based on the actual state of the tie switch and the latest measurements, regenerates the topology version, and returns it to the coverage gap identification module to calculate the new candidate control chain.

[0137] When making a new decision, the failed sample version will no longer participate in control support under the current topology. The system selects an alternative connection path that still meets the protection, capacity, power supply path, and interlocking requirements, issues subsequent control commands sequentially, and completes feedback verification after each action.

[0138] When all executable control chains are completed, the device outputs self-healing control results, which include fault isolation status, restored power area, unrestored power area, actual action sequence, guard determination for each action, changes in sample confidence, and the latest sample library version.

[0139] In this embodiment, sample compensation does not replace the protection and interlocking of power distribution equipment, nor does it directly use unverified analog outputs as batch control commands. Instead, it establishes traceable support for single state transitions that lack direct real evidence, and the real feedback determines whether subsequent actions continue.

[0140] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. 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. A self-healing control method for power distribution equipment faults based on sample compensation, characterized in that, include: Obtain the power distribution equipment measurements, protection actions, switch status, and current topology version at the time of fault triggering to form the current fault status and current electrical role path; Based on the current topology, candidate self-healing control chains containing fault isolation and power restoration actions are generated, and the topology versioned action transfer sample library is matched according to the action prefix and guard state to determine the sample coverage gap. For the sample coverage gap, the electrical role path of the historical anchor point sample is mapped to the current electrical role path, and the pre-action state, control action and post-action state are projected within the physical security envelope to generate a compensation sample chain with source labeling and credibility. The target self-healing control chain is determined based on the support of real action transfer samples, the credibility of the compensation sample chain, and the sample dependency relationship. The control instructions of the target self-healing control chain are output one by one, and the feedback state transition of each control instruction is compared with the corresponding sample. If the guard condition is met, the execution continues. If the condition is not met, the credibility of the corresponding compensation sample is reduced, the unexecuted control instructions that depend on the sample are invalidated, and the sample coverage gap is re-determined until the self-healing control result and the updated sample library version are output.

2. The self-healing control method for power distribution equipment faults based on sample compensation according to claim 1, characterized in that, The topology-versioned action transfer sample includes a topology version identifier, electrical role path, pre-action state vector, single control action, guard condition, expected post-action state vector, source marker, credibility, and before-and-after sample links. The source marker is used to distinguish between real action transfer samples and compensated action transfer samples, and the before-and-after sample links are used to record the action sequence in the same self-healing control chain.

3. The self-healing control method for power distribution equipment faults based on sample compensation according to claim 2, characterized in that, The determination of the sample coverage gap includes: forming a gap signature from the device role sequence of the current electrical role path, the fault evidence interval, the current action prefix, and the guard status; Retrieve the set of real action transfer samples that directly correspond to the gap signature; When the set of real action transition samples is empty, the number of samples is lower than the support threshold, or the dispersion of the expected post-action state exceeds the dispersion threshold, a sample coverage gap descriptor is generated. The support threshold is determined based on the minimum number of repetitions of the same state transition in the historical action log that has been verified by the control results, and the dispersion threshold is determined based on the allowable bandwidth of the corresponding measurement.

4. The self-healing control method for power distribution equipment faults based on sample compensation according to claim 3, characterized in that, The electrical role path includes equipment type, power direction, electrical distance level of the equipment relative to the suspected faulty equipment, upstream and downstream relationship, control capability and interconnection power supply margin. The mapping of the electrical role path includes first matching the fault side boundary and the power side boundary, then mapping the controllable switches in sequence according to the electrical distance level, and determining that the corresponding historical anchor point sample cannot be mapped when there is a conflict in power direction or inconsistent control capability.

5. The self-healing control method for power distribution equipment faults based on sample compensation according to claim 4, characterized in that, The generation of the compensation sample chain includes: converting the continuous state quantities of the historical anchor point samples into per-unit state quantities based on the equipment rating and the current operating benchmark. Replace the device identifier and target state corresponding to the control action based on the mapping result of the electrical role path; Project the mapped post-action state onto the physical security envelope; When the projection changes the discrete target state of the control action or does not meet the equipment interlocking relationship, the corresponding mapping result is discarded. Otherwise, the initial credibility is determined according to the verification status and role mapping integrity of the historical anchor point sample and the compensation sample chain is formed. The physical safety envelope is jointly determined by the protection setting, equipment capacity, node voltage allowable range, branch current allowable range, power supply path constraints and equipment interlocking relationship.

6. The self-healing control method for power distribution equipment faults based on sample compensation according to claim 5, characterized in that, The screening of the target self-healing control chain includes: eliminating candidate self-healing control chains whose state does not satisfy the physical safety envelope after any action; The remaining candidate self-healing control chains are sorted in the following order: larger recovery priority weighted load, higher proportion of real action transfer samples, higher minimum confidence of the dependent samples, fewer dependent compensation action transfer samples, and fewer switching operations. If the sorting results are still the same, the target self-healing control chain is determined according to the lexicographical order of the unique identifier of the first device in the candidate self-healing control chain.

7. The self-healing control method for power distribution equipment faults based on sample compensation according to claim 6, characterized in that, The comparison of the feedback state transition includes: forming a state vector after the actual action is performed after receiving confirmation of the position of the controlled object and detecting that the sliding measurement window has reached a stable condition; Verify the discrepancies between the discrete switch states, physical safety envelope determination results, and continuous state quantities in the state vector after the actual action and the expected state vector after the action. When the discrete switch states are consistent, the physical safety envelope determination is passed, and the deviation is within the guard condition limit, the credibility of the corresponding sample is increased and the current state transition is written into the real action transition sample set. If any condition is not met, control instructions time out, or the current topology version changes during execution, unexecuted control instructions are frozen, the credibility of the corresponding compensation samples is reduced, unexecuted control instructions that depend on the corresponding compensation samples become invalid, and the sample coverage gap is re-determined based on the latest feedback status and the latest topology version.

8. A self-healing control device for power distribution equipment based on sample compensation, characterized in that, include: The current state construction module is used to form the current fault state and the current electrical role path based on the power distribution equipment measurement, protection action, switch status and the current topology version; The coverage gap identification module is used to generate candidate self-healing control chains and determine sample coverage gaps based on action prefixes, guard states, and a topology versioned action transfer sample library. The sample compensation module is used to map historical anchor point samples to the current electrical role path and generate a compensated sample chain with source tags and credibility within the physical security envelope; The control chain decision module is used to determine the target self-healing control chain based on the support of real action transfer samples, the credibility of compensation sample chains, and sample dependencies. The feedback execution module is used to output control commands one by one, update the sample confidence and sample dependency based on the feedback state transition of each control command, and output the self-healing control result and the updated sample library version.

9. The self-healing control device for power distribution equipment faults based on sample compensation according to claim 8, characterized in that, It also includes a sample version storage module, a power distribution automation interface, and a safety envelope configuration interface. The sample version storage module is used to save topology versioned action transfer samples, source tags, credibility change records, and links between previous and subsequent samples. The power distribution automation interface is used to receive the status of power distribution terminals and protection devices and output control commands to controllable switches. The safety envelope configuration interface is used to read protection settings, equipment capacity, allowable operating range, power supply path constraints, and equipment interlocking relationships.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the self-healing control method for power distribution equipment faults based on sample compensation as described in any one of claims 1 to 7.

Images