A method and system for multi-level topology analysis of power systems

Abstract

The application discloses a kind of power system multistage topological analysis method and system, it is related to power system automation and distribution network operation control technical field, the application is based on real-time measurement calculation electrical node coupling strength and with threshold generation coupling partition, make partition boundary update with operating state, the data range relied on across border cooperation is converged to relevant area with coupling relationship, to improve the adaptability of topological analysis boundary to operating condition change;In the situation that measurement quality and link state fluctuation lead to coupling calculation and topological integration to be influenced by noise, missing, late in midday reverse sending etc.Senario, introduce the confidence index determined by measurement quality code, link quality and acquisition equipment type, and make confidence monotone attenuation with time interval, and attenuation intensity is updated with link statistical window self-adapting, so that outdated or low-trust data is reduced in weight or removed in key criterion such as partition edge retention, tie branch merging, to improve the consistency and stability of topological result.

Description

Technical Field

[0001] This invention relates to the field of power system automation and distribution network operation control technology, and in particular to a multi-level topology analysis method and system for power systems. Background Technology

[0002] As the scale of distributed energy sources such as distributed photovoltaics in distribution networks continues to expand, the power flow pattern of distribution networks is gradually changing from the traditional unidirectional radial power supply to an operating state that may have bidirectional power flow. In energy management, distribution automation, and advanced application computing, network topology processing is used to generate electrical connectivity models for calculation based on the connection relationships of primary equipment and switch states, and to provide a topological basis for state estimation, power flow analysis, protection setting, and operation control. Existing topology processing is usually based on physical connection diagrams, combined with the switch states of circuit breakers, disconnectors, tie switches, etc., to traverse and aggregate the equipment connection relationships, convert fine-grained connection node models into abstract topologies such as bus-branch, and further identify objects such as electrical islands and feeder segments, thereby forming a multi-level topology view.

[0003] In high-penetration distributed photovoltaic (PV) scenarios, the operating conditions of the distribution network exhibit stronger time-varying and coupling characteristics. Public research indicates that when the output of distributed PV increases significantly during periods of light load, the distribution feeders may experience reverse power flows back to the upstream grid. These reverse power flows can lead to operational problems such as feeder voltage rise, mismatch of traditional voltage regulation strategies, changes in short-circuit current levels, and disruptions to protection desensitization and protection coordination. Under such conditions, the power flow path and node voltage distribution may change significantly in a short period of time, resulting in a stronger real-time requirement for identifying critical lines, weak nodes, and control / protection areas.

[0004] Existing topology analysis and parallel computing frameworks often employ pre-defined regional boundaries for partitioning, with common partitioning criteria including geographical location or administrative scope. Related research indicates that partitioning based solely on geographical or administrative boundaries may overlook electrical interconnections between network segments, failing to reflect differences in coupling strength caused by changes in operating conditions. In situations where distributed photovoltaic systems experience high output during midday, low load, and potential backflow of power, the electrical coupling area may dynamically shift with changes in power flow and sensitivity distribution, potentially leading to inconsistencies between fixed partition boundaries and actual coupling areas. Furthermore, when cross-regional protection coordination, weak point analysis, or collaborative control is required, cross-regional information exchange and synchronization mechanisms may increase collaborative processing latency, impacting response to rapidly changing operating conditions.

[0005] On the other hand, some methods rely on historical statistical features or static network parameters in the identification of critical paths or weak nodes, or mainly rely on switch status and static connection relationships for bus aggregation and island identification in topology identification. It is difficult to map the real-time coupling strength changes caused by power fluctuations, reverse power flow, and voltage sensitivity changes. In the above cases, the critical paths or weak nodes obtained based on static indicators may not be consistent with the actual reverse power flow carrying paths, thus making the control and protection strategies formed based on the results insufficiently adaptable in specific periods.

[0006] To mitigate the aforementioned impacts, existing engineering measures include shortening the topology processing and analysis cycle, enhancing communication link capabilities, and increasing the conservative margin of protection and control strategies. While these measures have improved availability to some extent, they still suffer from insufficient perception of dynamic coupling changes in operating conditions and difficulties in adaptively adjusting partition boundaries and analysis granularity according to operating conditions. Therefore, a multi-level topology analysis method and system is needed that can perceive changes in electrical coupling during operation and adaptively adjust the analysis boundaries and granularity based on coupling characteristics to improve real-time performance and coordination in scenarios such as high-penetration distributed photovoltaics. Summary of the Invention

[0007] This application provides a multi-level topology analysis method and system for power systems, which solves the problems of bidirectional power flow caused by high-penetration photovoltaics, and the difficulty in adapting fixed partitions and static indicators to real-time coupling and protection coordination.

[0008] In a first aspect, embodiments of the present invention provide a method for multi-level topology analysis of a power system, comprising:

[0009] Step S1: Obtain real-time measurement data of the power grid, equipment status information, and power grid connection relationships, and record the corresponding timestamps;

[0010] Step S2: Calculate the real-time electrical coupling strength of the electrical node pair based on the real-time measurement data, and determine the confidence index by combining the measurement quality code, communication link quality and acquisition equipment type. The confidence index decreases monotonically with time interval, and the decrease intensity is adaptively adjusted according to the change of link status within a preset statistical window.

[0011] Step S3: Before generating electrical coupling partitions or performing global topology integration, the real-time electrical coupling strength or topology results within the partition are weighted based on the confidence index, and abnormal or missing data are processed according to a conservative level.

[0012] Step S4: Generate an electrical coupling partition set based on the weighted real-time electrical coupling strength and the first threshold; perform a topology search in parallel within each electrical coupling partition to obtain the topology results within the partition;

[0013] When an electrical node voltage exceeds the limit or the voltage exceedance risk indicator reaches the trigger threshold, the electrical coupling partition containing the exceeded node is identified as the partition to be coordinated. A versioned snapshot of the barrier synchronization is performed on the partition to be coordinated and a version identifier is generated. Topology search and inter-division connection branch status confirmation are completed on the data bound to the version identifier.

[0014] Step S5: Based on the topology results within the partition and the status information of the connecting branches between the partitions, perform global topology integration and output the electrical island division and energized status information of the entire network.

[0015] In some embodiments, generating the electrical coupling partition set includes:

[0016] Construct a coupled graph with electrical nodes as vertices and real-time electrical coupling strength as edge weights; retain edges with edge weights not less than the first threshold; and determine the connected components in the coupled graph as the electrical coupling partitions.

[0017] In some embodiments, the real-time electrical coupling strength is determined by voltage and power measurements, including:

[0018] A linearized sensitivity model is established based on the voltage and power measurements of the electrical nodes. The sensitivity index of the voltage of the first electrical node to the power change of the second electrical node is obtained, and the amplitude or normalized value of the sensitivity index is determined as the real-time electrical coupling strength between the two electrical nodes.

[0019] In some embodiments, it also includes:

[0020] The timestamp is used to record real-time measurement data and the real-time electrical coupling strength obtained from it;

[0021] The initial confidence level is determined based on the measurement quality code, communication link quality, and acquisition device type. The confidence index is obtained according to the time interval between the analysis time and the timestamp, following a monotonically decaying rule. The decay intensity is adaptively adjusted according to changes in link packet loss or end-to-end delay within a preset statistical window and is subject to upper and lower limits. When the initial confidence level experiences an isolated drop in a single analysis period and recovers to the original grading range in adjacent analysis periods, the initial confidence level of the previous analysis period is maintained. The initial confidence level is updated only when the drops occur consecutively.

[0022] Before generating electrical coupling partitions or performing global topology integration, the real-time electrical coupling strength or the topology results within the partition are weighted according to the confidence index. Missing or abnormal data are handled with a conservative level. If the confidence level is continuously lower than the lower limit, the partition edge is not retained or the partition is merged.

[0023] In some embodiments, when an electrical node voltage over-limit is detected or a voltage over-limit risk indicator reaches a trigger threshold, a set of over-limit nodes is determined.

[0024] For each node exceeding the limit, its associated nodes with an electrical coupling strength not lower than the second threshold are merged into the same partition to be coordinated. Synchronous snapshot acquisition is performed on the partition to be coordinated. Within the snapshot session, the measurement and switch status data are aggregated in the snapshot time window using the barrier synchronization method and a version identifier is generated. After the version is submitted, it enters a read-only state. On the data bound to the version identifier, the topology search within the partition and the status confirmation of the inter-partition connection branches are completed. The status of the connection branches is prioritized according to the same version, and multi-source consistency is decided by combining proximity priority and confidence priority. Late data is not written back to the submitted version, and when the conflict cannot be resolved, it is handled with a conservative strategy of not merging or triggering resampling. When the snapshot cannot be submitted within the barrier waiting period, the previous submitted version is rolled back or the boundary of the previous analysis cycle is maintained and a pending confirmation mark is output.

[0025] In some embodiments, batch-stream collaborative processing is also included:

[0026] The data stream formed by real-time measurement data is dynamically partitioned and topology searched within the partition according to the analysis cycle to obtain the first topology result; typical working condition coupling relationships are generated for historical batch data and pre-partitions are formed, and topology search is performed within the pre-partitions to obtain the second topology result; consistency verification is performed on the first topology result and the second topology result within a preset time window, and the final topology result is selected or merged based on the verification result.

[0027] In some embodiments, setting partition state stability conditions includes:

[0028] Calculate the degree of change in the electrical coupling partition for two adjacent analysis periods;

[0029] When the partition change rate is less than the third threshold for N consecutive analysis periods, the electrical coupling partition remains unchanged; when the partition change rate is not less than the third threshold, the electrical coupling partition is updated.

[0030] In some embodiments, the inter-regional connecting branch is a line or switch branch whose endpoints belong to different electrical coupling zones;

[0031] The global topology integration includes: when the connecting branch is in a closed state, merging the bus equivalence classes within the corresponding partition and updating the electrical island boundary; when the connecting branch is in a disconnected state, keeping the electrical island boundary of the corresponding partition from being merged.

[0032] Secondly, embodiments of the present invention provide a multi-level topology analysis system for power systems, comprising,

[0033] The data sensing and preprocessing module is used to access real-time measurement data, equipment status information and grid connection relationships and generate standardized data.

[0034] The dynamic electrical coupling analysis module is used to calculate the real-time electrical coupling strength and generate a set of electrical coupling partitions;

[0035] The parallel topology analysis engine is used to perform topology searches in parallel within each electrically coupled partition and output the topology results within the partition.

[0036] The global topology integration module is used to generate the electrical island division and energized status information of the entire network based on the topology results within the partition and the status information of the connecting branches between the partitions;

[0037] The external interface module is used to output the network topology results and their version identifiers.

[0038] In some embodiments, a topology snapshot and version management module is also included, which is used to generate a snapshot with a unique version identifier for each global topology integration; and to provide a consistent read and write interface bound to the version identifier for the dynamic electrical coupling analysis module, the parallel topology analysis engine and the global topology integration module.

[0039] Through the above technical solution, the present invention can achieve at least the following beneficial effects:

[0040] In situations where electrical coupling changes rapidly with operating conditions, causing misalignment between fixed partition boundaries and actual coupling areas, the coupling strength of electrical node pairs is calculated based on real-time measurements, and coupling partitions are generated using thresholds. This allows partition boundaries to be updated with the operating status, and the data range upon which cross-boundary collaboration depends to converge to the relevant region as the coupling relationship develops. This improves the adaptability of topology analysis boundaries to changes in operating conditions.

[0041] In scenarios such as midday data delivery, where fluctuations in measurement quality and link status affect coupled computation and topology integration due to noise, missing data, or delays, a confidence index determined by the measurement quality code, link quality, and acquisition device type is introduced. The confidence index is made to decrease monotonically with time intervals, and the decrease intensity is adaptively updated with the link statistics window. This causes expired or low-confidence data to be downweighted or removed in key criteria such as partition edge retention and link tributary merging, thereby improving the consistency and stability of topology results.

[0042] Parallel topology analysis is prone to boundary jitter and version misreading when dynamic partitions change frequently. In the case of limit-over triggering, the versioned snapshot of the partition to be coordinated containing the limit-over node is collected and a version identifier is generated. This allows the topology search and connection branch status confirmation within the partition to be completed on the same version of read-only data, thereby reducing integration conflicts and duplicate corrections caused by cross-time data misuse.

[0043] In situations where there are multiple sources reporting the status of the connecting branch and there may be conflicts, a multi-source consistency decision rule is adopted that prioritizes the same version, combined with proximity priority and confidence priority. If the conflict cannot be resolved, the processing path is either not to merge or to trigger resampling. This ensures that the electrical island boundary update remains conservative and consistent when the status is uncertain, thereby reducing the risk of boundary erroneous merging spreading to the entire network topology.

[0044] In cases where a snapshot session cannot be submitted within the barrier waiting period or where critical data coverage is insufficient, a fallback path is adopted, which rolls back to the previous submitted version, maintains the boundary of the previous analysis cycle, and outputs a pending confirmation mark. This ensures that the topology output maintains traceable and interpretable version continuity when data is unavailable, thereby avoiding a single point of data anomaly that could cause a complete process blockage.

[0045] In the analysis of real-time data streams and historical typical operating conditions, the consistency verification and fusion of real-time dynamic partitioning results and typical operating condition pre-partitioning results are performed within a time window through stream-batch collaborative processing. This allows the real-time topology to obtain the verification basis under the constraints of historical operating conditions when it is unstable in the short term, thereby improving the continuity and verifiability of the output topology.

[0046] In situations where dynamic partitioning may cause frequent repartitioning overhead and result jumps, the partitioning scheme is maintained by utilizing the partition change degree and continuous periodic stability conditions. This ensures that partition boundary updates have a stable threshold and traceable records, thereby improving the operational stability and result readability of parallel topology analysis under fluctuating conditions. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation on the scope of this application.

[0048] Figure 1 This is a flowchart of the power system multi-level topology analysis method in the embodiment.

[0049] Figure 2 This is a framework diagram of the power system multi-level topology analysis system in the embodiment. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0051] All terms used in this application (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0052] Example 1:

[0053] like Figure 1 As shown, this embodiment proposes a multi-level topology analysis method for power systems, including the following steps:

[0054] Step S1: Acquire real-time measurement data of the power grid, equipment status information, and power grid connection relationships, and record the corresponding timestamps. Real-time measurement data refers to the set of electrical measurements collected under the same time reference. The measurement set includes at least one of the electrical node voltage measurement value and branch or node power measurement value. For each measurement, record the associated timestamp, measurement source identifier, and measurement quality code. The measurement quality code is a discrete code set characterizing the validity and reliability level of the measurement. Equipment status information refers to the set of operating statuses of primary equipment. The set of operating statuses includes at least one of the circuit breaker status, disconnector status, and tie switch status. For each status, record the associated timestamp and status source identifier. Power grid connection relationships refer to the set of port connection data of primary equipment. The set of port connection data includes at least the equipment identifier, port identifier, electrical node identifier to which the port belongs, and port connection pair identifier, used to establish the association between electrical nodes and branches. Real-time measurement data, equipment status information, and grid connection relationships are written to the measurement table, status table, and connection table, respectively. The measurement table uses the electrical node identifier and timestamp as the alignment key and records the measurement source identifier and measurement quality code. The status table uses the primary equipment identifier and timestamp as the alignment key and records the status source identifier. The connection table uses the equipment identifier and port identifier as the index key and records the electrical node identifier to which the port belongs and the port connection pair identifier. When multiple records appear at the same timestamp for the same index key, a conflict flag is generated based on the consistency of the measurement quality code and the source identifier.

[0055] An electrical node refers to a computing node used in topology processing to carry out connections to measurement and equipment ports. The connection relationship between electrical nodes and primary equipment ports is determined through the power grid connection relationship. Power status information refers to a set of descriptions of the energization attributes of electrical nodes and branches. This set of descriptions includes at least whether the node is reachable from the power source, whether it is associated with an electrical island, and its corresponding version identifier.

[0056] Step S2: Calculate the real-time electrical coupling strength of electrical node pairs based on real-time measurement data. Combine the measurement quality code, communication link quality, and acquisition device type to determine a confidence index that monotonically decays with time intervals. The decay strength adaptively adjusts as the link status changes within a preset statistical window. The decay strength is a set of parameters characterizing the rate at which the confidence index decays with time intervals. The preset statistical window is a time interval used to statistically analyze the communication link packet loss rate and end-to-end delay. The acquisition device type is a set of data acquisition terminal categories that generate real-time measurement data or device status information. Real-time electrical coupling strength refers to a scalar representing the degree of correlation between voltage and power changes between two electrical nodes; this scalar is jointly determined by the real-time measurement data and the grid connection relationship. When calculating the real-time electrical coupling strength, time alignment is performed on the measurement records involved in the calculation. Time alignment includes mapping the measurement timestamps to the analysis cycle boundary and generating an aligned measurement window. The analysis cycle is the cyclic processing time granularity of the topology analysis. Missing measurements are recorded with a missing identifier, and their contribution is reduced in subsequent weighted processing. Abnormal measurements are removed or downweighted based on measurement quality codes or threshold rules. Abnormal measurements include out-of-bounds values, abrupt changes, and duplicate values. For example, the default analysis period is 1 second, adjustable from 0.2 to 10 seconds. Time alignment maps the measurement timestamp to the nearest analysis period boundary and writes it to the alignment timestamp field. The aligned measurement window uses the alignment timestamp as the window index key. Missing measurements within the window are marked with a missing identifier, and the previous valid value is retained as a fallback value. The fallback value is also marked with a fallback identifier to participate in subsequent weighting and conservative leveling.

[0057] The calculation result of real-time electrical coupling strength is represented by a set of coupling edge records. The set of coupling edge records includes at least the first electrical node identifier, the second electrical node identifier, the coupling strength value, the calculation timestamp, the identifier of the set of participating measurement records, and the result quality identifier. The result quality identifier is used to indicate whether the calculation is based on complete measurement, partial measurement, or alternative data.

[0058] Step S3: Before generating electrical coupling partitions or performing global topology integration, the real-time electrical coupling strength or topology results within the partition are weighted based on the confidence index, and abnormal or missing data are processed according to the conservative level. The conservative level processing is to place the confidence index corresponding to the missing or abnormal data in the preset low confidence level, and make the corresponding coupling edge not participate in the partition edge retention or partition merging.

[0059] Step S4: Generate an electrical coupling partition set based on the weighted real-time electrical coupling strength and the first threshold; perform a topology search in parallel within each electrical coupling partition to obtain the topology results within the partition;

[0060] When an electrical node voltage exceeds the limit or the voltage exceedance risk indicator reaches the trigger threshold, the electrical coupling partition containing the over-limit node is identified as the partition to be coordinated. A versioned snapshot of the barrier synchronization is performed on the partition to be coordinated and a version identifier is generated. The topology search and the status confirmation of the inter-division connection branch are completed on the data bound to the version identifier.

[0061] Step S5: Perform global topology integration based on the topology results within the partition and the status information of the connecting branches between partitions, and output the electrical island division and energized status information of the entire network;

[0062] In this embodiment, generating the electrical coupling partition set includes:

[0063] Construct a coupled graph with electrical nodes as vertices and real-time electrical coupling strength as edge weights; retain edges with edge weights not less than a first threshold; and determine the connected components in the coupled graph as electrical coupling partitions.

[0064] A coupled graph refers to an undirected or directed graph structure with electrical nodes as the set of vertices and coupled edge records as the set of edges. The edge set includes edge weights and edge quality identifiers. When retaining edges with weights not less than a first threshold, a hysteresis rule is applied to the first threshold. The hysteresis rule includes two levels of judgment: an entry threshold for adding edges and an exit threshold for deleting edges. When a connected component is determined as an electrically coupled partition, a partition identifier and a set of electrical nodes belonging to the partition are generated for each partition, and the partition generation timestamp and corresponding version identifier are recorded.

[0065] When an isolated electrical node occurs, it is recorded as a single-node partition, or it is incorporated into an adjacent partition with a physical connection and closed switch, based on the grid connection relationship. To limit the fluctuation of partition size, upper and lower limits are set for the number of partition members; when the number of partition members exceeds the upper limit, the set of edges within the partition is cut according to edge weight from low to high until the constraint is satisfied; when the number of partition members is below the lower limit, adjacent partitions are merged according to edge weight from high to low until the constraint is satisfied.

[0066] In this embodiment, the real-time electrical coupling strength is determined by voltage and power measurements, including:

[0067] A linearized sensitivity model is established based on the voltage and power measurements of electrical nodes. The sensitivity index of the voltage of the first electrical node to the power change of the second electrical node is obtained, and the amplitude or normalized value of the sensitivity index is determined as the real-time electrical coupling strength between the two electrical nodes.

[0068] A linearized sensitivity model refers to an approximate mapping relationship constructed around an operating point within a given analysis period, used to output an index of the response of voltage changes to power changes. The operating point is determined by aligned voltage and power measurements, including node injected power, branch power, or a combination of both. The input set of the sensitivity model includes electrical node voltage and power measurement vectors, and the output set includes an estimate of the voltage change at the target electrical node.

[0069] The sensitivity model can be obtained by combining network parameters and operating point states. Network parameters include at least one of the following: branch impedance, transformer turns ratio, and switch on / off state. Operating point states are obtained through measurement or state estimation. The sensitivity model is updated on an analysis cycle basis. When network connections or equipment status information change, a new set of valid branches participating in the calculation is generated, and the sensitivity model is updated.

[0070] When measurement sparsity makes it impossible to directly obtain sensitivity metrics from the model, a data-driven estimation rule is used to generate sensitivity indices. The training data sources for these rules include historical measurement sequences, historical topology state sequences, and historical power flow calculation results. Labeling methods include using sensitivity obtained from power flow calculations as supervised labels or using the incremental ratio within the perturbation window as self-supervised labels. Training objectives include minimizing sensitivity estimation error and minimizing sensitivity fluctuations between adjacent periods. Inference inputs include the aligned measurement window and device status information for the current period, and inference outputs include the sensitivity indices for the target electrical node pair. Update methods include periodic incremental updates or updates triggered during topology version switching.

[0071] The data-driven estimation rule maintains training data using a training sample table. This table includes at least an alignment timestamp, a subset identifier of device status information, a grid connection version identifier, an input measurement window index key, and a supervision annotation value field. Supervision annotation values ​​are obtained from power flow calculations or the increment ratio of disturbance windows and written into the annotation source identifier. The loss function categories corresponding to the training objective categories include mean squared error loss or Huber loss. During training, training parameters and model version identifiers are written into a model version table, which includes at least a model version identifier, training data version identifier, training timestamp, applicable topology version range, and inference output quality identifier. During online inference, the current aligned measurement window and device status information are used as input, and the inference output is written into the coupling edge record set, carrying the model version identifier and result quality identifier.

[0072] This embodiment also includes:

[0073] The timestamp is used to record real-time measurement data and the real-time electrical coupling strength obtained from it;

[0074] The initial confidence level is determined based on the measurement quality code, communication link quality, and acquisition device type. The confidence index is obtained according to the monotonic decay rule based on the time interval between the analysis time and the timestamp. The decay intensity is adaptively adjusted according to the changes in link packet loss or end-to-end delay within the preset statistical window and is subject to upper and lower limits. When the initial confidence level shows an isolated drop in a single analysis period and the adjacent analysis period recovers to the original range, the initial confidence level of the previous analysis period is maintained. When the drop occurs continuously, it is updated.

[0075] Before generating electrical coupling partitions or performing global topology integration, the real-time electrical coupling strength or topology results within the partition are weighted according to the confidence index. Missing or abnormal data are handled with a conservative level. If the confidence level is consistently below the lower limit, the partition edge is not retained or the partition is not merged.

[0076] The confidence index is a scalar measure of the usability of measurement records or calculation results within the current analysis period. This scalar is associated with the time interval, measurement quality code, and source stability identifier. The time interval refers to the difference between the measurement timestamp and the current analysis period's time base. Attenuation rules are represented by a rule table, which includes at least an initial confidence level, attenuation trigger conditions, attenuation step rules, and a minimum confidence lower limit identifier. Attenuation trigger conditions include at least one of the following: the time interval exceeding a threshold and the measurement quality code being an anomaly.

[0077] The rule table records a set of rule entries categorized by data item type. Data item categories include electrical node voltage measurement, power measurement, switch status, and tie branch status. Each rule entry includes at least an initial confidence mapping key, an attenuation trigger condition key, an attenuation step key, a minimum confidence lower limit key, and upper and lower limit constraint keys. Communication link quality is maintained in the link statistics table, which includes at least the link identifier, statistical window start and end timestamps, packet loss rate statistics, end-to-end delay statistics, statistical sample count, and statistical validity identifier. When a statistical validity identifier is invalid, the link statistics are written to a conservative level, and a statistical missing marker is generated.

[0078] When generating electrical coupling partitions, confidence weighting is applied to the edge weights of the coupling edge records. The weighted edge weights are used for threshold determination and connected component generation. When integrating global topology, confidence priority processing is applied to the merging decision of topology results within the partition. If there is a conflict, the status information of the connecting branch with a lower confidence level will not participate in the merging or trigger re-acquisition.

[0079] In one implementation, the confidence index is calculated using initial value weighting, time decay, and conservative saturation logic, and is used as a unified weight for the real-time electrical coupling strength before it enters the partitioning and integration criteria. The calculation of the confidence index specifically includes:

[0080] (I) Sources and Quantification of Initial Confidence Values

[0081] For each pair of electrical nodes The real-time electrical coupling strength, given its initial confidence level:

[0082] ,

[0083] in, Represents node pairs The initial confidence level, with a range of values ​​of . ; This represents the quality confidence component obtained from the measured quality code, with a value range of [value range missing]. ; This represents the link confidence component obtained from the communication link quality, with a value range of [value range missing]. ; This represents the type confidence component obtained from the type of acquisition device, with a value range of [value range missing]. Link confidence component The data is obtained by combining the packet loss rate component and the delay component within the statistical window. The packet loss rate component decreases in different tiers as the packet loss rate increases, and the delay component decreases in different tiers as the end-to-end delay increases. Take the smaller of the packet loss rate component and the delay component, and truncate the result to the range of 0 and 1. The grading boundaries are determined according to the engineering indicators of the communication link. The packet loss rate grading can be 0%–1%, 1%–5%, 5%–10%, and ≥10%, and the delay grading can be ≤50 ms, 50–200 ms, 200–500 ms, and ≥500 ms. When the packet loss rate or delay statistics are missing, the corresponding component is taken according to the conservative grading.

[0084] Measurement quality code can be mapped to Four levels; when the measurement quality code is an unrecognized code, an out-of-bounds code, or is marked as invalid, the quality confidence component will be... The value is taken from the lowest possible level or a conservative level, and truncated within the range of 0 and 1; the quality confidence component does not increase as the quality code deteriorates. Communication link quality is calculated by adjusting the packet loss rate within the statistical window and the end-to-end delay. (Components decrease as packet loss rate or latency increases); Acquisition device type is specified as synchronous measurement device / high refresh rate terminal / conventional telemetry. Tiered classification.

[0085] Direction and boundary constraints: When any of the three types of components varies, Subsequently, the value will not increase; when the quality code or link statistics are missing, the corresponding component will be set to a conservative level (e.g., 0.5) to avoid raising unreliable measurements to high confidence.

[0086] When the quality confidence component, link confidence component, or type confidence component experiences an isolated drop within a single analysis period, and the adjacent analysis period returns to the original grading range, the initial confidence value is not immediately lowered with the isolated drop, but maintains the initial confidence value of the previous analysis period; when the drop occurs consecutively for no less than two analysis periods, it is updated according to the minimum value rule to reduce the impact of short-term anomalies on the partitioning results.

[0087] (ii) Computational skeleton that decays monotonically with time interval

[0088] At the moment of analysis Calculate the decay confidence level for timestamped data:

[0089] ,

[0090] in, Indicates the analysis time. Indicates the analysis time Node pairs The confidence index, with a value range of ; Represents node pairs The initial confidence level; Represents node pairs The attenuation coefficient, with a value range of . The unit is ; This indicates the lower limit of the attenuation coefficient. This indicates the upper limit of the attenuation coefficient. Indicates the analysis time The time interval between the timestamps bound to this coupling strength, in seconds.

[0091] Monotonicity and Saturation: When it increases, Monotonically decreasing; when When a negative value occurs due to clock bounce, Treat as 0; when When the confidence level is below the lower limit threshold (e.g., 0.2–0.4), a conservative strategy is adopted, which involves truncating the data at the lower limit threshold or directly classifying it as not participating in the partitioning and not triggering the merge.

[0092] When timestamps have alignment errors or cross-device clock deviations, the time interval is processed according to a conservative rule of not less than zero, and data that is closer to the snapshot reference time is given priority in the calculation within the snapshot time window; when the time interval exceeds the upper limit of the snapshot time window, the corresponding data is treated as expired and does not participate in the partitioning or merging determination, so as to avoid false low confidence or false high confidence caused by clock errors.

[0093] The confidence threshold is determined based on the data source type and business risk; a higher value is used for scenarios triggering exceeding the limit, and a lower value is used for normal scenarios; when If the node's position falls below the lower confidence threshold for at least one consecutive analysis period, the corresponding edge is removed from the coupled graph; if the position falls below the lower confidence threshold only once, the node is removed from the coupled graph. The calculation is truncated to the lower limit threshold.

[0094] (iii) Attenuation coefficient Adaptive update

[0095] Triggering conditions: If the link packet loss rate reaches the packet loss trigger threshold or the end-to-end latency reaches the latency trigger threshold within a preset statistical window (e.g., the last 30–120 seconds), then triggering occurs. Increase; if both remain below the recovery threshold for an extended period, trigger... The thresholds are lowered. The packet loss trigger threshold is higher than the packet loss recovery threshold, and the delay trigger threshold is higher than the delay recovery threshold, to form a hysteresis interval. The trigger and recovery thresholds are determined according to the communication link level. The packet loss trigger threshold can be 5%–10%, the packet loss recovery threshold can be 2%–5%, the delay trigger threshold can be 200–500 ms, and the delay recovery threshold can be 100–200 ms. When the packet loss rate or delay increases and exceeds the corresponding trigger threshold, Take the upward adjustment direction, but not exceeding its upper limit; when both packet loss rate and latency are simultaneously below the corresponding recovery threshold and remain below for at least one statistical window, Take the downward adjustment direction, and do not go below its lower limit.

[0096] Update frequency / window: Updated in line with the calculation cycle of real-time electrical coupling strength, and the window is refreshed by scrolling.

[0097] Upper and lower limits / saturation: Limited to Inside (e.g.) Take values ​​between 0.005 and 0.02. Take values ​​between 0.05 and 0.2, in units of... Each update uses a fixed step size or tiered jumps to avoid drastic changes with a single jitter. When the packet loss rate or end-to-end latency fluctuates alternatingly above and below the trigger threshold and recovery threshold within the same statistical window, the attenuation coefficient maintains the value of the previous statistical window and is not adjusted upwards or downwards. When the trigger condition is met for at least one statistical window, a second update in the same direction is performed to reduce the amplification effect of link indicator jitter on adaptive updates.

[0098] Fixed step size press The value range is defined as 5%–20% of the interval width, or... Discretized into no fewer than 3 levels and no more than 10 levels; within each analysis period Adjustments will be made at most once; once the upper or lower limit is reached, saturation will be maintained and no further adjustments will be made. If a single peak occurs in the statistic within a window but does not persist for more than one statistical window, the statistic will be maintained as it was in the previous analysis period. Values.

[0099] Conservative strategy for abnormal operating conditions: When link statistics are unavailable or fluctuate excessively within the window, By directly selecting a larger tier and maintaining at least one window length, the confidence level decays more quickly, thereby reducing the impact of expired data on partitioning and integration. Excessive jitter is determined by the fluctuation range of packet loss rate or latency within a statistical window. When the difference between its maximum and minimum values ​​exceeds a preset proportion of the corresponding trigger threshold, or when multiple consecutive sampling points alternately cross the trigger threshold, it is considered excessive jitter and is maintained within at least one statistical window. Take the larger range. The preset ratio is set to 20%–100% relative to the trigger threshold; when the fluctuation amplitude increases and exceeds the preset ratio, it is judged as excessive jitter; when the fluctuation amplitude decreases and remains below the preset ratio for at least one consecutive statistical window, the excessive jitter judgment is lifted.

[0100] (iv) Specific operators involved in the confidence level in partitioning edge weights and merging decisions

[0101] The confidence level is used as the weight for the real-time electrical coupling strength before threshold comparison and ranking.

[0102] ,

[0103] in, This represents the weighted real-time electrical coupling strength; Indicates the confidence index at the time of analysis; This represents the unweighted real-time electrical coupling strength.

[0104] Before entering the weighted calculation, the unweighted real-time electrical coupling strength undergoes outlier suppression according to a preset statistical window. When its change relative to the previous analysis period exceeds a preset upper limit, it is weighted according to the value of the previous analysis period. The preset upper limit is calibrated based on the relative change amplitude, with 30%–80% for normal scenarios and 20%–50% for scenarios triggering exceeding the limit. When the relative change amplitude increases and exceeds the preset upper limit, a conservative processing method of weighting according to the value of the previous analysis period is triggered, and the preset upper limit is limited to the range of 0%–100%. When multiple consecutive analysis periods show changes in the same direction and no outlier judgment is triggered, it is allowed to participate in the weighting according to the real-time value to reduce the impact of noise spikes on the edge weight ranking and threshold comparison.

[0105] Weighting before threshold comparison: The edges are compared with a first threshold to determine whether to retain the coupled graph edges, thus preventing strongly coupled but outdated / low-quality edges from entering the partition.

[0106] Edge weight sorting: When it is necessary to prioritize candidate edges, sort them by weight. Descending order; when Within the gray area close to the first threshold, priority is given to selecting Those who are of higher caliber.

[0107] When a candidate edge alternates between retention and removal within the gray area, the retention status of the edge from the previous analysis cycle is maintained until the edge stably falls outside the gray area for at least two consecutive analysis cycles before being updated, in order to reduce the impact of boundary oscillations within the gray area on the partitioning results.

[0108] The gray area is calibrated according to the relative deviation from the first threshold, and the gray area range can be ±5% to ±15% of the first threshold; when When falling into the gray area, priority is given to using Higher edges are added to the retain set; when At the same time, even if it is below the confidence threshold, Edges located in the gray zone are not retained; the gray zone range is 5%~15%, used to determine the boundary uncertainty interval of the weighted real-time electrical coupling strength near the first threshold; coupling edges falling into the gray zone participate in the retention determination according to the confidence index priority. The snapshot time window is an implementation parameter, defaulting to 2s, adjustable from 1 to 4s; the snapshot time window and the barrier waiting period use the same time base and are aligned with the snapshot reference time. The confidence lower limit threshold is an implementation parameter, defaulting to 0.3, adjustable from 0.2 to 0.4; when the confidence level is continuously lower than the confidence lower limit threshold, the corresponding coupling edge is removed from the coupling graph retention set, and the corresponding connection branch does not trigger partition merging in global topology integration.

[0109] Conflict resolution priority: When a conflict occurs during partition generation or global topology integration, where the same node can be assigned to multiple partitions / multiple merge paths, the path with the higher cumulative confidence is retained first.

[0110] When the number of edges in competing paths differs significantly, the confidence level of the edge with the lowest confidence in the two paths is compared first, and the cumulative value is compared when the edges with the lowest confidence are close. When the edge with the lowest confidence is below the lower confidence threshold, the path is not used to participate in merging or attribution determination, so as to avoid bias caused by the accumulation of long paths.

[0111] If the confidence difference is insufficient (below the preset difference threshold), the partition or electrical island boundary of the previous analysis period is maintained as a conservative strategy to be consistent with the strategy of maintaining the stability condition of the partition state. The difference threshold is determined by the difference in cumulative confidence or cumulative weighted coupling strength of the candidate paths, and the difference caliber can adopt the relative difference caliber. The difference threshold is calibrated to 5%–20% based on historical operating data. When the difference is lower than the difference threshold, the partition or electrical island boundary of the previous analysis period is maintained. When the difference is higher than the difference threshold, the path with the larger difference is selected. The difference threshold is truncated within the comparable interval corresponding to the range of 0 and 1.

[0112] In this embodiment, when an electrical node voltage exceeds the limit or the voltage exceedance risk indicator reaches the trigger threshold, the set of nodes that exceed the limit is determined.

[0113] For each node exceeding the limit, its associated nodes with an electrical coupling strength not lower than the second threshold are merged into the same coordination partition. Synchronous snapshot acquisition is performed on the coordination partition. Within the snapshot session, a barrier synchronization method is used to aggregate measurement and switch status data within the snapshot time window and generate a version identifier. After version submission, it enters a read-only state. The snapshot session is a data aggregation and version submission process established around a versioned snapshot acquisition. Snapshots are written to the snapshot table with the version identifier as the key. The snapshot table includes at least the version identifier, snapshot reference time, snapshot time window identifier, identifier of the electrical node set of the coordination partition members, index key of the real-time measurement data subset within the partition, index key of the device status information subset within the partition, index key of the tie branch set, and version submission status identifier. The index key is used to reproduce the same version of the read set in read-only state. The decision result for the tie branch status confirmation is written to the decision record table. The decision record table includes at least the branch identifier, version identifier, decision conclusion, identifier of the participating source set, conflict marker, and conservative strategy marker. The decision record table serves as one of the inputs to global topology integration and is associated with the version identifier of the output topology view record set.

[0114] On the data bound to the version identifier, perform topology search within the partition and confirm the status of the connecting branches between partitions. The status of connecting branches is prioritized according to the same version, and multi-source consistency is decided by combining proximity priority and confidence priority. Late data is not written back to the committed version, and when the conflict cannot be resolved, it is handled with a conservative strategy of not merging or resampling is triggered. When a snapshot cannot be committed within the barrier waiting period, the previous committed version is rolled back or the boundary of the previous analysis cycle is maintained and a pending confirmation mark is output.

[0115] Voltage exceeding the limit refers to the state where the measured voltage value of an electrical node exceeds the preset upper voltage limit or falls below the preset lower voltage limit. The voltage exceeding the limit risk index is a scalar measure of the probability or severity of a voltage exceeding the limit. This scalar is generated jointly by the measured voltage value of the electrical node, the recent voltage change trend, and a confidence index. The recent voltage change trend is characterized by the direction and magnitude of the voltage sequence change within the aligned measurement window. The exceeding-limit node set refers to the set of electrical node identifiers that meet the exceeding-limit conditions or risk triggering conditions, and records the set generation timestamp and corresponding version identifier.

[0116] A partition to be coordinated refers to a partition obtained by expanding the set of nodes exceeding the limit. Expansion includes incorporating associated nodes with a coupling strength of not less than a second threshold with any node exceeding the limit into the same partition, and recording the coupling edge record set identifiers on which the expansion is based. A synchronization snapshot refers to a data view frozen under the same version identifier. This data view includes at least: the set of electrical nodes of the partition to be coordinated, a subset of device status information within the partition, a subset of real-time measurement data within the partition, and a set of connection branches connected to the partition to be coordinated, along with their status information. Snapshot acquisition employs consistency barrier rules, which include refusing to write to older versions after the version identifier is generated and prohibiting mixed reading of different versions before the barrier is released. Late measurements are recorded with a late identifier, and their timestamps determine whether they are incorporated into the current version or moved to the next version.

[0117] In one implementation, synchronous snapshot acquisition is performed on the partition to be coordinated, and topology search within the partition and confirmation of the status of inter-partition connection branches are completed on the data version corresponding to the snapshot. This is performed according to the steps of versioned snapshots for barrier synchronization, read-only calculation of the same version, multi-source consistency adjudication, and failure degradation, avoiding the mixing of data from different times in the same topology inference. Specifically, this includes:

[0118] 1) Engineering semantic definition and process for synchronous snapshot acquisition

[0119] The consistency semantics of a synchronous snapshot can be defined as follows: assigning the same version identifier to all data items (measurements, switch / circuit breaker status, status quantities at both ends of tie branches, and necessary connection fragments) that participate in topology calculations within the coordinating partition, and satisfying the barrier synchronization constraint: only data falling within the same snapshot time window is accepted within this version, and data outside the window or late data is not included in this version.

[0120] Version locking and two-phase commit semantics implementation:

[0121] Barrier Synchronization: When the trigger threshold is triggered and a partition to be coordinated is formed, a snapshot session is initiated within that partition, and a snapshot collection request is issued, carrying the snapshot reference time and snapshot time window width. The snapshot time window width is set with a minimum and a maximum value. The minimum value is not less than the sum of the maximum reporting latency and timestamp alignment error of the critical data source, and the maximum value is not greater than twice the reporting cycle of the critical data source or not greater than two analysis cycles. When rapid changes occur in operating conditions under the trigger scenario, the snapshot time window width is adjusted towards the minimum value.

[0122] Preparation phase: Each data source returns the latest frame of data requested, with a timestamp falling within the snapshot time window, along with its own sequence number or counter (which can be considered a vector clock component used to determine if it belongs to the same round of acquisition). The acquisition end writes this data to a temporary storage area, but does not allow the topology engine to read it.

[0123] Submission Phase: When the critical data coverage of the partition to be coordinated reaches a threshold within the preset barrier waiting time (e.g., critical switch status coverage and critical measurement coverage reach a set ratio respectively), a version identifier is generated and submitted;

[0124] In addition to meeting the critical data coverage threshold, version submissions must also ensure that there are no contradictory closure or disconnection conclusions for the critical states of the connection branches within the version. When contradictions exist and cannot be resolved according to the adjudication priority, the version identifier is not submitted and the process enters the downgrade path, or the connection branch is treated as disconnected and marked as pending review before submission, in order to reduce the risk of inconsistent data solidifying the version results.

[0125] After submission, this version enters read-only mode, and topology search and connection branch confirmation only allow reading data from this version. The barrier waiting time is configured in conjunction with the snapshot time window width. The barrier waiting time is no less than the statistical upper bound of data aggregation and database entry delays, and no greater than the snapshot time window width. If the barrier waiting time reaches its upper limit and the submission conditions are not met, the waiting time will not be extended, and a degradation path will be entered to avoid blocking subsequent analysis cycles. The critical switch status coverage threshold can be 90%–99%, and the critical measurement coverage threshold can be 70%–95%. In scenarios triggering exceeding limits, the critical switch status coverage threshold should be higher; in normal scenarios, a lower value should be used. When coverage is insufficient, a version identifier will not be submitted. Higher coverage makes triggering a submission easier, and the coverage threshold is truncated within the range of 0 and 1.

[0126] Key data coverage is calculated as the percentage of valid entries out of the total number of key data entries. Valid entries must simultaneously meet the following conditions: their timestamp falls within the snapshot time window, their quality code is not invalid, and their corresponding data source has been returned in this snapshot session. If key data coverage does not reach the threshold, a version identifier is not committed, and the process either enters a downgrade path or triggers a resampling snapshot session.

[0127] Version locking: A version identifier is bound at the start of the topology search, and cross-version readback is prohibited during the calculation process; if external data updates arrive, it only enters the preparation stage of the next version and does not write back the committed version.

[0128] 2) Snapshot Time Window and Late Data Processing

[0129] Snapshot time windows are used to constrain the freshness of data within the same topology calculation. In one implementation, a time window encompassing the envelope before and after the reference time or a time window looking back from the reference time can be used, but both must be consistent across the entire network.

[0130] Reference / calibration criteria: Coordinated partitions are typically triggered by exceeding limits. Snapshot time windows can be configured to match data types: smaller windows for synchronous measurements (such as high refresh rate voltage / phasors), and larger windows for regular telemetry / switching states; for mixed scenarios, windows covering the reporting cycle of critical states should be used, but an upper limit should be set to avoid cross-condition splicing. In engineering, the upper limit of the time window can be configured in conjunction with the barrier waiting time (the larger the time window, the higher the barrier waiting time, but both are constrained by the upper limit).

[0131] Direction and boundary constraints: A larger time window can improve coverage, but it will increase the time difference of data within the same version; a smaller time window can improve consistency, but it will increase the probability of missing data. When configuring, you can set minimum / maximum boundaries, and in scenarios that exceed the limits, you should prefer a more conservative configuration. A conservative configuration means that it is better to degrade due to missing data than to expand the time window to a range that crosses significant changes in operating conditions.

[0132] Late data rules: Data arriving after a committed version, whose timestamps should have fallen within that version's time window, is uniformly marked as late data and enters the next version candidate; no patch-style write-back is performed on the committed version. If late data pertains to a critical state of a connection branch and may change the merging conclusion, a fast resampling snapshot can be triggered to generate a new version, which will then overwrite the output. When late data only affects a single connection branch and its decision is contrary to the committed version, conservative handling is prioritized for that connection branch, i.e., it remains unmerged in the current output and retains the "awaiting review" mark, while a resampling confirmation is triggered for that connection branch in the next version; when late data simultaneously affects multiple connection branches, a fast resampling snapshot is triggered to generate a new version that overwrites the output, in order to reduce the risk of boundary erroneous merging caused by late critical states.

[0133] When late data involves switch position information or terminal charge measurement of a connection branch, and its conclusion is contrary to the corresponding conclusion in the submitted version, a fast resampling snapshot is triggered. The barrier waiting time for fast resampling is taken as the conservative upper limit of the original waiting time. If the submission conditions are still not met, the connection branch is treated as disconnected and marked as pending review until the new version is submitted in the next analysis cycle.

[0134] 3) Multi-source consistency rules for confirming the status of connecting branches

[0135] The status confirmation of the connecting branch can be limited to closure and merging only when using the same version and multiple sources are consistent; if conflicts exist, a conservative strategy is adopted not to merge, and an executable decision priority is given:

[0136] Version priority: Only data consistent with the version identifier will be accepted; cross-version data will not be included in this ruling.

[0137] Prioritize proximity: If multiple points report within the same version, prioritize the switch / circuit breaker status and endpoint measurement that is physically closer to the endpoint of the tie branch. Proximity is determined by the topological distance from the tie branch endpoint to the corresponding switch interval of the data acquisition device, with the smaller distance taking precedence; when the distances are the same, prioritize the one whose timestamp matching the version identifier is closer to the snapshot reference time.

[0138] Confidence Priority: When multiple sources conflict within the same version, the source with higher quality code, higher link stability, and more reliable acquisition device type shall be given priority; if the conflict cannot be resolved, the connection branch shall be treated as disconnected / not merged and marked as pending review.

[0139] When the mapping relationship between the data source and the primary device fails to match the grid connection segment in the current version, the data source will not be used in the nearest priority or confidence priority decision, and the connection branch will be treated as disconnected and marked as pending review until the mapping relationship is restored to consistency in subsequent versions.

[0140] Consistency Threshold: A closure determination can be set to require consistency between at least two independent sources (e.g., consistency between primary switch position information and endpoint live voltage measurement); otherwise, partition boundary merging will not be performed. This threshold remains stricter in over-limit triggering scenarios to reduce the propagation of electrical island boundary errors caused by erroneous merging. Independent sources refer to different data sources from different acquisition links or device entities, including at least two types: switch position information sources and endpoint live voltage measurement sources. When two sources originate from the same acquisition link or the same device, they are not considered independent sources and included in the consistency threshold.

[0141] 4) Downgrade path when snapshot fails

[0142] When a snapshot cannot be committed due to a timeout of the barrier or insufficient coverage of critical data, a degradation strategy without adding new modules can be adopted:

[0143] Degradation Strategy A (Weak Consistency Snapshot): For missing items, roll back to read the corresponding data from the previous committed version and mark it as a low-confidence source; during global topology integration, merging actions involving low-confidence connection branches are not performed, and only the topology results within the partition and the pending confirmation status of connection branches are output.

[0144] Degradation Strategy B (Conservative Boundary Freeze): Maintain the partition boundaries and electrical island boundaries of the previous analysis cycle, and do not merge or update areas outside the partitions to be coordinated that have exceeded the limits; resume normal process after the next cycle snapshot is successful.

[0145] Degradation Triggering and Frequency: Degradation only takes effect within the snapshot session of the partition to be coordinated and automatically exits after the next snapshot session attempt is successful. When the number of consecutive failures reaches the threshold, the partition to be coordinated can be expanded to a data source set with higher coverage or the barrier wait time can be extended, but it is still subject to the upper limit to avoid long-term blocking of the entire network output. The number threshold is calibrated to 2 to 5 times based on consecutive analysis cycles. When the number of consecutive failures reaches the threshold, the resampling frequency of the partition to be coordinated will be increased to once per analysis cycle, or the barrier wait time will be increased to its upper limit, and the above adjustments will be restored to the calibrated value after a version identifier is successfully submitted.

[0146] This embodiment also includes batch processing:

[0147] The data stream formed by real-time measurement data is dynamically partitioned and topology search within the partition is performed according to the analysis cycle to obtain the first topology result; typical working condition coupling relationship is generated for historical batch data and pre-partitioned, and topology search is performed within the pre-partition to obtain the second topology result; consistency verification is performed on the first topology result and the second topology result within a preset time window, and the final topology result is selected or merged based on the verification result.

[0148] Historical batch data refers to a collection of measurement and status sequences aggregated across multiple historical periods. This collection must include at least one of the following: voltage sequence, power sequence, equipment status sequence, or topology version sequence. Typical operating condition coupling relationships refer to a set of coupling edge records aggregated from historical data according to operating condition categories. These categories must include at least one of the following: light load with high output, peak load, fault switching, and interconnection transfer. Pre-partitioning refers to a set of partitions generated based on typical operating condition coupling relationships, with a corresponding pre-partition version identifier recorded for each operating condition category.

[0149] The first and second topology results adopt a unified representation structure, which includes at least a set of electrical islands, a set of electrical nodes within each island, a set of branches within each island, energized status information, and a version identifier. Consistency verification includes comparing the differences in the number of electrical islands, island boundaries, and critical connection branch status between the two types of topology results, and generating verification identifiers. The fusion output includes selecting the first topology result based on the verification identifier, selecting the second topology result, or generating a fused topology result. The fused topology result includes at least a list of conflict locations and conflict resolution basis identifiers.

[0150] In this embodiment, the partition state stability conditions are set, including:

[0151] For electrical coupling partitions between two adjacent analysis periods, the partition variability is calculated. Partition variability is a scalar representing the degree of difference between partition sets in adjacent analysis periods. This scalar is generated from one or a combination of partition member migration ratio, partition overlap, and partition number change. Partition member migration ratio refers to the percentage of electrical nodes migrating from the original partition to other partitions within an adjacent period; partition overlap refers to the intersection-union ratio of member sets between partitions with the same name or matching partitions in adjacent periods; partition number change refers to the difference in the number of partitions between adjacent periods. The count for N consecutive analysis periods is based on the version identifier sequence; the count is reset when the version identifiers are discontinuous. While maintaining the electrical coupling partitions unchanged, the number of maintenance periods and the partition variability sequence that triggered maintenance are recorded for subsequent verification and traceability.

[0152] When the partition change is less than the third threshold for N consecutive analysis periods, the electrical coupling partition remains unchanged; when the partition change is not less than the third threshold, the electrical coupling partition is updated.

[0153] In this embodiment, the inter-regional tie branches are line or switch branches whose endpoints belong to different electrical coupling zones; the tie branch status information refers to a set of records describing the on / off status of the tie branch, and the record set includes at least the branch identifier, endpoint electrical node identifier, on / off status, status timestamp, status source identifier, and status quality identifier. The bus equivalence class refers to the set of electrical nodes formed by connecting through closed switches under given equipment status information, and the set is used as a merging unit in global topology integration.

[0154] When multiple source status records exist for a connecting branch and conflicts occur, the status records to be integrated are selected according to the rule of prioritizing version identifier consistency and confidence index. When the conflict cannot be resolved, the corresponding connecting branch is marked as a branch to be confirmed and a snapshot is triggered for re-acquisition. Branches to be confirmed do not participate in the merging of electrical island boundaries. For cases where the status of a connecting branch is flipped, the number of flips and the flip timestamp sequence are recorded and used as one of the inputs for partition stability judgment.

[0155] Global topology integration includes: when the connecting branch is closed, merging the bus equivalence classes within the corresponding partition and updating the electrical island boundaries; when the connecting branch is open, keeping the electrical island boundaries of the corresponding partition from being merged.

[0156] Example 2:

[0157] Based on Example 1, such as Figure 2 As shown, this embodiment provides a multi-level topology analysis system for power systems, including:

[0158] The data sensing and preprocessing module is used to access real-time measurement data, equipment status information and grid connection relationships and generate standardized data.

[0159] The dynamic electrical coupling analysis module is used to calculate the real-time electrical coupling strength and generate a set of electrical coupling partitions;

[0160] The parallel topology analysis engine is used to perform topology searches in parallel within each electrically coupled partition and output the topology results within the partition.

[0161] The global topology integration module is used to generate the electrical island division and energized status information of the entire network based on the topology results within the partition and the status information of the connecting branches between the partitions;

[0162] The external interface module is used to output the network topology results and their version identifiers;

[0163] The entire network topology results are output as a topology view record set. This set includes at least: a version identifier, a generation timestamp, a set of electrical islands, a set of bus equivalence classes within islands, a set of branches within islands, energized status information, a list of tie branches, and result quality identifiers related to measurement quality. Result quality identifiers include at least a measurement coverage identifier, a status consistency identifier, and a conflict location list identifier, used to describe the usability of the topology results in the current version. The external interface output also includes incremental change records for the topology results. These records include at least the addition of electrical islands, merging of electrical islands, splitting of electrical islands, changes in branch on / off status, and changes in the energized status of electrical nodes.

[0164] This embodiment also includes a topology snapshot and version management module, which is used to generate a snapshot with a unique version identifier for each global topology integration; and to provide a consistent read and write interface bound to the version identifier for the dynamic electrical coupling analysis module, the parallel topology analysis engine and the global topology integration module.

[0165] A version identifier is a unique identifier that binds data and calculation results consistently within the same analysis period. This unique identifier is generated by a logical clock and associated with a generation timestamp. Snapshots are stored using the version identifier as the key, and the snapshot content includes at least a subset of input data, a set of partitions, topology results within each partition, and global integration results. The consistency read / write interfaces bound to the version identifier include read and write interfaces. The read interface returns the snapshot content corresponding to a given version identifier, while the write interface only allows writing to the currently active version and rejects overwriting historical versions.

[0166] When multiple processing stages concurrently access a snapshot, a version barrier rule is used to ensure that the same processing stage uses the same version identifier. If an inconsistent version identifier is detected in a processing stage, the current stage's output is terminated, and processing is restarted from the snapshot with the latest version identifier. Version reclamation is executed using retention window rules, which include at least an upper limit on the number of versions to be retained, an upper limit on the retention duration, and a mandatory retention identifier for abnormal versions.

[0167] Example 3:

[0168] This embodiment provides an application for situations where high output from distributed photovoltaic systems during midday hours leads to reverse power flow and local voltage exceedance risks in the distribution network.

[0169] The application targets distribution networks containing multiple feeders and tie branches. These networks contain numerous electrical nodes and primary equipment switches, and the network topology may change due to tie switch operations. Operational constraints include: topology analysis is performed cyclically according to an analysis cycle (default 1s, adjustable from 0.2 to 10s); in scenarios triggering exceeding limits, a snapshot time window is used for consistency aggregation (default 2s, adjustable from 1 to 4s); low-confidence data is removed using a confidence threshold (default 0.3, adjustable from 0.2 to 0.4).

[0170] In this embodiment, the system operates on a multi-level power system topology analysis system, organizing input data into measurement tables, status tables, and connection tables. Real-time measurement data originates from electrical node voltage measurements and branch or node power measurements. Equipment status information comes from circuit breaker status, disconnector status, and tie switch status. Grid connection relationships are derived from primary equipment port connection data. Measurement tables use electrical node identifiers and timestamps as alignment keys, status tables use primary equipment identifiers and timestamps as alignment keys, and connection tables use equipment identifiers and port identifiers as index keys. When multiple records appear with the same alignment key, a conflict flag is generated based on the consistency of the measurement quality code and source identifier. Timestamps use the same time base. Time alignment maps the measurement timestamps to the nearest analysis cycle boundary and writes them into the alignment timestamp field. The aligned measurement window uses the alignment timestamp as the window index key. Missing measurements are marked with a missing flag, and the previous valid value is retained as a fallback value. The fallback value is written into a fallback flag to participate in subsequent weighting and conservative level processing. The quality of the communication link is maintained in the link statistics table. The link statistics table records the start and end timestamps of the statistics window, the packet loss rate statistics, the end-to-end delay statistics, the statistical sample count, and the statistical validity identifier. When the statistical validity identifier is invalid, the link statistics are written according to the conservative level and a statistical missing marker is generated.

[0171] For example, during continuous daily operation, the system reads the measurement window and status window falling into the same aligned timestamp from the measurement table and status table in each analysis cycle, and determines the current set of valid branches in conjunction with the connection table. Then, it calculates the real-time electrical coupling strength of electrical node pairs and writes the calculation results into the coupling edge record set. The coupling edge record set carries the first electrical node identifier, the second electrical node identifier, the coupling strength value, the calculation timestamp, and the result quality identifier. The confidence index is generated within the same analysis cycle. The initial confidence value is determined by the measurement quality code, communication link quality, and acquisition device type, and is obtained according to the monotonically decaying rule based on the time interval between the analysis time and the timestamp. The decay strength is adaptively adjusted according to the packet loss rate or end-to-end delay changes in the link statistics table within the preset statistical window and is constrained by upper and lower limits. When the initial confidence value experiences an isolated drop in a single analysis cycle and recovers to the original range in adjacent analysis cycles, the initial confidence value of the previous analysis cycle is maintained. Updates are made only when consecutive drops occur. Weighting is performed before generating electrical coupling partitions or performing global topology integration. The real-time electrical coupling strength or topology results within the partition are weighted according to the confidence index. Missing or abnormal data are handled with a conservative level. When the confidence level is continuously lower than the lower confidence threshold, the corresponding coupling edge is removed from the coupling graph retention set, and the corresponding connection branch does not trigger partition merging in global topology integration.

[0172] Furthermore, the system constructs a coupled graph with electrical nodes as vertices and real-time electrical coupling strength as edge weights, retaining edges with weights not less than a first threshold. Connected components in the coupled graph are then defined as electrical coupling partitions. The gray area range of the first threshold is 5% to 15%, used to determine the boundary uncertainty interval of the weighted real-time electrical coupling strength near the first threshold. Coupled edges falling into the gray area participate in the retention determination according to the confidence index priority and maintain the retention status of the previous analysis cycle until the update conditions are met. After generating electrical coupling partitions, a topology search is performed in parallel within each electrical coupling partition to obtain the topology results within the partition. The topology results within each partition include at least the intermediate representations of the bus equivalence class and the electrical island boundaries within the partition, and carry the corresponding version identifier. Subsequently, based on the topology results within the partition and the status information of the connecting branches between partitions, global topology integration is performed, outputting the electrical island division and energized status information of the entire network. The output is in the form of a topology view record set, including version identifier, generation timestamp, electrical island set, equivalence class set of buses within the island, branch set of the island, energized status information, list of connecting branches, and result quality identifier. The external interface synchronously outputs the incremental change record of the topology results and associates it with the version identifier.

[0173] During the midday high-output photovoltaic period in this embodiment, when the measured voltage value of an electrical node exceeds the preset upper voltage limit or falls below the preset lower voltage limit, or when the voltage over-limit risk indicator reaches the trigger threshold, the system determines the set of over-limit nodes. Subsequently, for each over-limit node, its associated nodes with an electrical coupling strength not lower than the second threshold are merged into the same coordination partition, and a snapshot session is initiated within the coordination partition to perform synchronous snapshot acquisition. The snapshot session aggregates measurement and switch status data within the snapshot time window and generates a version identifier. After the version is submitted, it enters a read-only state. The snapshot is written to the snapshot table with the version identifier as the key. The snapshot table records the snapshot reference time, snapshot time window identifier, identification of the set of electrical nodes in the coordination partition, index key of the subset of real-time measurement data within the partition, index key of the subset of equipment status information within the partition, and index key of the set of connecting branches. On the data bound to the version identifier, perform intra-partition topology search and confirm the status of inter-partition connection branches. The status of connection branches is prioritized according to the same version, and multi-source consistency adjudication is performed by combining proximity priority and confidence priority. The adjudication results are written to the adjudication record table, which records the branch identifier, version identifier, adjudication conclusion, participating source set identifier, conflict flag, and conservative strategy flag, and serves as one of the inputs for global topology integration. Late data is not written back to the committed version and enters the next version candidate. When conflicts cannot be resolved, connection branches are handled according to the conservative strategy of not merging or resampling is triggered. When a snapshot cannot be submitted within the barrier waiting period, the previous committed version is rolled back or the boundary of the previous analysis cycle is maintained and a pending confirmation flag is output.

[0174] Similarly, if a short-term packet loss or end-to-end latency jitter occurs in the midday communication link, causing the valid identifier in the link statistics table to be invalid, the system will write the link statistics value to a conservative level and adaptively adjust the attenuation intensity towards a more conservative direction, causing the corresponding confidence index to decrease faster. When the confidence level is continuously lower than the lower confidence threshold, the corresponding coupled edge will not participate in the edge retention or partition merging of the coupled graph, thus maintaining conservative consistency in the partition generation and global integration process. If the measurement table has missing identifiers for several consecutive analysis periods, the system will use the previous valid value as the rollback value and write the rollback identifier. The rollback value will be used to maintain the continuity of the coupled edge record set after weighting. When the rollback continues to exceed the upper limit of the snapshot time window or triggers the conservative level processing condition, the result quality identifier of the corresponding coupled edge record set will be updated to low confidence and will be preferentially removed in the threshold determination to avoid boundary misretention caused by missing measurements.

[0175] Through the above-described process, this embodiment can generate electrical coupling partitions that vary with operating conditions, and perform global topology integration after parallel topology search within each partition, ensuring consistency between the topology analysis boundary and the electrical coupling region under midday reverse power flow. By weighting the initial value of the confidence index, applying time decay, and adaptively attenuating the link state strength, the weighted coupling strength undergoes unified weighting before the first threshold determination and partition merging criterion. This allows low-quality, late, or missing data to be downweighted or removed during partition edge retention and link tributary merging, thereby suppressing partition boundary jitter and erroneous merging caused by data fluctuations.

[0176] Furthermore, in scenarios involving voltage exceedances or risk triggering, the versioned snapshot acquisition of barrier synchronization binds data items participating in topology calculations within the partition to be coordinated to the same version identifier and enters a read-only state. Topology searches and connection branch status confirmations within the partition are completed on data of the same version, maintaining the consistency of global topology integration input. Based on the configuration of a default 2-second snapshot time window and a barrier waiting period not exceeding the snapshot time window, version updates of topology output are controlled by version submission conditions when exceedances are triggered. When data does not meet consistency conditions, recovery and continuation can be achieved by rolling back the previously submitted version or maintaining the boundary of the previous analysis cycle and outputting a confirmation flag, thereby maintaining the traceability and reproducibility of topology output.

[0177] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0178] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of this application and form different embodiments. For example, all the embodiments above can be used in any combination. The information disclosed in this background section is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.

Claims

1. A multi-level topology analysis method for power systems, characterized in that, include: Step S1: Obtain real-time measurement data of the power grid, equipment status information, and power grid connection relationships, and record the corresponding timestamps; Step S2: Calculate the real-time electrical coupling strength of the electrical node pair based on the real-time measurement data, and determine the confidence index by combining the measurement quality code, communication link quality and acquisition equipment type. The confidence index decreases monotonically with time interval, and the attenuation strength is adaptively adjusted according to the change of link status within a preset statistical window. Step S3: Before generating the electrical coupling partition, the real-time electrical coupling strength is weighted based on the confidence index, and abnormal or missing data is processed according to the conservative level. Step S4: Generate an electrical coupling partition set based on the weighted real-time electrical coupling strength and the first threshold; perform a topology search in parallel within each electrical coupling partition to obtain the topology results within the partition; When an electrical node voltage exceeds the limit or the voltage exceedance risk indicator reaches the trigger threshold, the electrical coupling partition containing the exceedance node is identified as the partition to be coordinated. A versioned snapshot of the barrier synchronization is performed on the partition to be coordinated and a version identifier is generated. Topology search and inter-division connection branch status confirmation are completed on the data bound to the version identifier. Step S5: Based on the topology results within the partition and the status information of the connecting branches between partitions, perform global topology integration and output the electrical island division and energized status information of the entire network. The generation of the electrical coupling partition set includes: constructing a coupling graph with electrical nodes as vertices and weighted real-time electrical coupling strength as edge weights; retaining edges with edge weights not less than the first threshold; and determining the connected components in the coupling graph as the electrical coupling partitions. The method also includes: recording timestamps for real-time measurement data and the real-time electrical coupling strength obtained therefrom; The initial confidence level is determined based on the measurement quality code, communication link quality, and acquisition device type. The confidence index is obtained according to the time interval between the analysis time and the timestamp, following a monotonically decaying rule. The decay intensity is adaptively adjusted according to changes in link packet loss or end-to-end delay within a preset statistical window and is subject to upper and lower limits. When the initial confidence level experiences an isolated drop in a single analysis period and recovers to the original grading range in adjacent analysis periods, the initial confidence level of the previous analysis period is maintained. The initial confidence level is updated only when the drops occur consecutively. Before generating electrical coupling partitions, the real-time electrical coupling strength is weighted according to the confidence index. Missing or abnormal data is handled with a conservative level. If the confidence level is continuously lower than the lower limit, it will not participate in the partition edge retention or partition merging. Before performing global topology integration, the topology results within a partition are weighted according to the confidence index. The status information of the connecting branch with a confidence level that is consistently below the lower limit will not trigger partition merging.

2. The power system multi-level topology analysis method as described in claim 1, characterized in that, The real-time electrical coupling strength is determined through voltage and power measurements, including: A linearized sensitivity model is established based on the voltage and power measurements of the electrical nodes. The sensitivity index of the voltage of the first electrical node to the power change of the second electrical node is obtained. The amplitude of the sensitivity index is taken, and the normalized amplitude is obtained according to the preset normalization rule. The normalized amplitude is determined as the real-time electrical coupling strength between the two electrical nodes.

3. The power system multi-level topology analysis method as described in claim 1, characterized in that, When an electrical node voltage exceeds the limit or the voltage exceedance risk indicator reaches the trigger threshold, the set of nodes that exceed the limit is determined. For each node exceeding the limit, its associated nodes with an electrical coupling strength not lower than the second threshold are merged into the same partition to be coordinated. Synchronous snapshot acquisition is performed on the partition to be coordinated. Within the snapshot session, the real-time measurement data and switch status data of the partition to be coordinated are aggregated in the snapshot time window using a barrier synchronization method and a version identifier is generated. After the version is submitted, it enters a read-only state. On the data bound to the version identifier, the topology search within the partition and the status confirmation of the inter-partition connection branches are completed. The status of the connection branches is prioritized according to the same version, and multi-source consistency is decided by combining proximity priority and confidence priority. Late data is not written back to the submitted version, and when the conflict cannot be resolved, it is handled with a conservative strategy of not merging and resampling is triggered. Before the resampling is completed, the connection branches do not participate in the partition merging. If a snapshot cannot be submitted during the barrier waiting period, roll back to the previous submitted version; if the previous submitted version is unavailable, maintain the boundary of the previous analysis cycle and output a pending confirmation flag.

4. The power system multi-level topology analysis method as described in claim 1, characterized in that, It also includes batch processing: The data stream formed by real-time measurement data is dynamically partitioned and topology search within the partition is performed according to the analysis cycle to obtain the first topology result; typical working condition coupling relationship is generated for historical batch data and pre-partitions are formed, and topology search is performed within the pre-partitions to obtain the second topology result; consistency check is performed on the first topology result and the second topology result within a preset time window, and the first topology result is output when the consistency check meets the preset consistency conditions; When the consistency check does not meet the consistency conditions, the first topology result is checked based on the second topology result and the fused topology result is output.

5. The power system multi-level topology analysis method as described in claim 1, characterized in that, Set partition state stability conditions, including: Calculate the degree of change in the electrical coupling partition for two adjacent analysis periods; When the partition change rate is less than the third threshold for N consecutive analysis periods, the electrical coupling partition remains unchanged; when the partition change rate is not less than the third threshold, the electrical coupling partition is updated.

6. The power system multi-level topology analysis method as described in claim 1, characterized in that, The inter-district connecting branch is a branch whose endpoints belong to different electrical coupling zones, and the branch includes at least one of line and switch branch; The global topology integration includes: when the connecting branch is in a closed state, merging the bus equivalence classes in the corresponding partition and updating the electrical island boundary; When the connecting branch is disconnected, the electrical island boundaries of the corresponding zone are kept separate.

7. A power system multi-level topology analysis system, based on the power system multi-level topology analysis method according to any one of claims 1 to 6, characterized in that, include: The data sensing and preprocessing module is used to access real-time measurement data, equipment status information and grid connection relationships and generate standardized data. The dynamic electrical coupling analysis module is used to calculate the real-time electrical coupling strength and generate a set of electrical coupling partitions; The parallel topology analysis engine is used to perform topology searches in parallel within each electrically coupled partition and output the topology results within the partition. The global topology integration module is used to generate the electrical island division and energized status information of the entire network based on the topology results within the partition and the status information of the connecting branches between the partitions; The external interface module is used to output the network topology results and their version identifiers.

8. The power system multi-level topology analysis system as described in claim 7, characterized in that, It also includes a topology snapshot and version management module, which generates a snapshot with a unique version identifier for each global topology integration; and provides a consistent read / write interface bound to the version identifier for the dynamic electrical coupling analysis module, the parallel topology analysis engine, and the global topology integration module.

Images