Intelligent medium voltage ring network switch cabinet emergency protection method

Abstract

The application belongs to the technical field of smart grid protection, and discloses an emergency protection method for a smart medium-voltage looped network switch cabinet, which comprises the following steps: obtaining real-time electrical state data of each switch cabinet in the looped network topology, performing multi-parameter electrical coupling analysis on the data, and obtaining a dynamic electrical correlation graph; identifying a fault source node based on the dynamic electrical correlation graph, and constructing a fault propagation path prediction matrix by analyzing the position of the fault source node and the distribution characteristics of the electrical coupling strength; calculating the breaking time sequence priority of each switch cabinet in combination with the diffusion time sequence probability and the load importance, obtaining a linkage breaking instruction sequence; and dynamically adjusting and real-time incrementally updating the mechanical response time delay and arc extinction time of the linkage breaking instruction sequence; the application significantly improves the accuracy of fault isolation and the coordination of protection actions, and provides high-quality technical support for looped network power supply protection and intelligent operation and maintenance.

Description

Technical Field

[0001] This invention relates to the field of smart grid protection technology, and more specifically, to an emergency protection method for a smart medium-voltage ring network switchgear. Background Technology

[0002] With the in-depth development of smart grid technology, medium-voltage distribution network protection is undergoing a transformation and upgrade from traditional fixed protection configurations to adaptive intelligent protection. Traditional ring network protection relies on preset values ​​and fixed logic, which faces the risks of expanding protection blind spots and frequent malfunctions against the backdrop of increasingly complex power grid structures and the large-scale integration of new energy sources. In recent years, dynamic topology sensing and artificial intelligence technologies have achieved significant breakthroughs in the field of power system protection. Combined with the multi-dimensional data analysis capabilities of deep learning, they have provided technical possibilities for intelligent protection of ring network power supply.

[0003] However, existing intelligent ring network protection technologies face multi-dimensional technical bottlenecks in complex fault identification, particularly in the issue of feature conflicts and weight mismatches encountered when processing heterogeneous measurement information during multi-parameter electrical data fusion. With the increase in the types of monitoring devices, the physical dimensions, sampling frequencies, and noise characteristics of different electrical parameters vary significantly, making traditional fixed-weight fusion strategies unable to adapt to data quality fluctuations in actual operation. In complex fault scenarios, when a measuring device malfunctions or communication is interrupted, the system still forcibly fuses low-quality data according to preset weights, leading to severe distortion in electrical condition assessment. To maintain basic protection reliability, the system is forced to rely on a single dominant parameter (usually current data), artificially ignoring complementary information from other dimensions. This simplification strategy results in significant loss of fault characteristics, especially the complete inability to capture key fault signs such as changes in the zero-sequence component of ground faults and the directional characteristics of voltage drops. Maintenance experts often point out that protection actions appear similar but lack specificity.

[0004] Meanwhile, existing methods generally employ a single-time-scale feature analysis architecture, failing to consider both transient mutations and continuous evolution across time domains. Traditional sliding window algorithms, limited by a fixed time window length, either focus on short-term mutations and lose gradual trends, or focus on long-term evolution and ignore transient features. In particular, microsecond-level fault transient processes and second-level load change trends are mixed in multi-scale signals. Existing filtering schemes use global time-domain processing, which, while suppressing noise interference, also eliminates the key transient components that reflect the essence of the fault.

[0005] Furthermore, in terms of fault propagation path modeling, existing technologies often treat single-point fault detection and network-wide impact assessment separately, lacking holistic modeling of propagation characteristics such as electrical coupling relationships and power flow transmission. This leads to unreasonable phenomena in protection strategies, such as over-cutting and power supply islanding. Existing solutions, such as adding backup protection or improving sampling accuracy, not only fail to address the fundamental problem of insufficient understanding of fault mechanisms but also introduce additional equipment investment and setting complexity, severely limiting the practical application value of the technology in the field of ring network protection.

[0006] In view of this, the present invention proposes an emergency protection method for intelligent medium-voltage ring network switchgear to solve the above problems. Summary of the Invention

[0007] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution:

[0008] An emergency protection method for intelligent medium-voltage ring network switchgear includes:

[0009] The real-time electrical status data of each switchgear in the ring network topology is obtained, including the load current, bus voltage and zero-sequence current of each switchgear.

[0010] A dynamic electrical association graph of the ring network topology is constructed based on the electrical state data. The dynamic electrical association graph uses switchgear as graph nodes and the electrical coupling strength between switchgear as the edge representation between graph nodes.

[0011] Based on the dynamic electrical correlation diagram, fault identification is performed on the switchgear to obtain the fault source node. Then, based on the location of the fault source node in the ring network topology, the distribution characteristics of electrical coupling strength, and the load distribution status of the current ring network topology, a fault propagation path prediction matrix is ​​constructed.

[0012] Based on the propagation timing probability of each propagation path in the fault propagation path prediction matrix and the load importance of each switchgear on the corresponding path, the disconnection timing priority of each switchgear is calculated; and a linkage disconnection instruction sequence is generated based on the disconnection timing priority, the linkage disconnection instruction sequence including the disconnection time and disconnection order of each switchgear.

[0013] Based on the aforementioned linkage disconnection command sequence, and combined with the mechanical response delay and arc extinction time characteristics of each switchgear, the disconnection time is dynamically adjusted to form a disconnection execution sequence; disconnection control signals are sent to the corresponding switchgear according to the aforementioned disconnection execution sequence.

[0014] During the disconnection process, the dynamic electrical association diagram and the fault propagation path prediction matrix are incrementally updated by monitoring the changes in electrical status data of each switchgear in real time, and the generation process of the linkage disconnection command sequence is re-executed based on the updated fault propagation path prediction matrix.

[0015] Furthermore, the process of constructing a dynamic electrical connection diagram includes:

[0016] The load current of each switchgear is decomposed into phasors to extract the current amplitude component and current phase component of the load current, and the absolute value of the phase difference between adjacent switchgear is calculated based on the current phase component.

[0017] Gradient analysis is performed on the bus voltage between adjacent switchgear. By performing differential operations on the voltage sampling points over time, the instantaneous and average values ​​of the voltage drop gradient are extracted. The abnormal fluctuation range of the voltage drop gradient is identified by the deviation between the instantaneous and average values.

[0018] Based on the absolute value of the phase difference and the voltage drop gradient, and using the circulating characteristics of the zero-sequence current as an auxiliary coupling parameter, the cross-correlation coefficient of the zero-sequence current between adjacent switch cabinets is calculated to identify potential ground fault propagation paths.

[0019] The absolute value of the phase difference, the abnormal fluctuation range of the voltage drop gradient, and the cross-correlation coefficient of the zero-sequence current are weighted and fused to generate the electrical coupling strength between switch cabinets, and this strength is used as the graph edge between graph nodes.

[0020] A dynamic electrical association graph is constructed based on the ring network topology and electrical coupling strength, and edges with electrical coupling strength greater than a preset coupling threshold are marked as strongly coupled edges.

[0021] Furthermore, the process of identifying the fault source node includes:

[0022] Multi-scale time window analysis was performed on the electrical status data of each switchgear, including extracting the current change amplitude characteristics in the short time window, the voltage drop duration characteristics in the medium time window, and the zero-sequence current cumulative increment characteristics in the long time window.

[0023] A dynamic baseline for the magnitude of current fluctuations is constructed based on the statistical distribution of historical electrical condition data, and a threshold adaptive judgment is performed on the extracted current fluctuation magnitude features based on this baseline to identify a set of suspected fault nodes.

[0024] The voltage drop direction is determined for the set of suspected fault nodes, and the propagation direction vector of the voltage drop is calculated by analyzing the voltage phasor angle difference between the suspected fault nodes and adjacent switch cabinets. The center node of the voltage drop is determined based on the propagation direction vector.

[0025] Zero-sequence current verification is performed on the voltage dip center node. By analyzing the correlation between the cumulative increment of zero-sequence current and the magnitude of current change, the fault type is determined. Based on the fault type determination result, the fault confidence score of the voltage dip center node is calculated.

[0026] The node with the highest fault confidence score is identified as the fault source node. The uniqueness of the fault source node is determined by the score difference between the fault source node and the second highest confidence node. When the score difference is less than a preset threshold, multiple high confidence nodes are marked as suspected multi-point fault source nodes.

[0027] Furthermore, a fault propagation path prediction matrix is ​​constructed, including:

[0028] Based on the location of the fault source node, a breadth-first traversal is performed within the dynamic electrical association graph to traverse all candidate fault propagation paths originating from the fault source node.

[0029] The electrical coupling strength of each graph edge on the candidate fault propagation path is accumulated by the path accumulation calculation, and the propagation impedance coefficient of each suspected fault propagation path is obtained based on the weighted geometric average of the electrical coupling strength of each graph edge.

[0030] The propagation impedance coefficient is subjected to time-series probability mapping. By introducing the diffusion rate parameter and path length parameter of the fault current, the estimated time delay of the fault propagation from the fault source node to the end node of the path is calculated, and the inverse of the estimated time delay is normalized to generate the diffusion time-series probability.

[0031] Based on the ratio of the load current of each switchgear to the average load current of the ring network along each suspected fault propagation path, the diffusion acceleration factor is calculated, and the diffusion timing probability is nonlinearly adjusted based on it.

[0032] By organizing all the nonlinearly adjusted diffusion time-series probabilities into a matrix form, the fault propagation path prediction matrix is ​​obtained.

[0033] Furthermore, the disconnection sequence priority of each switchgear is calculated, including:

[0034] Based on the fault propagation path prediction matrix, the diffusion time sequence probability corresponding to each switch cabinet as the target node is extracted as the fault spread probability.

[0035] The load importance of the corresponding switchgear is obtained based on the type and capacity of the load it carries.

[0036] The weighted difference between the fault spread probability and the load importance is used as the initial disconnection priority of the switchgear.

[0037] Obtain the impact of disconnecting the tie switch cabinet in the ring network topology on the reliability of the ring network power supply; and correct the initial disconnection priority of the tie switch cabinet according to the impact to obtain the final disconnection timing priority.

[0038] Furthermore, the generation of the linked breakout instruction sequence includes:

[0039] Based on the disconnection priority, the switchgear that needs to be disconnected in the ring network is arranged in descending order to obtain the preliminary disconnection sequence.

[0040] The initial disconnection sequence is verified by topology constraints to identify unreasonable disconnection combinations that lead to power islanding or load islanding, and the sequence of unreasonable disconnection combinations is adjusted.

[0041] The adjusted initial segmentation sequence is further refined by allocating segmentation times to generate a segmentation time distribution.

[0042] The adjusted initial segmentation sequence and segmentation time distribution are then encapsulated into a linked segmentation instruction sequence and subjected to timing verification.

[0043] The process of further timing verification of the linkage breakout instruction sequence includes:

[0044] A global timing check is performed on the linked interruption instruction sequence, and the total time required to complete all interruption operations is calculated. When the total time exceeds the preset protection time limit, a timing compression operation is performed.

[0045] Furthermore, the process of obtaining the segment execution timing includes:

[0046] Based on the type of operating mechanism of the switchgear, the energy storage state of the spring, and the response delay data of historical disconnection actions, a prediction model for mechanical response delay is constructed; the confidence level of the output of the prediction model is evaluated to obtain the prediction confidence level, and the adjustment range of the delay compensation is determined based on the prediction confidence level.

[0047] The arc extinction time characteristics of the switchgear are extracted from the pre-stored equipment parameter database; the real-time load current of the switchgear is obtained according to the electrical state data, and the actual arc extinction time of the switchgear under the current load is calculated by interpolation and used as an additional item for time delay compensation.

[0048] The predetermined interruption time in the linkage interruption instruction sequence is adjusted according to the determined delay compensation adjustment range and additional items to obtain the interruption execution sequence after delay compensation.

[0049] Furthermore, the dynamic electrical correlation diagram and fault propagation path prediction matrix are incrementally updated, including:

[0050] During the disconnection process, for the switchgear that has been disconnected, the voltage and current data on both sides are collected in real time to verify the disconnection effectiveness of the corresponding switchgear.

[0051] For switchgear that has been disconnected, the corresponding graph node and its associated graph edge are removed from the dynamic electrical association diagram to form an updated topology.

[0052] For the remaining graph nodes in the updated topology, the electrical coupling strength between them and their neighboring nodes is recalculated, and the fault propagation path prediction matrix is ​​locally updated based on this strength.

[0053] Based on the locally updated fault propagation path prediction matrix, the disconnection sequence priority of the remaining switchgear is reassessed. For switchgear whose priority has changed significantly, a disconnection sequence adjustment operation is performed. The disconnection sequence adjustment operation includes reordering and reallocating disconnection times. The significant change is defined as the change in priority exceeding a preset change threshold.

[0054] Furthermore, the method also includes:

[0055] In the ring network topology, multiple protection levels are preset, each corresponding to different fault severity and disconnection range;

[0056] The severity of the fault is identified based on the fault current amplitude of the fault source node and the number of high-probability paths in the fault propagation path prediction matrix.

[0057] The appropriate protection level is selected based on the severity of the identified fault; and the corresponding protection parameter set is called from the preset multi-level protection strategy library based on it.

[0058] When performing a linkage disconnection operation, the set of switchgear that needs to be disconnected is selected according to the disconnection range limit in the protection parameter set; and the current operating status of switchgear that is not within the disconnection range limit is maintained and monitoring is strengthened at the same time.

[0059] During the disconnection process, the fault isolation effect is judged by monitoring whether the zero-sequence current in the area surrounding the fault source node decreases and whether the voltage in the healthy area is stable.

[0060] When the fault isolation effect does not meet expectations, a protection level upgrade operation is performed. The protection level upgrade operation includes: expanding the disconnection range, increasing the disconnection priority threshold, and shortening the protection time limit. Based on the upgraded protection level, a new linkage disconnection instruction sequence is generated, and more stringent protection disconnection is performed on the remaining undisconnected switchgear.

[0061] The technical effects and advantages of the emergency protection method for intelligent medium-voltage ring network switchgear of the present invention are as follows:

[0062] This invention scientifically integrates heterogeneous electrical data from multiple parameters, ensuring that information from current, voltage, zero-sequence components, and phase angles truly contributes to fault characteristic representation. This significantly improves the accuracy of fault identification and the precision of isolation operations, enhancing the reliability of ring network power supply and the coordination of protection actions. Furthermore, in practical applications, even when facing complex and variable high-resistance grounding faults (such as single-phase grounding and intermittent faults), it maintains the integrity of cross-time-domain characteristics from transient changes to long-term evolution, eliminating the need for manual setting or simplification, and making the protection action sequence more intelligent and efficient. In particular, the dynamic electrical correlation diagram modeling breaks through the limitations of traditional static configuration, eliminating unreasonable phenomena such as protection dead zones and malfunctions.

[0063] In terms of data quality management, the adaptive weight fusion mechanism improves the system's robustness to fluctuations in the operating environment. When the measuring device fails or communication is interrupted, there is no longer obvious feature distortion. It performs stably in long-term operating scenarios and provides technical support for the continuous and reliable protection of complex ring network topologies.

[0064] In terms of power grid protection value, multi-scale fault identification technology has for the first time achieved intelligent differentiation between transient fault characteristics and normal disturbances, truly preserving the core electrical features that reflect the essence of the fault, enabling the protection system to accurately respond to various fault types. Combined with fault propagation path prediction matrix-driven disconnection optimization, the system can automatically correlate the propagation patterns of fault impacts, supporting quantitative evaluation of the effectiveness of disconnection strategies, and providing a scientific and precise technical tool for smart distribution network protection. Attached Figure Description

[0065] Figure 1 This is a schematic diagram of an emergency protection method for an intelligent medium-voltage ring network switchgear according to the present invention. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Example 1

[0068] Please see Figure 1 As shown in this embodiment, an emergency protection method for an intelligent medium-voltage ring network switchgear includes:

[0069] Real-time electrical status data of each switchgear in the ring network topology is acquired. This data includes key operating parameters such as load current, bus voltage, and zero-sequence current for each switchgear, collected in real-time through a distributed intelligent sensor network. Load current reflects the current load status and power transmission characteristics of the switchgear, bus voltage reflects power quality and voltage stability, and zero-sequence current is an important indicator of grounding faults and system imbalance.

[0070] A dynamic electrical correlation graph is constructed based on real-time electrical status data to represent the ring network topology. The dynamic electrical correlation graph uses switchgear as graph nodes, and the electrical coupling strength between switchgears is represented by the edges between nodes, forming a weighted directed graph structure. The electrical coupling strength comprehensively considers multiple dimensions such as current phase difference, voltage drop gradient, and zero-sequence current cross-correlation coefficient, quantifying the degree of electrical correlation between switchgears. Graph nodes carry attributes such as timestamps, location information, and real-time electrical parameters, and the weights of the graph edges dynamically reflect changes in the electrical coupling strength between nodes.

[0071] Fault identification of switchgear is performed based on dynamic electrical correlation diagrams to identify fault source nodes and construct a fault propagation path prediction matrix. Fault identification utilizes multi-scale time window analysis to extract multi-dimensional fault features such as current surges, voltage dips, and zero-sequence current anomalies. Combined with topological information from the dynamic electrical correlation diagram, the fault source node is accurately located. The fault propagation path prediction matrix is ​​constructed based on the location of the fault source node, the distribution characteristics of electrical coupling strength, and the current load distribution. Candidate propagation paths are generated through breadth-first traversal, and the propagation time-series probability of each path is calculated. This matrix comprehensively describes the possible propagation direction, speed, and impact range of the fault, providing predictive basis for subsequent fault isolation strategy formulation.

[0072] The tripping sequence priority of each switchgear is calculated based on the fault propagation path prediction matrix, and a coordinated tripping command sequence is generated. The tripping sequence priority comprehensively considers two core factors: fault spread probability and load importance. Weighted difference calculations reflect the urgency and necessity of node tripping. The coordinated tripping command sequence includes the tripping time and order of each switchgear, and is verified through topology constraints to ensure no islanding effect occurs. Optimal protection timing is achieved through fine-grained time allocation. The coordinated tripping command sequence embodies a globally optimized protection strategy, ensuring rapid fault isolation while maximizing the protection of power supply continuity in healthy areas.

[0073] Based on the linkage tripping command sequence, the tripping time is dynamically adjusted by combining the mechanical response delay and arc extinction time characteristics to form the tripping execution sequence and send control signals. Delay compensation is a key step in ensuring the accurate execution of the tripping action. The response delay of the operating mechanism is estimated through a predictive model, and the arc extinction time under the current load is obtained through interpolation calculation, and feedforward compensation is performed on the predetermined tripping time. The tripping execution sequence completes global timing verification, and when the total protection time exceeds the time limit, a timing compression operation is automatically performed. Control signals are synchronously transmitted through a high-speed communication network to ensure the time consistency of multi-point coordinated actions.

[0074] During the disconnection process, the dynamic electrical association graph and fault propagation path prediction matrix are incrementally updated by monitoring changes in electrical status data in real time. This incremental update mechanism is the core function of the adaptive adjustment of the protection strategy. It verifies the effectiveness of disconnected nodes, removes the corresponding nodes and associated edges from the graph structure, and recalculates the electrical coupling strength and propagation probability of the remaining nodes. Based on the updated matrix, the disconnection priority is reassessed, and timing adjustments are performed on nodes that have undergone significant changes. This closed-loop feedback mechanism enables the protection system to dynamically optimize its strategy based on actual execution results, effectively addressing the uncertainties brought about by fault evolution and topology changes.

[0075] In embodiments of the present invention, the detailed implementation steps for constructing a dynamic electrical correlation diagram include:

[0076] Phasor decomposition is performed on the load current of each switchgear to extract the current amplitude and phase components, and the absolute value of the phase difference between adjacent switchgear is calculated. Phasor decomposition refers to converting the time-domain waveform into a frequency-domain phasor representation through Fourier transform. The decomposition process uses a sliding window fast Fourier transform to extract the amplitude and phase information of the fundamental component and filter out harmonic interference. For a three-phase system, the current phasors of phases A, B, and C are extracted separately, and the positive-sequence components are calculated as the main analysis objects. The absolute value of the phase difference between adjacent switchgear is obtained by directly subtracting the phase angles and taking the absolute value, reflecting the phase change characteristics of the current in the transmission path; a significant increase in phase difference usually indicates the existence of impedance abrupt changes or fault points between nodes, and is an important parameter for electrical coupling analysis. This feature provides a phase-domain perspective for identifying the tightness of electrical connections and potential fault propagation paths.

[0077] Gradient analysis is performed on the bus voltage between adjacent switchgear to extract the instantaneous and average values ​​of the voltage drop gradient and identify abnormal fluctuation intervals. Gradient analysis calculates the voltage change rate through time series differencing. The analysis process first calculates the first-order difference on the sampling point sequence to obtain the instantaneous value of the voltage drop gradient; then, it calculates the average gradient using moving average filtering to reflect long-term trends; finally, it calculates the deviation between the instantaneous value and the average value, marking abnormal fluctuation intervals when the deviation exceeds a dynamic threshold. The dynamic threshold is determined based on the 3σ criterion of historical statistical data to adapt to different operating conditions. Abnormal fluctuation intervals typically correspond to transient processes such as fault occurrence, load surges, or switch actions, and are critical time windows for fault propagation path identification.

[0078] Based on phase difference and voltage drop gradient, combined with the circulation characteristics of zero-sequence current, the cross-correlation coefficient of zero-sequence current between adjacent nodes is calculated. The cross-correlation coefficient is a classic statistic for evaluating the similarity and temporal relationship between two signals; its application in zero-sequence current analysis can effectively identify the conduction path of ground faults. The calculation process first extracts the time series of zero-sequence currents from adjacent nodes and eliminates amplitude differences through normalization; then, it calculates the correlation coefficients under different time lags to form a cross-correlation function; finally, it extracts the maximum correlation coefficient and its corresponding time lag as the correlation characteristics of zero-sequence currents between nodes. A cross-correlation coefficient close to 1 indicates that the zero-sequence currents of the two nodes are highly synchronized, usually meaning they are both affected by a ground fault; a coefficient close to 0 indicates independent changes, possibly indicating they are in different fault-affected regions. This parameter, as an auxiliary coupling parameter, together with phase difference and voltage gradient, constitutes a multi-dimensional electrical correlation characteristic system.

[0079] The electrical coupling strength is generated by weighting and fusing the absolute value of the phase difference, the abnormal fluctuation range of the voltage drop gradient, and the cross-correlation coefficient of the zero-sequence current. Weighted fusion is a key step in integrating multi-source electrical characteristics, achieving a coordinated expression of different physical quantities through reasonable weight allocation. The fusion process first normalizes each feature, mapping the absolute value of the phase difference to the [0, 1] interval, obtaining the voltage gradient deviation corresponding to the abnormal fluctuation range, transforming it using the sigmoid function, and directly using the absolute value of the cross-correlation coefficient; then, weight coefficients are set according to the fault type and system characteristics; finally, the comprehensive coupling strength is calculated by weighted averaging. A larger coupling strength value indicates a stronger electrical correlation between nodes, making faults more likely to propagate between nodes.

[0080] A dynamic electrical correlation graph is constructed based on the ring network topology and electrical coupling strength, and strongly coupled edges are marked. Graph structure construction is the final step in organizing discrete nodes and coupling relationships into a network model. The construction process first establishes a node set, with each switchgear corresponding to a graph node, assigning attributes such as node number, type, location, and real-time electrical parameters. Then, graph edges are established based on topological connections, with edges only created between nodes that are physically connected. Next, the calculated electrical coupling strength is assigned as a weight to the corresponding graph edge. Finally, a coupling threshold is set, and edges exceeding the threshold are marked as strongly coupled edges, displayed in the graph visualization using different colors or line widths. Strongly coupled edges represent tightly coupled transmission paths and are high-risk channels for rapid fault propagation. The dynamic electrical correlation graph is updated in real time with electrical status data, accurately reflecting the current operating status and potential risk distribution of the ring network.

[0081] In an embodiment of the present invention, the process of obtaining the fault source node includes:

[0082] Multi-scale time window analysis was performed on the electrical status data of each switchgear to extract features such as current surge amplitude, voltage dip duration, and zero-sequence current cumulative increment. Multi-scale time window analysis is a core method for capturing fault characteristics at different time scales, comprehensively describing fault characteristics through the collaborative analysis of short, medium, and long time windows. The short time window is used to detect current surges, identifying the timing and magnitude of the surge by calculating the rate of change of amplitude between adjacent sampling points; the medium time window is used to assess voltage dips, quantifying the severity of the dip by the duration of a sustained voltage drop below a fixed percentage (e.g., 85%) of the rated voltage; and the long time window is used to accumulate zero-sequence current, obtaining the cumulative amount of zero-sequence current through integration, reflecting the sustained impact of grounding faults. The features at the three scales complement each other: short-time features reflect the instantaneous nature of the fault, medium-time features reflect the duration of the fault, and long-time features reflect the cumulative effect of the fault, together forming a multi-dimensional fault feature vector.

[0083] A dynamic baseline for current fluctuation amplitude is constructed based on the statistical distribution of historical electrical condition data, and an adaptive threshold determination is performed to identify a set of suspected fault nodes. The dynamic baseline serves as an adaptive reference standard to distinguish between normal fluctuations and abnormal fluctuations, with a reasonable threshold determined through statistical learning from historical data. The construction process first collects recent current fluctuation amplitude data, eliminating known fault periods; then, the mean and standard deviation of the corresponding current fluctuation amplitude data are calculated, and the upper bound of the confidence interval is determined as the initial baseline based on the normal distribution assumption; next, a time-varying factor is introduced, adjusting the baseline according to the daily and weekly cycle characteristics of the load curve, appropriately increasing the threshold during high-load periods and decreasing it during low-load periods; finally, a sliding window online update mechanism is used to ensure that the baseline always reflects the latest operating status. When the measured current fluctuation amplitude exceeds the dynamic baseline, the corresponding node is marked as a suspected fault node. This adaptive determination method effectively avoids the false alarms and false negatives caused by fixed thresholds, improving the accuracy and robustness of fault detection.

[0084] Voltage sag direction discrimination is performed on a set of suspected fault nodes. By analyzing the voltage phasor angle difference between nodes, the propagation direction vector is calculated, and the voltage sag center node is determined. Voltage sag direction discrimination is a key technology for accurately locating fault sources, based on the physical law that voltage phasors exhibit a radial distribution near the fault point. The discrimination process first extracts the voltage phasors of the suspected fault node and its adjacent graph nodes, and calculates the phasor angle difference between node pairs. Then, based on the sign and magnitude of the angle difference, the direction of voltage sag from the high side to the low side is determined. Next, a direction vector field is constructed, with each node corresponding to a vector pointing in the direction of voltage decrease. Finally, through vector convergence analysis, the convergence center pointed to by all direction vectors is identified, i.e., the node with the most severe voltage sag, which is determined as the voltage sag center node. The voltage sag center usually highly coincides with the location of the fault source, providing a spatial geometric criterion for fault location. This step significantly narrows the candidate range of fault sources, providing a precise target for subsequent zero-sequence current verification.

[0085] Zero-sequence current verification is performed on the voltage dip center node. Fault type is determined by correlation analysis between the cumulative increment of zero-sequence current and the amplitude of current surge, and a fault confidence score is calculated. Zero-sequence current verification is a crucial step in distinguishing between ground faults and phase-to-phase faults, based on the characteristic that ground faults generate significant zero-sequence current. The verification process first extracts the cumulative increment of zero-sequence current at the voltage dip center node and performs Pearson correlation analysis with the current surge amplitude at that node. When the correlation coefficient is significantly positive (usually greater than 0.6) and the cumulative increment of zero-sequence current exceeds a threshold, it is determined to be a single-phase ground fault. When the current surge is significant but the zero-sequence current increment is small, it is determined to be a phase-to-phase short-circuit fault. When the correlation between the two is weak and both values ​​are small, it is determined to be a suspected fault or load surge. Based on four dimensions—fault type, current surge amplitude, voltage dip depth, and zero-sequence current level—a fault confidence score is calculated by weighted summation using the following formula:

[0086] ;in, Score the confidence level of the fault. This represents the normalized amplitude of the current surge. To normalize the voltage drop depth, To normalize the zero-sequence current level, The fault type coefficient is 1 for ground faults, 0.8 for phase-to-phase faults, and 0.5 for suspected faults. , , and The weighting coefficients are used. The confidence score comprehensively reflects the reliability of fault determination and provides a quantitative basis for identifying the fault source.

[0087] The node with the highest fault confidence score is identified as the fault source node, and its uniqueness is determined by the score difference. The final determination of the fault source node is the output of fault identification, based on the ranking and difference analysis of confidence scores. The determination process first sorts all suspected fault nodes in descending order of confidence score, selecting the node with the highest score as the primary fault source candidate; then, the difference between the highest score and the second highest score is calculated. When the difference is greater than a preset threshold, the uniqueness of the fault source is considered high, and a single node is identified as the fault source node; when the difference is not greater than the threshold, it indicates that there may be multiple faults or ambiguity caused by the proximity of fault locations. In this case, the top N high-confidence nodes (N is an integer) are marked as suspected multi-point fault source nodes. For the case of a single fault source, subsequent processes predict the propagation path based on this node; for the case of multiple faults, propagation paths are constructed starting from each suspected source node, and the protection strategy is finally determined through comprehensive evaluation. This differentiated processing mechanism improves the system's adaptability to complex fault scenarios.

[0088] In an embodiment of the present invention, constructing a fault propagation path prediction matrix includes:

[0089] Based on the location of the fault source node, a breadth-first traversal is performed within the dynamic electrical association graph to generate candidate fault propagation paths. Breadth-first traversal is a classic algorithm in graph theory for systematically exploring node connectivity, simulating the step-by-step propagation process of a fault through hierarchical expansion. The traversal process starts with the fault source node, which is marked as level 0; then it expands to all neighboring nodes directly connected to the source node, marked as level 1; then it continues to expand to the unvisited neighbors of the level 1 nodes, marked as level 2; and so on until all reachable nodes are traversed or a preset maximum level is reached. Path information is recorded during each expansion, and the complete sequence of nodes from the source node to the current node constitutes a candidate propagation path. The traversal algorithm considers the directionality of graph edges, expanding only along the possible direction of current flow to avoid physically impossible paths. The final set of candidate paths comprehensively covers all areas that the fault may affect, providing a complete candidate space for subsequent path evaluation and selection.

[0090] The propagation impedance coefficient is calculated by performing path accumulation on the electrical coupling strength of each edge along the candidate propagation path. The propagation impedance coefficient is a comprehensive indicator that quantifies the propagation capability of a path, reflecting the ease with which a fault propagates along a specific path. The calculation process first extracts the electrical coupling strength of all edges along the path; then, a weighted geometric mean method is used for path accumulation, which better reflects the weakest link effect (the limiting effect of the weakest coupling link on the overall propagation capability) compared to the arithmetic mean; simultaneously, a path length correction factor is introduced to compensate for the cumulative attenuation effect of long paths. The formula for calculating the propagation impedance coefficient is:

[0091] ; Let P be the propagation impedance coefficient. Let be the electrical coupling strength of the edge along the path from graph node i to graph node j. For path length, This is the attenuation coefficient (usually 0.05-0.1). This is a length correction factor; a larger propagation impedance coefficient indicates a stronger path propagation capability, making it easier for faults to spread along that path; a smaller coefficient indicates the presence of electrical isolation or impedance barriers, limiting fault propagation. This indicator provides a quantitative basis for path priority ranking based on electrical characteristics.

[0092] A time-series probability mapping is performed on the propagation impedance coefficient, introducing parameters such as the fault current propagation rate and path length to calculate the estimated time delay and propagation time probability. This time-series probability mapping is a crucial step in transforming the static impedance coefficient into a dynamic propagation time prediction. The mapping process first calculates the physical time delay of the fault signal propagating along the path based on the fault current propagation rate (determined according to line impedance and voltage level, typically 0.1-0.9 times the speed of light). Then, considering the path length, it is converted into propagation distance, and combined with the propagation rate to obtain the propagation time. Next, the propagation impedance coefficient is used as a probability correction factor; paths with higher impedance coefficients have shorter propagation times, and paths with lower impedance coefficients have longer propagation times. Finally, through reciprocal normalization, the estimated time delay is converted into a propagation time probability; paths with shorter delays have higher probabilities, indicating they are affected by the fault more quickly. This mapping method organically combines spatial topology, electrical coupling characteristics, and physical propagation laws, realizing the transformation from structural analysis to time prediction, and providing a time-dimensional reference for fault timing arrangements.

[0093] The diffusion acceleration factor is calculated based on the ratio of the load current at each node along the path to the average load of the ring network, and a nonlinear adjustment is applied to the diffusion time series probability. The diffusion acceleration factor is a correction parameter that considers the influence of load distribution on fault propagation speed, based on the physical phenomenon that faults propagate faster in high-load areas. The calculation process first extracts the real-time load current at each node along the path and calculates the average load current of the path; then, it calculates the ratio of this average load current to the average load current of the ring network as a preliminary acceleration factor; next, it is mapped using a nonlinear function (such as the hyperbolic tangent function), significantly increasing the acceleration factor under high load and moderately decreasing it under low load, reflecting the nonlinear load influence characteristics; finally, the acceleration factor is applied as a multiplicative correction to the diffusion time series probability. The nonlinear adjustment formula is: ;

[0094] in, The adjusted path P diffusion time series probability. The initial diffusion time series probability. Let P be the average load current. This represents the average load current of the ring network. To adjust the strength coefficient, It is the hyperbolic tangent function.

[0095] The adjusted diffusion timing probability more accurately reflects the fault propagation time characteristics after considering load distribution, providing a refined probability input for fault priority calculation.

[0096] The adjusted propagation time-series probabilities of all paths are organized into a matrix to obtain the fault propagation path prediction matrix. Matrix organization is the final step in transforming discrete path information into structured data, facilitating subsequent matrix operations and algorithmic processing. The organization process adopts an extended form of the node pair adjacency matrix, where the rows and columns represent switchgear in the ring network, and the matrix element M(i,j) represents the fault propagation probability from node i to node j. For node pairs connected by multiple paths, the maximum propagation time-series probability among all paths is taken as the element value; for node pairs without propagation paths, the element value is set to 0. The diagonal elements of the matrix are set to 1, indicating that the fault ripple probability of the node itself is a fixed value. The final fault propagation path prediction matrix is ​​an N×N asymmetric matrix (N is the total number of switchgear), which fully describes the fault propagation relationship and time-series characteristics between any node pairs in the ring network, providing a global predictive view for decision-making.

[0097] In an embodiment of the present invention, the process of calculating the disconnection sequence priority of each switchgear includes:

[0098] Based on the fault propagation path prediction matrix, the propagation time-series probability of each switchgear node as the target node is extracted as the fault spillover probability. The fault spillover probability is a quantitative indicator of the likelihood of a node being affected by a fault, extracted directly from the propagation prediction matrix. The extraction process involves column-wise operations on the matrix. For target node 'a', the probability values ​​from all rows to column 'a' in the matrix are extracted, representing the probability of propagation from each source node to node 'a'. Then, the comprehensive fault spillover probability of target node 'a' is obtained through probability fusion (maximum value method or probability sum method). The maximum value method takes the maximum value of all propagation probabilities; the probability sum method uses the independent event summation formula from probability theory to reflect the comprehensive impact of multiple paths. The fault spillover probability range is [0, 1]. A larger value indicates that the node is more likely to be affected by the fault, and the interruption requirement is more urgent. This indicator provides a risk dimension input for interruption priority calculation.

[0099] The load importance is calculated based on the load type and capacity carried by the switchgear. Load importance is an evaluation index of the power supply value of a node, reflecting the importance of the node to system security and user services. The calculation process first classifies the loads according to their type, including primary loads (such as hospitals and data centers), secondary loads (such as commercial users), and tertiary loads (such as general industrial loads), and assigns them corresponding load importance coefficients; then, it is weighted according to the load capacity, with nodes with larger capacities having higher importance; finally, the importance is mapped to the [0, 1] interval through normalization. Nodes with high load importance should be kept powered as much as possible, and their disconnection priority is relatively low; nodes with low importance can be disconnected first to protect critical loads. This index provides a value dimension input for the disconnection priority calculation.

[0100] The weighted difference between the probability of fault spread and the importance of the load is used as the initial disconnection priority. Weighted difference calculation is an optimization decision-making method that balances risk and value, achieving the strategic goal of prioritizing high-risk, low-value disconnections through difference operations; a positive priority value indicates a higher priority for disconnection, while a negative value indicates that power supply should be maintained. This weighted difference mechanism achieves the dual objectives of minimizing risk and loss in the protection strategy.

[0101] The impact of disconnecting tie switchgear on the reliability of the ring network power supply is assessed, and its initial disconnection priority is adjusted. Tie switchgear is a critical node in the ring network topology, and its disconnection status directly affects the network's connectivity and the flexibility of power supply paths. Impact assessment employs topology connectivity analysis, simulating changes in connectivity components after tie switch disconnection to calculate the number of affected nodes and load capacity. Tie switches with larger impact ranges have higher impact and their disconnection priority should be reduced to maintain topology flexibility. The adjustment process reduces the initial priority of tie switches based on the impact, with the adjustment magnitude proportional to the impact. The final disconnection sequence priority, after adjustment, considers both the risk and value characteristics of the nodes themselves and the global impact at the topology level, ensuring the systematic and rational nature of the disconnection strategy.

[0102] In an embodiment of the present invention, the generation of the linkage breakout instruction sequence includes:

[0103] The nodes to be dismantled are sorted in descending order according to the dismantling sequence priority to obtain a preliminary dismantling order. The sorting process first selects nodes with positive priority values, representing the set of nodes to be dismantled; then, they are sorted from high to low priority values ​​to form a linear dismantling order list; nodes with the same priority are sorted secondary based on the probability of fault spread.

[0104] The initial disconnection sequence is verified for topological constraints to identify unreasonable disconnection combinations and adjust the sequence. Topological constraint verification is a crucial step to ensure that disconnection operations do not create islanding effects. The verification process uses a graph connectivity analysis algorithm to simulate the topological changes after disconnecting nodes sequentially according to the initial order. Connectivity component detection is performed on the graph after each disconnection to identify whether isolated power nodes (power islands) or load nodes (load islands) have been created. When islanding risk is detected, the current disconnection combination is marked as unreasonable, and a sequence adjustment operation is performed. Adjustment strategies include: swapping the disconnection order of adjacent nodes, delaying the disconnection time of critical tie switches, or inserting topology reconstruction operations (such as closing backup tie switches) into the disconnection sequence. The adjusted sequence ensures that basic topological connectivity is maintained for each disconnection operation, preventing protection actions from triggering new power outages.

[0105] The adjusted initial disconnection sequence is then finely allocated to determine the disconnection time distribution, generating a disconnection time distribution. Time allocation is a crucial step in transforming a discrete sequence into a continuous time series. The allocation process first determines the protection start time (usually the time when fault identification is completed); then, according to the disconnection sequence, a relative time is allocated to each node, considering the minimum time interval for disconnection actions; next, the times are fine-tuned based on the fault sweep probability and estimated delay of each node, with high-sweep-probability nodes disconnected earlier and low-sweep-probability nodes disconnected later; finally, a complete time series is formed, with each node corresponding to a precise disconnection time. This disconnection time distribution ensures both the speed of protection and avoids system oscillations caused by excessive concentration, achieving optimized configuration in the time dimension.

[0106] The adjusted disconnection sequence and disconnection time distribution are encapsulated into a coordinated disconnection instruction sequence. Instruction sequence encapsulation is the final step in generating executable commands. The encapsulation process organizes information such as the node number, disconnection time, disconnection type (three-phase or single-phase disconnection), and confirmation requirements of each node to be disconnected into structured instruction frames. These instruction frames are arranged in chronological order to form an instruction sequence. Simultaneously, checksums, timestamps, and priority identifiers are added to the instruction sequence. Finally, a coordinated disconnection instruction sequence conforming to the communication protocol is formed, which can be directly sent to the intelligent controllers of each switchgear for execution. The instruction sequence embodies a globally optimized protection strategy and serves as the direct output and execution basis for the system's intelligent decision-making.

[0107] In an embodiment of the present invention, the process of obtaining the segmented execution timing includes:

[0108] Based on the type of operating mechanism, spring energy storage status, and historical response delay data of the switchgear, a mechanical response delay prediction model is constructed. Mechanical response delay is the time delay from issuing a disconnection command to the actual contact separation, and is significantly affected by the characteristics of the operating mechanism and its current state. The prediction model adopts a data-driven approach, training a neural network or regression model based on historical execution data. Input features include operating mechanism type (spring mechanism, permanent magnet mechanism, etc.), spring energy storage status (energy storage voltage or position), ambient temperature, and historical average delay, etc.; the output is the predicted response delay value. Model training uses recently accumulated actual execution data, optimizing parameters through supervised learning to achieve adaptation to individual differences and aging effects. The prediction results are evaluated for confidence level; confidence intervals are calculated based on the historical distribution of prediction errors. Predictions with high confidence levels use a smaller safety margin, while those with low confidence levels use a larger margin. The adjustment range of delay compensation is determined based on the confidence level to ensure that compensation is both sufficient and not excessive, improving the accuracy of disconnection timing control.

[0109] The arc extinction time characteristic is extracted from the equipment parameter database, and the actual arc extinction time is calculated based on real-time load current interpolation. The arc extinction time is the time it takes for the arc to completely extinguish and achieve electrical isolation after contact separation, and it is closely related to the magnitude of the breaking current. The extraction process reads the arc extinction time characteristic curve of the switchgear from the pre-stored equipment parameter database. The arc extinction time characteristic curve describes the extinction time under different breaking currents. Then, the real-time load current value of the switchgear is obtained as the expected breaking current. Next, the arc extinction time corresponding to the current load is calculated on the characteristic curve using linear interpolation or spline interpolation algorithms. Finally, this is used as an additional item for time delay compensation. This additional item ensures that the setting of the breaking time fully considers the physical process of the arc and avoids coordination failure caused by subsequent operations before the arc is extinguished.

[0110] Based on the adjustment range and additional items of the delay compensation, the predetermined breaking times in the linkage breaking command sequence are adjusted to obtain the breaking execution timing. Timing adjustment is the final step in achieving precise control. The adjustment process performs feedforward compensation on the predetermined breaking times of each node in the command sequence, deducting the mechanical response delay and arc extinction time from the breaking times in advance to obtain the actual command transmission time; at the same time, considering communication delay, the command transmission time is further advanced; after comprehensive compensation, the breaking execution timing is formed, which ensures that each node completes actual electrical isolation at the expected time, achieving precise coordination of multi-point breaking.

[0111] In an embodiment of the present invention, the process of performing timing verification includes:

[0112] A global timing verification is performed on the linkage tripping command sequence. The tripping time of all tripping nodes in the sequence is extracted, the total time required to complete all tripping operations is calculated, and compared with the preset protection time limit. When the total time exceeds the protection time limit, a timing compression operation is performed. The verification process first extracts the predetermined tripping time of each node to be tripped from the linkage tripping command sequence, identifies the earliest and latest tripping times in the sequence, and the difference between the two is the time span of the tripping operation. Then, the arc extinction time and mechanical response delay of the last tripping node are extracted as additional time, and the time span is added to the additional time to obtain the total time. Next, the preset protection time limit is determined according to the fault type, fault current amplitude, and ring network protection configuration requirements. Finally, the total time is compared with the protection time limit, and a timing compression operation is triggered when the total time exceeds the time limit. The timing compression operation includes two collaborative strategies: First, the timing interval of the breakout is compressed by traversing the instruction sequence to calculate the timing interval of adjacent breakout nodes, identifying compressible intervals that are greater than the minimum safe interval, and reducing the interval through proportional compression or priority compression strategies to ensure that the compressed interval is not lower than the safe lower limit. The breakout times of each node are then redistributed from the start of the sequence. Second, the parallel breakout group is identified and organized by calculating the electrical isolation between nodes to be broken out based on a dynamic electrical correlation diagram. This indicator comprehensively considers topological distance and electrical coupling strength.

[0113] When the electrical isolation is greater than a preset threshold (usually 3-5), the nodes are determined to be electrically independent. An independence graph is constructed, and a graph coloring algorithm is used to organize the mutually independent nodes into parallel disconnection groups. Nodes within a group can perform disconnection operations synchronously at the same time. The capacity of the parallel disconnection groups is verified to assess the energy storage capacity requirements of the operating mechanism and the short-term current withstand capability of the system, ensuring that parallel disconnection will not cause voltage drops or equipment overload. Finally, based on the compressed time interval and the organization result of the parallel disconnection groups, the linkage disconnection command sequence is regenerated, and the linear disconnection sequence is converted into a grouped sequence. The timing between groups is maintained according to the priority relationship and the compressed interval is applied. Nodes within a group share the same time, forming a time-compressed disconnection execution timing sequence.

[0114] In an embodiment of the present invention, the incremental update process includes:

[0115] During the disconnection process, for the switchgear that has been disconnected, voltage and current data on both sides are collected in real time to verify the effectiveness of the disconnection. The verification process continuously collects voltage and current data on both sides of the switchgear within a set time window after the disconnection command is executed. Criteria for effective disconnection include: a significant voltage difference between the two sides (e.g., greater than 10% of the rated voltage), current dropping to near zero, and the disappearance or significant reduction of zero-sequence current (for ground faults). When these conditions are met simultaneously, the disconnection is confirmed as effective, and the node is successfully isolated; otherwise, it is marked as a disconnection failure, triggering the backup protection strategy or alarm mechanism. Effectiveness verification ensures the accuracy of subsequent diagram structure updates, avoiding decision adjustments based on erroneous assumptions.

[0116] For nodes where the breakout is effective, the corresponding graph node and its associated edges are removed from the dynamic electrical association graph, forming an updated topology. The update operation first deletes the node object of the broken node in the graph data structure, releasing related resources; then, it iterates through all associated edges of the node, removing these edges from the edge set, and simultaneously updates the adjacency lists of adjacent nodes, deleting references pointing to the broken node; finally, it recalculates the graph's connectivity and key topology parameters, such as node degree distribution, shortest path, and connected components. The updated topology accurately reflects the actual state of the ring network after the breakout operation, providing a correct structural basis for re-evaluating the relationships of the remaining nodes.

[0117] For the remaining graph nodes in the updated topology, the electrical coupling strength between them and their neighboring nodes is recalculated and locally updated. The recalculation process involves re-extracting the latest electrical state data for the remaining nodes adjacent to the broken nodes. Since topology changes may alter the load current and voltage levels of nodes, the corresponding phase differences and voltage gradients also change. Based on the updated electrical data, the electrical coupling strength is recalculated, and the weights of the corresponding graph edges are updated. For regions far from the broken nodes, if the electrical state changes are not significant, the original coupling strength values ​​can be maintained to reduce computational burden. This local update strategy balances accuracy and real-time performance, ensuring that the graph model always reflects the current electrical interconnection state.

[0118] Based on the locally updated fault propagation path prediction matrix, the timing priority of the remaining nodes is reassessed, and timing adjustments are performed on nodes with significant changes. Priority reassessment is the core mechanism for adaptively adjusting the fault propagation strategy. The assessment process first re-executes fault propagation path prediction based on the updated graph structure and electrical coupling strength, generating a new propagation matrix; then, the fault sweep probability of the remaining nodes to be terminated is extracted and compared with the sweep probability used in the original priority calculation; when the change exceeds a preset significance threshold, it is marked as a significantly changed node; the termination priority of these nodes is recalculated, which may lead to a rearrangement of the termination order and a redistribution of termination times. The timing adjustment operation includes updating the termination times of the remaining nodes in the instruction sequence, and, if necessary, re-verifying the topology constraints to ensure that the adjusted sequence still meets safety requirements. This dynamic adjustment mechanism enables the protection system to continuously optimize its strategy based on actual performance and environmental changes, improving the adaptability and effectiveness of the protection.

[0119] In embodiments of the invention, the method further includes:

[0120] Multiple protection levels are preset in the ring network topology, each corresponding to different fault severity and coverage areas. Multi-level protection is a hierarchical strategy framework for responding to different risk levels. The preset process designs a three-level protection system based on the ring network's importance, load characteristics, and historical fault data: Level 1 protection corresponds to minor faults, with coverage limited to the fault source and its directly adjacent nodes; Level 2 protection corresponds to moderate faults, extending coverage to all nodes with a propagation probability greater than a threshold; and Level 3 protection corresponds to severe faults, covering the entire affected area or even the entire ring network. Each protection level has a preset set of protection parameters, including key parameters such as coverage area limits, coverage priority thresholds, protection time limits, and the number of parallel coverage groups. Parameter settings are based on simulation analysis and field experience to ensure that the protection strategies at different levels are both effective and economical, avoiding over-protection and under-protection.

[0121] The severity of a fault is identified based on the fault current amplitude of the fault source node and the number of high-probability propagation paths, and the corresponding protection level is selected. The identification process first extracts the fault current amplitude of the fault source node and compares it with a preset severity threshold; a larger current amplitude indicates a more severe fault. Then, the number of high-probability paths in the fault propagation path prediction matrix with a propagation time-series probability greater than a preset probability threshold is counted; a larger number indicates a wider impact range. Finally, the fault severity level is determined by a comprehensive score of the current amplitude and the number of paths. The scoring method combines the normalized values ​​and weighting coefficients of the two factors, mapping the scoring interval to the corresponding protection level. Based on the selected protection level, the corresponding protection parameter set is retrieved from a preset multi-level protection strategy library to provide parameter basis for subsequent tripping decisions.

[0122] Based on the disconnection range defined in the protection parameter set, a set of switchgear that needs to be disconnected is selected, and nodes outside the range are subject to enhanced monitoring. Node selection is an execution step for implementing differentiated strategies according to protection levels. The selection process, based on the disconnection range definition parameters, selects a subset of nodes within the range from all candidate nodes to be disconnected; the disconnection range can be defined based on geographical distance, electrical distance, or propagation probability thresholds; the selected set of nodes serves as the target for the actual disconnection operation, generating a corresponding sequence of linkage disconnection commands.

[0123] For nodes outside the scope of the power outage, the system maintains its current operating status but increases monitoring frequency and alarm sensitivity, along with data sampling rate and anomaly detection intensity, to ensure timely detection of potential signs of fault propagation. This tiered approach avoids unnecessary power outages while maintaining vigilance against potential risks.

[0124] During the disconnection process, the fault isolation effect is judged by monitoring the decrease in zero-sequence current around the fault source and the voltage stability of the healthy area. The isolation effect judgment is a real-time monitoring mechanism for evaluating the effectiveness of protection actions. The monitoring process continuously collects the zero-sequence current of the fault source node and its surrounding area, comparing the changes before and after the disconnection. Effective isolation should result in a significant decrease in zero-sequence current, indicating that the ground fault is under control. Simultaneously, the bus voltage of the healthy area (the undisconnected area) is monitored; effective isolation should restore the voltage of the healthy area to its normal range and remain stable. When the monitoring results simultaneously meet both the conditions of decreased zero-sequence current and stable voltage, the fault isolation effect is determined to have met expectations; otherwise, the effect is determined to have failed to meet expectations, and a protection level upgrade operation is required.

[0125] When the fault isolation effect fails to meet expectations, a protection level upgrade operation is executed to expand the disconnection range, increase the priority threshold, and shorten the protection time. Level upgrade is an emergency response mechanism to address protection failure or fault propagation. The upgrade operation first raises the current protection level by one or more levels, determining the upgrade magnitude based on the degree of deviation in isolation effectiveness; then, it calls upon a higher-level protection parameter set to expand the disconnection range to cover more potentially affected nodes, increase the disconnection priority threshold to allow more nodes to enter the disconnection sequence, and shorten the protection time to accelerate the disconnection execution speed; based on the upgraded protection level and parameters, the complete process of fault propagation path prediction, disconnection priority calculation, and linkage disconnection command sequence generation is re-executed; finally, more stringent protection disconnection is performed on the remaining undisconnected switchgear, and, if necessary, full-ring network disconnection is used to isolate the fault. The level upgrade mechanism ensures that the system can quickly respond to the upgrade strategy when initial protection fails, minimizing the scope of fault impact and ensuring power grid safety.

[0126] This invention achieves end-to-end emergency protection for intelligent medium-voltage ring network switchgear through real-time data acquisition, dynamic electrical correlation diagram construction, accurate fault source identification, propagation path prediction, disconnection timing optimization, and incremental adaptive updates. The graph-based modeling method of this invention can accurately locate fault sources and propagation paths. The multi-dimensional optimized linkage disconnection strategy quickly isolates faults while maximizing the protection of intact areas. Incremental updates and multi-level protection mechanisms significantly improve the system's adaptability to complex fault scenarios, providing an intelligent protection solution for the safe and reliable operation of medium-voltage switchgear.

[0127] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0128] All formulas in this manual are dimensionless and calculated numerically. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0129] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. An emergency protection method for intelligent medium-voltage ring network switchgear, characterized in that, include: The real-time electrical status data of each switchgear in the ring network topology is obtained, including the load current, bus voltage and zero-sequence current of each switchgear. A dynamic electrical association graph of the ring network topology is constructed based on the electrical state data. The dynamic electrical association graph uses switchgear as graph nodes and the electrical coupling strength between switchgear as the edge representation between graph nodes. Based on the dynamic electrical correlation diagram, fault identification is performed on the switchgear to obtain the fault source node. Then, based on the location of the fault source node in the ring network topology, the distribution characteristics of electrical coupling strength, and the load distribution status of the current ring network topology, a fault propagation path prediction matrix is ​​constructed. Based on the propagation timing probability of each propagation path in the fault propagation path prediction matrix and the load importance of each switchgear on the corresponding path, the disconnection timing priority of each switchgear is calculated; and a linkage disconnection instruction sequence is generated based on the disconnection timing priority, the linkage disconnection instruction sequence including the disconnection time and disconnection order of each switchgear. Based on the aforementioned linkage disconnection command sequence, and combined with the mechanical response delay and arc extinction time characteristics of each switchgear, the disconnection time is dynamically adjusted to form a disconnection execution sequence; disconnection control signals are sent to the corresponding switchgear according to the aforementioned disconnection execution sequence. During the disconnection process, the dynamic electrical association diagram and the fault propagation path prediction matrix are incrementally updated by monitoring the changes in electrical status data of each switchgear in real time, and the generation process of the linkage disconnection command sequence is re-executed based on the updated fault propagation path prediction matrix.

2. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 1, characterized in that, The process of constructing a dynamic electrical connection diagram includes: The load current of each switchgear is decomposed into phasors to extract the current amplitude component and current phase component of the load current, and the absolute value of the phase difference between adjacent switchgear is calculated based on the current phase component. Gradient analysis is performed on the bus voltage between adjacent switchgear. By performing differential operations on the voltage sampling points over time, the instantaneous and average values ​​of the voltage drop gradient are extracted. The abnormal fluctuation range of the voltage drop gradient is identified by the deviation between the instantaneous and average values. Based on the absolute value of the phase difference and the voltage drop gradient, and using the circulating characteristics of the zero-sequence current as an auxiliary coupling parameter, the cross-correlation coefficient of the zero-sequence current between adjacent switch cabinets is calculated to identify potential ground fault propagation paths. The absolute value of the phase difference, the abnormal fluctuation range of the voltage drop gradient, and the cross-correlation coefficient of the zero-sequence current are weighted and fused to generate the electrical coupling strength between switch cabinets, and this strength is used as the graph edge between graph nodes. A dynamic electrical association graph is constructed based on the ring network topology and electrical coupling strength, and edges with electrical coupling strength greater than a preset coupling threshold are marked as strongly coupled edges.

3. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 2, characterized in that, The process of identifying the fault source node includes: Multi-scale time window analysis was performed on the electrical status data of each switchgear, including extracting the current change amplitude characteristics in the short time window, the voltage drop duration characteristics in the medium time window, and the zero-sequence current cumulative increment characteristics in the long time window. A dynamic baseline for the magnitude of current fluctuations is constructed based on the statistical distribution of historical electrical condition data, and a threshold adaptive judgment is performed on the extracted current fluctuation magnitude features based on this baseline to identify a set of suspected fault nodes. The voltage drop direction is determined for the set of suspected fault nodes, and the propagation direction vector of the voltage drop is calculated by analyzing the voltage phasor angle difference between the suspected fault nodes and adjacent switch cabinets. The center node of the voltage drop is determined based on the propagation direction vector. Zero-sequence current verification is performed on the voltage dip center node. By analyzing the correlation between the cumulative increment of zero-sequence current and the magnitude of current change, the fault type is determined. Based on the fault type determination result, the fault confidence score of the voltage dip center node is calculated. The node with the highest fault confidence score is identified as the fault source node. The uniqueness of the fault source node is determined by the score difference between the fault source node and the second highest confidence node. When the score difference is less than a preset threshold, multiple high confidence nodes are marked as suspected multi-point fault source nodes.

4. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 3, characterized in that, Construct a fault propagation path prediction matrix, including: Based on the location of the fault source node, a breadth-first traversal is performed within the dynamic electrical association graph to traverse all candidate fault propagation paths originating from the fault source node. The electrical coupling strength of each graph edge on the candidate fault propagation path is accumulated by the path accumulation calculation, and the propagation impedance coefficient of each suspected fault propagation path is obtained based on the weighted geometric average of the electrical coupling strength of each graph edge. The propagation impedance coefficient is subjected to time-series probability mapping. By introducing the diffusion rate parameter and path length parameter of the fault current, the estimated time delay of the fault propagation from the fault source node to the end node of the path is calculated, and the inverse of the estimated time delay is normalized to generate the diffusion time-series probability. Based on the ratio of the load current of each switchgear to the average load current of the ring network along each suspected fault propagation path, the diffusion acceleration factor is calculated, and the diffusion timing probability is nonlinearly adjusted based on it. By organizing all the nonlinearly adjusted diffusion time-series probabilities into a matrix form, the fault propagation path prediction matrix is ​​obtained.

5. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 4, characterized in that, Calculate the disconnection sequence priority for each switchgear, including: Based on the fault propagation path prediction matrix, the diffusion time sequence probability corresponding to each switch cabinet as the target node is extracted as the fault spread probability. The load importance of the corresponding switchgear is obtained based on the type and capacity of the load it carries. The weighted difference between the fault spread probability and the load importance is used as the initial disconnection priority of the switchgear. Obtain the impact of disconnecting the tie switch cabinet in the ring network topology on the reliability of the ring network power supply; and correct the initial disconnection priority of the tie switch cabinet according to the impact to obtain the final disconnection timing priority.

6. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 5, characterized in that, The generation of the linked breakout instruction sequence includes: Based on the disconnection priority, the switchgear that needs to be disconnected in the ring network is arranged in descending order to obtain the preliminary disconnection sequence. The initial disconnection sequence is verified by topology constraints to identify unreasonable disconnection combinations that lead to power islanding or load islanding, and the sequence of unreasonable disconnection combinations is adjusted. The adjusted initial segmentation sequence is further refined by allocating segmentation times to generate a segmentation time distribution. The adjusted initial segmentation sequence and segmentation time distribution are then encapsulated into a linked segmentation instruction sequence and subjected to timing verification.

7. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 6, characterized in that, The process of timing verification of the linkage breakout instruction sequence includes: A global timing check is performed on the linked interruption instruction sequence, and the total time required to complete all interruption operations is calculated. When the total time exceeds the preset protection time limit, a timing compression operation is performed.

8. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 6, characterized in that, The process of obtaining the segment execution timing includes: Based on the type of operating mechanism of the switchgear, the energy storage state of the spring, and the response delay data of historical disconnection actions, a prediction model for mechanical response delay is constructed; the confidence level of the output of the prediction model is evaluated to obtain the prediction confidence level, and the adjustment range of the delay compensation is determined based on the prediction confidence level. The arc extinction time characteristics of the switchgear are extracted from the pre-stored equipment parameter database; the real-time load current of the switchgear is obtained according to the electrical state data, and the actual arc extinction time of the switchgear under the current load is calculated by interpolation and used as an additional item for time delay compensation. The predetermined interruption time in the linkage interruption instruction sequence is adjusted according to the determined delay compensation adjustment range and additional items to obtain the interruption execution sequence after delay compensation.

9. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 8, characterized in that, Incremental updates are performed on the dynamic electrical correlation diagram and fault propagation path prediction matrix, including: During the disconnection process, for the switchgear that has been disconnected, the voltage and current data on both sides are collected in real time to verify the disconnection effectiveness of the corresponding switchgear. For switchgear that has been disconnected, the corresponding graph node and its associated graph edge are removed from the dynamic electrical association diagram to form an updated topology. For the remaining graph nodes in the updated topology, the electrical coupling strength between them and their neighboring nodes is recalculated, and the fault propagation path prediction matrix is ​​locally updated based on this strength. Based on the locally updated fault propagation path prediction matrix, the disconnection sequence priority of the remaining switchgear is reassessed. For switchgear whose priority has changed significantly, a disconnection sequence adjustment operation is performed. The disconnection sequence adjustment operation includes reordering and reallocating disconnection times. The significant change is defined as the change in priority exceeding a preset change threshold.

10. The emergency protection method for intelligent medium-voltage ring network switchgear according to claim 1, characterized in that, The method further includes: In the ring network topology, multiple protection levels are preset, each corresponding to different fault severity and disconnection range; The severity of the fault is identified based on the fault current amplitude of the fault source node and the number of high-probability paths in the fault propagation path prediction matrix. The appropriate protection level is selected based on the severity of the identified fault; and the corresponding protection parameter set is called from the preset multi-level protection strategy library based on it. When performing a linkage disconnection operation, the set of switchgear that needs to be disconnected is selected according to the disconnection range limit in the protection parameter set; and the current operating status of switchgear that is not within the disconnection range limit is maintained and monitoring is strengthened at the same time. During the disconnection process, the fault isolation effect is determined by monitoring whether the zero-sequence current in the area surrounding the fault source node decreases and whether the voltage in the healthy area is stable. When the fault isolation effect does not meet expectations, a protection level upgrade operation is performed. The protection level upgrade operation includes: expanding the disconnection range, increasing the disconnection priority threshold, and shortening the protection time limit. Based on the upgraded protection level, a new linkage disconnection instruction sequence is generated, and more stringent protection disconnection is performed on the remaining undisconnected switchgear.

Images